You are on page 1of 456


Arbeiten 210
Herausgegeben von Hans Altmann, Herbert E. Brekle, Hans Jürgen Heringer,
Christian Rohrer, Heinz Vater und Otmar Werner

Brought to you by | UCL - University College London
Download Date | 5/9/18 6:49 PM

Brought to you by | UCL - University College London
Download Date | 5/9/18 6:49 PM

Understanding the Lexicon
Meaning, Sense and World Knowledge
in Lexical Semantics

Edited by
Werner Hüllen and Rainer Schulze

Max Niemeyer Verlag
Tübingen 1988
Brought to you by | UCL - University College London
Download Date | 5/9/18 6:49 PM

CIP-Titelaufnahme der Deutschen Bibliothek

Understanding the lexicon : meaning, sense and world knowledge in lexical seman-
tics / ed. by Werner Hüllen and Rainer Schulze. — Tübingen : Niemeyer, 1988.
(Linguistische Arbeiten ; 210)
NE: Hüllen, Werner [Hrsg.]; GT

ISBN 3-484-30210-0 ISSN 0344-6727

Max Niemeyer Verlag Tübingen 1988
Alle Rechte vorbehalten. Ohne Genehmigung des Verlages ist es nicht gestattet,
dieses Buch oder Teile daraus photomechanisch zu vervielfältigen.
Printed in Germany. Druck: Weihert-Druck GmbH, Darmstadt.

Brought to you by | UCL - University College London
Download Date | 5/9/18 6:49 PM




Section 1
Ernst Burgschmidt

Horst Geckeier

Dirk Geeraerts

MichaZ Post

Section 2
Hans Ulrich Boas

Peter Bosch

D. Alan Cruse

Richard A. Geiger
OF REFER 85 - 96

Bart Geurts

Brought to you by | UCL - University College London
Download Date | 5/9/18 6:49 PM


Ekkehard König, Elizabeth C. Traugott

Bernd Kortmann

Barbara Lewandcwska-Tomaszczyk

Arthur Mettinger

Edgar W. Schneider

Pieter A. M. Seuren

Section 3
Dieter Kastovsky

Günter Rohdenburg

Bruno Staib

Section 4
Martin Durrell

Wolfgang Kühlwein
LEXICOLOGY 242 - 251

H. Joachim Neuhaus

Brought to you by | UCL - University College London
Download Date | 5/9/18 6:49 PM

Section 5
Rosemarie Gläser

Karl Sornig

Section 6
George L. Dunbar, Terry F. Myers

Johannes EngeUcaitip

Section 7
Hubert Cuyckens

Rene Dirven

Use Karius

Leonhard Lipka

Peter Rolf Lutzeier

Günter Radden

Rainer Schulze

Werner Wblski

NAME INDEX 426 - 431
KEY WORD INDEX 432 - 443

Brought to you by | UCL - University College London
Download Date | 5/9/18 6:49 PM

Brought to you by | UCL - University College London
Download Date | 5/9/18 6:49 PM


This volume contains thirty-three papers read at a symposium on problems of
lexical semantics at the University of Essen from November 19 to 21, 1987.

The contributions mirror different trends in current linguistics, comprising
research work done, for example, on word-field theory, structural semantics
(with componential analysis in particular), reference semantics, and cognitive
grammar. The editors believe that the collection provides readers with a repre-
sentative sample of current investigations into the field. Apart from the
renewed discussions on meaning and sense relations such as antonymy, hyponymy,
taxonymy and others, and on the undetermined status of semantic (distinctive
or criterial) features and selectional restrictions, many researchers have
initiated and established a largely cognition-based approach. They predomi-
nantly focus on the investigation of conceptual and mental structures, question
the traditional distinction between semantic and encyclopedic knowledge, and
describe semantic structures as products of conventionalized conceptual con-

The contributions not only show borderlines between semantics and lexicology
or lexicography, but also those between semantics and psycholinguistics, prag-
matics, artificial intelligence, and especially the newly developing area of
cognitive linguistics. The editors hope that the papers collected in this
volume will pave the way for further interesting research to be done in the
near future.

The editors would like to express their gratitude to the University of Essen
which supported the symposium as well as this publication with generous funds.
Moreover, they wish to thank Christa Bohrhardt, once more, for her untiring
work in preparing the publication and producing the typescript. Thanks are also

Brought to you by | UCL - University College London
Download Date | 5/9/18 6:49 PM

Werner Hüllen. to include the chair-persons of the sessions. Vfe wish to thank everybody for this joint effort and. All its contributors were speakers and discussants at the symposium. Richard Nate.due to Roland Aley. This volume is the product of the talents and efforts of a large number of people. Rainer Schulze Essen. Michael Isermann.University College London Authenticated Download Date | 5/9/18 6:49 PM . particularly. DC Richard Brunt deserves our thanks for reading the manuscripts and occasionally giving them the final native-speaker touch. Monika Sieburg and Gaby Sikora for their work in putting the two indices together. May 1988 Brought to you by | UCL . As in previous cases.

SECTION 1 Brought to you by | UCL .University College London Authenticated Download Date | 5/9/18 6:49 PM .

It will be claimed that this combination contributed to analytical tendencies of word definitions in the description of general vocabulary as well as scien- tific terms. for example. As a considerable number of participants of the conference had decided either to defend structural ' componential' semantics (mainly in the Coseriu tradition) or to attack traditional 'analytical1 or 'purely paradigmatic1 semantics (taking up and developing ideas used in frame theory. and syn- tagmatic behaviour) and word meaning in everyday language use contrasted to technical. etc. the 'Advanced Learner's Dictionary'. Oxford). ERNST BURGSCHMIDT SOME REMARKS ON THE TRADITION AND APPLICATION OF ANALYTICAL SEMANTICS 1. a few thoughts on the development of traditional analytical and componential seman- tics in the context of the history of (English) semantics and lexicographic practice since the 17th century did not seem out of place in the scope of discussion of this conference. 'Collins English Dictionary'. prototype and stereotype semantics. Modern English lexicography (as. and artificial languages. Preliminary remarks The author had originally intended to contribute a lecture/paper on the rela- tionship of word form (including word class. philosophical. After reading the abstracts of the lectures handed in for the conference the author decided to shift the emphasis of his own contribution. and lexicographic practice in unilingual dictionaries (from the first published by Cawdrey in 1604 to Dr Johnson's monumental dictionary of 1755). England from the 17th century onwards is characterized by a marked combination of scientific progress (Royal Society. archetype semantics.).University College London Authenticated Download Date | 5/9/18 6:49 PM . even the 'Collins Cobuild English Language Dictionary') Brought to you by | UCL . categorial approaches towards language de- scription. the 'Dictionary of Contemporary English1. Although comparable tendencies can be observed in several European countries (especially France and Italy). morphological structure.

this tradition of presenting word meanings has to be seen in its historical context and cannot simply be disqualified as lacking any connection with modern lexical semantics and its progress. Dic- tionaries also often change their definition pattern even in the description of semantically related words. many words (polysemes. unilingual dictionaries do not seem to conform to the prin- ciples of paradigmatic componential semantics. Feature hierarchies (from the more general to the more specific components). 2. The semasiological alphabetic structure of such a dictionary obscures the user's view of the fact that most analytic sentences used to define words stem from similar attempts at com- parison as in 'word field1 semantics. and syntagmatic features are used in order to avoid a mere addition of fea- tures in semantic descriptions. entailment. diagrams. word formations. referential hints. Feature lists or feature configurations in componential semantics are usually presented a-syntactically. or the com- pilers of the 'Oxford Qrglish Dictionary1. thus arriving at essential distinctive semantic features by contrasting word meanings. At first glance. and cannot devote the same amount of time and methodological consideration as can the author of a study on a particular word field. Traditional semanticists have usually tried to differentiate word meanings by comparing semantically related words. lack con- sistency in the use of feature types. periphrases (especially for word formations). semantic relations such as hyponymy and hyperonymy. Analytical semantics and general lexicography Modern semantic theory of the componential or feature type and lexicographic practice have more in common than is normally assumed. non-verbal means like drawings. phrases) are difficult to treat in a consistent manner. of course. Ihus. include circular definitions. syn- thetic sentences.still relies en principles developed in the 17th and 18th centuries and re- fined by such influential lexicographers as Dr Johnson. either near-synonyms or word groups in the paradigmatic context of 'word fields'. even encyclopedic explanations. Webster. Moreover.University College London Authenticated Download Date | 5/9/18 6:49 PM . With Qiglish being the most fre- quently used defining language for users of Qiglish as a second language and for technical English. synonyms. etc. or combine different approaches meant to com- plement each other and with the intention of giving the user a somewhat redun- dant but more easily understandable bundle of synonyms and analytic sentences to grasp the exact word meaning. or pictures can be found together with analytic Brought to you by | UCL . Most dictionaries. The analogy to feature analysis in phonological theory is obvious. semantic relations like hyponymy.

archetypal semantics (Wildgen). prototype semantics (Rösch et al. This kind of analytical semantic procedure has recently been attacked by a number of psycholinguists and semanticists as being too narrow. Apart from the reproach that componential semantics imitates phonological theory. etc.sentences. Lexicographic practice has hardly been tackled in this way so far (but cf. Attacks. whereas lexicography often uses analytic sentences. The tradition of analytical description In order to assess the importance of analytical semantics it is necessary to see its increased use for general vocabulary and its refinement from the 17th century onwards. opponents of this approach have claimed that it wrongly employs ways of de- scribing well-defined words of science or standardized technologies for less well-defined. Often infinite clauses are used (verb in- finitives) . The similarity of these approaches has been described in 2. Aristotelian principles of defining have always been used for scientific descriptions but from that time on these principles were also adopt- ed for and adapted to the unilingual lexicography of individual languages and Brought to you by | UCL . Frame theory (Fillmare). with nouns modifier structures (adjective/apposition plus noun) can be regularly found. stereotype semantics (Putnam). 4. and as neglecting the varying importance of possible features of words in certain situations.University College London Authenticated Download Date | 5/9/18 6:49 PM . traditional semantics usually provides feature configurations. Other papers from this conference deal with these approaches extensively (Wegner 1985. Wildgen 1985: 4).). 3. These analytic sentences are in fact feature lists for segments of meaning in sentence or clause form. on analytical semantics In order to describe lexical meaning. as not cov- ering the 'gestalt' and interdependence of phenomena. perhaps opaque words of everyday communication (Wiegand 1985: 52. Wildgen 1985: 9-58) without pointing out the neces- sity of differentiating between the scientific and general use of words and recognizing the fact that a large amount of the vocabulary of an individual language today has to be described in a scientific 'analytical1 way. 56. This cannon prac- tice is followed in the description of general vocabulary and for technical word lists. These opponents of traditional semantics demand that atomistic attempts ought to be replaced by holistic descriptions. could be mentioned among those opponents. Negative analytic sentences are used for words with a 'negative1 meaning but also for reasons of contrasting words. Wierzbicka 1985).

Dolezal 1985). 5. Semanticists. orthoepists. medicine. Knowlson 1975. poets. Condillac. such as scientists. Bishop V&lkins1 attempt (1668) to establish a categorial system of semantic classification ('universal philosophy1) clearly shows the combination of linguistic aims with principles of scientific adequacy. (Aarsleff 1967. to avoid ambiguous words. Descartes. polysemes. universal. Kant. etc. Especially a-*priori languages tried to arrange their vocabulary on a planned categorial basis. readers of journals and literature. and language planners in these centuries often aimed at an educated group of users and readers with philosophical interests. zoology. He asked scientists (some of them founding members of the Royal Society 1662) to draw up some of his tables. Knowlson 1975. Blanke 1985. Language standardization and the development of nomenclatures in the rising branches of science (especially botany. and Germany. BS 3669 1963). and to use short analytical definitions (Couturat/Leau 1903. etc. the work of standardizing committees at the beginning of this century clearly illu- strates the analytical approach to the description of word meaning in individ- ual languages (cf. Such Brought to you by | UCL . such as Bacon.) from the 18th century onwards supported these tendencies towards the 'scientif- ic1 description of words relying on categorial components used as features in more complex words and concentrating on the central features of a word thought to be prevalent in any occurrence of the word in any situation. and philosophical languages. and he used concepts from his categorial tables as a lexicographic metalanguage in the dictionary part of his monumental work (Funke 1929. large 1985). all these attempts and projects also tried to facilitate international communication. and homonyms.University College London Authenticated Download Date | 5/9/18 6:49 PM . Leibniz. Land 1974). and schoolitasters were encouraged in their analytical approach by the work of many empirical and rational philoso- phers in England. Although many early nomenclatures were organized on a neo-Latin basis (Crosland 1962). France. Lexicogra- phers. general linguists. Cue famous example may be mentioned. Clearly. the principles em- ployed in devising them (over 400 between the 17th and 19th century) clearly reflect categorial and analytical methods of description. 'virtuosi1. DIN 2330 1961. lexicographers. etc.especially to artificial. Universal and artificial languages Although artificial languages and universal language schemes never attracted a large group of users (with the exception of Esperanto). chemistry. For English lexicography it took about 150 years (from Cawdrey 16O4 to Dr Johnson 1755) to establish an exemplary and consistent method of defining words in a mainly analytical way.

Despite all optimism many scientists. Some scientists and philosophers (such as Leibniz. of the long debate on the relationship of objects to ideas and word meanings (mainly in Prance). Condillac. This com- bination enables these languages to avoid grammatical and derivational homonymy as well as syntagmatic and syntactic idiosyncrasies. etc. The fact that many frames and prototypes resemble categorial elements of traditional semantics very closely is no surprise at all .and Bishop Wilkins and Dr Johnson are not as old-fashioned as seme linguists may believe. 8 a-pviori language schemes were mainly developed in the 17th and 18th centuries. Brought to you by | UCL . componential semantics of the 2Oth century should occasionally remember the background of its tools.but that does not mean that de- scriptions even of general vocabulary achieved in this way must necessarily be inadequate or wrong. especially when attacked by modem lexical semanticists. On the other hand. It has to be admitted that traditional semantics and lexicographic practice owe much to 'scientific1 approaches to word definition . philosophers. whereas in the 19th century a-posteriori languages (usually en an Indo-Euro- pean basis) were developed. Artificial languages are similar to most scientific nomenclatures and termino- logies in dividing their vocabulary into a basic body of free or root morphemes with unambiguous meanings and a very subtly devised and highly productive de- rivational component used to enlarge the vocabulary almost at will. It had to exclude many complex problems of the relationship of language description and mathe- matics in the 18th and 19th centuries. and that most categorial systems were still as inadequate as had been those of Aristotle two thousand years ago. having as their basis a compact categorial feature inventory to define and delimit the range of their morphemes.) hoped that universal language schemes and analytical methods in general could approximate a calculus-like precision. Analytical description and science In this paper only a rather simplistic picture of the influence of science on language description and analytical methods could be given. and linguists realized that words in general use and even nomenclatures lacked the clarity they were aiming at. and of the influence of writing systems and sign languages on language description. This precision would then clear scientific thought from ambiguous concepts and language use and would further progress on the whole. They mainly rely on analo- gy.University College London Authenticated Download Date | 5/9/18 6:49 PM . 6. Nevertheless. linguistic description did not only borrow analytical meth- ods from science and philosophy.

Large. Salmon. Vivian. University of Toronto Press. 1974. 15-100. Longman. Wierzbicka. 1963. Fredric. Khowlson.).). James. Allgemeine Grundsätze. Hyldgaard-Jensen. Internationale Plansprachen.18OO. Toronto. 1984. Princeton University Press. Blanke. Proceedings of the second international symposium on lexicography May 16-17. Eine Einführung. Karcma Publishers. Jahrhundert. Longman. Berlin. repr. 1975. Brought to you by | UCL . Wiegand. Louis. Niemeyer. A study of his writings in the intellectual context of the seventeenth century. Tübingen. Historical studies in the language of chemistry.186O. Stephen K.V. Salmon. Niemeyer. 1961. The artificial language movement. 1985. 1972. Tübingen. Andrew. Dolezal. London. 1985. Frame-Theorie in der Lexikographie. British Standards Institution. lexikographischen Definition.REFERENCES Aarsleff. The works of Francis Lodwick. Beuth-Verlag. Hildesheim. Amsterdam. Princeton. Ann Arbor. From signs to propositions. Land. formation and definition of technical terms. Arne Zettersten (eds.University College London Authenticated Download Date | 5/9/18 6:49 PM . Akademieverlag. Symposium on lexicogra- phy II. Olms 1979. 1985. 1967. Begriffe und Benennungen. 1903. Forgotten but important lexicographers: John Wilkins and William Lloyd. Hans. Histoire de la langue universelle. 1979. Blackwell. Detlev. Niemeyer. Berlin. (1982). DIN 233O. Wegner. Funke. BS 3669. Arne Zettersten (eds. Recommendations for the selection. London. Winter. Leopold Lsau. London. In: Karl Hyldgaard-Jensen. Otto. Paris. at the University of Copenhagen. John Benjamins B. Couturat. The study of language in 17th-century England. A modern approach to lexicography before Johnson. 1962. Crosland. Heidelberg. Lexicography and conceptual analysis. 1985. Karl. The concept of form in eighteenth-century semantic theory. 1985. Maurice P. Anna. The study of language in England 1780 . Eine neue Auffassung der sog. Ittmo. Vivian. 1929. Oxford. Herbert E. 1985. Universal language schemes in England and France 1600 . 1985. Ausschuß Normierungstechnik im Deutschen Normenausschuß. Heinemann. Zum Weltsprachenproblem in England im 17. Tübingen. London.

10 Wildgen. Hölfgang.1800. An essay towards a real character. (English Linguistics 1500 . (1968). Vol. 1668. Grundlagen für eine dynamische Semantik auf der Basis der Katastrophentheorie. John.University College London Authenticated Download Date | 5/9/18 6:49 PM . Wilkins. 199. Tubingen. and a philosophical language. Narr. Brought to you by | UCL . 1985. Archetypensemantik.. Scholar Press.) Menston.

not to pass over in silence its limitations. by the way. is no longer a novelty of the year 1987 since it was outlined by Coseriu as early as the sixties and has been developed further by him and by several of his disciples . the epithet 'struc- Brought to you by | UCL . This type of structural semantics has actually met with a lively echo not only in Germany but also abroad. are following orientations different from ours. 11 HORST rc MAJOR ASPECTS OF THE LEXEMATICS OF THE TÜBINGEN SCHOOL OF SEMANTICS 0. The aim of the present contribution is not to give an integral presentation of the theory of structural semantics of the Tubingen School .University College London Authenticated Download Date | 5/9/18 6:49 PM . i.2. the following years up to the present. to a considerable number of colleagues working in lexical semantics.e.2 The title of our contribution contains two notions which call for an of whan is the author of this con- tribution . but to work out the advantages it has over other ap- proaches. the lexematics of the Tübingen School of Semantics. partly. to call our colleagues1 attention to the various applications of this type of structural semantics and. especially in Southern Romance countries and in Japan (Geckeler 1981b: 407). but who. Space restrictions im- pose a rather brief presentation of this particular type of structural seman- tics.this would be a far too extensive task -. In contrast to other usages.1 'Lexematics1 is the term used by Coseriu to designate the type of structural (lexical) semantics developed and applied by the members of the Tübingen School of Semantics. finally.1 The colloquium on lexical semantics held at the University of Essen seems to us an excellent platform for presenting a particular approach to structural semantics. 0. which.

almost simultaneously with. 1968 ). Lipka (1980. we will refer to most of them later on. Pottier and Greimas. 1983 ). especially Coseriu 1964.the concept of 'Tübingen School of Semantics1 no longer denotes a geographical unity of place. the outlines of modern descriptive semantics along European structuralistic lines (cf. 1.University College London Authenticated Download Date | 5/9/18 6:49 PM . most of his disciples have left Tübingen for profes- sional reasons and are now teaching at other universities in Germany or abroad. 1967. dimensions. 1981. Japanese semanticists first used this name when referring to Goseriu and the circle of students interested in structural semantics whom he gathered around him after having moved from Montevideo to Tübingen in the early sixties. it was in the sixties that Goseriu developed. 1982: 66-15O) belong basically to this same orientation. In fact.2. 12 tural' in structural semantics is understood in this school as 'structural' in an analytical respect. but independently of.2 The denomination "Tubingen School of Semantics' also needs a comment.e. Brought to you by | UCL . Nowadays. the content of a lexeme results from the structure of its semantic features (archi- sememe. It is not possible within the scope of this paper to mention all the linguists belonging more or less directly to the Tübingen School of Semantics. 0. classemes). Thus. Although the master's continuing to teach in Tübingen still ensures the correct- ness of the denomination. In this primarily para- digmatic type of lexical semantics. also our joint publications Coseriu/ Geckeler 1974. 1981 ). the analysis of lexical meaning is carried out by the decomposition of content into smaller elements (situated below the sign-threshold). semes. 1982a. i. for example. Coseriu 1976a. Also in later years he continued to work and to publish in the domain of semantics (cf. 1976b. into relevant meaning-differentiating features. His structural semantics has a twofold aim: In spite of slightly different views on certain topics. The writer of these lines was the first at Tübingen to carry out a large-scale synchronic study of a lexical field according to Coseriu's lexematic theory (Geckeler 1971) and in his publications he elaborated certain concepts within the same theoretical framework (cf.0 Up to now Coseriu has elaborated and published the most comprehensive conception of a semantics of lexematic structures. 1981) and Kastovsky (198O. 1966. 1977. as referred to the organization of the content-level of language by means of functional lexical oppositions.

13 1) by precisely delimiting the object of his semantics. This system comprises the paradigmatic structures of vocabulary as well as the syntagmatic structures ('syn- tagmatic1 understood as combinatorial on the seme-level). although it is provisional. the domains eliminated by the distinctions were sometimes considered as being of minor importance for semantic studies. But it must be emphasized that the founder of this type Brought to you by | UCL . leads at the same time to a considerable reduction in the enormous quantity of lexi- cal material to be analysed. This reduction. it may suffice to give a schematic representation of the hierarchy of the preliminary distinctions (for detailed information. cf. necessary from a theoretical point of view. objects meta- s language language ^^ diachrony \ primary repeated language / discourse historical *\ synchrony ^ language technique type Designation of d^scourse / ^^ ~~~^. As we cannot pre- sent and explain these distinctions within the framework of this paper. 2) he wants to integrate all the questions of a structural semantics into a single and coherent system.functional-system language\ ~^*^ \^ norm \ norm signification 3 J discourse Only the domain of (each) distinction whose designation is written in italics is retained because it is amenable to a structural treatment. Initially. it constitutes one of the limitations of this ap- proach. infra his seven preliminary distinctions) . The application of the seven distinctions. is a methodological one. 1. in fact. Certain of our own formulations went in the same direction in the early years of the Tubingen School of Semantics. Coseriu (1966: 181-210)). which can only then be subjected to a structural semantic analysis. he intends to avoid the inadequacies of the approaches of other types of semantics (cf . Coseriu arrives at the desired homogeneous object of investigation.University College London Authenticated Download Date | 5/9/18 6:49 PM .1 ad 1 ) : By means of a succession of seven necessary preliminary distinc- tions.

we would like to emphasize that the Tübingen School of Semantics is firmly convinced that the first task of structural semantics is to establish a paradigmatic type of seman- tics. Coseriu (1982b) pleaded for the reintegration of the donains provisionally eliminated into the centre of linguistic interests.Development . on the contrary.e. Staib (1980) applied the structural semantic analysis to diatopical varieties of a histori- cal language. a word semantics. sentence semantics or text semantics (although there is a substantial proposal for text semantics on the part of Coseriu (198O)). he regarded it as a temporary neglect due to the priority given to the structural study of language. from the point of view of the history of linguistics. 1. Coseriu (1964) outlined the principles of a structural diachronic semantics . i.e. Brought to you by | UCL .Lexical class .3 Before passing on to comment on the lexematic structures. others have not yet been assigned their proper place in the realm of linguistics. delivered on numerous occasions.Composition . Several of the themes removed in favour of those amenable to structural analysis ware studied by Goseriu or by seme of his disciples. Thun (1978) analysed the section of repeat- ed discourse in his doctoral thesis on problems of phraseology.Lexical field .Affinity . At the very beginning of the elaboration of his semantic theory. i.Implication 1.2 ad 2): It is only after the careful application of the seven prelimi- nary distinctions that we can start with the structural analysis of the lexern- atic structures.University College London Authenticated Download Date | 5/9/18 6:49 PM .14 of semantics never considered the elimination of the domains in question as a definitive one. which Coseriu conceives of in the following manner (1968: 7): Lexematic Structures Paradigmatic Structures Syntagmatic Structures (oppositional) (combinatorial) (= Lexical Solidarities) Primary Secondary Structures Structures . a very interesting parallelism with Jakobson's reflections on diachronic phonology at the beginning of the development of phonological theory.Selection . and that any attempt at working seriously on com- binatorial semantics. must turn out to be premature unless linguists have first worked out a solid basis for word semantics. In a lecture entitled 'Au-dela du struaturalisme1.this is.Modification .

also able to function outside lexical fields or throughout a series of lexical fields. This method is a synthesis of the German tradition of lexical field research with the functional principles of European structuralism.University College London Authenticated Download Date | 5/9/18 6:49 PM . Clas- semes are general determinations in the lexicon. .A classeme is a content-feature by which a lexical class is determined. .A dimension .0 The limited space of this paper does not permit us to present in detail the different lexematic structures which Goseriu has established in the scheme presented above. 2. The basic constitutive elements of the lexical field are these: lexemes. these words standing in immediate opposition to each other by means of simple content-differentiating features. .An archilexeme is a lexeme whose content is identical with the content of the whole lexical field. within a dimension. The archilexeme may be realized as a lexical unit in a determinate language.this is a concept that has been introduced into the discussion of structural semantics by ourselves following a suggestion of Ißunsbury - is interpreted as a criterion of lexical 'articulation* ('Gliederung') which is operative in a lexical field and which . they are a specific kind of seme.furnished the scale for the oppositions functioning between determinate lexemes of the field (comparable to Greimas' 'axe semique'). the concept 'pole can be very usefully incorporated and applied. Coseriu (1976b) even developed a typology of the lexical fields. which is constituted by the partition of a lexical content- continuum into different unities corresponding to words in a given language. In fact/ such a presentation is no longer necessary because the lexematic structures have been explained in numerous publications and are supposed to be known to the majority of the semanticists. dimensions/ semes. Brought to you by | UCL . archi- lexemes. classemes.1 The structural semantics of the Tübingen School possesses a fully elab- orated method for the study of lexical fields. the semes are reduced to specific differences within a dimension. What we intend to do is to insist upon the strong points and the special traits of this type of structural to speak . 15 2. If we work with dimensions in lexical analysis. The lexical field is defined by Coseriu (1967: 294) as follows (in our trans- lation) : From a structural point of view a lexical field is a paradigm of the lexicon. .A lexeme is a lexical unit functioning within a lexical field. . but need not be.A seme is a content-differentiating feature functioning in a lexeme.

16 So the lexical content of a lexical item is composed of the archilexematic content (archisememe). devoted her thesis to the investigation of the lexical field of the adjectives designating cleanness and dirtiness in present-day French. i. nouveau. The word-class (pars oration-is) does not change.2. This study. 1976: 304-329.) leads to the classem- atic analysis of the lexical field.0 Another particularity of the structural semantics of the Tübingen School is the integration of word-formation into the domain of lexematics. studied the lexi- cal field of the verbs designating locomotion in contemporary French. etc.e. the schematic representation supra). 5. In this ap- proach. According to the respective grammatical determination of the primary lexical unit involved. Sum- maries of the results of the investigation of this particular lexical field can be found in various publications (Geckeler 1973: 52-70. realized by means of the ccmnutation test. 1979a: 199-217).ge3 anoien. of the dimension(s) and the semes and of the classeme(s).2. moderns} neuf. cf. In general. 2. As to the applications of the lexical field theory of the Tübingen School: Up to now the most extensive study of a lexical field on the basis of the lexem- atics of the Tübingen School is our synchronic analysis of the lexical field of the adjectives structuring the semantic area 'age' in present-day French (Geckeler 1971). Azem (forthcoming). was carried out in two phases: 1) The syntagmatic (distributional) study of the collocations of the respective adjectives (e. one of our students. word-formation is treated from a strictly semantic point of view.g. based on a rich corpus of examples taken from contemporary French texts. Coseriu distinguishes three types of 'secondary1 paradigmatic lexematic structures: 2. for example. In her doctoral thesis. Another one.University College London Authenticated Download Date | 5/9/18 6:49 PM . Krassin (1984). This evaluation of word-formation is very different from current classification of this lin- guistic domain (Coseriu 1982a). or Brought to you by | UCL . 2) the paradigmatic analysis.. to a determination which does not imply any sentence-function of the modi- fied primary lexical unit. jeune.. in modification we are dealing with a quantification of the primary lexical element.1 "Modification1 corresponds to an 'inactual1 grammatical determination. vieux. reveals the semic structure of the field. the different types of word-formation are considered as secondary paradigmatic lexematic structures (cf. diminutive and collective formations. In both theses the structural semantic analyses were realized according to the principles of the Tübingen School of Semantics.

Spanish. discret ->· discretion. or 'lexematic' composition.g. e. 2) The 'generic1. The prolexematic composition-type of this new semantically defined classification of word-formation is traditionally referred to as 'derivation' . quereia -> queraeto. . Ettinger (1974) thoroughly studied diminutive and augmentative word-formation with respect to Italian. or 'pronominal'. and Stein's (1971) thesis. Fr. it is always accom- panied by a change of the word-class of the primary term. This type of composition corresponds to traditional word-composition. beau + predicative function ·* beaute ('le fait d'Stre beau1) .ier.g.University College London Authenticated Download Date | 5/9/18 6:49 PM . tousser -> toussoter . and Rumanian substantives. Span. Brought to you by | UCL .2 'Development1 corresponds to a grammatical determination which com- prises a sentence-function of the primary lexical unit. In the donain of modification.2. 2. 2. It. 17 prefix formation in the case of verbs. where one of the combined A A ~~"* B A A — B elements (B) is not identifiable with a lexeme existing in the language in question. mur ·+ muret. Coseriu distinguishes two types of composition: 1) The 'specific1.g. two extensive studies of prolexematic composition in Spanish and French have been realized by disciples of Coseriu (Laca 1 986 . Staib forthcoming) .. calculer -+ oalculx . Rohrer 's (1967) doctoral thesis. Thus. station-service. can be considered as a very valuable contribution to the study of the 'lexematic composition '-type. a subject which refers main- ly to the 'development '-type. poire ·+ pair .noiratre. Fr. cf . oaballo -+ oaballito. and Spanish. Fr. etonner -*· etonnement. noir ->.atvice . In the domain of development we must make mention of Ludtke's (1978) voluminous doctoral thesis on predicative nominalizations by means of suffixes in French. Fr. although not in total accordance with Coseriu1 s principles of word-formation. Fr.2. Fr. Recently. or 'prolexematic1 composition.g. in which she deals with the derivational possibilities of adjectival word-formation in French and English. e. wagon-restaurant. or 'nominal1. e. for example. Portuguese. where both combined elements repre- sent really existing lexemes.3 'Composition1 always implies the presence of two basic lexical elements which stand in a grammatical relation to one another. wir -> revoir. Catalan. Fr. e.

3 A further characteristic trait of the structural semantics of the Tü- bingen School is the integration of what Porzig called 'essential meaning- relations' ('wesenhafte Bedeutungsbeziehungen1) or 'elementary semantic fields' ('elementare Bedeutungsfelder1) into the domain of lexematics. leaher (i.2. the lexemes dent and langue function as semes in the lexi- cal content of mordre and leoher respectively). Coseriu 1967 ) . cf. Fr. Coseriu distinguishes three types of lexical solidarities: 'affinity'. Kindergärtnerin (corresponds to a combination of lexematic composition (Kinder- garten) and a prolexematic composition (K. 1973) distinguishes five different types of lin- guistic content: 'lexical'. Germ. word-formation is a sort of granmaticalization of the lexicon and that is why word-formation con- stitutes a special domain within the lexicon: word-formation is regarded as a 'grammar of the lexicon1 (Coseriu 1982a: 12). cf.g. 1972.3 A connection can also be found between our type of structural seman- tics and the one represented by Lyons (1968. . 'syntagmatic' being understood as combinatorial on the semic level. cf. Eventually. There is quite a number of other phenomena within the domain of seman- tics that have been studied by linguists of the Tübingen School. 2. 'designation1. 3. and 'ontic' content (as regards the hierarchy of lexical· and gram- matical· content. elaborated by Coseriu (e. 1977). and 'implication1 (for further information. and 'sense1. 3. Geckeler forthcoming b ).1 The distinction between 'signification1. 3. With regard to the status of the implied element. 3. 'instrumental·'.18 The two types of composition may also appear contained.e.University College London Authenticated Download Date | 5/9/18 6:49 PM . particularly by Coseriu himself. to the increased Brought to you by | UCL . 'categorial'. cf. Here we refer first of all· to the study of 'antonymy' (sensu lato) and. 'selection'.2 Coseriu (e.. 198O). we want to insist upon a special view that characterizes this approach to word-formation: In Coseriu's conception. is fundamental for the theory of this type of structural semantics. mordre or of 'langue' in Fr. "structural· (syn- tactic)1. ouvre-boftes (as to the interpretation of this type of compound. Coseriu 1977 ). Coseriu named them 'lexical solidarities' and classified them as 'syntagmatic lexematic struc- tures'. As examples may be quoted the implication of 'dent1 in Fr. quite generaUy. for exampie.g.

198O) and the doctoral thesis by Nellessen (1982). for example. . our considerations on this subject (Geckeler forthcoming a). But a lot of work remains to be done. cf. we tried to establish a tentative classification of lexical gaps..4 Another question arising within the framework of structural semantics is that of lexical gaps. More than what can be said about the 'state of the art1 of our type of structural semantics in this limited contribution is to be found elsewhere (Geckeler 1981a). 19 interest of semanticists in this meaning-relation. the question of the non-existence of antonymic structures in certain areas of the lexicon was dealt with (Geckeler 1983a. 4. In Geckeler (1974. b).University College London Authenticated Download Date | 5/9/18 6:49 PM . This interesting domain has so far not been systemati- cally investigated in linguistic research.. We formulate our invitation to join this semantic theory by using the telling simile of St. Information on translations of the major publications of this orientation in structural semantics is available in Geckeler (1981b: 4O6-4O7) and in the two Festschriften in honour of Ooseriu. our articles on antonymy (Geckeler 1979b. Finally. 37-38: "The harvest truly is plenteous. Further perspectives as well as other limitations of this semantic theory have been presented in Coseriu/Geckeler (1981: 66-69) and Geckeler (1981b: 4O8-4O9)." Brought to you by | UCL . Matthew 9. but the labourers are few. 1985). A considerable amount of theoretical and practical work in the field of struc- tural semantics has been done by the Tübingen School so far and this type of semantics has proved its practicability. for the relations between synonymy and antonymy.. 3. We also published a rather sketchy study of the intralingual paradigmatic gap-type in the system of word-formation of modern French (Geckeler 1977). cf. which is reflected in the numerous publications on this topic in recent years.

). Bouvier. In: Thomas A. 5-15. Coseriu. Eugenio.). Eugenio. Dieter Kastovsky (eds. Eugenio. Eugenio. Die Lage in der Linguistik. Brekle. 3-16. Diss. Eugenio. Travaux de Linguistique et de Litterature 2. 1974. Tübin- gen. 1966. Benjamins. 198O. Juli 1976. Eugenio. Minster. Geburtstags von Hans Marchand am 1. Beiheft zur Zeitschrift für französische Sprache und Literatur 1.139-186. Perspektiven der Wbrt- bildungsforschung. 1964. 1981. Semantik und Grammatik. Pour une semantique diachronique structurale. Coseriu. 175-217. Pour et centre l'analyse semique. Eugenio. Narr. 77-89. Inhaltliche Wsrtbildungslehre (am Beispiel des Typs 'coupe-papier*). Das Wartfeld der Sauberkeitsadjektive im heutigen Französisch. In: Herbert Ernst Brekle. Cahiers de Lexicologie 27. Linguistica e Letteratura 7. Beiträge zum Wuppertaler Wörtbildungskolloquium von 9. Die funktioneile Betrachtung des Wortschatzes. Eugenio. Coseriu. Amsterdam. Horst Geckeier. Actes du premier collogue international de linguistique appliquee. 293-303.). Textlinguistik. Laure. Sebeok (ed. 7-25. Coseriu. Poetica 1. Kazuko Inoue (eds.1. Integrale Linguistik. Coseriu. Coseriu. Brought to you by | UCL . Eugenio. Eine Einführung. Jahr- buch 1975 des Instituts für deutsche Sprache. 137-148. 1973. Coseriu. Cahiers Ferdinand de Saussure 35. Trends in structural semantics. Fest- schrift für Helmut Gipper. 3-16. 1972. 1982b. Coseriu. especially functional. Vers une typologie des chanps lexicaux. Jahrbuch 1971 des Instituts für deutsche Sprache.University College London Authenticated Download Date | 5/9/18 6:49 PM . Eugenio. Coseriu. 1982a. Au-delä" du structuralisme. forthcoming. 1979. Peter Schmitter (eds. Bülow. Coseriu. 1977. Nancy. 1967. 1968. 1976a.). 9-16. Horst Geckeier. Oktober 1977. 1983. Coseriu. Eugenio. Bonn. semantics. Structure lexicale et enseignement du vocabulaire. lexikalische Solidaritäten.-1O. Linguistics and semantics . 3O-51. Eugenio. Innsbrucker Beiträge zur Sprachwissenschaft. Eugenio. Herbert Ernst. Coseriu. Eugenio. 1976b. 1977. Coseriu. Les structures lexematique. Coseriu.20 REFERENCES Azem.). Anläßlich des 70. Narr. 48-61. Tübingen. Eugenio. Dieter Kastovsky (eds. 103-171. In: Shirö Hattori. Coseriu. Vorträge 9. Edeltraud.Linguistic. Lss precedes semantiques dans la formation des mots.

). Geckeler. Semantica estructural y teoria del campo lexico. Logos Semantikos. Geckeier. 1982. Cahiers de Lexicologie 25. Geckeier. Die Antonymie im Lexikon. 1976. 1981. Boringhieri. Horst. Actös· du Xle collogue international de linguistique fonctionelle. Madrid. (eds. New York. de Gruyter. Considerations sur les relations entre la synonymie et l'antonymie. 1981.). Fink. 1974. Horst. worlds. Geckeler. Horst. Observations sur l'absence de l'antonymie dans certaines sections du lexique. Horst. In: Hans-Jürgen Eikmeyer. Tübingen. Niemeyer. Horst et al. In: Edeltraud Bülow. Gredos. August 29 . Geckeier. 1983a. Untersuchungen zur Gliederung des Wortfeldes 'alt-jung-neu' im heutigen Französisch. In: Dieter Kastovsky (ed. Lexeme ohne Antonyme. Brought to you by | UCL . Hannes Rieser (eds. 1981a. Vol. 1983b. Geckeier. Geckeier. Geckeler. Progrös et stagnation en semantique structurale.University College London Authenticated Download Date | 5/9/18 6:49 PM . Horst. 71-79.). Ettinger. Nouveaux regards sur les lacunes lexicales. 381-413.September 4. de Gruyter. Hattori. Horst. Kazuko Inoue (eds. Shirö. Geckeier. 1985. Berlin.). 42-69. (eds. Horst. Berlin. 1984. 98-1O6. Horst. Geckeier. Hans-Jürgen. Diminutiv. Zur Wartfelddiskussion. Madrid. Geckeler. Geckeier. In: Helmut Stürm. 31-45. Horst. 1979b. La semantics strutturale. 198O. In: Horst Geckeier et al. III. Padova. and contexts. Horst. Horst. Considerations sur la hierarchie des faits semantiques dans la graimaire et dans le lexique.).). 21 Eikmeyer. Horst. Wolfgang Raible (eds. 1974. 7O-82.).). Narr. Proceedings of the Xlllth inter- national congress of linguists. Horst. Quaderni di Semantica 4. 1983. 1977. New York. Hannes Rieser (eds. Tübingen. Antonymie und Wortart. Tokyo. Geckeler. CLESP. Words. The Committee. 1973. 248-252. Horst. 53-69. New approaches in word semantics. 1971. Studia linguistica in honorem Eugenio Coseriu 1921-1981. Dieter Kastovsky (eds. 1981b. Credos. Strukturelle Semantik des Französischen. Horst. Peter Schmitter (eds. Geckeler. 1979a. Zur Frage der Lücken im System der Wortbildung.und Augmentativbildung: Regeln und Restrik- tionen. Torino. Geckeier. Le problems des lacunes linguistiques. Stefan. Geckeier. forthcoming b.). 455-482. Structural semantics. München. In: Herbert Ernst Barekle. forthcoming a.

The Hague. 93-114. 198O. Cambridge University Press.). Gabriele. Dieter (ed. Staib. XII. Funktionelle Untersuchungen zum Französischen und Spanischen. LUdtke.). Primäre und sekundäre Adjektive im Französischen und Englischen. Bei- träge zum Wuppertaler Semantikkolloquium von 2. Pädagogischer Verlag Schwann. John.). Francke.).). forthcoming. Probleme der Phraseologie. 1982. 1974. (ed. Helmut.22 Kastovsky. Stein. Stürm. Niemeyer. Niemeyer. 1980. 1984. Introduction to theoretical linguistics. Krassin. Düsseldorf. Harald. Tübingen. 1977. Perspektiven der lexikalischen Semantik. Staib. Vol. Die Antonymie im Bereich des neufranzösischen Verbs. 70-92. Lipka. Semantics. Mathcdology and representation in the study of lexical fields. Narr. Tübingen. 429-445. Tübingen. Niemeyer. Nellessen. Itouton. Die Wortbildung als Granmatik des Wortschatzes. Prädikative Nominalisierungen mit Suffixen im Französischen. Bouvier Verlag Herbert Grundmann. 1986. 198O. Kastovsky.1971. Tübingen. Die Wortzusammensetzung im modernen Französisch. Dieter. In: Horst Geckeier et al. Das Wortfeld der Fortbewegungsverben im modernen Französisch. Generische Komposita. New York. Semantik und Sprachgeographie. Lyons. Tübingen. Wortbildung und Semantik. Paris.). Kastovsky/ Dieter. Wiesbaden. Wolf gang Raible (eds. On the interrelation of syntagmatic modification and paradigmatic lexical structuring in English. Narr. Laca.. 1981. Sebeok. Cambridge. 1968. etc. 1978. 1978. Brenda. Bern. Bruno. Gudrun. Zur Semantik des Französischen. Kastovsky.. 373-383. Steiner.-3. Bruno. In: Dieter Kastovsky (ed. Jens. 1 and 2.). Dieter. 1982. Narr. Narr. Current trends in linguistics. (eds. etc. Lang. Tübingen. Bern. 1977.University College London Authenticated Download Date | 5/9/18 6:49 PM . (eds. 1983. Leonhard. Vol. Brought to you by | UCL . Thun. München. Cambridge. 1980. Rohrer. Lipka. in: Dieter Kastovsky (ed. Niemeyer. Lyons. Bagel. Frankfurt/taain. Thomas A. Dezember 1977. In: Horst Geckeier et al. Lexical fields and word-formation. Tübingen. John. Horst. Cambridge University Press. Christian. Leonhard. Katalanischen und Spanischen. Bonn. 1967. Tübingen. 1981. Selectional restrictions and lexical solidarities.

Brought to you by | UCL . For a more extensive treatment. figure 1). The history of lexi- cal semantics can only be treated briefly in the context of this paper. The view of the history of lexical semantics that lies at the basis of this paper is the same as the one presented in Geeraerts (forthcoming) (cf.University College London Authenticated Download Date | 5/9/18 6:49 PM . Dealing with this problem is not just an end in itself. I will try to make it clear that it is a way of uncovering seme major lines in the history of lexical semantics. Katz/Fodor's "The structure of a semantic theory1 (1963) formed a reference point for studies in lexical semantics. Here. most of the attention goes to- wards the relationship between post-structuralist cognitive semantics and pre-transformationalist diachronic semantics. ASPECTS OF ΊΗΕ HISTORY OF LEXICAL SEMANTICS All through the seventies. Taking a stand in lexical semantics meant taking a stand with regard to ccnponential analysis of the Katzian brand. and more particularly towards the transitional role of Katzian semantics. see Geeraerts (1986). and more often than not the stand was negative: a rejection of one aspect or another of the Katzian approach of- ten seemed to be the preliminary ritual parricide of lexical-semantic studies. and that it provides a framework for a new appreciation of the present-day situation of the discipline.if not partly identical . What I propose to do here is to address precisely the question of why Katz/ Fodor were so tremendously successful and why they were so tremendously un- successful. 23 DIRK GEERAERTS KATZ REVISITED. The wording of this question is deliberately paradoxical: I will try to show that the reasons for their success are intimately connected . together with the subsequent elaborations and revisions of the model in the work of Katz (1972). In the latter article.with the reasons for their failure. the attention goes towards the middle period of the development.

Katz was able to ride along the waves of transformationalist enthusiasm. 24 So why did Katzian semantics draw the attention of so many linguists? One rea- son is clearly extrinsic . of course. For Katz/Fodor's incorporation of a semantic component into the Chomskyan model clearly filled an unappealing and counter-intuitive gap that characterized the earlier. There is. and its mentalist conception of language. because the mentalist and the formalist stance of Katzian semantics derive from the Chomskyan emphasis on precisely these aspects of doing linguistics. Incidentally. that is. Syntactic Structures-type of grammatical model. this volume). and we should not forget that Katz/ Fodor took a decisive step towards the formulation of that model.University College London Authenticated Download Date | 5/9/18 6:49 PM . but who was bound to remain relatively isolated. Brought to you by | UCL . The attractiveness of Chomskyanism was based to a large extent on the elegance of the Aspects-model. There are three of them to be mentioned: its basically structuralist methodology. On the other hand. its attempt at formal representation. But we should give Katz the credit he deserves: he not only profited from the appeal of generative grammar. A fortiori. whose 1963 and 1964 articles present a type of semantic theory that is intrinsically in many ways similar to Katz's. con- 2 centrated round generative grammar. to the kind of lexical-semantic theory put forward by Katz. it is easy to see that the singular combination of the three characteristics itself was appealing: a major tradition of lexical semantics (indeed. this applies to European structuralist semantics of the Coseriu-brand (cf. Geckeler. mainly because he did not link up with the mainstream of linguistic discussions. But the appeal of Katzian semantics did not rest on extrinsic factors alone: the intrinsic characteristics of the model are equally important. a connection between these intrinsic characteristics and the extrinsic link with Chomskyanism. But each of the three characteristics itself was also attractive. the importance of this extrinsic reason for Katz's success may be brought out clearly by comparing his fate to that of Pettier. the major one in the period innediately preceding the generative era) was linked to the vanguard of linguistics. By incorporating semantics into generative grammar. Granting for a moment that Katz's methodology indeed brings to- gether the main strands of structural semantics (but I will presently have to come back to this). he also contributed to it.extrinsic. Katz's methodological position links up with the pre-transformationalist structural tradition of lex- ical semantics.

Katzian semantics inherited the mentalistic self-conception of Chomskyan- ism. we should go back to the origins of structuralist semantics. the practical realization of this attempt to develop a synchronic. it should be synchronic instead of diachronic. structural theory of semantics will depend on the way in which the notion of semantic structure is conceived. First. Primo. linguistic mean- ings should not be studied from a psychological perspective (as was usual in pre-structuralist historical semantics). semantics came to share the promises of explanatory adequacy that constituted so much of the appeal of generative grammar. For obvious reasons. Katz took over the Chomskyan requirement that linguistic analyses be rigidly fomalized.University College London Authenticated Download Date | 5/9/18 6:49 PM . Why can we say that Katz's method is a cabnination of structural semantics? To appreciate that this is indeed the case. Because the meaning of a linguistic sign is determined by its position in the linguistic structures in which it takes part. the first theoretical and methodological expose of the new approach is to be found in Weisgerber's Die Bedeutungelehre . and a formal descriptive apparatus. In practice. according to the model furnished by the mathematical sciences. Formalisation signified.ein Irrweg der Spracfojissensehaft? (1927). the study of meaning should not be atomistic but should be concerned with semantic structures. the study of linguistic meaning should proceed in an autonomously linguistic way. but while Trier's monograph may indeed be the first major descriptive work in structuralist semantics. the methodology of lin- guistic semantics will be autonomous. By defining the subject matter of semantics as the competential 'ability to interpret sentences' of the language user. The third aspect of Katz's approach needs some elaboration. there are mainly three distinct definitions of semantic structure that have been employed by structuralist semanticians. a guarantee of scientific seriousness. Weisgerber criticizes pre-structuralist historical semantics on three points. Second. Oomponential analysis was at the same time a method of descriptive analysis. too. 25 First of all. a formal apparatus that seemed indispensable to comply with the requirements of conceptual rigidi- ty and falsifiability that. any respectable scientific endeavour had to live up to. More particularly. Rather. Second. three distinct kinds of structural relations among lexical items have been singled out as the proper methodological basis of lexical semantics. because the subject matter of semantics consists of autonomous linguistic phenomena. These are customarily attributed to Trier (1931). there is the relation- ship of semantic similarity that lies at the basis of semantic field analysis. And third. in other words. non-psychological. Brought to you by | UCL .

and hyponomy. syntagmatic relationships are the subject of selectional restrictions and projection rules. the paradigmatic lexical relations high- lighted by Lyons. and while componential analysis is probably taken over from anthropological linguistics. these were for the first time systematically selected as the methodo- logical basis of structural semantics by Lyons (1963).26 inaugurated by Trier and ultimately leading to oomponential analysis in the work of anthropological linguists such as Goodenough (1956) and Lounsbury (1956). Brought to you by | UCL . To begin with. There are two remarks to be made with regard to the observation that the Katzian approach to semantics brings together the three structural types of semantic relations. it will be clear from what was said above that Katzian semantics is not a straightforward chronological development out of the struc- turalist tradition. They are not discussed in 1963. While his incorporation of lexical relations such as hy- ponymy and antonymy most certainly derives from Lyons. In the first place. they are brought together As far as I know. as a result of its syntactic.University College London Authenticated Download Date | 5/9/18 6:49 PM . the notion of selectional restrictions. Katz has never explicitly and unambiguously acknowl- edged such an indebtedness. Secunck). antonymy. at least. but in 1972 (apparently as a result of the publication of Lyons's book). there are unanalyzed lexical relations such as synonymy. seems to be an independent innovation. This becomes obvious when we consider. In fact. Tevtio. and hyponymy. antonymy. Katz points out explicitly that semantic theory should describe lexical relations such as synonymy. the paradigmatic elaboration of the model is to a large extent the result of later additions. what makes Katzian semantics so interesting from a structuralist point of view is the fact that it brings together the three types of semantic relations that may lie at the basis of structuralist semantic theories. transformational background. On the one hand. In the second place. syntagmatic lex- ical relations were identified by Borzig (1934) under the name of 'wesenhafte Bedeutungsbeziehungen'. only the first of these three kinds of semantic relationships received considerable attention in pre-transformation- alist semantics. In actual practice. Further. the 1963 article is mainly con- cerned with these syntagmatic phenomena. paradigmatic similarity relations show up in Katz/Fodor's adoption of componential analysis. it should be noted that Katz draws together the three types of struc- tural relations in two distinct ways. Now. in the third place.

What lexical semantics should try to describe. semantics moves away from the structuralist pole of the Katzian conjunction towards one of the other two poles. the set-theoretical elaborations and refinements of classical logic were introduced into linguistic semantics. this combination of distinct elements is at the same time the very cause of its gradual decline. his generalization of componen- tial analysis suggests a unified method for describing those distinct relational types. but Bierwisch (1969) and others then made clear that this was only a halfhearted and still insufficient Brought to you by | UCL . There are two major tendencies to be cited. cognitive orientation in se- mantic studies. Katz (1966) tried to remedy this by introducing complex markers. I think it can be shown that the Katzian amalgam contained the germ of a number of internal tensions that finally made the model fall apart. Katzian semantics appears to be a singular combination. he adds to the structuralist appeal of his unified observational basis for semantics by suggesting a uniform method of description. Weinreich (1966) among others pointed out that Katz's sentential readings did not have internal structure. On the other hand. The important thing is to see that this evolution was a 'logical' step.University College London Authenticated Download Date | 5/9/18 6:49 PM . 27 into the observational basis of lexical semantics. Let us now have a closer (though inevitably brief) look at these two lines of development. attempts to take the mentalist position of Katzian semantics seriously led to a straightforward psychological. While componential analysis had previously been restricted to the analy- sis of field relations. deriving as it did from certain inadequa- cies in Katz's formalizations. is precisely the total sum of those different kinds of structural relationships. Thus. Sumtiarizing. The evolution towards logical semantics took place in two steps. Next to this. mainly in the framework of Montague grammar. a mentalist philosophy of language and a formalized descriptive apparatus. the men- talist point of view contained an exciting promise of explanatory adequacy. Thus. After a primary transition from Katzian formalizations towards a notational system couched in the tradition of predicate logic. within the framework of generative grammar. the demands of formalization diminished the struc- turalist influence in semantics in favour of logical approaches to meaning anal- ysis. 1. for instance. in each case. Katz says. Thus. of a basic structuralist methodology. then. However. and the formalization seemed to establish the model's scientific respectability. This combination explains the attractiveness of the model: the structuralist basis seemed to ensure its descriptive adequacy and a large empirical scope. selectional restrictions are couched in componential terms. In this sense. Katz uses it to describe the other relations as well.

28 move towards the formalism of predicate logic.g. that a decompositional approach would need semantic axioms in any case. This is. The development towards psychologically rather than logically orientated forms of semantics can best be traced in the discussion between the componential method advocated by Katz. Instead of structural relations. This full step towards predicate logic was then taken precipitately by Generative Semantics. truth conditions became the primary empirical point of interest of semantics. Second. In the context of a methodological history of lexical semantics. As such. In 1977.theoretical apparatus needed to ensure a formal interpretation of the representational language completed the evolu- tion towards logical semantics. Thus. 2. e. e. Truth condi- tions are properties of sentences. while in the same period Dowty (1979) showed that it was possible to incorporate decomposi- tional definitions of lexical items into a ttontagovian semantic theory. Brought to you by | UCL . The subsequent intro- duction into linguistics of the model . The whole shift is awkwardly reflected by the fact that the terra 'struc- tural semantics' may now be used to denote semantic analyses dealing with the logical structure of sentences rather than the relational structure of the lexicon (cf. not of lexical items. in order to represent redundancy rules. however. a fusion of both approaches became possible. and the axiomatic method for semantic representation. Garnham 1985:99). also because quanti- fiers played an important role in the discussion with Interpretative Semantics. It gradually became clear. The next stage in the development was reached when logicians remarked that symbolic representations of meaning had to be formally interpreted if seman- tics was to became truly formal. First.University College London Authenticated Download Date | 5/9/18 6:49 PM . the initial phase of the discussion (Bar-Hillel 1967) appeared to be just another aspect of the tension between Katz's linguistic approach and existing logical forms of analysis. the gist of Lewis's well- known critique with regard to Katzian 'markerese' (1972). the observational basis of semantics is changed.g. As the use of semantic postulates originated in logical grammar. the evolution towards logical semantics means two things. Thcmason's state- ment (1974) that a semantic theory need not specify the way in which items such as walk and fun differ in meaning is typical for the shift in interest from 4 lexical relational structures to sentential truth conditions. of course. Katz tacitly introduced postulates into his componential approach. the emphasis moves away from lexical semantics towards sentential semantics.

but a substantive one. Collins/Loftus(1975). non-autoncmous option that gradually appeared was able to link up with existing work on comprehension processes and meaning representation. How- ever. an attitude that opposes this methodological extrapolation of the mentalist self-conception of Katzian semantics is supported by two other aspects of Katz's position. using psycholinguistic data about per- formative processes is the ultirtate consequence of the mentalist position of Katzian semantics. 29 But when the fornal compatibility of the decompositional and the axiomatic ap- proaches is established. psychological. who recorded the reaction times needed to verify sentences containing either overtly marked or implicit lexical negations. and it can be solved by using experimental evidence for the psychological reality of decotpositional representations. including data with regard to the way in which sentences are processed. It characterized the object of the investigation as something that is psychologically real. but it did not influence the method used to determine the nature of that object. it is methodologically acceptable to use all kinds of psychological evi- dence. In actual practice. provided by Fodor/Fodor/Garrett (1975). That method is structural in the sense that it is based on the static relations among linguistic elements. it should be clear that Katz's mentalism was not a methodological position. Such results were. Brought to you by | UCL . rather than on actual on-line processes of interpretation.University College London Authenticated Download Date | 5/9/18 6:49 PM . The closest point of contact between both approaches is when psycho- linguists try to devise formal models for semantic memory. In this kind of research. as in the work of Smith/Shoben/Rips ( 1 9 7 4 ) . This is no longer a formal question. whereas Katz's deconpositional approach explicitly wanted to be a competential theory of se- mantics (Katz 1981). On the other hand. At this point. On the one hand. On the other hand. If one remembers Weisgerber's charges against psychological approaches in linguistic semantics. Both the generative notion of competence and the structuralist attempt to define an autonomous method for linguistic semantics preclude using psychological data pertaining to the actual performative proc- esses involved in interpreting sentences. for instance. This work can be found in two overlapping areas: psycholinguistics and artificial intelligence. Katz's present-day (highly isolated) Platonistic position (1981) clearly continues his earlier competential autonomous point of view. these experiments clearly involved performative processes. the 'performative'. the question becomes what degree of deconposition is necessary. and Glass/ Holyoak (1976). a particular tension within the Katzian ap- proach became apparent. If semantic descriptions do indeed have psychological real- ity.

the visual mode of human memory. I have tried to make it clear that the two innovations added by Katz to a more traditional structuralist methodology eventually led to forms of semantics that went far beyond the initial structuralist position. Just as taking the formalization of semantic analysis seriously entailed the adoption of a full-fledged logical semantics. 7 Basic references are Rosch (. Lakoff (1987). Summarizing. the autonomistic method- ological ideal of structuralism is abandoned.University College London Authenticated Download Date | 5/9/18 6:49 PM . the distinction between semantic and encyclopaedic knowledge is discarded: there is no longer believed to be an autonomous level of semantic organization. but the meaning of another set of items may have to be represented in close connection with. Langacker (1987). Research- ers in psycholinguistics and artificial intelligence generally do not estimate that the linguistic capacities of man can be studied in isolation from his other cognitive capacities. In linguistic lexical semantics. this psychological reorientation has led to the school of Cognitive Semantics of which Langacker (1987) and Lakoff (1987) are currently the leading figures. but an adequate picture of language comprehension and production is expected to involve the cooperation of distinct forms of knowledge. in close connection with the prototypical theory of categorial structure developed in psycholinguistics by Rosen (1978). require a representation that has access to perceptual information. separated from the general conceptual capacities of the human being. then. one kind of lexical item may require a traditional conceptual representation. A satisfactory model of semantic knowledge may e. taking the nentalist claims seriously entailed a shift towards a psychologically orientated cognitive type of seman- tics. Brought to you by | UCL .g. say.1978). Both the logical and the cognitive approach may take many forms. not just by using 'performative1 psychological data. In short. but they 6 Smith/Medin(1981).3O the original question with regard to the alternative between componential and axiomatic approaches is transcended into the more general question 'What does an adequate model of man's use and knowledge of language look like?". but also more generally by incorporating the study of lin- guistic semantics into the broader field of cognitive science. Linguistic knowl- edge is no longer considered to be an autonomous form of knowledge that can be studied in isolation.

Second. whereas the development towards cognitive semantics was not triggered by such formal considerations of logical adequacy. In this sense. the cognitive tradition might also be called 'ecological semantics 1 . then. competential versus performative) that have to be formalized. pre-transformationalist structuralism reacts by imposing an autonomous relational method. following Seuren (1986). figure 1 appears. Some additional remarks with regard to the figure are necessary. Katz represents the grandeur et d&cadznee of structural semantics. First. the progression represented by the figure does not imply that the older traditions are extinct. Third. The point made in figure 1 is merely that the development towards logical semantics resulted from a require- ment of formal rigidity with regard to Katzian formalization s. but merely that they are no longer the main centre of attention and the main locus of progress in the field. approaches that were in- trinsically more coherent than Katz's eclectic approach. historical semantics starts off with a thoroughly psychological. this cognitive/ecological approach most certainly does not oppose formalization (e. adding to it the mentalist and formalist emphasis of trans- formational grammar. however. If we bring together these historical lines of development. Katz takes up the method. But there is an additional step that has to be taken if the historical position of Katz is to be fully appreciated. but the nuances he adds to it eventually incite semantics to move away from the structuralist framework. The latter gave way to other approaches mainly as a result of the dynamism inherent in its own constitution. Brought to you by | UCL . First. in the form of AI models of natural language comprehension). one of formalization versus the absence of formalization. But we also saw that structural semantics is related in a particular fashion to the pre-structuralist. but has something to do with the kind of facts (roughly. the Katzian amalgam dissolves. He resumes the whole endeavour.University College London Authenticated Download Date | 5/9/18 6:49 PM . let us briefly review ο the stages of the development. The distinction between both approaches as they exist now is not. One might say that the very diversity that lay at the basis of the appeal of Katzian semantics. 31 do constitute the major paradigms of contemporary semantic research. but rather by substantive considerations of psychological adequacy. historical-philological approach to lexical semantics.g. As we saw earlier. Third. Second. And fourth. Katz links pre-trans- formationalist structuralism to contemporary linguistics. non-autonomous method. led to the subsequent adoption of different approaches.

Brought to you by | UCL . Specifi- cally with regard to the present-day situation. proponents of such an approach should try to determine whether the structural or the logical point of view best suits their purposes. If this is correct.1985).takes up again the interests and methods of pre-structuralist historical semantics. i. of course. but there is one point that can easily be appreciated. what is truly important is to see how present-day cognitive semantics links up with pre-structuralist historical-philological semantics. historical context are not taken into account. Semantic phenomena are no longer studied as self-sufficient. cf. The first question is relevant only if an autonomous approach is adopted. whether more of structural semantics should be in- corporated into logical semantics than is usually the case. and autonomous approaches such as logical and structural semantics.University College London Authenticated Download Date | 5/9/18 6:49 PM . but dealing with them would go well beyond the scope of this article. cultural. but they are seen as non-autonomous phenomena that can only be understood against their contextual background. This is. while the attention devoted to socio-cultural phenomena followed straightforwardly from the historical outlook: a lot of diachronic semantic changes can hardly be understood if changes in the social. I do not 9 have the opportunity here to make the comparison in great detail. or alternatively. Both aspects were conspicuously present in historical-philological semantics. par- ticularly true of prototype theory. the major dividing line in the history of lexical semantics appears to be between non-autonomous approaches such as historical and cognitive seman- tics. purely linguistic data. socio-cultural on the other. if post-structuralist cognitive semantics. there are two major questions about doing lexical semantics that arise from our historical sketch. the references mentioned in footnote 1. In this respect. the earlier interest in semantic change is reflected by the attention given by cognitive semanticists to the dynamic flexibility of human categorization. Furthermore. The present paper will have served its purpose if it has become clear how the questions just mentioned arise from a historical overview of lexical semantics. of course. This context is twofold: it is psychological on the one hand. to a large extent. can and should cognitive seman- tics be combined? I have my own answers to these problems.e. whether the autonomous approach should be adopted at all.32 But new. The other question is. The psycho- logical approach has already been touched upon in connection with Weisgerber's denunciation of it. Geeraerts (. 10 In particular. 9 Cf. The cognitive approach again takes up the psychological approach that characterized historical semantics. structural.

Loftus. Langacker. 380-405. Geeraerts. Fodor. The structure of a semantic theory. Elizabeth F. university Press. Collins. 1969. Wbordbetekenis. Psychological Seview 6. Jerrold J. Jerrold J. Semantic theory. Katz. 1985. 1981. Katz.. Stanford University Press. Componential analysis and the study of meaning. 1976. Leuven. Goodenough. Alan. 313-339. Harper & Row. Explorations into a paradigmatic theory of meaning and its epistemological background. Jerrold J.University College London Authenticated Download Date | 5/9/18 6:49 PM . 1987. 1977. fire. Glass. Allan M. 17O-210. Foundations of language 3.. Lakoff. forthcoming. Semantics of natural language Dordrecht. Alternative conceptions of semantic theory. In: Brygida Rudzka (ed. 1979. London. Ronald W. A spreading activation theory of semantic processing. Richard Barman (eds. Keith Holyoak. Geeraerts.). 1987. Dowty. The psychological unreality of semantic representations. Stanford. Methuen. Janet. Dirk. 1975. 1975. 195-216. New York. Garnham. 643-673. 1966. David. 1986. Bierwisch. 1956. 515-531. Dictionaries and meaning rules. Linguistic Inquiry 6. Paradigm and paradox. Blackwell. Cognitive graintar and the history of lexical semantics. 1972. Women. Reidel. Harper & Row. and dangerous things. Manfred. Donald D. Davidson. Jerry A. Fodor. Katz. Katz.. Language and other abstract objects. The advantage of semantic theory over predicate calculus in the representation of logical form in natural language. Cal. 407-428. Jerrold J. Foundations of cognitive grammar. On certain problems of semantic representations. Psycholinguistics. 1985. Chicago university Press. Foundations of language 5. New York. Acco. London. What categories reveal about the mind. Dirk. Central topics.). 33 REFERENCES Bar-Hillel. Word meaning and Montague grammar. Language 39. Ihe philosophy of language. Dirk. 1967. Volume 1: Theoret- ical prerequisites. Yehoshua. 1972. Oxford. Een overzicht van de lexicale semantiek. 409-414. Dordrecht. Leuven. 1963. Arnold. Ward H. Jerry Fodor. Jerrold J. Brought to you by | UCL . Chicago. Language 32.. The Monist 6O. George. Katz. 53-184. Cognition 3. Reidel. Merrill Garrett. Geeraerts.

Leo.. etc. Seuren. (ed. Topics in cognitive grammar. 1986. Beiträge zur Geschichte der Deutschen Sprache und Literatur 58. Lawrence Erlbaum. Brought to you by | UCL . Selected papers of Richard Montague. Bernard. Mass. Sebeok. Faculte de Lettres et Sciences Humaines. Nancy. Thomason. 1-18. 214-241. 1934. Pottier. 27-48. In: Donald D. Amsterdam.). 1974. Rudzka. 169-218. The Hague. Language 32. Barbara B. Eleanor. Lloyd (eds. Principles of categorization. Eleanor. Structural semantics. Pieter A.). Edward. Current trends in linguistics. Harvard University Press. Edward. 158-194. Brygida (ed. Pottier. Uriel. 1927. 1978. Volume 3. New York. Formal theory and the ecology of language. 1956. forthcoming. Winter. Walter. Formal philosophy. Der deutsche Wortschatz im Sinnbezirk des Verstandes. Richmond (ed.). 1972. Wesenhafte Bedeutungsbeziehungen. Heidelberg. Porzig. 1963. Trier. General semantics.). Theoret- ical Linguistics 13. Blackwell. M. 1966. Rosch.). 1931. Vers une semantique moderne. John.ein Irrweg der Sprachwissenschaft? Germanisch-Rcmanische Monatsschrift 15. Benjamins. Bernard. New Haven. Lance Rips. 1966. 1978. Lloyd (eds. 1981. Floyd. 7O-97. 107-137.). Rösch. Mouton. 395-477. David. Edward Shoben. Yale University Press. Barbara B. Cognition and categorization. Oxford. 1974. Categories and concepts. Structure and process in seman- tic memory: A featural model for semantic decisions. 1963. Smith. Travaux de Linguistique et de Utterature 2. Psychological Review 81. Davidson. lounsbury.). Cambridge. Thomas A.University College London Authenticated Download Date | 5/9/18 6:49 PM . Lyons. 161-183. Smith. Sebeok (ed. Recherches sur l'analyse semantique en linguistique et en traduction mecanique. 1964. In: Thomas A. In: Eleanor Rosch. 34 Lewis. Die Geschichte eines sprachlichen Feldes. Die Bedeutungslehre . Weinreich. Jost. Richard Barman (eds. Weisgerber. A semantic analysis of Pawnee kinship usage. Explorations in semantic theory.. Douglas Medin.

35 Figure 1: HISTORICAL- PHnJODOGICAL SEMANTICS meaning should be studied from a synchronic. autonomous point of view STRUCTURAL SEMANTICS paradigmatic and syntagmatic relations reveal the autonomous semantic structure of language TRANSFORMATIONAL SEMANTICS mentalism formalization a psychologically a formally adequate adequate study of representation of meaning requires meaning requires a a non-autonomistic model-theoretic approach interpretation COGNITIVE LOGICAL SEMANTICS SEMANTICS Brought to you by | UCL .University College London Authenticated Download Date | 5/9/18 6:49 PM . non- psychological .

F. an outstanding Polish sociologist. The idea born in a particular period and milieu. The passage quoted suggests three questions. thus giving a beginning to what L. this idea is novel owing to its new contents. In a way.) In what follows I will argue that Fillmore's scenes-and-frames semantics (hence- forth SAF) can be regarded as such a new idea which is a continuation.University College London Authenticated Download Date | 5/9/18 6:49 PM . I have in mind here the version of lexical field theory developed by such linguists as Jost Trier and Leo Weisgerber. what is SAP's contribution to field theory? Thirdly. Firstly. Krzywicki labelled some time ago as "the wandering of ideas in time and space 1 . I prefer to concentrate on indicating the essential fea- tures shared by the two theories. the closing remarks to the present paper will offer a ten- tative answer to it. again responding to the changed conditions and needs. this new idea is a continuation of the old one which had earlier disclosed a research problem. As regards the third question. This old idea to which 2 SAF relates is lexical field theory (henceforth LFT). why should the idea of lexical fields appear again? As regards the first two questions. Krzywicki. Simultaneously. in response to specific con- ditions and requirements of the development of science. 36 POST SCENES-AND-FRAMES SEMANTICS AS A NED-LEXICAL FIELD THEORY 0. social activist and journalist. what is it that makes SAF a continuation of LFT? Secondly. (M.P. Brought to you by | UCL . reappears in a new period and in a new milieu. of an old idea. in new circumstances and intellectual context. Such an arrangement is in agreement with the general purpose of showing that SAF is indeed a continuation of LFT. and the entire new intellectual and social context. Ludwik J. (1859-1941). Introduction The Polish philosopher Adam Schaff in his book Jezyk a poznanie (1964: 11) ('language and cognition1) observes that One of the more interesting methodological problems is the analysis of the reception of earlier research ideas under new historical circumstances. the discussion of new elements being sub- jected to the former aspects.

the recognition of polysemy versus the formulation of core meanings. Fillmore (1975: 124) also admits that he can ". also see it (the scene-and-frame idea. as is evidenced by the above quotation. In his 1975 paper he even suggests that SAF can be an alternative to LET: "Other areas in which scene-and-frame approaches could give sensible alternatives to traditional accounts are: . I will now indicate those aspects of the two theories in question that motivate the treatment of SAF as a continuation of LET. (ii) the characterization of lexical meaning in LET and SAF.) in the work of the European semantic field theorists". (iii) the organization of the lexicon in LET and SAF. In what follows I will concentrate on the following issues relating to this claim: (i) the nature of conceptual fields in LET and SAF. 37 My implicit assumption is that LET and SAF are comparable.. The human colour schema identifies the semantic field of colour terms. a comparison of these two theories should be carried out. The schema and the associated linguistic frames for the human body provide for the field of body part names and the vocabulary of body positions and body movements. Such an identification would be wrong. As is usually the case one encounters contrary views which regard the juxta- position of LET and SAF as ill-conceived. Brought to you by | UCL . and he adds: The concept of semantic field can be captured by appealing to the notion of schema. (1977: 13O-1) Fillmore is clearly aware of the affinity of the twa theories.University College London Authenticated Download Date | 5/9/18 6:49 PM . the commer- cial event schema underlies the vocabulary field of buying and selling. It is the aim of the present paper to offer such a ten- tative comparison. And that before one of the opposing views is finally accepted.... and the allied concept of vocabulary field can be identified with the notion of frame and with various linkages among frames. synonymy. metaphor. And so on. selection restrictions. Verschueren (1981: 339) says that One might be tempted to identify lexical frames with what the structuralists called semantic or lexical fields. In my opinion. Both theories make the strik- ingly similar claim that groups of thematically related words in the lexicon are mapped on conceptual structures. M. For instance. the nature of semantic fields" (1975: 129). these opposing views on the relationship between LET and SAF indi- cate that their juxtaposition is a non-trivial matter. That they are some- how parallel has been hinted at by Fillmore in some of his papers.P.

due to which they seem to be objects of the same kind. they reflect the expe- riences of language users. too. is determined environmentally. As a result. Fillmorean scenes are chunks of knowledge of varying size. representing a common sense understanding of real world situations. i. i. This is naturally reminiscent of Humboldt's view that in languages the spirit of the nation (Volksgeist) is reflected. scenes are also experientially-based.University College London Authenticated Download Date | 5/9/18 6:49 PM . army ranks. kinship terminology. and it came to be exhaustively and differently articulated by kunst. Trier's discussion of intellectual terms shows that changes occurred in the representation of a fragment of the reality of the entire population. philosophy and theology came to be distinguished. IfT and SAF are based on the assumption that words are mapped on the continuum of thought. or rather their structuring. Conceptual fields in IfT and SAF As will be indicated below. Fillmore's SAF is concerned with the individual's view of the world rather than that of the nation. science. there Brought to you by | UCL .e. conceptual structures rather than directly on the external world or on external reality. What is important in the context of my discussion is the fact that the conceptual structures of LET. etc. canton conceptual structure (Brown 1967). A hundred years later. that for each nation there exists a permanent. An example of such a reflection of a part of external reality is constituted by Trier's description of German intellectual terms (1931). sim- plified. wizzen and wtsheit. stereotypical. conceptual fields of LET and scenes of SAF have a number of essential attributes in cannon.e. They are colour terms. Scenes are intended to be mental representations of the language user's real world experiences. i. feudalism and universality. However. Trier's and Weisgerber's fields are reflections of external reality both in their totality and their parts (fields and subfields). the structure of the entire conceptual field of knowledge altered.e. i. kunst and list em- bodied two fundamental principles of medieval civilization. In the relevant literature there are also other examples illustrating the role of conceptual fields in the organization of one's experience. 38 1. and sets them in opposition to Matore (1953). who would rather assume that sets of words are mapped directly on external reality. due to social changes and a breakdown in the medieval system.e. This assumption places IfT and SAF in the line of the continuators of Humboldt and de Saussure. At one period wtsheit.

For example. the words covering them are organized into closely-knit sectors. both theories are in principle capable of accamodating the nation's and the individual's world-views. their contours fitting one another like pieces of different shapes in a mosaic. Another support for the plausibility of this suggestion comes from the evolu- tion of LFT itself. which holds for the entire Australian population. cognitively linked with other scenes and other frames in such a way that in their totality they characterize perceived reality and the entire stock of lexical resources for talking about it (Fillmore 1977b: 72). frames are associated in memory with other frames by virtue of their shared linguistic material. In LFT. Wie mosaic metaphor. Due to this mosaic organization and the strict delimitation of individual conceptual structures. Brought to you by | UCL . deriving from Ipsen (1924). As a corollary. the conceptual structure. or relations or substances in them. In Cognitive Linguistics. or their context of occurrence (Fillmore 1975: 124). contemporaries and followers attempted to weaken some of its claims in various ways. in American Qiglish. and his analysis of anger. One should remember that SAF is a forerunner of Cognitive Linguistics in which something akin to a Volksgeist can also be detected. the difference between them with respect to the scope of conceptual structure is quantitative rather than qualitative. covering the entire world. and scenes are associated with other scenes by virtue of the sameness or similarity of the entities. which holds for American speakers. Oksaar (1958) suggested that each speaker has his own indi- vidual vocabulary. Trier's field theory is the 'strongest1 version of LFT.University College London Authenticated Download Date | 5/9/18 6:49 PM . Fillmorean scenes are organic wholes. It seems that SAP's theoretical assumptions do permit con- ceptual structures to be shared by larger populations or even by whole nations. Like LFT's conceptual fields. the variability in question is also reflected by a frequently made observation that speakers need to negotiate meanings. I can see it in Lakoff's (1987) discussion of categorization in Dyirbal. holds for the entire lexicon and the entire conceptual structure. Furthermore. signalling thus the shift from the uniformity of conceptual structures to their variability. In short. is an organic whole expressing the peculiar way in which a nation attempts to reflect reality It is argued that conceptual structures of higher rank consist of conceptual structures of lower rank. and his predecessors. 39 are reasons to believe that SAF is capable of incorporating the conceptual structures characterizing the warId-views of large populations.

stemming fron Humboldt's work. The above is a belief that IFT theoreticians adhered to. the conceptual structure becomes one of the principal devices of organization in the lexicon. or rather attempts to substantiate it. they are organic wholes. were undertaken by Sapir and Whorf in their empirical studies. under- standing a real world situation involves the evoking of a scene that best fits this situation. According to the view adopted by Fillmore from Artificial Intelligence (Minsky 1975). they were interested in how a priori given conceptual areas become fields. Conceptual structures do not only reflect the ideas. how they are articulated by the lexical resources of a given language. In this respect ΙίΤ links up with the Sapir-Whorf hypothesis on the influence of language upon thought. But Fillmore is concerned neither with philosophical implications of the problem nor with the empirical confirmation of it. There is a flexibility here. He focuses solely on the psychological mechanism of perception and comprehension. 40 The above indicates that despite the different principles by which the entire conceptual structure is arranged in ΙίΊ and SAF. then it may be verbalized with the lexical resources from the evoked scene. making a system of them. as the situation is allowed to fit the scene 'to a degree1. scenes are conceptual representations of real world experiences. The details of organization will be discussed in section 3 below. Sapir.University College London Authenticated Download Date | 5/9/18 6:49 PM . In actual practice. which is in keeping with the views of Humboldt. SAF also follows the view that conceptual structures play a role in dealing with reality. is that conceptual struc- tures. which is again reminiscent of Humboldt's and de Saussure's systemic nature of language. with a particular world-view encoded in them. however. The confirmation of the hypothesis. LET provides an approach to the influence of language on thinking. Brought to you by | UCL . In this way. on the other. Another aspect of IfT. values and outlook of a society but also hand them down to the later generations as a ready-made analysis of reality through which the world will be viewed (Ullmam1957). Whorf and field theoreticians. The systenaticity of mental structures in both theories imposes sys- tematicity on the words napped on them. they impose their structure on new world experiences. If it does fit. On the one hand. In other words. determine the perception of reality by the users of a particular language.

In other words. In the former theory there are two complementary views of what constitutes the mean- ing of the word. The meaning of a word is a relatively small conceptual area within a wider conceptual field. a list of necessary and sufficient conditions that an entity must satisfy to be named by a given word. on the contrary. lexical meaning in LET is a checklist in Fillmorean sense. or more precisely by the struc- ture of the field. That means that a word in any language is not an isolated carrier of meaning. The meanings of the words in a field fill up all the space in that field like the tiles of a mosaic. The characterization of lexical meaning in LIT and SAF LET and SAP operate with apparently different conceptions of meaning. at- tributed by Lehrer (1974) to Lyons (1977). Trier's discussion of intellectual terms is an illustration of this. This view was held by Trier and Weisgerber.e.University College London Authenticated Download Date | 5/9/18 6:49 PM . 41 2. and was a corollary of an absolute parallelism between conceptual structures and word fields (isomorphism). This approach recognizes sharply marked boundaries among the conceptual areas repre- senting the meanings of individual words in a given field. i. the meaning of a word is a function of its relationships to the other terms in the lexical field.the theory concentrates on the focal parts of conceptual areas. According to the prototype approach. Brought to you by | UCL . According to the other view. LFT's view of word meaning is determined by an explicit adherence to the struc- turalist assumption that each separate element of the language derives its essence (value) fron its relation with other elements. SAF assumes that the meaning of the word is a prototype (Coleman/Kay 1981) which for an adequate description of the meaning of a word requires the giving of only the typical conditions under which the word can be used appropriately. the mean- ing of words is determined by the entire field. and any conceptual area that is associated with a word as its meaning is a concept. all parts receive their meaning only from the whole. saying that in the system. Due to the rigidity of boundaries. Trier completely rejected the meaning of a word as an independent unit. dis- regarding their boundaries. With respect to the lexicon of a language this means that the meaning of each word depends on the existence of other words in the same field. Despite these different approaches to what constitutes the meaning of a word. the non-existence of sharp boundaries between the conceptual areas covered by words in a particular field causes no problem . each has a meaning only because there are others adjacent to it (Trier 1931: 43O). Let us consider LFT. LIT and SAF are surprisingly similar as to the way they define the meaning of a word.

Fillmore says that someone who understands a word can be thought of as activating a scene and pointing to a certain part of that scene: If we know in one way or another what the commercial event is. He adhered to the idea of a totally and precisely delimited lexicon. conceptual fields and scenes. we can know exactly what the vocabulary pertaining to that semantic domain means. integrated and fully articulated system. the two theories involved require that the meaning of individual words be defined within experientially-based mental structures. Thus conceptual structure becomes one of the principles of organization in the lexicon. too. From the dis- cussion in the relevant literature (Trier 1931. The whole vocabulary is a single. It is hierarchically segmented into two types of fields which are parallel to each other: (i) conceptual fields (Begriffsfelder. Trier adopted de Saussure's structuralist principle that language is an or- ganized whole whose elements delimit one another and derive their significance (value) from the general framework in which they are placed. which in its entirety represents the reality of a given speech community (nation) or of an individual. In short. they reflect the worlds as experienced by the users of languages. i. Öhman 1953. 3. sell. Gordon 1982) the fol- lowing views of the lexicon in LET emerge. the mental structures in both theories. pay or cost. I can believe of myself that I know exactly what is meant by such words as buy. given that knowledge. making a system of them. i. Lutzeier 1982. 42 Trier's remark that all the parts receive their meaning only from the whole echoes in FiHraore's slogan that "meanings are relativized to scenes" (Fillmare 1977b: 84). the whole lexicon consists of fields. Each field is a discrete entity within a totally structured lexicon. Weisgerber 1954. Lehrer 1974. whose structure in its totality constitutes a mental representation of reality. then. Firstly. Brought to you by | UCL . are experientially-based. Lyons 1977.e. The system- aticity of mental structures in both theories imposes systematicity in the words mapped on them. Secondly. LFT and SÄF are mentalistic frameworks. (1977b: 60) Recapitulating the present section. conceptual fields constitute a system.University College London Authenticated Download Date | 5/9/18 6:49 PM . Vassilyev 1974. I will argue that.e. Ihe organization of the lexicon in LFT and SAP From the above outline of the nature of conceptual fields and scenes certain common properties of LFT and SAP seem to follow. According to this view. Thirdly. despite the different conceptions of meaning. Both assume the existence of an intermediate mental structure between words and the real world. without requiring of myself that I have a complete and correct checklist description of the commercial event itself.

in turn. cycle. The part-whole relationship which holds between individual lexemes and the lexical field within which they are interpreted is identical or at least similar to the part-whole relationship which holds between the lexical fields and the totality of the vocabulary. i. From the time perspective it can be seen that the evolution of IFT proceeded from Trier's 'stronger1 version to a 'weaker1 variety.University College London Authenticated Download Date | 5/9/18 6:49 PM . This idea of the total delimitation of the lexicon is absent from the work of Ipsen and Porzig. contrast sets. etc. partonymy. defined as a lexical set whose members index portions or aspects of some conceptual or actional whole. As suggested above. he maintains that there are various kinds of structures that characterize different semantic domains: contrast set. network. Trier's UT is a 'strong1 version of this approach. Fillmore's most explicit statement on the issue in question can be found in another article (1978). Fields occupy an intermediate position between the totality of the vocabulary and its minimal dependent units . 43 Sinribezirke). and as a corol- lary. Porzig (1934) indicated that it is not only para- digmatic relations that are the organizing principle in the lexicon but also syntagmatic relations as illustrated by his theory of syntagmatic fields. Fill- more 's views on the structure of the lexicon can be placed within the 'weaker1 variety of IFT. and (ii) word fields (Wortfelder). Each of these fields. Ipsen's and Porzig's criticism was evidently motivated by the inadequacy of Trier's homogeneous lexicon. Furthermore. explicitly thus manifesting his adherence to the 'weaker' version of IFT. taxonomies and other kinds of structures. there can be found sets of words that form paradigms. the essence of which is that (i) words in the lexicon are related in a variety of ways. chain.e. is subdivided into elementary units. accord- ing to which every field is a discrete entity within a totally structured lexi- con. (1978: 165) In a frame. paradigm. concepts and words. where the components of the word field cover completely the corresponding conceptual one. most central and powerful of all the domains. taxonomy. Fillmore arrived at his heterogeneous lexicon not through the criticism of classical IFT but rather on the basis of his own work Brought to you by | UCL .words. (ii) there is no single semantic description of lexical meaning. In it he claims that there is no such thing as a uni- form semantics for ordinary language. Among the different semantic domain types there is a frame.

to type. In his book (1977a) Fillmore pre- sents the description of a writing scene. and the product of a writing act. the verb to cost will be preferred. to copy. but both are different from to paint. Still another way of grouping words in the lexicon is via natural categorization.e. to pencil. the implement with which the individual writes. not necessarily representational. given by a husband to his former wife. the different viewpoints that they offer on the same scene. to autographf to scribble. the verb to spend is selected. the word alimony is linked to an entity. Cne of the ways in which words may be linked in a frame is via perspectivization. Perspectivezation is thus an impor- tant relationship holding among the words in a frame. a discussion of about twenty English verbs follows. Brought to you by | UCL . to address. In his work (1977a) he discusses the noun alimony which activates a sequence of scenes associated with legal acts and with the marriage relationship (1977a: 114-5). which comprises the individual who does the writing. The word divorce. i. If the perspective of the buyer and the money is taken. that marriage ends and at the same time a legal agreement is made between the two participants to the effect that one would pay money to the other. to scrawl. An account similar to the Fillmorean one can be found in Kalisz's article (1981). to draw is more or less like to sketch.e. associated with the English verb to write. that is a linguistic conflgurätiDn of marks on the sur- face. i. he says that to sketch adds the information that the product is something repre- sentational and the implement is of a particular kind . to jot down. 44 on lexical semantics and the work of others. the surface on which the writing is done. and so on. since to paint is to produce an artistic object. clearly implying that what might relate them is the relation of family resemblance. i. In these scenes. For example. This has led people to give up the view of the lexicon as an unordered list of words and to recognize a rich forml and semantic structuring within it. money and goods can be taken into perspective. Fillmore occasionally indicates what he means by indexing a portion or aspect of some conceptual whole. In his papers. where the lexical field of obscenities and terms per- taining to sexual intercourse are discussed explicitly in terms of deviation from the prototype scene of sexual intercourse.a brush. If one wishes to take the perspective of goods and money. to take another example. a fixed sum of money. to pen. in the commercial event scene buyer and seller. to transcribe.University College London Authenticated Download Date | 5/9/18 6:49 PM . in terms of deviation from the prototype and the relationship of family re- semblance holding among the words from the set. For example. The other verbs discussed in a similar manner are to print. would be linked to a portion of the sane large scene. Then.e.

then. results of research in other relevant disciplines. can be characterized as follows: his heterogeneous lexicon can easily accommodate the proposals of Trier and other field theo- reticians . Trier and other field theoreticians should fall within the scope of SAF. One should recall that SAF falls within the province of Cognitive linguistics. namely why the ideas of LET should emerge again in the form of SAF. I have in mind here in particular Artificial Intelligence with its proposal of knowledge representation systems. It is not surprising. as compared with liT's related view. as demonstrated above. however. I would. for example. that the proposals of Weisgerber. We find this concept incorporated and elaborated in SAF. To answer this question is not an easy matter. Fillmore's view on the organization of the lexicon. Conclusion There remains to be answered the third question asked at the very outset of the present paper. either imposing structure on an event or on the conceptualization of an event in a fixed way and with a given perspective. Fillmore introduced the con- cept of case frame. who were either directly or indirectly influenced by the ideas of Herder and Humboldt (Brown 1967). Brought to you by | UCL . a continuator of the anthropological linguistics of Boas. Sapir and Whorf. and Cognitive Psychology with its natural categorization and prototypes. Finally. Fillmore's preoccupation with lexical semantics and the lexicon ever since his Case Grammar has evidently enabled him to undertake the idea of LET in a nat- ural way. and the encouraging intellectual atmosphere. The discovery of the complexity of the lexicon in the last fifteen years has also played an important role. 4.University College London Authenticated Download Date | 5/9/18 6:49 PM . Research in several fields has made it possible to undertake and interpret the essential problems of LFT in a new light. In the grammar mentioned. the encouraging intellectual atmosphere has made it possible to artic- ulate SAF publicly. 45 In a nutshell. understood as associated with a predicating word. sug- gest that an explanation should consider the interaction of subjective and ob- jective factors such as Fillmore's own work on the lexicon.

On the organization of semantic information in the lexicon. Lakoff. A history of semantics. The Hague. What categories reveal about the mind. In: Johannes Friedrich et al. Terence. 1977b.). 148-176. Proceedings of the first annual meeting of the Berkeley Linguistics Society. Paul Kay. Word 9.). Adrienne. Current issues in linguistic theory. Der Alte Orient und die Indogermanen. Didier. Theories of the linguistic field. 111. Oksaar. tartan. Coleman. 1967. Fest- schrift für Wilhelm Streitberg. Stockholm. Matore. Lingua 56. 1982. In: Dieter Metzing (ed.. Scenes-and-frames semantics. In: Adrian Zampolli (ed. Semantic fields and lexical structure. Chicago. 1977. Amsterdam. Linda. Almqvist & Wiksell. In: Roger W. Cole. Wilhelm von Humboldt's concept of linguistic relativity. Lyons. 1977. Indiana University Press. Peter R. 1924. Fillmore. Els. Cambridge university Press. Frame semantics and its validity for linguistic descrip- tion. Fillmore.). Roman. 1-42. 1974. fire and dangerous things. Marvin. A framswork for representing knowledge. Bloomington.). An alternative to checklist theories of meaning. 1-25. Gordon. 1987. London. Linguistica et Anglica Gedaniensia 2. Paris. 76-138. New York. Topics in lexical semantics. 83-95. 1924. W.University College London Authenticated Download Date | 5/9/18 6:49 PM . (ed. Friedrich. Johannes et al. Amsterdam. Charles F..). Cole (ed. Kälisz. Cambridge. Charles F. Roger. Suzanne. Semantics 1. North Holland. Metzing. 1977a. Karol Todrys (eds. Mouton. Farkas. Wesley M. Charles P. Jacobsen. 1981. 1953. Dieter (ed. 1982. John. London. Fillmore. Boger W. Berlin. Chicago Linguistic Society. Ipsen. Stand und Aufgaben der Sprachwissenschaft. In: Donka Farkas. Chicago. 1981. 1953. Benjamins. Günther. 1978. de Gruyter. Minsky. La methode en lexicologie. Heidelberg. Charles F. Wesley M.). Brought to you by | UCL . Prototype semantics: The English verb lie. 1978.). Winter. Papers from the parasession on the lexicon. 26-44. 1974. Lehrer. The notion of lexical field and its application to English nouns of financial income. Jacobsen. Karol Ibdrys (eds. Lutzeier. Fillmore. Language 57. Women. 123-34. University of Chicago Press. Frame conceptions and text understanding. Semantische Studien im Sinnbereich der Schnelligkeit. 46 REFERENCES Brown. 1974. 1958. George. 200-237. 1975. 123-131. Donka. 55-81. George.

Winter. 1974. Mam. 1977. 1967. Ihe principles of semantics. Jackson. Leon. Jost. Beiträge zur Deutschen Sprache und Literatur 58. EWN. 47 Borzig. Adrian (ed. Leo. North Holland. Zampolli. Jef. Jezyk a poznanie. Weisgerber. Glasgow. 70-97. 1957. 79-93. Problems of lexical semantics.University College London Authenticated Download Date | 5/9/18 6:49 PM . Hei- delberg. 1934. 1931. Amsterdam. Meisenheim. Stephen. Brought to you by | UCL . Hain. Linguistics 137. Walter. Die Sprachfelder in der geistigen Erschließung der Welt. Lingua 53. Schaff. 1964. New York. The theory of semantic fields. Vassilyev. Warszawa. 1954. 1981. Trier. Verschueren. Der deutsche Wortschatz im Sinnbezirk des Verstandes. Ullmam. Linguistic structures processing. Oxford.). Wesenhafte Bedeutungsbeziehungen. 317-351.

University College London Authenticated Download Date | 5/9/18 6:49 PM .Brought to you by | UCL .

49 SECTION 2 Brought to you by | UCL .University College London Authenticated Download Date | 5/9/18 6:49 PM .

at any rate. therefore. plau- sibility of their theoretical constructs. Coseriu mentions an empirical difficulty that arises if one considers the vocabulary of such a functional language to constitute a sin- gle system or. Geckeler (1981). Due to the very large number of units it seems impossible to deal with the whole vocabulary of a language from the beginning. a 'system of systems' (1978: 1O5).University College London Authenticated Download Date | 5/9/18 6:49 PM . Brought to you by | UCL . Coseriu. I will try to show that the structural semanticists' analytical procedure of setting up minimal lexical oppositions in order to arrive at meaning-differentiating semantic features is inadequate to describe complex semantic information inherent in lexical items. suggests starting out by setting up relatively simple partial systems that may be integrated later into For a summary of Coseriu's version of structural semantics. Such de- scriptions require approaches which in modelling the semantic competence of native speakers give priority to the psychological reality or. but not to a historical language as a whole (1978: 91). Having emphasized at the outset that structural methods of investigation can only be applied to the individual 'functional language1 in the sense of more or less uniform lin- guistic systems within a historical language. cf. a proposal on which his theory of structural semantics seems to be mainly based. the references given there and his paper in this volume. One of Coseriu's early treatments of the parallelism between phonology and lexi- cal semantics can be found in his article of 1964/1978.50 HANS ULRICH BOAS THE INTERNAL STRUCTURE OF LEXICAL ENTRIES: STRUCTURAL AND/OR 'DEFINITIONAL·' SEMANTICS This paper will be concerned with a critical evaluation of Coseriu's proposal to transfer the structural method developed in phonolgy as a model to the analysis of lexical meaning. at least.

i. by applying the well-known types of dis- covery procedures of segmentation and classification to the speech flow of an 2 unknown language. (1978: 107/8) Notice already at this point that it seems objectionable to speak of 'perfect comparability1 of the phonic substance and substance as the experience of re- ality. 3 Cf. the object language. denn in diesem Sinne sind die Organisation der Er- fahrung der Wirklichkeit durch lexikalische Einheiten und die Organisation der phonischen Substanz durch Phoneme vollkommen vergleichbar. are different. be arrived at without having access to its semantics. in principle.e. The use of capitals in writing down semantic features may signal their intended metalinguistic status. no way of defining semantic features as ele- ments of the metalanguage by means of extralinguistic correlates. there is no finite list of minimal mean- ing-differentiating features in terms of which all lexical items of a language 2 Cf.e. There is. i. versteht.University College London Authenticated Download Date | 5/9/18 6:49 PM . He then tries to substantiate this claim: Wenn man unter "Struktur" die Abgrenzung und Organisation einer Substanz mittels funktioneller Einheiten.e. correlates in terms of which all phonemes of all languages are supposed to be exhaustively characterizable. die in verschiedenen Sprachen verschieden sind. can only be determined if the language under consideration and the meanings of its lexical items are known by the analyst such that minimal meaning differences can be named and identified. von einer "lexikalischen Struktur" zu sprechen. Lexical structures.e. In addition. imme- diate paradigmatic oppositions between lexical items. i. or without recourse to. moreover. i. but does not really solve the problem of defining them independently of the object language. the mono- morphemic words of a language. 51 more oonplex ones (1978: 1O6). Being aware that whole vocabularies cannot be claimed to be organized like systems of phonemes. i. ist man zweifellos berechtigt.e. Brought to you by | UCL . Newmeyer's (1981: 6) remarks on structuralist methodology. The elements of the metalanguage in which phonological de- scriptions are couched such as distinctive features have pretty clear extra- linguistic. The phonemes of a language can. In all these respects the primary units of the vocabulary. Sprengel (198O) and Kastovsky (1982: 1O6). articulatory and/or acoustic. inde- pendently of. he puts forth the weaker claim that the vocabulary of a language contains structures that are similar to phonological ones and can thus be dealt with in an analogous fashion.

which again form a minimal opposition constituting the dimension MATURITY specified by the features ADULT and YOUNG. Kastovsky's (1982a) attempt to justify the use of 'privative oppositions' as developed by Praguian phonologists for the analysis of lexical meaning.(Kastovsky 1982a: 32) Neither the semantic features MALE and FEMALE. Take. 4 Coseriu's views on this point seem to be somehow inconclusive: "Daher kann man auch nicht wie in der Phonologie ein Inventar kleinster di- stinktiver Einheiten aufstellen und behaupten. The residual meanings which remain after the extraction of these features are equivalent to the mean- ings of the lexical items adult and child. Brought to you by | UCL . Die Methoden und Ergebnisse der Phonologie sind also nicht ohne Modi- fizierung auf die Semantik übertragbar. die man zwischen phonologischen Einheiten und Lexemen feststellt. ist in dieser Hinsicht nicht weniger evident. for example. it retrains totally obscure how such analyses could be carried out given that it is impossible for native speakers to have intuitions about that part of the meaning of a lexical item that is left after the ex- traction of one or more semantic features. But even if unanalysed residual meanings are admitted in principle. der Wortschatz bestünde aus einer feststellbaren beschränkten Anzahl dieser kleinsten Elemente. 5 This is suggested in Coseriu (1973: 15). SEX. kann man ebenfalls von 1 lexikalischen Strukturen" sprechen. This is obviously incompatible with: "Wenn man unter "Struktur 1 die Tatsache versteht. daß die funktionellen Einheiten restlos in differentielle Elemente ('unterscheidende Züge 1 ) zerlegt werden können. 6 Cf. It appears that the same types of problems persist even if the analytical and descriptive machinery of structural phonology is modified and applied only to small partial systems of the vocabulary. viz. They result in the semantic features MALE and FEMALE specifying this dimension. nor the residual meanings that allegedly remain after the extraction of these features are independently moti- vated." (1978: 111-112). denn die Analogie. In surmarizing this theory he discusses lexical oppositions: Man:woman3 boy:girl represent minimal oppositions and involve only one semantic axis or dimension.52 4 could be described exhaustively. He claims that "any semantic analysis of lexical items must ultimately be based on a notion of opposition in the Saussurean sense" (1982a: 31) and adopts as a starting point a slightly modified version of Coseriu's theory of lexical fields. They are based solely on the semantic intuitions of linguists qua lin- guists. also Sprengel (198O: 158/9)." (1973: 15).University College London Authenticated Download Date | 5/9/18 6:49 PM .

apart fron semantic features such as MALE and FEMALE or ADULT and YOUNG. the nature of which is. in turn. without any doubt. it is also the criteria for distinguishing these levels in the first place (Kastovsky 1982: 1O9) and for applying other principles of structural semantics that remain mysterious. the semantic feature ADULT which. besides the feature ADULT. to which.and object-linguistic level that is far from clear. Kastovsky discusses at least one problematic aspect of this type of analysis: Obviously. for example. just in case there exist primary lexical items in the language whose meanings are judged to be equivalent to the respective residues (Kastovsky 1982a: 33). Adult involves. Brought to you by | UCL . that such lexical items as person or baby are not included in this analysis. items which might prove the notion of minimal opposition to be too simplistic? In a footnote. if so. The metalinguistic use of lexical items as designations of semantic features must therefore not be confused with the object-linguistic meaning of these designations. etc. must be assumed to be different from the lexical item adult (as a noun) seems to be identical with the object linguistic adjec- tive adult except for being capitalized (but cf. an inherent relationship between these two levels. For the status of semantic features. Lipka (1972: 42 f f . however.University College London Authenticated Download Date | 5/9/18 6:49 PM . and should therefore be excluded from the primary paradigmatic structures of lexical fields. 391). Why is it. namely the distinction between primary and secondary vocabulary which. This follows from a well-established principle of structural semantics. the feature HUMAN. one has to assume a difference between the lexical item adult (as a noun) and the semantic feature ADULT. 53 There are also no established discovery procedures or transfer rules to find out whether the residual meanings are equivalent to the meanings of any lexi- cal items and. then this noun is obviously derived from the underlying adjec- tive adult. cf. Thus. Kastovsky 1982: 1O8). although there is. also the feature HUMAN. besides the feature ADULT. The method of setting up immediate oppositions can be recursively applied to the residual meanings. Geckeler 1981: 382-383. In this way it yields. such semantic dimensions as SEX. according to Kastovsky.. (1982a: 32) It is not only the relationship between the meta. which are said to be specified by the relevant semantic features. ) . But if it is true that the noun adult also involves. far from clear. is among the seven preliminary distinctions drawn by Coseriu in the attempt to delimit the object of his semantics. MATU- RITY. to reduce the great number of lexical items and to arrive at a homogeneous object of investi- gation (Coseriu 1973: 16ff.

that a syntactic form must be imposed on the semantic features contained in the dictionary. He showed that there are numerous Simplexes "which are not definable without a configuration of features" (446). that nest of the work done within Coseriu's brand of struc- tural semantics is concerned with the internal structure of lexical fields and word-fields. The 'sentencehood1 of dic- tionary definitions demonstrates that he is not conmitted to the structuralist discovery procedure of setting up minimal oppositions between morphemes. 'sememes' and 'archisememss'. i.e. A reason for this could be that Coseriu seems to analyse the meanings of pri- mary lexical items as simple heaps or clusters of semantic features (Ooseriu 1973: 54).". (Weinreich 1966: 447) C f . but must take place by trial and error. which play a decisive role in the overall structuring of the vocabulary in terms of lexical fields or word fields (Kastovsky 1982a: 33-34). that of oat) by means of any feature mechanism is also a good reason to search for alternatives. sentences as analytic. Cf.e. the analysa- bility of functional units into differentiating features] bedeutet allerdings nicht. at the sane tine. delimiting 'lexemes'. Since the definitional status of a sentence. "The awkwardness of representing some meanings (e. antonymy. converseness. . to indicate meaning relations between lexical items such as hyponymy. oder daß sie durch Zusammensetzung von schon gegebenen Merkmalen entstehen. Coseriu (1976: 19): "Dieses Korollar [i. for example.University College London Authenticated Download Date | 5/9/18 6:49 PM . however. daß Einheiten aus Merkmalen bestehen. or even as definitions of morphemes.g. 'semes'. 54 In Kastovsky's approach. 'archilexemes'. however. Weinreich's chief result was the finding that "complex semantic information is stored in the dictionary of a language in forms of the same type as it assumes in sen- tences" (1966: 471). the isolation of definitional sentences cajinot be reduced to a procedure. i.e. or even its analyticity. is not self-evident from its structure." (Weinreich 1966: 473). Brought to you by | UCL . semantic features are thus taken to characterize the internal semantic structure of individual lexical items and. The internal structure of individual items themselves has not received much attention. Weinreich's concept of the internal structure of morphemes is thus incompatible q with the one advocated by structural semanticists. Notice. 8 Biis view is at variance with the results of Weinreich's (1966) investigation into the semantic representations of lexical items within the framework of a standard-theoretical model of generative transformational grammar. Wein- ipeichis interested in reconstructing and explicating the ability of the ordi- nary native speaker to recognize.

'langue' as a ccrtmunity possession (de Saussure 1916." (1966: 447).e. and an investigation of reducible circularities and irre- ducible. While the lat- ter believe in the applicability of the sane discovery procedures of permuta- tion and Genmutation to different idealized substances. i. I would like to propose that the differences between the kinds of approaches to linguistic semantics advocated by Ooseriu as against Weinreich are ultimately due to distinct sets of primary generalizations in the sense of basic assumptions of the general linguistic theory they subscribe to (Boas 1984: 7O).. to the "Substanz der Ausdrucksebene" as well as to the "Substanz der Inhaltsebene" (Geckeier 1971: 195). Thus. 3Off.he still subscribes basically to de Saussure's view that "dans la langue il n'y a que des differences sans termes positifs" (1916: 166) and to his conception of 'langue' which emphasizes the social as- pect of language.Coseriu distinguishes four levels of struc- turing: "type1. Ihe two most basic parameters have to do with the choice of the subject matter or domain of theories and with the types of ideali- zations that are adopted. 55 As problems of his theory.University College London Authenticated Download Date | 5/9/18 6:49 PM . a proof that every morpheme has a unique optimal definition. Weinreich mentions "a proof that every morpheme in a language has at least one definition. 'norm1. i. . Brought to you by | UCL .e..). c'est une grande illusion de ΙΟ Cf. It should be obvious that this program of research is different in principle from the structural semanticists' one. La collect!vite est necessaire pour etablir des valeurs dont 1'unique raison d'etre est dans I 1 usage et le consentement general.. mutually interdependent sets of semantic primitives in the network of definitions. Ihis conception of 'langue' as a social phenomenon is logically linked to other cornerstones of Saussurean structuralism such as Ί 'arbitraire du signs' and 'la valeuv Itnguistique': . 65) and Kastovsky (1982: 2 3 ) .. and 'discourse1 where 'norm' is the level of what is traditionally (or socially) fixed and not necessarily functional (Geckeler 1981: 388-389) . "system1. To express it in more fashionable terms. Weinreich admits that "there is no known discovery procedure for correct semantic descriptions" (1962: 26) and suggests "Ideally a description is adequate if it supplies us with overt means for approximating the intuitions of native speakers about the semantic relationships of words in their language" (1962: 26). le fait social peut seul creer un Systeme linguistique.'parole' . despite Coseriu's modifications of the Saussure- an dichotomy 'langue' . also Coseriu (1973: 58. the theories in question fix cer- tain parameters differently.

can be viewed as a series of con- tiguous subdivisions marked off on both. Fran such assumptions the analytical pro- cedure of setting up minimal pairs follows naturally.the linguistic competence of the idealized native speaker who is supposed to live in a homogeneous speech-comnunity (Boas 1984: 14-15) never- theless encompasses all these phenomena including elements traditionally fixed 11 Cf. In such theories analytical pro- cedures like the minimal pair principle cannot be derived. 389) . in turn. It also appears that the psychological premises of Saussure's conception of 'langue' mentioned above (cf. is the individ- ual aspect of language."notre pensee n'est qu'une masse amorphe et indistincte" (1916: 155) . for example. 12 C f . It is part of this ability to use his language as a metalanguage (1966: 447) as. alors qu'au contraire c'est du tout solidaire qu'il faut partir pour obtenir par analyse les elements qu'il renferme. on the other hand. (de Saussure 1916: 157) The linguistic system 'la langue'. While Coseriu constructs his homogeneous object of investigation 'functional language1 by excluding such phenomena as terminologies. for example. also "La langue elabore ses unites en se constituant entre deux masses amorphes" (de Saussure 1916: 156). would not belong to the func- tional system . the indeterminate level of thoughts . ce serait croire qu'on peut commencer par les termes et construire le Systeme en faisant la somme. 1916: 155) as well as the concept of 'Substanz der Inhaltsebene' are unten- able by today's standards in view of the results of psychological research in the last seventy or eighty years. ce serait l1isoler du syst me dont il fait partie. Differences in idealization are also evident fron the distinct assumptions concerning homogeneity. Jackendoff (1983) and the papers on 'prototypes' in this volume.and the equally vague level of sounds. who idealized 'langue' as a social institution by abstracting away from individual instances of 'parole' (Boas 1984a). In contrast to Saussure. 385.e. the linguistic theory underlying Weinreich1s approach idealizes the native speaker as an individual. .University College London Authenticated Download Date | 5/9/18 6:49 PM . Brought to you by | UCL . when he zeros in on the correct defi- nition of a lexical item (1966: 447). for example.56 considerer un terme simplement comme 1"union d'un certain son avec un certain concept.morphologically conditioned allomorphs of historical languages. metalinguistic usage of the primary language and the linguistic norm (Geckeler 1981: 382. i. Le definir ainsi. Rosch (1977). Ίΐιβ subject matter of Weinreich1 s approach. the linguistic ooqpetence of the more or less ide- alized native speaker.

semantic primitives. Cf. which state the semantic competence of native speak- ers. Wilks (1977).e. Ihis is particularly evident in a recent 'definitional1 approach to lexical semantics. i. 57 in this speech-community such as irregular plural and past tense forms of this historical language. they subscribe to some version of mentalism as against the alleged objec- tivity of discovery procedures applied to homogeneous objects of inquiry. But such an assumption is totally unfounded. It is therefore a structuralist fallacy to conclude. "Ihe presence or absence of a back support as in ahair vs. In his mentalistic theory." (1979: 16). 14 For a survey of arguments about 'semantic primitives'. for example. stool can be observed." (1985: 335). cf. namely Wierzbicka's (1985). Brought to you by | UCL . or at least plausibility.University College London Authenticated Download Date | 5/9/18 6:49 PM . Since her investigations demonstrate that the meaning of concrete concepts such as human artefacts must be described primarily in terms of their function and not of their outward appearance (1985: 335) she criticizes feature analysts like Pottier (1965) for using simplistic questionnaires of binary features which are assumed to be relevant to all the words in a given domain. 13 A case in point is. are expressed in a metalanguage " which is derived exclusively from nat- ural language and which is. in addition. i. for example. their 'distinctive' features. Cf. of their theoretical constructs. priority to the psycho- logical reality. therefore. Speaker-hearer-oriented theories give. Wierzbicka considers meaning to be a conceptual structure and not a sum of shared properties of denotata (1985: 260).e. also Sapir (1933). Hie lexicon of this metalanguage consists of a small set of indefinable 14 expressions" (1985: 9). In following Weinreich's program of research her definitions. but the intended function can't: it has to be discovered by assiduous mental experimentation. intuitively understandable and verifi- able. Stampe's (1979) theory of natural phonology. that the meaning of 'parent1 must be included in the meaning of 'mother' because the set of denotata for 'mother' is included in the set of denotata for 'parent1 (1985 : 26O). under- lying segments are considered to be ontologically of the same status as any segment in surface representation.. for example. He rejects de Saussure1s and the Prague School's view that "phonemes are functional entities characterized solely by those proper- ties which distinguish them from other phonemes in a language. the morphophonemic subcomponent of Chomsky's early generative models or the treatment in Chomsky/Halle (1968).

sentences that are part of the native speaker's ability to use his language. i. struc- ture. syllables.e. i. In autosegmental phonology segments are no longer conceived as unordered sets of specified features. analytic. explores the hierarchical organization of segments into larger units. Brought to you by | UCL . suprasegmental structure is postulated.e. Metrical phonol- ogy. The results of this paper can be summarized in the following way: There are systematic discrepancies between phonological and semantic structure concerning the feasibility of the main discovery procedure of setting up mini- mal oppositions. 'higher levels'. have prompted generative phonologists in recent years to modify and extend the standard view of phonological representations (Chansky/Halle 1968). i.58 Coseriu's postulation of archilexemes parallel to archiphonemes in phonology has also been called into question. Moreover. Thus. subsegmental. i. To deal with tonal phenomena which can spread over several segments. In concluding these arguments against using structuralist methods in lexical semantics it should be pointed out that the same kinds of phenomena which could not be analysed satisfactorily by structural phonologists. It requires definitional. even the 'phonic substance1 appears to be 'structured' in ways not accessible to simplistic discovery procedures. 'feet1 and phonological words and their relation to stress (van der Hülst/Smith 1982)." (1983: 69. the overlappings between lexemes are so com- plicated "and the fields merge so frequently into each other that any crite- rion of sharp delimitations would seem to run contrary to the linguistic pic- ture presented by the verbs.e. pitch and juncture. 15 A similar conclusion is arrived at by Gauger (1972). namely stress. and concerning the nature and number of units of the respective metalanguages as well as their modes of application. which is open to empirical investigation.University College London Authenticated Download Date | 5/9/18 6:49 PM . on the other hand. but as having internal. also Wierzbicka 1985: 74). Thus. Capturing complex seman- tic information inherent in lexical items requires more than heaps or collec- tions of semantically distinctive features that have been extracted fron the elements of an over-idealized object language by means of discovery procedures taken over from phonology.e. the results of Snell-Hornby's (1983) extensive study of descriptive verbs in English and German neither confirm that the basic unit of a field is necessarily an archilexeme corresponding to the entire semantic content of the field nor that all lexemes must be accommo- dated within a field. cf.

They may thus stand a better chance of being incorporated into models of artificial intelligence. the lin- guistic competence of speakers including their metalinguistic ability to de- fine the lexical items of their language. on the other hand.e. whose psychological premises are no longer tenable. try to capture and describe the individual aspect of language. i. Mentalistic approaches like Weinreich's and Wierzbicka's. Brought to you by | UCL . 59 Structural semanticists obviously still subscribe to a notion of 'Substanz der Inhaltsebene' or 'Inhaltskontinuum' which is logically related to Saussure's concept of 'tongue' as a social phenomenon and to the correlative concepts of 'substance1 and 'form 1 .University College London Authenticated Download Date | 5/9/18 6:49 PM .

New York. 198O. by Dieter Kastovsky. Mass. Semantics and cognition. Hannes Rieser (eds. Structural semantics. (R)REST . 1987. Geckeier. Hans-Martin. 9O-163. Hannes Rieser (eds. 1975.. Hans. Coseriu.). (Tübinger Beiträge zur Linguistik 40. Eugenio. New York.) ed. Harry. Strukturelle Semantik und Wbrtfeldtheorie. Dordrecht. Reflections on language. 1962. Niemeyer. Householder. Tübingen. Bedeutung als Semstruktur? Vox Romanica 31. (Linguistische Arbeiten 150.) Berlin. 1968. Fink. Bouvier Verlag Herbert Grundmann. 1981.-3.). 665-676. Benjamins. Problems in lexicography. Gauger. Hans-Josef Niederehe (eds.). 1973. Hans Ulrich. Eikmeyer. Hans Ulrich. Morris Halle. Coseriu. Strukturelle Bedeutungslehre. Ind.).). Band 2. Words. Jahr- buch 1975 des Instituts für deutsche Sprache. Formal versus explanatory generalizations in generative transformational grammar. Narr. Fred W. 1984a. rfoam.).. Probleme der Strukturellen Semantik.).) Bonn. Boas. de Gruyter. Kelly. Indiana University Linguistics Club. Amsterdam. worlds. 24-39.Dezember 1977. Bloomington. Geckeier. Fontana. Coseriu. Eugenio. Chomsky. Beiträge zum Wuppertaler Semantikkolloquium vom 2. (Research in text theory 6. The sound pattern of English. 1971. Foris. 1978. 19-23 August 1984. Hans-Josef Niederehe (eds.). Die funktionelle Betrachtung des Wortschatzes. louis G.Chomskys Annäherung an Saussures 'langue1- Konzeption?. Horst. 1981. Horst (ed. Schriftenreihe Linguistik. Papers in the history of linguistics. 1972. Perspektiven der lexikalischen Semantik. and contexts. Ray. München. (Wege der Forschung 426. 60 REFERENCES Aarsleff. Jackendoff. 1982. Noam. Chomsky. 381-413. In: Hans-Jürgen Eikmeyer.University College London Authenticated Download Date | 5/9/18 6:49 PM . Wissenschaftliche Buchgesellschaft. London. (Gesamthochschule Wuppertal. Kastovsky. Sol Saporta (eds. Louis G. 7-25. Harper & Row. Für eine strukturelle diachrone Semantik. 1978. Geckeier. Cambridge.) Tübingen. Philadelphia. Boas. Norval Smith (eds. Proceedings of the third international con- ference on the history of the language sciences (ICHoLS III). New York. MIT Press. 1976. Dieter (ed. Princeton. van der Hülst.).) Darmstadt. In: Horst Geckeier (ed. Hans-Jürgen. Eugenio. Horst. The structure of phonological representations (part I). In: Hans Aarsleff. Kelly. 1983. 1984.. Brought to you by | UCL .

. 198O. Anna. Communication & Cognition 10. 1977. Academic Press. Leonhard. Karoma. 22-39. Bernard. 1982. Neil (ed.). Oours de linguistique generale. Sol Saporta (eds. Householder. Pottier. Wilks.) Heidelberg. Lexicographic definition in descriptive semantics. Academic Press. Bagel. Yorick. Ferdinand de. 247-265. Good and bad arguments about semantic primitives. 29-45. Rasch. Saussure. David. Fink. Indiana University Linguistics Club. München. Semantic structure and word-formation. Stampe. Dieter. Volume 3. Lipka. Paris. Thomas A. Publie par Charles Bally et Albert Sechehaye avec la collaboration de Albert Ried- linger. Current trends in linguistics. Eleanor. In: Thomas S. Mouton. Konrad. 1916. Edward. Snell-Hornby. Sebeok (ed. 1983. Verb-descriptivity in German and English. Bloomington. Winter. etc.. 1979. Über semantische Merkmale. Warren. In: Dieter Kastovsky (ed. 1-49. Ann Arbor. 145-177.). 181-221. Mary.. Sapir. Volume 1. la definition semantique dans les dictionnaires. Wierzbicka. 1966. London. Nevireyer. Journal de Psychologie Normale et Pathologique 30. (Angli- stische Forschungen 158. 1982a.). Brought to you by | UCL . 395-477. Uriel. Düsseldorf. Studies in cross-cultural psychology. 1933.). Travaux de Linguistique et de Litterature 3. 1972. La realite psychologique des phooänes. Studia Anglica Posnaniensia 14.University College London Authenticated Download Date | 5/9/18 6:49 PM . 1977. Payot. In: Neil Warren (ed. Pädagogischer Verlag Schwärm. Explorations in semantic theory. A dissertation on natural phonology. Dieter. 25-44. 198O. 61 Kastovsky. etc. Sebeok. 'Privative opposition1 and lexical semantics. Sprengel. 1966. 1985. New York.). Ind.). 1962. Weinreich. Human categorization. 1977. 1971. Francke. Wortbildung und Semantik. Frederick. Weinreich. Kastovsky. 1965. Bern. In: Fred W. The Hague. Verb-particle constructions in contemporary English. München. Linguistic theory in America. Lexicography and conceptual analysis. Uriel. (ed.

letting not only the sen- tence contribute to this set of inferences. de- pending on the sentences and contexts in which it occurs. should comprehension of new senses be possible? Conversely. but also the context. A problem with this proposal for the notion of word meaning is that a word can take in principle an arbitrarily large number of such contextual meanings. or so it seems. and indeed many more things than a speaker could possibly ever have experienced. second. and perhaps impos- sible. A reasonably practical stipulation for what the mean- ing of a word could be in a particular context of use. Hence 'knowing a word' or 'knowing what a word means' must be much more. if you like. is this: The meaning of a word is the contribution the word makes toward the inferences that can be drawn from the sentence in which the word occurs in a particular context of use. as it were.62 PETER BOSCH ON REPRESENTING LEXICAL MEANING 1. one might want to argue that if indeed the representation of word meaning is rich enough to already 'contain'. however. the result of. Still. than just 'having a represen- tation of the sense(s) of the word that have already been experienced1. to make explicit. it only means that a word can mean very many things. any new senses a speaker may encounter. this multiplicity of word senses does not form an obstacle for ordinary cotrprehension. For how. Introduction The general notion of the meaning of a word is difficult. the extreme case just one case of successful Brought to you by | UCL . This is. This does not mean that a word can mean just anything. then the problem of arriving at these rich representations on the basis of a limited input . otherwise. and.University College London Authenticated Download Date | 5/9/18 6:49 PM . turning the truth condi- tions of a sentence into the more operational notion of the set of inferences that can be drawn from the sentence. aggravated. To the extent that the formal model must be finitely specifiable in terms of one or the other formal language. the very fact from which our problem arises. is a plain absurdity for the structuralist. It can do so.University College London Authenticated Download Date | 5/9/18 6:49 PM . For most structuralist approaches and for many formal accounts of lexical se- mantics. 63 usage . is constituted by the relations of similarity and contrast between the meaning of one word and others. only at the price of ignoring language use. The meaning of a word. we are. But it is precisely relations of reference that account for the potential infinity of the set of interpretations that can be given to a word in actual language use. or because it is not considered as something of a dynamic nature that links past linguistic experience with current and future use. perhaps surprisingly. not con- cerned with reference semantics at all but with translation semantics: expres- sions from a natural language are given a semantic interpretation by being trans- lated into another language. the problem is not a problem for current varieties of formal seman- tics and. This problem. the problem does not arise. concentrates on language use rather than on the language system. and corre- spondingly we may choose between poor and rich lexical representations. More specifically.however. is not a problem for all approaches to lexical meaning. but to a formal model that figures as a substitute for reality. Either because language use is never even considered anyway. Mainstream structuralist! thus escapes the problem by not considering reference. I had better first say what exactly I mean. for a very similar reason. in the creativity of re-application or in learning. not in an empirical sense. it considers reference only in a formal. the formal language that specifies the relevant model. And since this is the perspective from which I want to consider the problem.e. b. It considers language use Brought to you by | UCL . In other words: we may choose where we want to locate the problem. The reference is not to anything in the intersubjective world. The problem becomes real though for a cognitive science approach to natural language. Similarly. Although formal se- mantics is professedly reference semantics. for the structuralist. strictly speaking. the multiplic- ity of interpretations a word receives in actual use. The cognitive science approach to natural language a. i. Relations of reference would have no direct influence on any kind of word meaning.

white fish. white chocolate. I shall avoid the hassle of explicitly specifying contexts by simply placing the adjective together with different nouns which. 64 (i) as representational. To give some body to this notion and thereby to the problem that understanding a word in one context by no means implies understanding the same word in another context. 2. in the sense that all processes that form part of language use must be effectively realizable within finite time and with finite memory resources. it makes an attempt to account for the. Take the interpretation of the adjective white in a number of different contexts. And. white hair. it makes a serious attempt to set up func- tioning models of natural language use that could be models of psychological mechanisms. An absolutely minimal requirement for this is the ccnputability assumption (cf. What is it then that a white wall. white coffee. despite grave differences in outlook. and thus allowing for an empirical theory in linguistics. about the intersubjectively shared . in the sense that language/ at least in its primary uses. it is certainly not the case that all these things somehow are of (dif- ferent shades of) the same colour. white wine. It requires besides linguistic also psychological plausibility of its hypotheses. Furthermore. hint at particular types of contextual!zation.and in this sense 'real1 . functions in cxranunication about the world. no cognitive science approach to linguists would even be thinkable without the work of European as well as American Structuralism. and the white race have in common? . If white wine has the same colour as white Brought to you by | UCL . The ancestors of the cognitive science view are found in American Pragmatism and in varieties of Philosophical Behaviourism. b ii). c. white paint.Whatever it is. to my mind. Reasons for choosing this rather than any other perspective may be found in the fact that.University College London Authenticated Download Date | 5/9/18 6:49 PM . first and foremost. let us consider an example. as it were. apart fron concentrating on language use or linguistic behaviour. and of course logical and philosophical seman- tics. (ii) as The problem The problem that I am addressing becomes manifest in the contextual variation of word meaning. fundamental intuition that linguistic utter- ances are. as far as the kind of outlook is concerned.

The point is nicely illustrated.e. The point is not so much the ob- jects having a particular property. traditionally out of bounds for the structuralist. white chocolate. first of all. won't become any whiter than they already are. on the other hand. Note. that could be learnt one by one. or combination of properties. 65 coffee.the white wall. then a dirty glass is about the only conceivable explanation. i. and the notion of an indefinite number of interpretations is supported. The list can be extended ad nauseam (Bosch 1985a. by an experiment reported in Olson (1970). of course. by the same English adjective. And here. Table 1: goal contrast set description Δ Z7 the triangle Δ Δ ^ the equilateral one Δ £i D^ the big one Δ A^ the small one Brought to you by | UCL . of the form white x. and the white man.University College London Authenticated Download Date | 5/9/18 6:49 PM . we are back with structuralist principles . white. all referred to. And as for white fish. Note further that we are not just concerned with a limited number of fixed expressions. that the problem here is not one of vagueness: we are not just concerned with different shades of white. in suit- able contexts. but the relevance of these properties for distinguishing the intended object from other objects in the same situation (Bosch 1985b).but new on a referential territory. there is nothing white about it until it has been cooked and the skin has been taken off . even when cooked and skinned. as it happens. but with the attribution of quite different properties that are. b).

by way of a Gedankenexperiment.University College London Authenticated Download Date | 5/9/18 6:49 PM . ana were asked to name that object. and even though one may concede that also a description by the adjectives big and small need not be contradictory. as in Table 2. looking at the properties of the object that were re- ferred to in the descriptions. goes much further: it also influences the way the object is conceptualized. here given by the objects in the contrast set. which do not have the same status of notorious exceptions. by the descriptions the big one vs. and hence influences our judgment as to whether or not the object has a particular property and not just whether or not that property is currently useful for purposes of the individuation of the referent. the context. don't straight- forwardly refer to object properties anyway.66 Subjects ware presented with oras particular object (the one in the first col- umn of Table 1) in combination with a differing set of contrasting objects. But this im- pression is wrong. a first impression may be that they would even- tually add up to something like a full description of the object. The influence of the contrasting objects.e. but hinge upon a standard of com- parison. the small one. being relative adjectives. first of all. depending on what other objects it had to be distinguished from. Nothing can be both big and small at the same time. This is illustrated. Suppose we extend Table 1. the way it is mentally represented. But then one may argue that big and small. i. Table 2: goal contrast set description Δ ^fk JXUA the white one Δ Δ Δ the black one [dotted and double lines are to be visualized as solid single lines of red and green colour respectively] Even though one should concede that an object may at the same time be a triangle and big and equilateral. because we are deal- Brought to you by | UCL . Not surprisingly the descriptions chosen for the goal object would differ. Let me try and make my point then with the help of the adjectives white and black.

This latter form of 'meaning' I refer to as a 'contextual notion'. Linguists studying English verb-phrase anaphora and the form so do. if so does is to have the same meaning as Brought to you by | UCL . do so. have insisted that.This set of inferences is here being used to make the notion of understanding an utterance operational. Table 2 shows that both descriptions of the same triangle are not only possible but would indeed be the most natural ones in the contexts given. but it has also played a manifest and operational role in more recent grammatical theory. 1985a. The latter is what he makes of it in a particular instance of use. Although the term 'contextual notion1 is new. etc. Now consider a sentence like Fred likes her car and so does Pete. it is identity of meaning that is required for VP-anaphora (cf. one would be less inclined to admit that a tri- angle may be white and black at the same time. This is roughly the distinction Hermann Paul was draw- ing between the 'usual meaning1 (usuelle Bedeutung) and the 'occasion meaning1 (oocasionelle Bedeutung) of a word. Not only is it more than just foreshadowed in Paul's notion of occasion meaning. A CN. Making meaning more precise In order to be able to discuss the above observations in a somewhat more pre- cise framework. Plainly. let me draw a distinction between what one may want to regard as the dictionary meaning of a word and the meaning the word receives in a par- ticular context of use. which thus is already the result of an interaction between the dictionary meaning of the word. while for pronominal anaph- ora referential identity of pronoun and antecedent +«=>τνΐ<? to be required. knowledge conferred by the preceding text. Bosch 1983: 8O-83. Intuitively speaking. then is the contribution that this linguistic form makes towards the in- ferences that may legitimately be drawn from the sentence in which it occurs in that particular context of use. and references cited there). Still. the notion of meaning that it refers to is not. 3. and more general world knowledge. . 67 ing with relative adjectives. associated with a particular linguistic form in a particular context of use. the former should be seen as the meaning a conpetent speaker knows when he knows the word.University College London Authenticated Download Date | 5/9/18 6:49 PM . for short: a CM. the dictionary meaning of words with which it combines in the sentence. The point brought home by these considerations is that the very notion of a particular object possessing a particular property makes no sense without ref- erence to a particular context (Winograd/Flores 1986). knowledge of the situation.

. with Katz/Fodor (1963). CNs are situation-specific products of cognitive processes. and Fodor et al. However many we were to put into the dictionary. So far we have seen that there is a definite need for fine-grained contextual 'meanings' like CNs. What is cannon to all three proposals is their fundamentally structuralist goal: they are designed to account for meaning relations between words. The above considerations illustrate this point. is CNs). we still wouldn't have enough.68 likes her oar.representation in terms of meaning postulates. again. CNs are not the sort of thing that one would want to put into a dic- tionary. Johnson-laird 1983: 205ff. . 4.e. for the sentence cannot be interpreted any other way than that Pete loves the same wcman's car as Fred. what is intended is of course not the same dictionary meaning. of the expression likes her car. depending on the situation. The problem we must then turn to is what kind of lasting mental representation of word meaning there is. at least if we want to have an explicit account of the role of word meaning in linguistic comprehension. and most of them would be entirely useless except in one single situation. such Brought to you by | UCL .University College London Authenticated Download Date | 5/9/18 6:49 PM . And the reference of the enclosed pronoun her must be part of the 'meaning1. or rather the CN. respectively.representation of word meanings via decomposition in terms of semantic markers. how the speaker gets from his experience (of CNs) in natural language use to mental representations of what may appropriately be called "knowledge of word meaning' and how he gets from these semi-permanent mental representations to what he needs in natural language use (which. but the same CN. The three types of proposals are associated. (1980). And we have also seen that their place is not the dictionary. and how it interacts with CNs. Quillian (1968). for discussion and further references): . express or refer to indefinitely many different CNs. Clearly. Proposals for mental representations of word meaning Three major types of proposals have been made in the literature for the repre- sentation of word meaning in the mental lexicon (cf. The CN thus plainly depends upon context-specific refer- ence and hence it follows that indeed one and the same word or expression may. i.representation in terms of semantic nets.

which is. If I say that my right hand is the one I write with. referential representations with the help of which rea- soning about the world becomes possible. in terms of meaning postulates. Even more blatently particularistic is the identification of my right hand as the one on which I have a scar.University College London Authenticated Download Date | 5/9/18 6:49 PM . or that it is on that side of my body on which my heart is not. A clear example of this. at least in the form we find distilled in CNs. I just had to imagine I was in front of my granny's dresser and look on what side the picture was. And when I was in any other situation. The same is true if I rely on the fact that my right is bigger than my left one.e. then this is true of me and many other people. as far as I can see.e.). i. Hilary Putnam (1975) pointed out what he called the 'indexical' nature of word meaning (at least for certain types of words): the fact that certain kinds of words can only be learnt in connection with particular experience. independent of experience. at least in this connection. This is in fact all we need. contrast. though perhaps not entirely in the spirit of Putnam. would be words like right and left. etc. and there is.they still need to be linked to representations of actual expe- rience in order to provide an account for linguistic comprehension. probably necessary . 69 as synonymy. but the fact I rely on in knowing which is my right hand is still en- tirely a matter of world knowledge. to the situationally relevant CNs that are interactionally shared. i. in each comnunication situation. Brought to you by | UCL . What is required eventually is a link between word meanings and mental models and. which is in principle different for every speaker. no essential need for any kind of intersub- jectively shared abstract or general meaning representation (e. What I am trying to say is this: although there are only personal and particular ways of tying our word meanings to experience there must also be a way of linking them.g. though for other than the above reasons. One inportant consequence of this is that if abstract meaning representations are assumed . semantic markers. via mental models. There is. with actual experience. no para- phrase for either of these t*ro words that would eventually be independent of particular facts in the world. And what is lacking from all three proposals is an idea of how these meanings would relate to anything else but other meanings. or the way that I myself learnt where right and left is when I was a toddler: I learnt that when I was standing in front of my grandmother's dresser then the picture of my great-grandfather was on my right. and subordination. Jonson-Laird criticised in particular the lack of any rela- tion between these meaning representations and what he calls 'mental models of the world1.

e. Integrating experience and meaning If the above considerations are correct and CNs are indeed required. i. and although this model prob- ably comes close to real human comprehension behaviour in 'hard cases'. such procedures. Then. In a sense. There may be exceptions. but a matter of relating linguistic expressions to repre- sentations of experience. Bosch 1983: Ch. in particular for meaning relations. They have the distinct practical advantage of greatest conceivable flexibility and the theoretical advantage of linking current understanding to past experience. Questions luce: Do I need a sweater when I go out. then. the above questions are answered right just supposing you come from a Celsius rather than a Fahren- heit culture. Cats are animals.. are unnecessarily cumbersome for most forms of everyday communi- cation. And without at least some such relations being estab- lished. or even a coat? Is this warm enough for sitting on the terrace? Will it be hot enough for the beach? remain unanswered until you have calculated that 8O Fahrenheit is the same as 27 centigrade. To the extent that such sentences are true. you may understand each word in the weather forecast. but more about how a culture finds it convenient to Brought to you by | UCL . i.University College London Authenticated Download Date | 5/9/18 6:49 PM . but they cannot straightforwardly account for many of the observations that have occupied lexical semantics in the past. 3) to representations of successful com- munication situations in the past (Bosch 1985b). Although comprehension processes can definitely be modelled as processes of 'reasoning from cases'.7O Consider for illustration a situation where you are travelling abroad and you hear in the weather forecast that the midday temperatures will be around 8O Fahrenheit. 5. but what would still be missing is some link to your experience.then this does of course not imply that more abstract representations are either useless or non-existent. Dogs are four-legged is like Boys don't aryi neither tells us a great lot about reality. Once this step is taken. which we disregard in the interest of generality and sim- plicity. such as the understanding of poetry. This is not a matter of meaning relations. there is nothing that one could reasonably call 'comprehension1. as far as they are currently understood.e. as processes that directly relate the utterance that is to be understood and the corresponding situation (dr 'context model*. for the generic truth of sentences like Dogs are four-legged. Bachelors are unmarried men. Vfe are here concerned with largely normative assumptions that can in no way be directly derived from the observation of individual cases. they are not true (and are not understood) merely by virtue of (the sum total of) individual cases in our past experience.

like CNs. How such abstract meanings are related to fairly immediate representations of linguistic experience.University College London Authenticated Download Date | 5/9/18 6:49 PM . if the study of lexical meaning is to be an empirical study of human linguistic behaviour. and how one can be derived from the other is a problem I have said hardly anything about in this paper. Geurts 1985). like CNs. Brought to you by | UCL . What I have done is only argue that abstract 'meanings' alone are insufficient and that they need to be related to representations of actual experience. 71 systematize those aspects of reality that it is concerned with (Bosch 1979.

New York. Dordrecht. Von einem logischen Standpunkt: Neun logisch- philosophische Essays.. Minsky (ed. Quillian. Against definitions. 263^367. Podor.University College London Authenticated Download Date | 5/9/18 6:49 PM . M. A study of the role of pronouns in syntax and discourse. 1979. Anton J. Aspects of a cognitive theory of semantics. Wblf. Wien. Geer A. Generics. 257-273. David R. 1968. Meaning and the lexicon. 1979. Rene Dirven (eds. Hoppenbrouwers. 1963. The ubiquity of metaphor. 1983. Ein Nachwort.). M. Semantic memory. Cambridge. Language and thought. Brought to you by | UCL . 1985. Quine. Fernando Flores. Jerrold J. 1983. Mind. 1985b.) London. 1985.72 REFERENCES Bosch. 1986. Foris.). Jerry A. Philip N. Johnson-laird. M. 247-255. Winograd. Bart. Ma. 1968. 251-258. Hoppen- brouwers. Cambridge uni- versity Press. Ablex. Cambridge university Press.). Seuren.). 170-210. Ullstein. In: Willard Van Orman Quine. Cognition 8. 227-270. et al. Putnam. Paprotte. N. Bodor. In: Marvin L. Amsterdam. J. Synonymie im Kontext. Rene Dirven (eds. Psychological Review 77.. M. In: Hilary Putnam. M.. Aca- deme Press. 215-271. 1970. Anton J. Agreement and anaphora. 141-176. Willard Van Orman. 1985. The meaning of meaning. Norwood. Language 39. MIT Press. Frankfurt/M. Mental models. Hilary.J. Lexical meaning contextual.. Pieter A. (Cognitive science series. Hilary. Seuren. Bosch/ Peter. Metaphor in language and thought. The structure of a semantic theory.).). In: Geer A. Context dependence and metaphor. Peter. language. (ed.. 1975. Cambridge. Terry. Peter. 1985a. Benjamins. 161-172. Bosch. Marvin L. M. Weijters (eds. Katz. and reality. J. Minsky. 198O. 1975. In: Wblf Paprotte. Olson. Putnam. Geurts. Berlin. Weijters (eds. Cambridge. Journal of Semantics 4. ized. Beter. Jerry A. M. Bosch. Semantic information processing. Pieter A. Ross. Understanding computers and cognition.

for him. For a mini- malist. For him. but so are what many would consider to be quintessentially semantic properties. but others will be excluded. none of what we knew about dogs properly belongs to the meaning of dog. These nay be labelled. There would appear to be two varieties of minimalism. Let us first look at the minimalist positions. according to Bierwisch. One form of intermediate posi- tion is that which attempts to draw a theoretical distinction between a dic- tionary and an encyclopedia. for a maximalist. The answers to this question that one finds in the literature can be roughly divided into three types. it all does. 73 D. The other variety of minimalism is exemplified by the views of Wierzbicka (198O) However. intermediate and maximalist positions. The intermediate positions are those which hold that seme items or aspects of our knowledge about dogs will form part of the meaning of dog. It will be further suggested that these three interpretations are not only perfectly natural ones. not only is encyclopedic knowledge excluded from word-meaning. such as sense-relations like synonymy. One represented by the views of Bierwisch (1983). hyponymy and antonymy. is distinct from meaning.University College London Authenticated Download Date | 5/9/18 6:49 PM . All these. for convenience of reference. In this paper it will be suggested that each of these positions is valid in relation to a particular interpretation of the expression 'the meaning of a word1 (a different interpretation in each case). Take the word dog. MAN CRUSE WORD MEANING AND ENCYCLOPEDIC KNOWÜDGE The topic I would like to discuss in this paper is the question of the proper relationship between word-ireaning and encyclopedic knowledge. but are also theo- retically interesting and ought to be recognised in any account of word-meaning. conceptual structure. Brought to you by | UCL . Meaning consists rather in the mapping relations between lexical items and concepts. not between meanings. she adopts this position only with respect to certain lexical items. the mini- malist. are relations between concepts. easily accessible to intuition.

All agree. These terms are claimed by Wierzbicka. to conceive of anything that might be discovered that would cause them to say that cats did not exist. however. So cats do not exist. I have tried out on several generations of students the following story. not only would they continue to be called cats. Both types of minimalism permit changes to occur in encyclopedic knowledge without consequential changes in meaning (conceived in the relevant way). It is clear. Wierzbicka glosses the meaning of cat as "animal of the kind that we call 'cats''.University College London Authenticated Download Date | 5/9/18 6:49 PM . to function rath- er like proper names. mosquito. in one sense at least. or even radically modify its internal strucutre. namely. then there would presumably be no change in its meaning. natural kind terms do exhibit an extraordinary denotational stability (and the range of words that behave in this way is probably much greater than Wierzbicka seems to think).) Furthermore. So cats are not what we believed them to be!'. then people would simply say that it was not a real cat. etc. Ordinary speakers assent readily to the proposition that if cats were discovered to be robots. I then ask them wheth- er this discovery would cause them to exclaim 'Aha. Fran the point of view of Wierzbicka's conception of the meaning of a natural kind term. the meaning of cat would not have changed. 74 namely. virtually any change whatsoever in knowledge or belief concerning the referents of a word could take place without affecting the word's meaning. (Of course. but unless there were changes in the mapping relations between a lexical item and its associated concepts. I ask them to imagine that one day cats are discovered not to be animals at all. which is similar to the gloss she gives for John. the so-called 'natural kind1 terms. following Kripke. if only one cat were discovered to be a robot. I am not fully convinced that Bierwisch's 'conceptual mapping' picture of word- meaning is a natural one. that the names of natural species belong to this category: horse. I am persuaded. but that. people find it extremely difficult. I suggest that the denotational Brought to you by | UCL . although it might be. The answer is always unanimously in favour of the latter. For Bierwisch. Exactly what is to be included in this category varies from scholar to scholar. In fact. provided only that it continued to denote the same referential class. 'the person whom I call 'John 11 . that Wierzbicka's denotational conception of meaning is natural.that on this view. but highly sophisticated self-replicating robots. or "Aha. for instance. however. after all!'. daffodil. the meaning of a natural kind term consists simply of its ability to denote a particular referential class. if not impossible. Thus. advances in scientific knowledge might enrich a con- cept.

Ihis is quite a differ- ent interpretation of the expression 'the meaning of a word" from the purely denotational one. If two items do not differ in this way. The notion of semantic distance. too. and one that descriptive lexical semantics needs to take on board. by no means all words exhibit the denotational stability of natural kind terms. if we wish to describe the dif- ferences between the meanings of two near-synonyms in one language. he does not offer a characterisation of meaning from which maximalism follows naturally . demands a generous conception of what meaning is. There is. for instance. non-restrictive view of meaning like that suggested by Haas. Every seirantically distinct lexical item has a characteristic profile of affinities and dis- affinities (what Haas calls a 'semantic field 1 ). and of compound nouns. but it is clear from their answers that they take 'meaning' to include encyclopedic matters. between a dictionary and an encyclopedia. and not only for natural kind terms. a more positive. Briefly. however. then for Haas they are identical in meaning. even in principle. Syntagmatic affinities show up as more normal or less normal collo- cations within grammatical expressions (basically sentences). represented. or two near translation equivalents in two different languages. Brought to you by | UCL .that is to say. Haitian's method is basically not so much to make a positive case for maximalism . radical type of maximalism. X and Y. But I believe that it. which has been much used in psychology. Haas views the meaning of a word as its pattern of syntagmatic and paradigmatic affinities (these are not his terms) with all other words in the language.University College London Authenticated Download Date | 5/9/18 6:49 PM . he scrutinises proposals for distinguishing dictionary meaning from encyclopedic meaning and finds them all wanting.) The maximalist position sometimes takes the form of a denial that a non-arbi- trary distinction is possible. paradigmatic affinities appear as the degree to which two words from the same grammatical paradigm share the same patterns of syntagmatic affinities. (However. Subjects have no apparent difficulty in under- standing this question. by the views of Haas (1964). 75 interpretation of the expression 'the meaning of a word1 is generally valid. Ihis is the line taken in Kaiman (198O). and the understanding of diachronic shifts in meaning all require us to adopt a comprehensive. is a natural one. for instance. I shall return to this point later. or X and Z? 1 .rather. Subjects are asked questions like "Which are the closest in meaning. The interpretation of metaphors. A considerable virtue of Haas's proposal is that it offers a satisfying way of characterising the most inclusive realm of word-meaning. We need to adopt this global picture of word-meaning.

a dictionary is not a mere arbitrary assemblage of miscellaneous facts. I would like first to review briefly some previous proposals for motivating the dictionary/encyclopedia distinction. Is it the job of the lexicographer to provide a specification of the criterial semantic features of the words? One difficulty here is in deciding what is criterial and what is not. Intuitively. which are true (or false) by virtue of the way the world is. We may not wish to go as far as Quine (1960). Suppose we take strict logical necessity. cookery recipes. (Of course. A theoretical dictionary is not necessarily the same as an ordinary dictionary. shown by entailment. Consider the following three examples. however.University College London Authenticated Download Date | 5/9/18 6:49 PM . and 1a to mean 'John stopped singing on one particular Brought to you by | UCL . as our criterion. as a volume would be that contained only. 76 What are here called the intermediate positions seem to be currently somewhat unfashionable. and indicate why they are unsatisfactory. and it ought to be possible to give an account of this. John stopped singing / entails? 1b» John did not continue singing 2a. However. They certainly raise some difficult problems. but we may legitimately ask whether Kempson's characterisation of the semantic content of the former might not also be valid for the latter. I believe we must. Lesley is female Example 1 has for me the ring of pure logic. who argued that no sharp distinction could be made in principle between analytic statements. And what we require of the theoretical account of lexical meaning is that this lexicon includes a specification for each lexical item of the contribution that lexical item makes to the truth conditions of all the sentences in which it occurs. if we interpret 1b to mean 'John aban- doned his singing career1. take account of Lyons's claim (1981: 89-90) that a number of what are commonly assumed in the literature to be entailments are in fact merely inferences in which we have a relatively high degree of confidence based on our experience of the world. We may begin by considering the following quotation from Kenpson (1977: 8O): What are listed in the lexicon (theoretical dictionary) are lexical items. which would probably by generally accepted as entailments: 1a. Celtic mythology and musical terms. but it does nonetheless have a certain internal coherence. Tiddles is an animal 3a. and makes me inclined to reject the extreme Quinean view. Tiddles is a cat / entails? 2b. Lesley is John's mother / entails? 3c. which are true by virtue of their meanings. I shall try to show that a distinction between dictionary meaning and encyclopedic meaning is necessary for an adequate descriptive lexical semantics. and synthetic statements. Before offering my own views. say. It is true that the theme of a dictionary is not immediately apparent. not words.

could. Lyons (1981) proposes that we should recognise. Would the man not then be the mother of the baby? It is at least arguable. which for me are in descending order of necessity (example 1 is repeated for convenience): 1a. of course. with advances in various forms of biological engineering. Con- sider the following examples. what are the grounds for maintaining that 'animal1 is a criteria! attribute of oat? Ihe logi- cal status of the relationship in 3 is even more dubious. that the denotative power of oat is independent of the truth or falsehood of 'Cats are animals'. be a product of the man's body. however. as it seems to be. X is a bar of pure platinum / entails? 4b. whether such a move would give us a satisfactory characterisation of dictionary meaning. does not seem to be a logical truth. X has four legs I nave accepted 1 as being logically necessary. 77 occasion'. what he calls 'natural necessity1. Perhaps even the original ovum. in addition to logical necessity. In Cruse (1986) this degree of necessity is labelled 'canonical1. It is surely not logi- cally inconceivable that one day. It is clear from these examples that strict logical necessity cannot be the selection criterion for semantic information in a dictionary: a dictionary devised on this principle would be severely impoverished. John stopped singing / entails 1b. the denotational sta- bility of terras like cat should make us hesitate to call this relationship en- tailmsnt. or which was born with only three legs. Example 5 represents an even weaker relationship: a dog which lost a leg in an accident.) Exairples like 2 are. Admitting examples Brought to you by | UCL . above which a particular feature would qualify for inclusion in a dictionary. although it might be the case that a universe in which pure platinum did not conduct electricity would be radically different from the one we inhabit. Perhaps a scale of necessity could be established. It is doubtful. although not a well-formed one. then there is no logical relation between them. a man would be able to carry a fertilized ovum in his body. Example 4. cxxnmonly used to diagnose hyponymy. 4 would fall into this category. for him. however. The point of these examples is this. to a baby. but an empirical one. with the help of biotechnology. but which would exclude 4. and eventually give birth. An alternative approach might be to relax the criteria for features of diction- ary meaning. There is no way of establishing a cut-off point on the scale of necessity which would allow 5 to qualify as dictionary meaning. also 2 and 3. However. If it is true. X conducts electricity 5a. John did not continue singing 4a. X is a quadruped / entails?? 5b. would still be a quadruped. and some deter- minate point on the scale be designated. perhaps by Caesarian section. presumably.University College London Authenticated Download Date | 5/9/18 6:49 PM .

either that or it would merely duplicate what was in the encyclopedia. form being iirposed on semantic substance differ- ently by each individual language. Haiman rejects this proposal on the grounds that there are no relations between the meanings of the world. to think of 'emic' and 'etic' facts. Haiman rejects this view on the grounds that what is essential varies from person to person and from culture to culture.knowledge which intuitively be- longs to the encyclopedia. The difference between a dictionary and an encyclopedia would then be something like the difference between a statement Brought to you by | UCL . whereas an encyclopedia contains information con- cerning the relations of words to extralinguistic facts. and that is. Haiman's objection this time is that there are no objective facts. peripheral informa- tion. One of these is the view that a diction- ary contains information concerning the relations of the meanings of words to the meanings of other words. or semantic form and semantic substance. he suggests. also rejected by Haiman. an encyclopedia would add that it also has three angles. Haiman gives the following illustration (1980: 339): For a dictionary. and an encyclopedia inessential. is that a dictionary contains subjective facts. A dictionary entry would thus appear to be something like a precis of an encyclopedia entry. under this proposal. There is a way of interpreting the subjective/objective distinction which Haiman mentions.78 like 4 to the dictionary would be an embarrassment: virtually the whole of natural scientific knowledge would qualify . and that the sum total of these angles is 18O degrees. Yet another view. indispensable semantic information. between the dictionary and the encyclopedia. Hainan (1980) mentions a number of other proposals for distinguishing diction- ary meaning and encyclopedic meaning. a triangle is sufficiently defined as a figure with three sides. we would end up with a dictionary that contained no semantic information whatsoever. On the other hand. accidental. If we took this proposal seriously. we want to include examples like 5: "animal with four legs' is surely a perfectly satisfactory dictionary entry for quadruped. Another proposal is that a dictionary contains essential. in the manner of the structuralists. so everything would have to go into the dictionary. but perhaps does not give enough weight to. On the Haasian view of meaning there would be no difference. There is certainly a problem in defining what is meant by 'indispensable'. one easily gets the impression that it means only 'indispensable for a good dictionary entry'.University College London Authenticated Download Date | 5/9/18 6:49 PM . and an encyclopedia objective facts. since every fact or belief concerning the extralinguistic world is reflected in the patterns of affinities between words.

Things are slightly less clear. whereas cats do not (hence The dog barked is semantically more normal than The oat barked). It seems undeniable that sate aspects of lexical knowledge belong more inti- mately to the words themselves. Word-specific aspects of lexical knowledge have not by any means all to do with meaning: among the least disputable are matters like orthography (if we want to know how to spell mosquito it is no use catching one and peering at it). which are the proper concern of the dictionary. grammatical category and etymology. its companion. It is not my concern to speculate here on what form ency- clopedic entries might take. Information on all these matters is of course provided by any reasonably comprehensive dictionary. Al- though this view of word-^neaning is perhaps closer in spirit to what I shall suggest than any of the others. while other aspects belong to words in a less direct fashion. become pregnant (surely not unrelated to the fact that My sister is pregnant is more normal than My brother is pregnant). what is in one sense the same piece of information might well appear in both components. and that only women. that dogs bark. I nonetheless believe that it makes the wrong selection of information for the dictionary. While a dictionary of the sort that I have in mind here would bear certain crucial resemblances to an everyday monolingual dictionary. and would contain a specification of all our concepts and their interconnections. however. The semantic facts proper to the dictionary would be estab- lished on the basis of distinctiveness. by and large. when we con- sider semantic properties. being more intimately related to the things that the words refer to (or to our conceptions of those things). Some facts conoerning word- meaning can be ascertained by observing things in the world. and with different consequences. This much is fairly clear. nor in the concepts which me- diate our dealings with referents. It is these word-specific aspects of lexical knowledge. Brought to you by | UCL . would look somewhat less like a conven- tional encyclopedia. pronunciation. tests of conmutation and the like. it would well be some sort of prototypic representation.University College London Authenticated Download Date | 5/9/18 6:49 PM . the repository of encyclopedic knowledge. for instance. but in a different form in each component. as well as being an essen- tial element of the descriptive semantics of a language. but for many concepts. 79 of the phonological structure of a language. But there are some aspects of lexical knowledge that can only be studied by making some sort of observation on words: the in- formation is not to be found in the referents. and a textbook of articulatory and acoustic phonetics. It would be a representation of our entire knowledge of the world. I suggest. I do not believe that semantic information needs to be strictly partitioned between dictionary and encyclopedia.

such information does not thereby became part of dictionary meaning: it is simply an ad hoc device to convey a vital piece of word-specific information. such as cat. Here. blonde = woman or girl with fair hair1. but it is not a definition in my sense. the second is the word's mapping onto concepts. I suggest. or he may select some items of encyclopedic information to perform the task. kitten = 'young cat1. such as the difference between complain and whinge. in the sense intended here. for example. He may simply use a picture. naturally necessary or canonically ncessary. furry domestic animal that purrs and miaows' may be sufficient to identify oat for most people. 80 Some semantic properties fall relatively ^problematically into the category of word-specific properties. such as violin. However. or joined in holy matrimony and married . and is encyclopedic in nature. such as the fact that pass away applies only to human beings. it does seem that the features of meaning which appear in a definition must be logi- cally necessary. perhaps. Natural kind terms. as we have seen. There are two further types of information that are clearly word-specific: the first is the class of referents which constitutes the denotation of the word. are not criterial. Nor do what Lyons (1981) calls 'cultural kind terms'. however. aspects of expressive mean- ing. telephone. I would like to suggest that a significant num- ber of words have. church. mother = 'female parent". such as the difference beta^een die and kick the bucket. Typical defi- nitions are: stallion = 'male horse'. or evocative or associative meaning arising from dialect or register affilia- tions. But they certainly do not have to be: the defining features of quadruped. 'a small. Information of this kind is given in any good dictionary. as one level of their meaning. For example. have no definitions. crocodile and kangaroo. or cry and blubber. because it is infor- mation that cannot be obtained from referents.University College London Authenticated Download Date | 5/9/18 6:49 PM . If he does use encyclopedic in- formation in this way. More controversially. a definition. Only some words have definitions in the intended sense. a lexicographer is free to use whatever means he finds effective to establish the correct coneptual and denotational connec- tions for a word. chair. is different from an adequate set of identificatory clues. horse. and which belongs in the diction- ary. which is dis- tinct from their encyclopedic specification. or arbitrary collocational restrictions. while die applies to any- thing that lives. quadruped = 'animal with four legs'. Consider. my point is simply that this is its proper and natural place. (Lyons suggests Brought to you by | UCL . kill = 'make dead1. A definition. Many of the features of meaning expressed in definitions are what would usually be considered criterial.

but this fact does not appear in its definition. it is the use of the term blonde that makes it so. Likewise. It might be wondered what reasons there are for claiming that definitions are word-specific properties. that 'parental female' would also define mother logically. there is a sense in which one can be held to have said "That is one of my parents'. is wrong. even though both are entailments. and is not derivable from observations on referents. and hence belong to the dictionary. consider mother. Recall that word-specific properties cannot be veri- fied by observing referents. it is probably a matter of natural necessity that a stallion is an animal. John is legally married to Ra. In other words. For instance. social necessity/ to account for the implication in."parent1 occupies a central position and 'female1 a subordinate position. The first is that the choice of a set of definitional features is in a sense arbitrary. but it. Notice. 'birth giver'. Wierzbicka (198O) rejects 'female parent' as a definition of mother. be at all salient. instead. There are often alternative ways of defining a word. too. a dog is an animal by natural necessity. since dog has no definition.University College London Authenticated Download Date | 5/9/18 6:49 PM . but neither of these features is defi- nitional for dog. The second reason for maintaining that definitions are word-specific is that they introduce a kind of differential focus. and four-legged by canonical necessity. in the semantic structure of the word mother. The hair of the person referred to by that blonde over there might not. not all features which satisfy these cri- teria are necessarily part of a word's definition. 81 another variety of necessity. It is possible that this type of necessity should also be available for definitions.Tba. Similarly. for instance. I would maintain that 'female parent' is the correct dictionary definition for mother. But surely stallions can be observed to be both horses and males. For instance. However. and offers.) Although features of definitional meaning must have sane degree of necessity. too. blondes can be observed to have fair hair. and that 'birth giver1 is encyclopedic (this does not necessarily apply to the equiv- alents of mother in other languages). but one cannot be held to have said "That is a woman'. and mothers can be observed to be both parents and females? This is perfectly true. in the absence of the lin- guistic designation. From a logical point of view both are equally good. but only one of them is correct. and this is reflected in the syntactic Brought to you by | UCL . Take the case of blonde.Ya> therefore he cannot be legally married to Susan. In an un- marked utterance of That is my mother. and these features will be associated with the appropriate words through the encyclopedia. defined as 'female parent1. of equal logical plausi- bility. I suggest two reasons why definitions are word-specific.

) We may cite. horse and stallion in respect of denotational stability. The notion of definitional meaning offers an explanation for a number of facts which are not easily explained if word-meaning is represented only with a pro- totypic structure. This focussing effect is a specific property of the word mother which the definition reflects. Stallion. Neither 'parental female1 nor 'birth giver1 would account for this fact. or 'Aha. Those for whom stallion has denotational stabil- ity. is for many people not denotationally stable at all. nor horses. then there could not be any stallions. then this will force a reconsideration of sense-relations. A suggestion with some plausibility is that those speakers for whom stallion dees not have denotational stability have only a definition as their representation of its meaning. both horse and stallion have denotational stability. If the notion of definitional meaning which I have suggested is a valid one. Stallions are not what we thought they were!'. whereas horse cannot. on the other hand. and this. It is not an objective property of the referents of the word. 'because that's what stallion means'. They actually belong to a different species. yet another interpretation of the expression 'the meaning of a word* is to equate it with definitional meaning. for instance.) For some speakers. since mares are par- thenogenetic. (Some of these have already been mentioned. for instance. and the activities they appear to engage in with mares has nothing to do with procreation. the fact that stallion may be succinctly paraphrased.) The notion of definitional meaning may also offer an explanation of the difference between. I then ask them whether they would exclaim on learning this fact 'Aha. I have asked students to imagine that stallions are one day discovered to be neither male.University College London Authenticated Download Date | 5/9/18 6:49 PM . to some extent autonomous concept of stallion. the fact that of several logically possible definitions. only one in a particular case seems intuitively correct. with a prototypic structure that re- sembles that of a natural kind term. the fact that clear-cut sense-relations can be associated with essentially indeterminate concepts. too. (It should be noted that facts such as these justify rather fewer features than commonly appear in a typical componential analysis. Sense-relations which arise from definitional meaning will have to be treated separately from sense-relations that are really relations between concepts. the former being Brought to you by | UCL . needs expla- nation. 82 structure of the definition. (In other words. on the other hand. A common reply is that if there were no male horses. have in addition to a definition a richer. So there are no such things as stallions!1. Horse is like oat in having a high degree of denotational stability.

which is diagnosed by the test-frame 'An X is a type/kind of Υ ' .University College London Authenticated Download Date | 5/9/18 6:49 PM . with some of their properties being attributable to their definitions. would be fundamentally a relation between concepts. and others to rela- tions between concepts. the descriptive task of specifying the semantic properties of words should be at least to some extent simplified if a clear separation can be made between word-specific properties and con- ceptual structure. since it would arise fron word-specific properties. whereas taxonymy. as would meronymy. For instance. in general. that is. the inclusion relation between horse and stallion would exemplify a purely lexical hyponymy. between dictionary and encyclopedia. Seme relations may have a dual nature. Complementary opposites would almost certainly be definitional in nature. although sane tasks are un- doubtedly made more complex. as would proportional series like man:woman::boy:girl. Brought to you by | UCL . the relation between horse and animal. From a wider perspective. while antonyms are probably conceptual. 83 more centrally 'linguistic1 in nature.

Dictionaries and encyclopedias.). Lunt. (studia granmatica XXII. Language. 1977. Fontana. 1960.. The semantics of natural language.University College London Authenticated Download Date | 5/9/18 6:49 PM . Cambridge University Press. Brought to you by | UCL . Rudolf.. 1964. 1O66-1072. Word and object. D. 198O. John.). Willard Van Oman. Academic Press. 1962. Semantische und konzeptuelle Repräsentation lexika- lischer Einheiten.). Cambridge. Kaiman. Mass. Wierzbicka. 1983. Gruse. Semantic theory. Sydney. RfiSicka. 1986. In: Rudolf RUSiSka. Semantic value. Cambridge. In: Horace G. 1980. Wolfgang Motsch (eds. Kempson. Cambridge University Press. 1964. Wolfgang Motsch (eds. 1981. The Hague.. Untersuchungen zur Semantik. 1983. Ruth M.) Berlin. 61-99.. Cambridge. (ed..84 REFERENCES Bierwisch. Mouton. Manfred. Anna. Akademie-Verlag. Haas. Lingua 5O. 329-57. meaning and context. John. Proceedings of the ninth international congress of linguists. MIT Press. Lingua mentalis. William. Lunt (ed.). etc. Lexical semantics. London. August 27-31. Alan. etc. Mass. Quine. etc. Horace G. Lyons. Cambridge.

why philosophers regard this verb as indeterminate. some of whose writing on the subject tends to be cryptic enough itself. to be less its inscru- tability than. since they all usually do so by means of the verb refer. three. despite the striking fact that philosophers and non-philosophers alike have little trouble talking about it. even after a hundred years of such grappling it is therefore still le- gitimate to ask. Quine (1971).University College London Authenticated Download Date | 5/9/18 6:49 PM . psychology and other disciplines have only fairly recently been sharing with philosophy. I will synopsize the copious literature on reference phenomena and also discuss the few brief at- tempts that have been made to describe the verb refer. GEIGER THE PROBLEM OF REFERENCE AND THE INDETERMINACY OF REFER O. Thus. Furthermore. calls it "inscrutable". The next time you pin it down. two. 85 RICHARD A. what reference is and. and others seem to agree with him: Black (1968: 226) finds reference "somewhat obscure". But as Searle (1969: 79) once pointed out. And lately Martinich (1984: vii). to put it metaphorically. why it is often called a prob- lem. again raises the old question: what is it? The notorious "problem of reference1 appears. lo and behold. you find a slightly different notion before you. one. "the philosophical literature really does reveal an extraordinary amount of confusion" on matters relating to ref- erence. while proclaiming it to be "the most widely discussed issue in the philosophy of language during the twentieth century". its downright recalcitrance: just when you believe you have finally pinned down the notion for analysis. however. I aim to demonstrate Brought to you by | UCL . In order to give partial answers to these questions. you discover it slipping away and have to grapple with it once again. it seems reasonable to ask. Reference is a topic of research which linguistics.

the so-called 'speaker reference' is the more fascinating concept to work with. it is not by chance that users of language go unmentioned: Once upon a time philosophers treated meaning as a relation between words and things. One plethora of notions of reference can be regarded as variations on three main views. Following Strawson (195O) and Searle (1969). 1. it is a shared conversa- tional process. one mentions other concepts of reference "only to get them out of the way". Where once philosophers took the things to be the meanings of the words. (1984: vii) Brought to you by | UCL . Now philosophers generally agree such a notion of meaning was a terribly bad mistake. In Hartman's (1984: 145) recent irreverent formulation of this view. And for still others. precisely because it is defined prag- matically. A statement by Martinich can be taken as representative of the attitude among the members of this group.2 For those more interested in language as a means of comunication than in language as an abstract system. As Geach (1968: 8) once said. Thereafter I will indicate ways in which the problem of indeterminacy is being treated in recent studies in lexical cognition and apply these to the analysis of refer.1 On the view of reference as a semantic relationship. one relating to what people do with words and not to what property words have in themselves. he maintains: the problem of reference cannot be solved by semantic methods. it is an abstract semantic relationship. For some. One of the more noteworthy aspects of these varied views is the degree of attention each pays to communicating human beings. it is a concrete pragmatic act. because they are not taken into account sufficiently. 1. human beings as language processors are of entirely negligible importance.University College London Authenticated Download Date | 5/9/18 6:49 PM . Ihus. It is rather a pragmatic phenomenon. and in the opinion of such philosophers it is this relationship that should be called 'semantic reference1 (Kripke 1977. Philosophers now treat reference as the relation between words and things that meaning was formerly supposed to be. for many philosophers of language the abstract relationship between words and things is primary. for example. for refer- ence is not a semantic phenomenon. 86 that there are connections between the problem of reference (in reference se- mantics) and the indeterminacy of refer (in lexical semantics) which. Donnellan 1978). irrespective of the particular intentions of speakers in context. philosophers now take the things to be the referents of the words. For others. this concept of reference is considered to be the only kind of reference philosophically worth talking about. On the extreme view. are at least partly responsible for the confusion noted by Searle and others. 1.

The concept on this view involves more than one active conraanicator and. at any given moment. reference is a cooperative process which takes place in verbal interaction between speech partners. it is not possible to recognize a distinction in the literature between reference semantics and the semantics of refer. Still. We want to know how information is transferred from the speaker to the listener. is a great deal more dynamic than its pragmatic counterpart.e. this is the central fact underlying most of my remarks here. reference in conversation is accomplished by two interacting individuals who.University College London Authenticated Download Date | 5/9/18 6:49 PM . This view has two important aspects which take it beyond the study of reference as a speech act: it ex- plicitly incorporates the phenomenon of feedback missing in pragmatic studies of reference and implicitly stresses the cognitive effort involved in step-by- step processing for both speaker and hearer. Accordingly. As Clark/Wilkes- Gibbs (1986) describe it.3 On the conversational view. different scholars represent. they have studied reference from the perspective of the addressee as well as from that of the speaker. However. finally. Brought to you by | UCL . formulate the concern of their approach to the problem of reference as: a problem of communication. The differences among the views of reference just described are indicative of the evolutionary process reference semantics is caught up in and they reflect changing intuitions about human language in general. while adhering to a principle of least collabora- tive effort. 87 So on what I am calling the pragmatic view of reference/ the focus of interest is on speech acts performed by an individual speaker. One would think that the semantic analysis of the verb refer would be a matter of concern for lexical semantics and not. 2. two of the first to do research on reference as a collaborative process. As I have already indicated. comprehension as well as production) of reference. since this view takes the speaker's relevant beliefs and intentions in context to be primary criteria for defining the concept of reference. since comparative assess- ments of their views are copiously available elsewhere. 1. As with all progress in the history of ideas. dif- ferent stages in the evolutionary process. take on mutual responsibility for the successful computation (i. for reference semantics. or at least not primarily. and our emphasis is on the psychological representation of the information and on the cognitive processes that are involved. as a two-way process. I will leave the topic at this point in order to discuss the analysis of the verb refer and the con- nections I see between it and the views on reference just described. Nbrman/Rumelhart (1975: 67).

2 As an exanple now of the predicate-structure approach. these senses can therefore be roughly characterized as. a) In (3) α can represent as the argument of the one-place predicate either an agent or a linguistic expression. Martinich does not spell out exactly how such an analysis of refer would look. Brought to you by | UCL . but presumably he means that it can have the following structures: (3) refer (a) (4) refer (a. and an expression. occur only in philosophical discourse and therefore do not really qualify as examples of the same 'language1. (1) refer = 'denote1.1 The polysem/ approach attempts to tease out the relevant meanings of the verb refer. In any case. for example. b) (5) refer (a. Cohen (1970: 259) is an example: Of course the English verb 'refer' has two senses in ordinary usage: one in which what is actually referred to need not be the same as what the agent intends to refer to. it appears that although it is often convenient and harmless to use "refer 1 as a one-. as in: She was referring to you just new.. Martinich (1984: 163) describes refer in this way: .) says essentially the same thing. semantic and speaker reference. b. 2. in the opening statement of one of Donnellan's (1978: 47) papers: (6) People refer and expressions refer. as in: The term bellis perennis refers to the daisy. These one-place predicate uses of refer. it should be noted. the notion to be analyzed is what Russell would call a tetradic relation between a speaker. as. a hearer. two-. (2) refer = 'mention'..) Cohen's statement is so brief that it at best indicates merely the direction which a description of refer in terms of polysemy might take. Meiland (1970: 114f. but even so it is detailed anough to shew a weakness symptomatic of all three approaches: that is. or three-place predicate. 88 2. and another in which a reference is made if and only if it is intended. Itore about this problem later. the senses of refer that Cohen seems to have in mind are directly related to the first two types of reference discussed above.University College London Authenticated Download Date | 5/9/18 6:49 PM . an object. (In his publication of the same year. it explains the verb refer in terms of reference.

b a referential object and a a linguistic expression. and expressions refer to things. In addi- tion. as in an expanded (quasi- ordinary-language) version of Donnellan's statement: (7) People refer to things.. It is reminiscent of the conversational view of reference inasmuch as it raises the addressee to the status of a theo- retical variable equal to the others. On this reading refer also has the following structure: (10) refer (a. respectively.University College London Authenticated Download Date | 5/9/18 6:49 PM . b as addressee and a as referen- tial object. We can take a sentence from Mailand (1970: 118) as an example of this philosophical usage: (9) He is trying to refer the hearer to the individual. as in the sen- tence: (8) Wags often refer to Noam Chomsky as 'Nim Chimsky1.. one occasionally finds a technical use of the verb with a slightly dif- ferent set of predicates. a hearer. then his statement is acceptable within the caimunicative framework he has chosen. and an expression"? Ihere is unfortunately no completely sure way of knowing whether the "notion" involved In this formula- tion is that of reference or of the verb refer. But what are we to understand by Martinich's statement that "the notion to be analyzed is . 89 In (4) b represents a referential object of the two-place predicate and α again represents either an agent or a linguistic expression.) warns against because of the "philosophical mischief". This is an ordinary-language use of refer as a three-place predicate. o. a as agent. If he is speaking about refer- ence per se (as I believe he is). ο a referential object. d) where a represents an agent. However. as in the following sentence from Martinich (1984: 163): Brought to you by | UCL . and the additional argument (here b) an addressee. Ihis much all seems clear enough. as he puts it/ which ensues. a practice which Hartman (1984: 166f. a tetradic relation between a speaker. Better examples for the two versions of the two-place predicate are the sen- tences cited in (2) and (1). an object. namely. d a linguistic expression. In (5) the arguments are such in the three-place predicate that α can represent an agent. if he should happen to be saying that the verb refer also represents a four-place predicate with the arguments named. b. then I think Martinich is again (as in the case of refer as a one-place predicate) illadvisedly commin- gling the technical and ordinary senses of the word.

) Leech then concludes from this test: Both ( 1 2 ) and (13) implicate that there is something wrong with the process of encoding or decoding . The goal here of course is to determine what kind of communicative activity the verb refer indicates. describe or refer to things without there being an addressee to take note of the fact. But for the verbs in ( 1 3 ) the hitch Brought to you by | UCL . etc. I am not speaking here of the 'classical1 version of speech- act theory as developed by Austin. Still. because it sets out primarily to account for (pragmatic) referring behavior and not for the se- mantics of the verb refer. but rather 'cognitive verbs'. (Leech's original numbering has been changed. verbs of referring.3 And now a few remarks about the speech-act approach to refer. should be made between those variables and the two or three variables in lexical semantics which account for the non-technical uses of the verb refer involved here.. As evidence. classifying. more attention should generally be paid to the difference between its technical and non-tech- nical uses.. he uses the context "try to VERB1: (12) I tried to thank/congratulate/pardon the driver. 2. I was no more successful in finding a detailed account of refer in terms of speech-act verbs than I was in finding one in terms of polysemy or predicate structure. like define.. Others. although they appear in lists of illocutions and performative verbs. Indeed. did not take any notice. (13) I tried to define/name/refer to the driver. They carry no implica- tion that s is communicating with an addressee: it is quite possible to define. A distinction. do not fit the category of illocutionary verbs according to present criteria ... stand for linguistic acts which are LOCUTIONARY . This four-place predicate use of refer is found only in the technical register characteristic of recent philosophical studies in reference semantics that are primarily concerned with the four (main) variables necessary for explaining cjommunicative phenomena. and predicating which. the hitch could be imagined as happening at h' s end: perhaps h was too far away to hear. Leech (1983: 2O7). rather than illocutionary. has the following rather surprising things to say about refer: There are . 90 (11) A speaker S . are not speech-act verbs at all. however... refers a hearer H to an object 0 with a definite description D... in a survey of speech-act verb classes. like classify and identify. But for the illocutionary verbs of ( 1 2 ) .University College London Authenticated Download Date | 5/9/18 6:49 PM .. Some of them. Searle and others. describing the human processing of thoughts and experiences.


must be located in s's encoding process. For example, J tried to define my
feelings suggests that I failed to find a suitable verbal representation
of my feelings, rather than that the addressee, because of (say) impaired
hearing, failed to make sense of my words.

To repeat, the assertion that the verb refer carries "no duplication that s is
ccnmunicating with an addressee: [so that] it is quite possible to ... refer to
things without there being an addressee to take note of the fact" is, in my
opinion, nothing less than astounding, coming from a pragnatician. After all,
pragmaticians usually assert just the opposite, as we see from Martinich's
(1984: 162) statement that "it does not make any sense to talk about referring
independently of any possible hearer". Parenthetically I might add in all fair-
ness, however, that Martinich is unwilling to go so far as to accept the view
that "referring is a perlocutionary act that cannot take place at all unless
the hearer reacts as intended".

Here, the reader may or may not have noticed that in the last few sentences
there has been a transition from discussing the verb refer to discussing ref-
erential behavior. It is not only easy to make such a transition; it is also
difficult to notice when one has done so. This, I think, must be the reason
for Leech's unexpected conclusion that one can "refer to things without there
being an addressee to take note of the fact". I believe Leech has fallen in
and out of a referential trap and not noticed it - in, by unwittingly leaving
his metapragmatic subject, the verb refer, to speak about the pragmatic activity
of referring (which of course we do without using refer), and out again, by con-
cluding something about the verb on the basis of the activity, a sort of re-
verse 'reference fallacy1 Searle might call it. Thus, I naturally find myself
also in disagreement with Hudson (1984: 158), who has written that he believes:

we have only a single conceptual structure corresponding to any given bit
of experience; the alternative would be to assume that we have one cognitive
representation for processing deeds themselves, and a different one for
processing the word deed.

The latter is precisely what I do assume. Empirical psycholinguistic evidence
for this alternative view can be found in the writings of Bigelkamp, Kintsch,
Paivio and their associates. Still, as I have emphasized, it is not easy to
maintain the distinction between the two cognitive domains even after one has
become aware of them. This subtlety, alluded to by Searle (1969: 141ff.), has
been a perennial trap in reference semantics all along and is now becoming one
in lexical semantics as well.

Brought to you by | UCL - University College London
Download Date | 5/9/18 6:49 PM

3. Each of the approaches to refer is, as we have seen, in its own par-
ticular way unsatisfactory, but all three are deficient in one theoretically
significant respect: they offer no solution for the 'indeterminacy problem1
of refer. The speech-act verb approach is probably the worst offender in this
respect, since it attempts to reduce the description of refer to the single
characteristic of indicating a locutionary or an illocutionary or a perlo-
cutionary activity.
The other approaches are not much better. While the one does indeed recognize
more than one sense for refer and the other, several predicate structures for
it, neither attempts to account for obviously related features of meaning in
a manner similar to unifying descriptions proposed recently for other poly-
semic lexical items. For example, the word school, according to Bierwisch
(1987), can be thought of as a family of concepts that includes 'building',
'process1 and 'principle', each related to the core concept 'social institution1
by a cognitive shift motivated by various aspects of contextual knowledge. Some
aspects of that knowledge are directly linguistic in nature; in his discussion
of the semantic preconditions of predicates in sentences containing polysemic
referential terms like school, Seuren (in this volume) provides evidence of
that linguistic nature.
Extralinguistic contextual knowledge may also be involved in cognitive shifts,
as Rcmnetveit (1987) has recently demonstrated with the following example in
which the same question asked of a person on two different occasions receives
different answers:
(14) Is your husband working this morning?
(15) No, he isn't working; he's mowing the lawn.
(16) Yes, he's working; he's mowing the lawn.
These contradictory answers, Itomrnetveit points out, can both be true. The ap-
parent paradox is resolved by considering the varying sense of work as 'pro-
fessional activity' or merely 'any useful activity' which the answerer inter-
prets as the intention of each questioner. Romnetveit views this use of con-
textual knowledge in making sense of language as a necessary "dialogical
As a preliminary step in describing the related senses of refer, consider again
a few typical (ordinary-language as well as technical) contexts for this verb:

Brought to you by | UCL - University College London
Download Date | 5/9/18 6:49 PM


(17) Speakers referι addressees to things with expressions.
(18) Somebody I know refers2 to certain people as 'zombies'.
(19) The name Venus usually refers3 to the planet.
(20) Speakers refer^.
(21) Expressions refers.
If we follow Bierwisch's model, a description of the various sensesof refer in
these sentences might look like this:

I I J [

Ordinary senses shown inside of broken line; technical senses shown
outside of broken line.

In other words, the term refer in ordinary usage has the core sense of 'activity1
(i.e. the activity of, say, indicating something) as in (18). It» this core an-
other sense, that of "general characteristic" as in (19), is closely related in
ordinary usage. And finally, three other non-ordinary, technical senses, 'defi-
nitions' as in (17), 'ability1 as in (20) and 'function' as in (21), are directly
or indirectly related to the core. Compared to Bierwisch's analysis of school,
the analysis of refer seems to require concepts that are somewhat more abstract.
To be honest, I am not entirely satisfied with the ones I have proposed. Never-
theless, I believe that the general approach is right, and the difference in
concreteness between the two analyses seems more or less appropriate to the
different orders of cognition required for processing these two lexical items.
Following Bierwisch"s example, I have, with some misgivings, identified a core
concept, but I feel that it would be wise to consider it a very open question
whether one is justified at this point to conclude that refer has one, two or
no core concept at all to which other senses are relatable in some kind of
derivational manner. Interestingly enough, in reference semantics there is a

Brought to you by | UCL - University College London
Download Date | 5/9/18 6:49 PM


parallel to the core-conoept question which has been discussed off and on at
least since Strawson's earlier work. It involves the notion of a 'primary1 or
'paradigmatic1 case of reference (Martinioh 1984: 161).

There are other analytical possibilities wiiich I have not considered here. For
example, a treatment of refer· in terms of case grammar (as a nascent cognitive
theory) is not unthinkable. Another approach is Rommetveit's (1987), in which
"default options' for lexical items rather than the more familiar 'default
values' favored by Johnson-Laird (1984, 1987) are the principal analytical
concepts. Until we have more empirically grounded information about the struc-
ture-processes activated in cognitive shift, e.g. of the types described by
Engelkamp (1987, and in this volume), perhaps the use of the more modest notion
of 'options', rather than that of a hierarchy of 'core1 and lesser-ranking con-
cepts, should actually be the preferred mode of discourse, at least for lexical
items as troublesome as refer. On the other hand, refer may turn out to be
sonething slightly different yet, perhaps an example of what Lakoff (1987: 82)
has called a 'radial category', i.e. a complex concept which has no one main
or core sense but 'central senses' instead which correspond to important sub-
categories. Though other noncentral or extended senses are not as a rule pre-
dictable in such categories, they are connected, according to Lakoff, to the
central senses üy nonarbitrary cognitive links.

At present, I find it premature to speak strongly in favor of Lakoff's or any
of the other cognitive models. No doubt something can be learned about refer
from each of them. In any case, further linguistic and psychological research
is obviously still needed in this area of semantics.

4. To conclude: I have attempted to show that often no distinction is
made between the pragmatics of referring and the semantics of refer. Reference
semanticists easily confuse the object of their attention, namely the referen-
tial use of language, with semantic aspects of (one of) the metareferential
means people use to speak about it, the verb refer. As a result, discussions
of both suffer from the confusion. Careful separation of the levels of dis-
course and attention to the cognitive shifting that refer undergoes might
alleviate some of the perplexity associated both with "the problem of refer-
ence1 and the indeterminacy of refer. Of course, just as with other facets of
the general problem of the relationships among words, things and thinking, for
the near future there will still be quite a lot left to be perplexed about con-
cerning reference. Perhaps Johnson-Laird's hunch (1984: 1O3) is right that "the
heart of the problem is ... in lexical semantics".

Brought to you by | UCL - University College London
Download Date | 5/9/18 6:49 PM



Black, Max. 1968. The labyrinth of language. Harmondsworth, Penguin.
Bierwisch, Manfred. 1987. Pragmatics and formal linguistics. Paper presented at
the 1987 international pragmatics conference, Antwerp.
Bornstein, Marc H. (ed.). 1984. Psychology and the humanities. (Psychology and
its allied disciplines, vol. I.) London, etc., Lawrence Erlbaun.
Clark, Herbert H., Deanna Wilkes-Glbbs. 1986. Referring as a collaborative pro-
cess. Cognition 22, 1-39.
Cohen, L. Jonathan. 197O. What is the ability to refer to things, as a compo-
nent of a language-speaker's competence? In: Linguaggi nella societa e nella
tecnica, 255-268.
Cole, Peter (ed.). 1978. Pragmatics. (Syntax and semantics 9.) New York, Aca-
demic Press.
Donnellan, Keith S. 1978. Speaker reference, descriptions and anaphora. In:
Peter Cole (ed.), 47-68.
Engelkamp, Johannes. 1987. Language and modes of thought. Paper presented at the
second international congress of applied psycholinguistics, Kassel.
French, Peter A., Theodore E. Uehling, Howard K. Wettstein (eds.). 1977. Studies
in the philosophy of language. (Midwest studies in philosophy 2.) Morris,
University of Minnesota Press.
Geach, Peter T. 1968. Reference and generality: An examination of some medieval
and modern theories. Ithaca, Cornell University Press.
Hartman, Ton. 1984. The heterodox interpretation of reference talk. Philosophical
Studies 46, 145-169.

Hudson, Richard. 1984. A psychologically and socially plausible theory of
language structure. In: Deborah Schiffrin (ed.), 15O-159.

Johnson-Laird, Philip N. 1984. Psychology and linguistics.In: Marc H. Bornstein
(ed.), 75-112.

Johnson-laird, Philip N. 1987. The mental representation of the meaning of
words. Cognition 25, 189-211.

Kripke, Saul A. 1977. Speaker's reference and semantic reference. In: Peter A.
French, Theodore E. Uehling, Howard K. Wettstein (eds.), 255-276.

Lakoff, George. 1987. Cognitive models and prototype theory. In: Ulric Neisser
(ed.), 63-1CO.

Leech, Geoffrey N. 1983. Principles of pragmatics. London, Longman.

Linguaggi nella societa e nella tecnica. Convegno prcmosso dalla Ing. C.
Olivetti & C., S. p. a. per il centenario della nascita di Camillo
Olivetti. Museo nazionale della scienza e della tecnica, Milano, 14-17
ottobre 1968. (Saggi di cultura contemporanea 87.) Milano, Edizioni di
Comunita. 1970.

Brought to you by | UCL - University College London
Download Date | 5/9/18 6:49 PM


Martin, Richard M. 1978. Semiotics and linguistic structure: A primer of
philosophic logic. Albany, State University of New York Press.
Martinich, Aloysius P. 1984. Coranunication and reference. Berlin, de Gruyter.
Meiland, Jack W. 1970. Talking about particulars. (International library of
philosophy and scientific method.) London, Routledge & Kegan Paul.
Neisser, Ulric (ed.). 1987. Concepts and conceptual development: Ecological
and intellectual factors in categorization. (Emory studies in cognition 1.)
Cambridge, etc., Cambridge University Press.
Norman, Donald A., David E. Rumelhart. 1975. Reference and comprehension. In:
Donald A. Norman, David E. Rumelhart (eds.), 65-87.
Norman, Donald A., David E. Rumelhart (eds.). 1975. Explorations in cognition.
San Francisco, Freeman.
Nunberg, Geoffrey D. 1978. The pragmatics of reference. Bloccnington, Indiana
University Linguistics Club.
Quine, Willard Van Qrman. 1971. The inscrutability of reference. In: Danny D.
Steinberg, Leon A. Jakobovits (eds.), 142-154.
Rommetveit, Ragnar. 1987. Psycholinguistics, hermeneutics, and cognitive
science. Paper presented at the second international congress of applied
psycholinguisties, Kassel
Schiffrin, Deborah (ed.). 1984. Meaning, form and use in context: Linguistic
applications. (Georgetown University roundtable on languages and linguistics.)
Washington D.C., Georgetown University Press.
Searle, John R. 1969. Speech acts: An essay in the philosophy of language.
Cambridge, etc., Cambridge University Press.
Steinberg, Danny D., Laon A. Jakobovits (eds.). 1971. Semantics: An inter-
disciplinary reader in philosophy, linguistics and psychology. Cambridge,
etc., Cambridge University Press.
Strawson, Peter F. 195O. On referring. Mind 59, 320-344.

Brought to you by | UCL - University College London
Download Date | 5/9/18 6:49 PM




The general problem that I want to discuss here has been ejqpressed in quite a
few different ways, but allow me to simplify historical reality sotiewhat by
distinguishing only two versions.The first is the Oldspeak version, which asks:
How should the meaning of words be represented?
The second is the Newspeak version, which is a little more convoluted, and goes
something like this:

What kind of long-term cognitive structures does the mental lexicon provide
to aid us humans in processing word tokens?
Of course, the Newspeak version is not really new. My reason for adopting
Orwell's terminology is that the Oldspeak idiom has for some time dominated
the semantic scene, and it is only since recently that Newspeak formulations
have begun to regain their former respectability.
I personally believe that the Oldspeak and the Newspeak questions address the
same problem, that the latter is by far the best way of posing that problem,
and should therefore supplant the former. But if you choose to disagree with
me on this point I will not quarrel provided you are prepared to grant that the
Newspeak question raises a legitimate problem. In that case I will be dicussing
a psychological issue that, for all its inherent interest, has nothing to do
with the traditional conundrums about meaning - even if the word meaning crops
up from time to time.
One of the reasons why the Newspeak question is a little less snappy than its
Oldspeak counterpart is that it presupposes a distinction between cognitive
structures that are long-term, or general-purpose, and short-term, special-
purpose structures that are built on the spot to interpret words in concrete

Brought to you by | UCL - University College London
Download Date | 5/9/18 6:49 PM

contexts. This distinction must be made in view of the fact that the way a word
is understood depends on the context in which it occurs. Take, for instance,
the disparate notions that the word table helps to convey in the following ex-
pressions :
kitchen table
billiard table
drawing table
tea table
high table
Even if you would prefer not to talk about different meanings or interpretations
here, it should be uncontroversial to observe that table conveys quite different
notions in each of these expressions. And, of course, such notions are not only
determined by modifier forms, as in the examples given, but also by the wider
linguistic context in which the word occurs as well as by the situational context.
Since there are a lot of different contexts in which a word can occur, there
must also be quite a large number of notions that it can express (Bosch 1985).
For all I know, each single word may express infinitely many distinct notions.
However, in the present context it suffices to observe that their number is so
large as to make it impracticable to have them all on store in the mental lexi-
con. Whatever gets stored there must be something more abstract and less context-
dependent. Hence the contextual notion evoked by a word can only be the result
of the interaction between, on the one hand, the concept that is more or less
permanently associated with that word, and, on the other hand, information about
both the linguistic and the situational context in which the word occurs.
I am interested here not in the contextual notions conveyed by words, but in
the general-purpose concepts that are associated with words - or, more specif-
ically, with nouns. And in this connection a methodological predicament arises.
We want to study nominal concepts, but of course there is no way we could ob-
serve these cognitive structures directly. All we can go by is their outward
manifestations, so to speak - the ways, that is, in which tokens of the words
they are associated with are understood in concrete contexts. And as we have
just seen, contextual notions are something different from the lexical concepts
that help them to come about. So, in short, the methodological problem is how
to get from the contextual notion to the nominal concept.
Fortunately, there is, at least in the case at hand, some methodological con-
solation, which derives from the observation that there is one class of

Brought to you by | UCL - University College London
Download Date | 5/9/18 6:49 PM


sentences that appears to have a special affinity with the semantics of nouns,
viz. generics. (1a) - (1c) exemplify this class of sentences:
(1) a. Beavers build dams.
b. The dodo is extinct.
c. A rabbit has four legs.
When you want to explain what the noun N means, or - what, in my view, comes
down to the same thing - what it is to be an N, then you will almost certainly
employ generic sentences, and not, for example, quantified sentences. For in-
stance, when you consult a dictionary or encyclopaedia to learn something about
poodles, the relevant entry will start off with something like (2a), not (2b)
or (2c):
(2) a. A poodle is a dog.
b. All poodles have four legs, a head, and a tail.
c. Schopenhauer had a poodle.
More generally, it is clear, if only on the intuitive level, that generics are
extremely important to language learning. Generics somehow appear to be uniquely
suited to explain what a word means, and it is probably for this reason that
they so often find their way into dictionaries and encyclopaedias. It is also
for this reason, presumably, that generics are so often employed as test mate-
rials in psychological experiments on categorization and the mental lexicon.
If, as seems reasonable to assume, there is a special affinity between the
semantics of generics and the semantics of nouns in the sense that speakers
commonly rely on generics to explain the meaning of a noun, then we may suppose
that this affinity derives from the fact that generics are somehow particularly
suited to explain what nouns mean. This would imply that it might be a useful
exercise to consider the semantics of generics and nouns together, and try to
develop their respective theories in parallel. This is, in short, what I intend

Although generics have certain formal characteristics associated with
them (e.g. a sentence with a bare plural in subject position will mostly,
though not invariably, have a generic reading), such characteristics are
neither separately nor collectively sufficient to classify sentences as
generic vs. non-generic. So, strictly speaking, instead of talking of
'generics' or 'generic sentences', we rather should say something like
'sentences that are being interpreted generically 1 .

Brought to you by | UCL - University College London
Download Date | 5/9/18 6:49 PM

fforeover. which at first are characterized in terms of their respective ontologies. and therefore unrealistic. starting out from the assumption that the semantics of generics and the seman- tics of nouns are mirror images of one another. A sparrow does not lay eggs. No doubt this assumption will turn out to be oversimplified. A sparrow lays eggs. but the following examples will help to give an idea of how it runs. unfortunately. Which quantifier would that be? (3a) seems to suggest that the universal quantifier must be the one. the sor-tal and the denotational level.against the background of the cognitivist philosophy that I presuppose .(3d). which in itself should make us suspicious about the pre- sumed relationship between genericity and quantification. but I believe that. nonetheless. and will at the same time provide us with some important observations about the semantics of generics: (3) a. Problem cases arise because generics do not constitute a homogeneous class. generics are quantified sentences in fact. as far as I can see.100 to do in this paper. which is based on the assumption that nominal concepts are stereotypes. is pretty definitive. The preferred quan- Brought to you by | UCL .University College London Authenticated Download Date | 5/9/18 6:49 PM . however. would appear to imply drastic ontological commitments. The paper winds up with a brief discussion of what the consequences are. superficially considered. Two types of problem cases are then considered more closely. Accordingly. d. it is tempting to suppose that.of employing descriptions that. several attempts have been made to uncover the quantifier underlying generics. But since it appears to work in the ma- jority of cases. I will sketch the first approximations to such theories. it is worth seeing where it leads to. I will not reiter- ate Carlson's argument here. A sparrow has feathers. Carlson (1978) presents an a priori argument against quantificational treatments of generics which. But all these at- tempts have failed. but as (3b) shows. this cannot be right. Suppose that there is a quantifier underlying each of (3a) . which I shall argue give rise to two additional levels of nominal-concept structure. b. After some remarks on the differences between quantified sentences and generics. A sparrow is a bird. So nominal concepts are organized in at least three distinct levels. I present an initial hypothesis about the semantics of generics. The remainder of this paper is structured as follows. although they lack the usual symp- toms. it will be claimed to be correct in principle. viz. c. this hypothesis turns out not to be generally valid. Given the comparatively successful theories about quantification that we have at the moment.

The notion of a default property. For if we adopt roughly half of all as the generic quantifier. Carlson (1978) speaks of 'fluctuating truth conditions'. Here the fundamental problems with the whole enterprise of finding the generic quantifier become visible. we would have to explain why we need different quantifiers for (3c) and (3d). on the other hand. 101 tificational 'paraphrase1 of (3b) would employ most rather than all. in its turn. which it evidently is not. First. and make ex- plicit two crucial assunptions that are being made. all sorts of quantifiers turn up. for the only quantifier that oomes into consideration for (3c) would be something like roughly half of all. Second. at the very least. in particular. it is perhaps appropriate as a provisional characterization. So on the one hand generic!ty and quantification should be kept apart. if we try to translate generics as quantified sentences. 2 My starting hypothesis about the semantics of generic sentences goes as follows: A generic sentence asserts that the default property that its predicate term expresses belongs to the stereotype denoted by its subject term. And (3d). Brought to you by | UCL . fron very weak ones up to the universal quantifier. and although I will take exception to this terminology later on. To make clear what this hypothesis actually says. I shall draw some preliminary conclu- sions. that there is more than one generic quantifier. But even then. To patch up these problems we would have to postulate. But most. Ihis might seem a strange situation at first. The theory I worked out (Geurts 1985) is a bit more refined than the version outlined here. but for the proof I refer to Carlson). we obtain interpretations for (3a) and (3b) that are patent- ly incorrect. would get an interpretation that makes it true. would be too strong in the general case. while on the other hand speakers do have clear intuitions about what the quantified counterpart of a given generic should be (if one exists). And so on. generics are not implicitly quantified sentences (this has not actually been proven here. In this connection. Without pursuing this argument further. is handled with more care in that paper than is possible here.University College London Authenticated Download Date | 5/9/18 6:49 PM . but I shall try to clarify it in the following. I will have to explain the basic underlying notions of a default property and a stereotype.

and if it is found to be less successful. It is a default property of ele- phants. We can now see wherein the fundamental difference between generic!ty and quan- tification lies. say. A stereotype does not constitute a set of direct assertions about the world. is a property that may be assigned to an individual i if a set of conditions C is satisfied. If that stereotype is thought to be successful in all cases. if needed. in that speakers often agree about what the quan- tified counterpart of a generic should be. generics express expectations. whereas a generic faces each. A generic sentence. there is also a peculiar relationship between the two types of sentences. A quantified sentence makes an assertion about a whole set of individuals at once. is to assess the overall success of the stereotypical expectation ex- pressed by the generic. which he must for this property to apply at all). as we have seen. somewhat metaphorically speaking. The idea is as follows. and there are no indications that i is not-P. as I intend the term to be understood here. if at all. It expresses an expectation which is to be applied. What speakers do when they are asked to "paraphrase1 a generic with the help of a quantified sen- tence. So if you know that Clyde is an ele- phant (and thus satisfies the condition of category membership. A stereotype (Bosch 1985) is a set of default properties that share applicabili- ty conditions (e. because they are associated with the same category). And even if a stereotype occasionally gives rise to false predic- tions. on a case by case basis. unless you have specific reasons to believe that he is not grey. a correspondingly weaker form of quantification will be chosen. Second.University College London Authenticated Download Date | 5/9/18 6:49 PM . but is rather a ccnplex of interlocking expectations. But. Brought to you by | UCL . for instance. First. white) is not enough reason for doubting that Clyde is grey. Thus we have explained the differences between generics and quantified sen- tences. doesn't make a direct claim about a state of affairs in the world. It is important that you must have specific reasons for withholding default inferences: knowing that some elephants are not grey (but. whereas quantified sentences make assertions about the world. that they are grey. you will assume that he is grey. This relationship can now be charac- terized (following up a suggestion made by Nunberg/Pan 1975) as being inferen- tial rather than paraphrastic. a quantified sentence faces the set of individuals that are relevant to its success at once. it can be a good stereotype in the sense that its predictions are reli- able most of the time. meant to apply in concrete situations. So we have two incisive differ- ences.102 A default property P. one at a time. universal quantification is preferred.g. however.


So what at the surf ace locks like paraphrase is in fact a form of inference
that heavily depends on world knowledge.
Two crucial assumptions underlie my hypothesis about generics. One is that the
predicate of a generic denotes a default property. The other is that the subject
of a generic denotes a stereotype. Or, in other words, it is claimed that
nominal concepts are stereotypes, to be represented, perhaps, somewhat like

sparrow (x) 'sparrow1

has wings (x)

This rather sober stereotype would correspond to the generic statements that
a sparrow is a bird and that it has wings. The quoted expression represents
the linguistic label that this stereotype happens to have.
The xs in this structure are variables of sorts: they are to be replaced by
individual constants when the stereotype is activated. In a sense, therefore,
they range over individuals (but not, as the preceding discussion should have
made clear, in the way quantified variables do) and I will refer to them as
arbitrary individuals.
It seems to me that the hypothesis about generics outlined in the foregoing
works fairly well. Nonetheless, there are problems with it. But although there
are several exanples that are clearly at odds with the hypothesis as presented
above, I still believe that, in essence, it remains valid, and that the problem
cases call for complementing it rather than giving it up. In the following I
will discuss two types of cases in particular.
The first class of problems is brought up by sentences like the following:

(4) a. The telephone was invented by Bell.
b. The whale is the largest manmal.
c. Of all animals the elephant has the longest nose.

I borrow this term from Fine (1984). The underlying notion, however,
is different from Fine's.

Brought to you by | UCL - University College London
Download Date | 5/9/18 6:49 PM


If, as we would expect on the basis of our initial hypothesis, (4a) would give
rise to the addition of the property 'was invented by Bell1 to our telephone
stereotype, we would have to explain how it can be that this property is not
now and will never be useful in any single case (because no single existing
or future telephone was or will have been invented by Bell), whereas sentence
(4a) is apparently not inappropriate. Similarly for (4b) and (4c). CHly one
individual mammal can be the largest one, and only one single animal can have
the longest nose. So it would be downright frivolous to assume (by default)
of every whale you encounter that it is the largest mammal, and of every ele-
phant you happen to run into that it has the longest nose. It is evident that
these are not good default properties, and, what is more, every speaker could
tell you that they are not. But the corresponding generics are, again, appro-
priate nonetheless.
In my opinion, the proper treatment of this type of case calls for the adoption
of kinds as entities for lexical concepts to refer to (this is just an informal
way of putting things, which eventually will be refined somewhat). This is what
untutored intuition tells us is needed to explain (4a) - (4c), and it is prac-
tically forced upon us by the observation that (4b) and (4c) involve a kind of
higher-order quantification. Intuitively speaking, only by admitting quanti-
fication over kinds - or sortal quantification, as I will call it - can we make
sense of these cases.
Sortal quantification might well be a much more common phenomenon than you
would think. Take a simple quantified sentence such as
(5) All animals have lungs.
Now think of ways to gather evidence for this statement. One way - the hard
one - would be to inspect each and every animal in the world, and convince
yourself that it has lungs. But another, more comfortable, way would be to
ask yourself, for each species of animal that you can think of, whether it
has lungs. Both verification methods reflect valid interpretations of (5), and
it is hard to escape the conclusion that while the first reads (5) as quan-
tifying over individuals in the world, the second interpretation understands
it to be quantifying over kinds.
If we have individual quantification and sortal quantification, it is thinkable
that we not only have generics that express stereotypical expectations about
individuals (i.e. the type of generic to which our initial hypothesis applies),
but also second-order generics, which would express expectations about kinds.
And I tend to believe that we have, because it seems to me that, e.g.

Brought to you by | UCL - University College London
Download Date | 5/9/18 6:49 PM


(6) Animals have lungs.
can have at least two interpretations, in much the same way as (5) has. To
describe such cases we would need, apart from the arbitrary individuals posited
earlier, arbitrary kinds. Be that as it may, it should be clear by now that
cases like (4a) - (4c) and (5) require kinds as referrable entities.
It now turns out that instead of equating nominal concepts with stereotypes,
what we should have said is that stereotypes represent one level of chunking
information within a nominal concept. For meanwhile we have seen that, beside
the stereotypical level, at least one further level should be postulated. This
sortal levelf as I will call it, is characterized by its referring to kinds,
rather than to arbitrary individuals, as the stereotypical level does.
One immediate advantage of recognizing the sortal level (apart from the fact
that it is needed anyway to interpret certain generics) is that it can help to
explain the observation that some nouns seem to have a strong preference for
sortal interpretations. Mammal and - albeit perhaps to a lesser extent - animal
would be cases in point. It seems to me that these nouns are typically used as
referring to kinds, not individuals or sets of individuals. For instance,
although it is clearly possible in principle to employ phrases like all mammals
in this cage, it is surely not customary to do so.
(7a) - (7c) illustrate the second type of cases that present problems to our
working hypothesis about generics, and, moreover, cannot be analysed satis-
factorily at the sortal level just introduced:
(7) a. Rabbits are very common.
b. Dodos became extinct in the 19th century.
c. Horses were brought to America by the Spaniards.
That these sentences cannot be handled in terms of arbitrary individuals should
be clear. For it is sinply nonsense to say of an individual that it is common or
extinct (cf. Quine 196O; Carlson 1978; Qeurts 1985). An interpretation along the
lines of our initial hypothesis is therefore out of the question.
What is less apparent, perhaps, is that these examples cannot be analysed in
terms of kinds either. For although it seems that we can say of a kind that it is
common or extinct, there is more to the interpretation of (7a), for instance,
than simply storing the property 'is very common1 with the mental representation
of the rabbit species. Any reasonable interpretation of (7a) should involve in-
ferences about the number of individual rabbits living at this moment in this

Carlson C1978) works out a theory in which all generics are interpreted
as involving reference to kinds.
Brought to you by | UCL - University College London
Download Date | 5/9/18 6:49 PM


world. And analogously for (7b) and (7c). To know that dodos became extinct in
the 19th century is to know, among other things, that since some tine in the
19th century there have been no dodos in this world any more. And to know that
horses were brought to America by the Spaniards is to know that, one time in
history, some individual Spaniards brought sons individual horses to America.
In order to represent such information we must resort to a third level, which
I will refer to as the denotational level. At this level, information is struc-
tured, not in terms of arbitrary individuals or kinds, but of real-world indi-
viduals. The denotational level has more of an assertional character than the
other two levels, and it comes closest to being quantificational in the classi-
cal sense. One way of viewing this level is as what I would like to call an
'induction buffer 1 . At the denotational level information about the world is
represented, which is used to construct more abstract concepts (i.e. to build
up information represented in stereotypical and sortal terms). Once these con-
cepts have been formed, denotational information can help to monitor them. If
our dog stereotype threatens to become obsolete, for instance, it can be up-
dated with the help of the relevant denotational information.

We have distinguished three levels of information structure within nominal
concepts: stereotypical, sortal, and denotational. Of course, these levels do
not stand alone. They are interdependent, and interact in various, and probably
numerous, ways. In discussing the denotational level I already indicated one
type of connection between this level and the other two, which I characterized
as an inductive relation. Another, more diffuse, connection was hinted at when
I suggested that the denotational information probably does not exhaust the
meaning of sentences (7a) - (7c), but should be regarded as complementing
sortal information. More generally, it seems to me that the contents that a
generic expresses is often pertinent to all three levels, although it may
focus on one of them. The statement that dodos are extinct, for instance, tells
us something about the kind as well as about the current denotation of the noun
dodo, and perhaps consequences should be drawn from it for the dodo stereotype
as well (in that we should not expect it to be instantiated by a real-world
individual). Such observations are as yet unsystematic, however, and if we
adopt the kind of layered structure that I propose a lot of work will have to
be done in order to map out the interrelations between the various levels.
I have characterized the three levels of nominal-concept structure in terms of
the entities they purport to refer to: denotational information is structured
around individuals, stereotypical information around arbitrary individuals,

Brought to you by | UCL - University College London
Download Date | 5/9/18 6:49 PM


and sortal information around kinds. Coning to the end of this paper, I want
now to make good my premise to refine this convenient but misleading way of
Quine (196O) suggests that a scientific theory is ontologically committed to
those entities that belong to the universe of values for bound variables of
quantification. According to this criterion of ontological commitment, my
characterization of nominal concepts commits us speakers to an ontology com-
prising, at least, individuals, arbitrary individuals, and kinds.Now while it
is fairly clear what individuals are, and that they are indispensable to any
scientific theory whatsoever, a host of philosophical conundrums can be raised
about the cntological status of arbitrary individuals and kinds.These are not
the kind of entities that live in the world like ordinary individuals do, or
at least I, for one, would not be prepared to say that they do. But then in
what sense do they exist? How do we tell any twa arbitrary individual s or any
two kinds, apart? And how do we know that they are the same? These embarrassing
questions can be multiplied at will, so shouldn't we reconsider our decision
to introduce arbitrary individuals and kinds?

I don't believe we should. Even if Quine's criterion applies to scientific
discourse, it doesn't begin to follow that it applies to natural language as
well. What I want to suggest is that, although quantification may be possible
on all three levels of nominal-concept structure, it is only the denotational
level that creates ontological commitments, because these must be allowed if
we want to explain how it is to function. If you want to prove that a certain
mode of structuring information is ontologically committed to the entities
(if that is still the right term to use) it quantifies over, you must demon-
strate that this commitment is crucial to its functioning. And I argue that
in the case of the stereotypical level this cannot be proven, and that for
the sortal level it remains to be proven.
When I characterized the denotational level as an 'induction buffer1, I was
actually claiming that the representation of real-world individuals is crucial
to the function of this level. You cannot have induction without something to
induce from. With the stereotypical level, however, it is different. Stereo-
types are sets of expectations, which, in concrete contexts, are in a sense bound
to (representations of) individuals. The existence of arbitrary individuals
is not necessary for stereotypes to function as they do, because every time a
stereotype is applied, references to arbitrary individuals are replaced by
references to individuals. Arbitrary individuals are, to speak metaphorically,

Brought to you by | UCL - University College London
Download Date | 5/9/18 6:49 PM


numbers in our mental calculus. Just as you can manipulate numbers in a sensible
way without having to assume that they exist, you can perform your mental com-
putations with the help of arbitrary individuals, and still consistently refuse
to acknowledge theiv existence (Fine 1984).
I have tried to make clear that you can pretend to refer to entities of sorts
without thereby committing yourself to their existence. I believe that the
notion of a stereotype is clear enough to warrant the conclusion that arbitrary
individuals don't have to be hypostasized, and I hope that the same can be accom-
plished for kinds. Unfortunately, the problem with kinds is that it is much
less clear how sortal-level information processing works. If we had a clearer
picture of how references to kinds are to function, how the sortal level relates
to the denotational and the stereotypical level, and so on, we would also know
whether or not they must be accorded the status of entities. But, at least, I
hope to have shown that, in principle, ontological commitment to kinds can be
avoided without giving up on kinds.

Brought to you by | UCL - University College London
Download Date | 5/9/18 6:49 PM



Bosch, Peter. 1985. Context dependence and metaphor. In: Wolf Paprotte, Rene
Dirven (eds.), 141-176.
Carlson, Greg N. 1978. Reference to kinds in English. Bloomington, Ind., Indiana
University Linguistics Club.
Fine, Kit. 1984. A defense of arbitrary objects. In: Fred Landman, Frank Veltman
(eds.), 123-142.
Geurts, Bart. 1985. Generics. Journal of Semantics 4, 247-255.
Landman, Fred, Frank Veltman (eds.). 1984. Varieties of formal semantics.
Proceedings of the fourth Amsterdam colloquium, September 1982. Dordrecht,
Nunberg, Geoffrey D., Chiahua Pan. 1975. Inferring quantification in generic
sentences. Papers from the eleventh regional meeting, Chicago Linguistic
Society, 412-422.
Paprotte, Wolf, Rene Dirven (eds.). 1985. Ihe ubiquity of metaphor. Metaphor in
language and thought. Amsterdam, Philadelphia, Benjamins.
Quine, Willard Van Oman. 1960. Word and object. Cambridge, Mass. MIT Press.

Brought to you by | UCL - University College London
Download Date | 5/9/18 6:49 PM





0. Introduction
Semantic change, in contrast to sound change or syntactic change/ is generally
assumed to be random, arbitrary and highly irregular. In the last fifteen years
or so, a large number of studies have challenged this traditional view and
shown that it is possible to formulate general tendencies and principles for
semantic change, both in the lexical area (WiUcins 1980; Sweetser 1984; Tuau-
gott 1986, forthcoming) and in the development of grammatical from lexical mean-
ing (Givon 1979; Lehmann 1982; Traugott 1982, 1986; Heine/Reh 1983). Some of
these studies have suggested discourse-pragmatic motivations for semantic
change, but very few attempts have been made to integrate Grice's theory of
pragmatics and further developments of that theory (Atlas/Levinson 1981; Horn
1984; Sperber/Wilson 1986) with the theory of regularity in semantic change.
The present paper takes up the idea, briefly mentioned in Grice (1975: 58), that
"what starts life as a conversational implicature may become conventionalized"
and examines the role of such changes for the development of connectives and
adverbs in a variety of languages. The idea that pragmatic properties of an
expression may become centrally semantic has been discussed before, of course.

Ekkehard König gratefully acknowledges the financial support received
from the Deutsche Forschungsgemeinschaft (Reisebeihilfe Ko 497/3-1).

Brought to you by | UCL - University College London
Download Date | 5/9/18 6:49 PM


Seme interesting suggestions and discussions of such phenomena can be found in
Geis/Zwicky (1971), Cole (1975), Abraham (1976), Fleishman (1983) and Manoliu-
Manea (1987). Our paper is partly based on these earlier studies, but differs
fron them insofar as an attempt will be made (a) to give a more systematic ac-
count of the domains in which such phenomena can be observed, (b) to charac-
terize the general properties of such changes and (c) to mark off such proc-
esses of semantic change from other processes, such as metaphorical transfer,
metonymic change, bleaching, etc.

1. Conversational implicature
The notion of 'conversational implicature1 was introduced by Grice (1975, 1978),
who opposed it to the truth-conditional content of an expression, i.e. to what
is 'said1 in the strict sense of the word. This notion is meant to provide an
explicit account of how it is possible to mean more than what is actually said.
The basis for this kind of pragmatic inference is seen in some general princi-
ples of cooperative interaction, in four basic maxims of conversation which
jointly express a cooperative principle.
Several types of maxims have been distinguished in the literature. The para-
meters used for such distinctions are (i) the role of the context (particular-
ized vs. generalized inplicatures), (ii) the question of whether the impli-
catures arise fron observing or flouting the maxims (standard implicatures vs.
exploitations) and (iii) the identity of the maxims that give rise to the im-
plicature (Q-based vs. R-based urplicatures).
The types of inplicatures that are of particular interest to us are the gener-
alized and the standard ones. 'Generalized1 implicatures do not require any
particular contextual conditions, in contrast to 'particularized' ones. And
in contrast to 'exploitations', which are based on an obvious flouting of some
maxim, 'standard' implicatures are amplifications of the literal content of an
utterance made on the basis of the assumption that the speaker is obeying the

2. The conventionalizing of conversational implicature

2.1 Inferred causation
One of the best known instances of a pragmatic inference that leads to an inter-
pretative enrichment of an expression is expressed by the classical fallacy
post hoc ergo propter hoc. Utterances expressing a temporal sequence of situa-

Brought to you by | UCL - University College London
Download Date | 5/9/18 6:49 PM

ever since1). has a tendency to precede the main clause. puisque (<lat. Sp. G. Note also that causal since. John has been very miserable. (temporal + causal) c. din moment ae 'from the moment on. Braunrnuller 1978). Since you are not ccming with me. After we heard the lecture. The following examples are instances of this semantic change from 'tenporal sequence' to 'causation1: (2) E. and with a purely causal (3) meaning: (3) a. It has often been pointed out that such causal inferences can become a perma- nent part of the meaning of an originally tenporal expression (Geis/Zwicky 1971: 564. I will have to go alone. post 'after'). Since Susan left him. Braunmüller 1978. The minute John joined our team. Eston. (3a. whereas an event identi- fies the time required for a temporal interpretation (cf. because'. because of. D. since. Temporal connectives that do not primarily denote a sequential relationship and connectives other than temporal ones may also develop a causal meaning. daher. The following examples illustrate such developments (cf. This tendency to preserve temporal iconicity in the word order is a reflex of the earlier temporal meaning. eftersom. infolgedessen. Fr. dal momenta ahe 'since. the temporal meaning that these connectives originally had is not lost immediately. we typically get a purely causal interpretation (cf.University College London Authenticated Download Date | 5/9/18 6:49 PM . I have done quite a bit of writing since we last met. pues (<Lat. Rum. want "for. Lsvinson 1983: 146): (1) a. ( 3 c ) ) . Abraham 1976): One of the factors that determines the exact interpretation of since is aspect. If the clause following since expresses a state. things started to go wrong. Abraham 1976. since can be used in a purely temporal sense. with a temporal meaning and causal implicatures. Abraham 1976. Swed. emedan (cf. in contrast to because. because' (<PGmc 'when + then1) In all of these cases. posteaquam "after. but is typically maintained alongside the causal meaning for a long time. (causal) The traditional formulation post hoc evgo pvoptev hoc does not fully capture the process that led from a temporal to a causal meaning. Brought to you by | UCL . because1. Note that E. b ) . medan 'during'). paräst 'after. It.112 tions typically give rise to the inference that there is a causal connection between these situations (Geis/Zwicky 1971: 564. we felt greatly inspired. b. consequently3 hence. (temporal) b.

Again it is possible to observe the relevant implicatures. I could not work when the television was on. since. quand 'when. dependant. so. au toate £a (. Turk.2 From concomitance to concessivity Among the major sources in the development of concessive connectives are ex- pressions whose earlier meaning can be described simply as expressing 'concomi- tance' or 'co-occurrence' (König 1985.University College London Authenticated Download Date | 5/9/18 6:49 PM . while. so that. d. nonetheless. yet. evenwel. Lat. also 'thus'. connectives originally expressing simple 'co-occurrence' or 'concom- itance1 : EWE withal (for all). because1. da). kun 'when. G. Fr. G. 'temporal overlap" or 'continuation': E. I can't sleep now that I am alone. bei all.'with all things that').g. Fr. c. Sp. because'. thus. D. Finn. dim 'when. etc. Fr. as («DE eal + swa 'all + so').as long as. 113 (4) E. gleichwohl. dennoch. Rum. adhuc). itaque 'therefore' (< 'and so") Ihese and analogous examples show that temporal connectives originally denoting 'temporal overlap1 or 'simultaneity' may also develop a causal sense and the same process of semantic change may affect markers of other adverbial relations such as 'manner' (e. while. I feel very comfortable like this. notwithstanding. so) and 'place' (e. Lat. Lat. nevertheless. b. ungeachtet The change from 'co-occurrence' to 'concessive1 that all of these expressions manifest can again be regarded as a case of a conversational implicature having Brought to you by | UCL . three subcategories can be distinguished: (6) a. negative specifications of co-occurrence (p does not prevent <$ : E. forthcoming). all/just the same. weil (<OHG dia wila so "as long as1). b. I slept very well at the seaside. G. which often have developed into conventional meaning. as. because1. indes (sen). still. da 'therefore'.g. n'errp&che que. bununla beraber ('together with this') c. Within the large group of concessive connectives that manifest this origin. non obstante. zugleich. G. aunque (lat. as pure implicatures in the synchronic use of many expressions of the relevant domains: (5) a. G. 2. connectives originally expressing 'simultaneity'. nichtsdestoweniger. encore que.

more quickly. where one situation does not normally co-occur with the other. Rather than manifests roughly the same etymology as sooner than . König 1985. and certainly all European ones. this year does not look too promising. I'd just as soon not go. b.but. while is still awkward in contexts expressing anteriority of one event to another shows that this conjunction has not lost its original meaning of temporal overlap: (8) While we were extremely successful last year. many of the concessive connectives dis- cussed in this section have retained their original meaning alongside the con- cessive implicature that has become part of the conventional meaning. As in the case of causal connectives. Heber (< lieben 'love') or Swed. in contrast Brought to you by | UCL .114 become conventionalized. immediately1 . Sooner than. but the preferential reading may also derive from a more basic temporal reading 'sooner. hellre. c. as soon asf rather than in English are clear cases of such developments.3 From temporal precedence to preference Many is a com- parative form of the adverb (h)rathe 'quickly. It is difficult to find a method that is effective and. c. (9) a. The fact that. at the same time. Harry walked to work rather than drive. forthcoming): (7) a. inexpensive. b. And one of these cases where co-occurrence is highly relevant and newsworthy is that where there is a general incompatibil- ity between two situations. I'd sooner die than marry you.University College London Authenticated Download Date | 5/9/18 6:49 PM . It is midnight and he is still working. That expressions of simultaneity or concomitance can be anplified and interpreted as expressions of concessivity is illustrated by exanples like the following (Abraham 1976. East Germany is sub- stituting cheap brown coal for imported oil. And this is exactly what concessive connectives express. Even as it admits serious pollution problems. There are so many things going on simultaneously that this is only worth pointing out in special cases. for instance. 2. Mere co-occurrence or concomitance of two situations is never highly relevant information. have adverbs or comparative forms of certain adverbs that are used to express preference. Such preference may be expressed directly as in expressions like G. earlier1.

' The so-called counterfactual use of before may also manifest the pre- ferential reading. D. 115 to sooner than. the earlier temporal meaning has been lost. (Swed. pika "at once). Brought to you by | UCL . "I'd rather sell my grandmother than this dog. Swed. It. piuttosto. ehe. skoree 'more quickly.University College London Authenticated Download Date | 5/9/18 6:49 PM . fremur en ad.' c. 'It is better to suffer than to lie. Russ. Human.) Ich werde eher zurücktreten als nachgeben. (ii) He died before he finished his autobiography. nai ourCnd 'earlier. 'He'd rather work himself to death than give up.v. He would sit in the dark before he'd watch television. rather1. rathe) The semantic development from temporal precedence to preference can again be regarded as an instance of the semantic change ve are investigating (Dieterich/ Napoli 1982: 164).) Plutöt souffrir que mentir. (OED. 1985: 1112. eerder.'before. rather1. Finn. but there is no essential connection between these two readings as the second of the following examples shows: (i) He left the party before he hit anyone. Examples of originally temporal expressions that have developed such a prefer- ential reading can be found in a wide variety of European languages.): (12) a./orr 'before. plutit (<plus tot). rather1. ennemin (14) a. The earlier mean- ing of rathe can still be observed in the following example: (10) The pilgrimage that now is used is a good meane to come rathe to grace. rather than'. kuin möisin toman koiran. ante 'before')./Span. antes (Lat.) Förr arbetar hon ihjäl sig an han ger upp. I'd buy a typewriter faster than I'd buy a television set. (Finn. (13) G. pikemmin (cf. note b. s. b. eher als3 bevor. (G.1 b. The relevant implicature can be spelt out as follows: (11) A generally does p before doing q + > A prefers doing p to doing q Examples like the following manifest the preferential reading merely as a con- versational implicature (Quirk et al.) Ennen mä mö-Csin nmmmoni. ennen 'before. rather1.' d.. (Fr. Port. Fr. will sooner resign than give in. Icel.

later are the targets of very different semantic changes. the former tend to develop into markers of preference. Languages differ as to whether the second formulation is presented as dispreferred or totally wrong. In English. (17) a.after or earlier . (G. b. the "preference1 reading al- ready mentioned and a reading often glossed as 'denial of assertion or assump- tion1 (Thompson 1972. we should note that the use of rather and related expressions as in- tensifiers can be analysed as a reduced case of a metalinguistic 'preference1 reading. b. Examples like (12) show that the conversational implicature is generally ob- servable in connection with the appropriate expressions in the synchrony of a language. How can we support the assumption also made in Dieterich/Napoli (1982) that the development from 'temporal precedence1 to 'preference' is an instance of the process of semantic change under discussion in this paper? The main support for this assumption comes from the fact that this process manifests all the prop<- erties that are typical of changes from conversational to conventional meaning. As far as the relative order in the development of these two readings is con- cerned/ it is very plausible to assume that the 'denial of assertion' reading is simply an extension of the 'preference' reading to metalinguistic contexts. or French examples like (16). Mary seduced John rather than be seduced by him. Whereas the latter typically develop into ex- pressions of causation. What is expressed in sentences like (15 b) is preference with regard to possi- ble formulations. Mary seduced John rather than was seduced by him.) H faisait plutdt froid. we still have genu- ine preference of one formulation over another. Converse expressions like before . In examples like (17). less appropriate one: (16) II recite plutit qu'il ne ahante. so that rather than is roughly equiva- lent on this reading to 'and not'. the latter is the case: metalinguistic preference has further developed into 'correction'.' Finally. Dieterich/Napoli 1982). 'It was cold if anything.116 It has often been pointed out that rather than and related expressions in other languages have at least two different readings.University College London Authenticated Download Date | 5/9/18 6:49 PM . these two readings may correlate with a syntactic difference: the verb following the comparative fonn may be finite ('denial of assertion1) or non-finite ('preference'): (15) a.) Er ist eher schüchtern als abweisend. Examples like (13) and (14) show that the change in question is a Brought to you by | UCL . (Fr. In English sentences like (15b).

even have the same form as emphatic reflexives and can be assumed to derive from such reflex- ives (G. a loss of selectional restrictions and an ex- tension of a specific meaning to more and more contexts. a) In a wide variety of languages. m£me. It. He asked me for money. stesso. but an augmentation to a more newsworthy and thus also more relevant meaning. zelf&j lat. in contrast to metaphorical changes. b. it is clear that one of their most elementary functions is the emphatic asser- tion of identity. None of these properties is observable in connection with the change discussed in this paper. the earlier temporal meaning of the expressions described above is not incompatible with the new preferential meaning. Note furthermore that. I myself will not be able to participate. Even though English differs from most Germanic languages in not employing a former emphatic reflexive as scalar focus particle. selbst. but I don't have any money myself. Whatever the exact number of meanings that need to be distinguished for such expressions.4 Some further domains In addition to the domains discussed above in some detail. 117 very general phenomenon. ipset Fr. Metaphorical change also typically involves a loss of semantic substance. just as superstition based on the inference post hoc ergo propter hoc. this change does not involve a loss of semantic substance. Such an emphatic assertion of identity is only required. but can only be mentioned very briefly here. there are a few other phenomena which can be explained in an analogous fashion. c. In contrast to bleaching. He always thinks of other people before he thinks of himself. in certain contexts: in cases of contrastive focussing. scalar focus particles like £. It also typically has the function of analyzing a complex domain in terms of a simpler one. how- ever. the historical development Brought to you by | UCL . d. Note that the inference expressed by (11) also underlies such expressions as can't wait1 and that it is often objected to as a maxim of behaviour by in- junctions like 'Erst die Arbeit. 2. dann das Vergnügen* ('First come your duties. observable in a wide variety of languages.University College London Authenticated Download Date | 5/9/18 6:49 PM . The President himself will open the cases where the identity of a referent is remarkable and is stressed by excluding possible al- ternatives: and in cases where it is remarkable that a referent satisfies a given prepositional schema: (18) a. Tr*fein). D. (You can go if you like). then the pleasure1).

True in English. Even started out with a relati- vely concrete meaning 'level. OE willan is assumed to have lost its volitional component and to have been ex- tended from human to non-human subjects. why will and shall could well be regarded as the result of a pragmatic inference that became conventionalized. how- ever.University College London Authenticated Download Date | 5/9/18 6:49 PM . As a result. This is more or less the view expressed by Bybee (1987). i. (= precisely the person) In the early fifteenth century. swarf schon. more concrete use. a new scalar meaning began to spread and in the end ousted the sense of 'exactly1. b) The development of a futurate msaning from an earlier meaning 'volition1 or Obligation' is often analyzed as a case of bleaching and generalization. true. horizontally1. If a convention of Brought to you by | UCL . indeed. this development can be regarded as an instance of the conventionalizing of a conversational implicature. Even she.118 of even closely parallels the development of emphatic expressions of identity to scalar particles found in other languages. extreme or unlikely value for the relevant context. however1 are cases in point (Konig 1988). oertes in French. This more abstract use is found in the following Shakespearian text and is still found in the collocation even as: (19) P. which was also from the earliest times on associated with the adver- bial and focussing use of this form can be regarded as a more abstract meaning derived from the earlier.e. In WE the focus particle even character- izes its focus value as an unlikely and thus remarkable value for a given prepo- sitional schema: (20) In Warre. straight. This analysis does not explain. visserligen in Swedish and quidem in Latin 'certainly. if the prepositional schemata in which the arguments occur are in some conflict. who argues that the present tense of modals leads to future meanings because they "imply or predict a future completion of an event or activity expressed by the main predicate. and is she not a heavenly saint. (1641) Again. An emphatic assertion of identiy of two argu- ments in two different contexts is only called for if the identity is in some way remarkable. Was this the idol that you worship so? - V." c) Che of the typical sources in the development of concessive connectives are expressions of emphatic affirmation. The meaning 'equally. exactly1. wohl in German. even the conqueror is commonly a loser. an assertion of identity suggests that the value fccossed on is a remarkable.

causal or concessive. Ibe inference leading from one meaning to the other is clear enough: If something will soon be the case. Interesting enough. 'certainly.' Brought to you by | UCL . he went back into the house. nearly'. 'Are you coming with us? I sure will. bald in German is also used in the sense of 'almost.g. Such participles (e. Not seeing me. assuming that) often develop into conjunctions or prepositions with a meaning that corresponds to a recurrently inferred value of the constructions in typical contexts (Stump 1985: 346). d) Free adjuncts in sentences like (21) are vague and unspecific in their mean- ing: (21) a.23) a. seeing that. it is almost the case at some reference point: (24) Iah warte jetzt schon bald eine halbe Stunde. it should not come as a surprise that an emphatic assertion of truth should give rise to certain implicatures. I entered a different country. never may also be used in this sense: (.' b. 1 I have been waiting for nearly half an hour now. In German allemal originally meant 'always' (ein für allemal Once and for all 1 ). f) In German bald 'soon' has a similar history as its English counterpart. As Stump (1985) and others before him have shown. an enphatic assertion of the truth of the former makes sense. Since concessive sentences imply that the situation described by the subor- dinate clause is generally inconpatible with the content of the main clause. immer and niemals as well as E. definitely': (22) Das schaffe iah allemal. since it changed its meaning from 'immediately1 to "in the near future from a given point of reference'. duringf providing. And it seems to be this use that has left an imprint on the historical development of such expressions. Kommst du mit? . temporal. but is now used in the sense of E. 119 truthfulness is generally taken for granted. He will never do that ('certainly not 1 ).University College London Authenticated Download Date | 5/9/18 6:49 PM . Crossing the street. b. 'I am certainly able to do that. e) Temporal expressions of frequency may develop into modal adverbs.Aber immer.' itote that G. a variety of semantic and pragmatic factors determine the exact interpretation of the relationship be- tween main clause and adjunct as conditional.

an important test. (i) All of the semantic changes examined and discussed above involve inter- pretative augmentations and cannot simply be described in terms of deseman- ticization or bleaching. in other words. bleaching or generalization. they are also observable in the synchrony of a language. dann der Vergnügen1 ('First come your duties. for the type of semantic change under discussion is whether the enrichment of the conventional meaning of an expression also shows up as a conversational implicature of related expressions. As pointed out above. Sperber/Wilson define 'relevance' as informativeness in relation to context. The preceding discussion has identified a certain type of semantic change and also shown how wide-spread it is. (iii) The patterns of inference upon which the interpretative enrichment is based are general ones and are often criticized and attacked by certain pre- scriptivists. (ii) Since the interpretative augmentations are conversational implicatures based on maxims of cooperative interaction which later become conventionalized. TMs enrichment is based on maxims of cooperative interaction and follows general patterns of inference. The inference post hoc ergo propter hoc gives rise to supersti- tion.120 3. Brought to you by | UCL . which increase the newsworthiness of an utterance and without which an 4 utterance is trivial in many cases. provides an adequate explana- tion of these augmentations is not so clear. then the pleasure1) is meant to counteract the inclination to do first what one likes best. Horn's R Principle (Horn 1984) and Atlas/Levinson's Principle of Informative^ ness provide a good basis for explaining most of these interpretative enrich- ments. What remains to be done is to take a more detailed look at the general properties of this type of semantic change which justify opposing it bo other types such as metaphorical change. which can be measured by the number of contextual implications. to which all Gricean maxims are reduced.and the German slogan 'Erst die Arbeit.University College London Authenticated Download Date | 5/9/18 6:49 PM . Whether the Principle of Relevance proposed by Sperber/Wilson (1986). Seme general properties of such developments. but the phenomena they discuss are very different from the ones under consideration in this paper.

Mstaphorical changes typically involve a desemanticization and a generalization of certain aspects of meaning to more contexts. The semantic changes discussed above can hardly be seen in this light. (v) Metaphorical change is generally assumed to play an important part in the development of grammatical meaning (Claudi/Heine 1986). In fact this meaning is often preserved after the semantic change in the form of an augmentation has taken place. Brought to you by | UCL . observable in a wide variety of languages. The function of such changes is generally seen in an analysis of conceptually complex phenomena in terms of less complex ones. Neither does this meaning provide a conceptually simpler analysis of a complex domain as genuine metaphorical changes do. As we have tried to show above.).University College London Authenticated Download Date | 5/9/18 6:49 PM . it is to be expected that the semantic change as a result of which implicatures become part of conventional meaning is a wide-spread phenomenon. 121 (iv) Since conversational implicatures and the maxims they are based on are a very general and perhaps universal phenaienon. metaphors are often defined as 'false or erroneous1 statements (Claudi/ Heine 1986: 298f. In none of the cases discussed above can the earlier mean- ing be described as 'erroneous1. by establishing striking analogies. Further- more. this assumption is borne out by our data.

San Francisco. Lille 1987. Bybee. Academic Press.id1. Heine. 297-335. Dieterich.). Berlin. London. 1975. 1981. 137-165. Nikolai Salnikow (eds. 1978.). Paul.University College London Authenticated Download Date | 5/9/18 6:49 PM . Jay D. 1983. Journal of Linguistics 18.). Nordic Journal of Linguistics 1. H.). London. Geis.). Logic and conversation. Fisiak. Suzanne. 1979. Morgan (eds. Hawkins. Historical word-formation. Talmy. Volume 3: Speech acts. Grarmaticalization and reanalysis in Afri- can languages. Hraunraüller. Lingua 6O. Cole. Fleishman. Amsterdam. San Francisco. Grice. Paul.. 11-72. Horn. On understanding grammar. pragmatics. Bernd Heine.. In: Peter Cole (ed. Explaining language universals. 1982. Claudi. The synchronic and diachronic status of conversational im- plicature. Werner. Michael L. H.). London. 561-566. Buske. Morgan (eds. Atlas. Toward a new taxonomy for pragmatic inference: Q-based and R-based implicature. Historical semantics. San Francisco. Academic Press. Cole. Linguistic Inquiry 2. 1976. Academic Press. Paper given at ICHL VIII. Brought to you by | UCL . Cole. Stephen C. 1987. Peter (ed. Morgan (eds. New in Language 10. 1985. Grice.). It-clefts. In: Heinz Dieter Bohl. Bernd. Mouton. In: Deborah Schiffrin (ed. 1975. 257-288.).). 1988. Donna Jo Napoli. Further notes on logic and conversation. Levinson. Laurence R. 1984. 1-61. 1983. Comparative rather. Kurt. Arnold M. Sti. On invited inferences. Jerry L. On the metaphorical basis of grarrmar. Radical. Peter (ed. New York. 99-120. 113-127. Syntax and semantics. Blackwell. Peter. 41-58. 1971. Cole.. 1978. London. Thomas G. Ulrike. Jerry L. informativeness and logical form: Radical pragmatics.). Syntax and semantics.122 REFERENCES Abraham. Academic Press. New York. Oxford. Jacek (ed. Mechthild Reh. Hamburg. 1975. The semantic development of past tense modals in Qiglish and other languages. New York. 11-42. 1986. In: Peter Cole (ed. Jerry L. New York. Remarks on the formation of conjunctions in Germanic languages. 1981. Die Rolle von Trugschlüssen in der Diachronie von Satz- konnektoren. Given. 183-214. Joan. In: Peter Cole. Peter. Zwicky. John (ed. San Francisco. 1978. From pragmatics to grammar. In: Peter Cole.). Volume 9: Pragmatics.

Elizabeth C.University College London Authenticated Download Date | 5/9/18 6:49 PM .). 1982. Schiff rin. Gregory. Christian. Manoliu-Manea. Traugott. In. Amsterdam. Washington. Pohl.. Concessive connectives and concessive sentences: cross- linguistic regularities and pragmatic principles. Georgetown University Press.. 539-55O. Sandra A. Amsterdam. form and use in context: Linguistic applications. 245-271. Proceedings of the twelfth annual meeting of the Berkeley Linguistics Society. London. 1983. etc. Nikiforidou. Nikolai Salnikow (eds. Festschrift für Alexander Issatschenko. Brought to you by | UCL . Pragmatics. Where do concessives come from? On the development of concessive connectives. König. Sperber/ Dan. Reidel.). A programmatic sketch. 1984. Berkeley. (eds. Lehmann. 1986. Lahmann. Fran conversational to conventional implicature: The Romanian pronouns of identity and their substitutes. Perspectives on historical linguistics.).). Winfred P.). Berkeley Thompson. Thoughts on gramnaticalization. 1972. 263-282. (Akup 48. Benjamins. Stephen C. Benjamins. 123 König. 1982. Berkeley Linguistics Society. In. Winfred P. Universität zu Köln. Croon Helm. Harvard University Press.). speech acts. 1987. New York. Opuscula Slavica et Lingui- stica. Ekkehard. Longman. D. (eds.C. The semantic variability of absolute constructions. Cambridge. Cambridge. 1986. February 15-17. Yakov Malkiel (eds. Lavinson. Deborah (ed. Meaning. Maria. Cambridge University Press. From polysemy to internal reconstruction. Diss. 1986. A comprehensive grammar of the English language. Randolph et al. Lehmann. forthcoming. Dordrecht. 1985. Sweetser. Elizabeth C. Traugott. Ekkehard. Heyn. 1985.).) Köln. 1985. Yakov Malkiel (eds. Mass. The meaning of focus particles: A comparative perspective. Eve E.. 1986. perception. Heinz Dieter. 1982. Ekkehard. Vassiliki et al. Relevance: Communication and cognition. In: Vassiliki Nikiforidou et al. 237-249. Quirk. Papers from the 7th International Conference on Historical Linguistics. Semantic structure and semantic change: A cognitive linguistic study of modality. From prepositional to textual and expressive meaning: Some semantic-pragmatic aspects of gramnaticalization.. In: Jacek Fisiak (ed. Stump. John Hawkins (ed. 1988. Deirdre Wilson. and logical rela- tions. London. König. Journal of Linguistics 8.). 1982. Instead of and rather than clauses in English. 1986. Klagenfurt.

University College London Authenticated Download Date | 5/9/18 6:49 PM . David. 198O.). Diss. In: Linda Waugh (ed. English speech act verbs: A historical perspective. New vistas in granmar. Towards a theory of semantic change. Elizabeth C.124 Traugott.infia (ed. forthcoming. forthcoming. Waugh.). Wilkins. Brought to you by | UCL . Canberra. T.

confirming and. for the first time. it is intended to furnish further evidence. namely lexical items belonging to at least one For detailed accounts of lexical oppositeness. particularly nouns.. which is standardly defined in the following way (Mlwcod/Andersson/Dahl 1981: 8): Given a certain universe of discourse U. It was here that. then we can also talk about the set of all members of U which are not members of A.. ) . This paper will not be concerned. especially complementarity (or: ungradable oppositeness). taken from both English and German.. complementation was not applied primarily to the analysis of the so-called 'content words'. etc. with a summary or evaluation of previous accounts of types of lexical oppositeness..University College London Authenticated Download Date | 5/9/18 6:49 PM .. The application of set operations in general and conplementation in particular to the study of sense relations forms indeed a well-known approach. cf. Instead. and a subset A of that set .. however. dead-alive.e. underlining the sig- nificance of complementation to studies in lexical semantics by following the lines first described in Konig/Kortmann (1987).. 125 BERND KDBTMANN OMPLEMENTATION AND ITS SIGNIFICANCE TO THE LEXICON 0. but to a special group of predominantly 'functional1 or 'grammatical words' of often fuzzy categorical status. indeed. and adjectives (Cruse 1986: 201). i. verbs. Brought to you by | UCL . Introduction One of the fundamental operations in set theory is complementation. ) or Cruse (1986: 1 9 7 f f . animate-inanimate. Lyons (1977: 2 7 O f f . antonymy and complemen- tarity. as is typically illustrated by pairs like male-female. This set is then called the complement of A with respect to U.

etc. The dimensions within which the expressions in (1) work are manifold. Three notional domains In a brief introduction to the notional domains investigated here the following two questions will be answered for each of the domains in turn: a) Which lexemes serve as prototypical instances?. so. "exception1 and "restriction1. or manner as in (8) and (9): Cf. More exactly. G.1 'Complementation" The notional domain which is most closely linked to this set operation and which has therefore also received the label 'complementation" represents the primary function of expressions as in (1): (1) E. so3 suah3 thus. or time as in (2) . identity with re- spect to person. König/Kortmann (1987: 173-182) on quality and manner deixis as two widely-neglected areas in traditional accounts of deixis.University College London Authenticated Download Date | 5/9/18 6:49 PM . direction. e. 126 of the notional domains of 'conplementation'. Of these two issues it is the former that will be focussed on in the present paper. quality as in (6) 2 and (7). it will be shown that not only do the primary meanings and uses of lexemes belonging to one of the above-mentioned notional domains lend themselves to an analysis in terms of complementation. place. solan.. etc. etc. Brought to you by | UCL . e-tse. sonst. anders.(5). which also forms the background of the present study. The major hypothesis put forward in Konig/Kortmann (1987).g. 1. with respect to a universe of discourse which is/ by definition (Allwood/Andersson/Dahl 1981: 62). othe?. but that this set operation may also serve as the basis of extended meanings and uses of the expressions in question. G. otherwise. ander-. typically provided by demonstratives or E. and the respective compounds These are expressions which pick out the complement of some deictic value given in the context. also restricted by the context. is the following one: Taking the set operation of complementation as an onomasiological starting-point provides the means for a coherent unifying account of many aspects of a) the complexity of uses of in- dividual lexical items and b) the relationships holding between different ex- pressions that fall into at least one of the three notional domains under con- sideration. dijf&rsnt. other· than. b) How can the respective domain be analyzed in terms of complementation? 1.

abgesehen von. (9) Wie sonst hätte er wohl seine Schulden bezahlen sollen? 1. (ii) the exception set E (denoted by the syntactic complement of except). or test-frames. (7) Paul ist ganz anders als gestern. Prom the point of view of set theory the proposition of (11) contains three sets: (i) the universe of dicourse £/ (denoted by the universal quantifier).(x same to the party)} E' = {x |x came to the party} It emerges Iran this analysis that the negation of a proposition.e. which has as its members all those persons that indeed came to the party. aside from. barring. ausgenommen. apart from. G. proper names. not only be in construction with universal quantifiers. however. bis auf. which consists of the one per- son that did not cone to the party. Ihus E and E' in (11) can be de- fined as follows: (12) E = {x j'v. Prepositional phrases headed by one of the prepositions expressing 'exception' may. (iii) the complement of E rela- tive to U3 E' (denoted by the whole subject-NP). außer. negative quantifiers and definite descriptions (e.fop). etc.g. (6) I would not have him other than he is.) are likewise possible. save (.) Du hättest es ihm ein andermal sagen sollen.) Geh woandershin! (5) (Heute war der falsche Zeitpunkt. Nigel. · . 127 (2) Ich brauche andere Arbeiter.) (3) Maet this person elsewhere! (Not here!) (4) (Zuhause kannst Du Dich nicht mehr sehen lassen. few. etc.e. i.2 'Exception1 The priitary function of (complex) prepositions as in (10) is to denote the no- tional domain of 'exception1: (10) E. NPs with numerals or determiners like some. (Diese/Solche kann ich nicht gebrauchen. of course. Why these expressions can be analyzed in terns of complementation becomes clear from an analysis of a sentence like (11): (11) Everybody except Nigel came to the party.University College London Authenticated Download Date | 5/9/18 6:49 PM . can also be regarded as the complement of a proposition p. i. (8) You evidently think otherwise. Brought to you by | UCL . bar. Thus the following three environments. except (for). which con- sists of all persons who might have come to the party. can be set up: 3 The brackets in (13) indicate the various positions that are available to the expressions in question. etc.

G. e. It should be noted that the prepo- sitions belonging to the latter class do not denote 'exception' but rather 'addition' (Quirk et al.3 'Restriction1 'Restriction' represents the primary function of a special class of focus par- ticles. E. 1985: 708). except (for). R has only one member. This set is the complement of R. comprising all persons who might have known the secret of Skullion's black eye. bloß. e. (ii) expressions permissible in and T2 only (the majority). nur. ausgenommen. abgesehen von-. in addition to. On the basis of these three environments three classes of lexemes belonging to the notional domain of "exception1 can be distinguished in English and German: (i) expressions that can be employed in all test-frames. allein. but. However.g. G. not counting. alone. mit Aus- nahme von. solely. 1.e. exalusi-oely. etc. so-called 'restrictive1 or 'exclusive' focus particles. bis auf. etc. lediglich. which fulfill the opposite predication. neben. other than. barring. R is opposed to the set of all those members of U. e. The proposition of (15) contains only one set. Let us analyze (15): (15) Zipser alone knew the secret of Skullion's black eye. Therefore sentences like (15) also lend Brought to you by | UCL . thereby implicitly selecting the complement R' of their focus relative to U. Nigel. whose focus denotes a set consisting exclusively of members with respect to which a certain predication can truthfully be made: (14) E. namely Zipser. Nigel) nobody ( Nigel) came to the party Nigel). ausschließlich. apart/aside from.g. namely the set R of all values that satisfy the prepositional schema 'x knew the secret of Skullion's black eye ' . i. außer. bar.University College London Authenticated Download Date | 5/9/18 6:49 PM . it is impossible to give an account of the notional domain 'exception1 without making mention of 'addition1 as the latter represents one possible use of the lexical items in (i).g. G. (iii) expressions permissible in T 3 only. merely. Jane and Tom came to the party. R'. This class of particles takes as focus a subset R of the universe of discourse U. only. purely. G. In an environment like T 3 the sets con- sisting of Nigel on the one hand and Jane and Tom on the other hand are not related to opposite predications. E. Implicitly.128 Nigel) everybody ( Nigel) came to the party Nigel).

and (28) as well as the corresponding examples: Note that the glosses in (18) and ( 2 8 ) are only roughly equivalent to examples as in (19-22) or (29-32). More exactly. the propositional variable p represents a demodalized version of the proposition it stands for.1 From 'complementation' and 'exception' to conditionality Three major types of conditionality can be distinguished into which some of the expressions belonging to either of these two domains develop. Extended meanings and uses Complementation as that set operation that divides a given basic set into two complementary subsets can also be shown to serve as the basis of extensions in the meaning and use of the lexical items exemplified in (1). adversative or concessive connectives. In all three notional domains a number of expressions developed a use as con- ditional. and (14). (10). the following two generalizations can be made: 'Complementation1 and 'exception' may serve as source domains of conditional!ty whereas 'exception' and 'restriction1 may serve as source domains of adversativeness and concessivity. conditionality exception adversativeness/ restriction * ' concessivity The remainder of this paper will be devoted a) to an investigation of both types of extension and b) to a demonstration of why complementation can justly be re- garded as a mechanism underlying both of them.University College London Authenticated Download Date | 5/9/18 6:49 PM . 129 themselves to an analysis in tents of the set operation complementation in that here an explicit set R contrasts with an implicit set R'. on the other hand. (23). 2. p is a sentential expansion of the NPs that the expressions in question are in construction with. In ( 2 8 ) . In (18). Consider the 4 glosses in (18). These two sets can be defined as in (16): (16) A = {x | χ knew the secret of Skull-ion's blaak eye} R' = {x |^ (x knew the secret of Skullion's black eye}} 2. as sketched in (17): (17) source domains extensions complementation · > ^^____^_^. Brought to you by | UCL .

University College London Authenticated Download Date | 5/9/18 6:49 PM . or else you've lost it. pick out its complement. its negation ^p. From now on no distinction will be made between the consequences of the antecedent p and the conclusions it leads to. the latter are sub- sumed under the former. This type of conditional ity seems to be almost exclusively reserved for expressions with the primary func- tion of denoting the notional domain of 'complementation1. It depends on the speaker's preferences concerning the consequent from a complex proposition 'unless p. q 1 follows 'therefore p' or Brought to you by | UCL .e. (19) and (20)) or to the con- clusions it gives rise to (cf . (21 ) and (22) ) . q = if -vp. daß Du in dieser Firma noch jemals Karriere machen wirst. außer Sie erfüllen ihm seine Bedingungen. barring accidents.130 (18) f otherwise else P q = : if p. otherwise/ (or) else you'll spend the rest of the day in your room. Ansonsten habe ich sie im Auto liegen lassen. The expressions in (18) perform the following task: They anaphorically relate back to the given proposition p. 'Exception'. which contains either a verb in imperative mood or a modal expression. (26) We'll arrive in the afternoon. (27) Meier wird zur Konkurrenz gehen. and relate the complement to its consequences (cf. außer / (24) I'm not going except you go with me. i. (21) The bag must be here. (22) Die Brille muß im Arbeitszimmer liegen. Anders /sonst glaube ich kaum. q unless (implication: (only) if p. (25) unless we shout nobody is going to hear us.e. on the other hand.^q) \ \ ansonsten (19) Do what Mutiny said. develops the following two types of conditional extension: (23) f except (that) excepting (that) save that p. i. (20) Du mußt Dich beim Chef entschuldigen.

(32) She would leave her husband except for the children.2 From 'exception1 and 'restriction1 to adversativeness and concessivity Let us consider the source domain of adversativeness first. we would have lost the match. Here the following lexical items can be listed: (33) exaept (that) ^ " but (that) only au er da abgesehen davon da > q Ξ ρ but q nur blo allein . they would have all died. What is entailed is that ρ is the cause of ^q. Quirk et al. namely ^q. is related to a counterfactual consequent. to be roughly distinguished according to whether the mood of the main clause is a factual (realis) or non-factual (irrealis) one. 131 Ihe above sentences can be analyzed thus: The lexical items under consideration pick out those circumstances that bring about the complement of q.^ρ. 1985: 7O9). Brought to you by | UCL . Also notice that German provides no instances of this type of meaning extension. 2. (31) Except for John. (30) But for our goalie. In these four sentences the ccnplement of ρ. and that both ρ and ^q are factual.University College London Authenticated Download Date | 5/9/18 6:49 PM . (28) fbut for/that1 ) = f ^ p q (except far/that} ' (entaillnt: (only) because p/vq) (29) He would have helped us but that he was short of money at the time. lediglich / In order to illustrate this analysis two groups of examples can be adduced. In other words. The following four examples illustrate the latter type: But for is almost exclusively used as a conditional connective of this kind whereas except for serves primarily as an expression denoting exception ( c f . p provides a necessary condition for the avoidance of q.

University College London Authenticated Download Date | 5/9/18 6:49 PM . except that/but that she can't keep her mouth shut. The decision in favour of or against the employment of this candidate is represented by r or ^ r respectively. Therefore the analysis of but sug- gested by Anscombre/Ducrot (1977: 28). applies to all of these expressions: (38) p but q: a. bloß wußte ich nicht wie. q carries greater weight An adversative reading of the latter kind is also possible as one of the two readings available for sentences like the following ones. abgesehen davon. however. q is used as argument for ^r c. bloß. etc. The adversativeness in all of these sentences can be explained thus: There is a hypothetical proposition p supporting as argument the conclusion r. as in (38). (36) Ich würde Dich gerne in München besuchen. Brought to you by | UCL . may serve either as adversative or concessive connectives. (40) Er ist fleißig. (37) Ich wollte ihm helfen. only my mother told me not to. indeed. which becomes clear from the analysis of these Q lexemes in (41): The analysis in(41) can be illustrated by applying it to a sentence like (39) in the context of a conversation in which the pros and cons of a particular candidate for employment are weighed against each other. er müßte nur/bloß/lediglich sorgfältiger sein. p is used as argument for r fa. (35) I would've asked you. In these sentences but. most probably leading to ^r. daß ich kein Geld für die Fahrkarte habe. nice. the lexi- cal items in (33) pick out the factual circumstance(s) supporting as argument and. 132 (34) I would ccrne except (that) it's too far. only. in which. both p and q are factual: (39) She's a great help in the kitchen.

then: although q» r To the last group of examples to be discussed here. too. namely a concessive one: (42) Except that I lost all my money. q is used as argument for^r I Cj. however. q außer daß I (presupposition: abgesehen davon (.. ^ r ( but q) c2. it is the presupposition that p normally serves as an argument for ^q which yields a concessive interpretation: (46) except that except Jfor \ . abgesehen von dem Desaster mit dem kalten Büfett. Brought to you by | UCL . (45) Es war eine gelungene Party. q although p. ^ > p.. the reason for their concessive reading is different from the one described in (41). concessivity resulted from the fact that p and q served as arguments of two opposite conclusions r and 'vr. . then: because lediglich ' q. there is only one reading available. (43) It was a fantastic outing except for the weather. In (41). Although the mood in all of these sentences if factual.University College London Authenticated Download Date | 5/9/18 6:49 PM . (44) Die Fahrt ist gut verlaufen. p is used as argument for r [ b.daß) ' normally if p. if q gets more weight. I had a good time. if p gets more weight. außer daß man uns zweimal das Auto aufgebrochen hat. In examples (42) to (45). 133 (41) pf except (that) q : a.

adversative and concessive uses of expressions otherwise primarily denoting the notional domains of 'complementation'. which is divided as to its consequences.ql IP.1 and 2. ( 2 3 ) . showing that the members of the two com- plementary sets E and E1 always relate to opposite predications. ( 3 3 ) . Brought to you by | UCL . that either p and q give rise to opposite con- sequences or that p gives rise to a consequence opposite to q. >p| χ it ' : ι χ ι|q The arrows indicate the implicated/entailed/presupposed relations holding between p and q as sketched in ( 1 8 ) . to describe it in a different way: A set of circumstances is projected onto a set of consequences and is di- vided into two complementary subsets on the basis of the fact that its members support as arguments complementary consequences (cf. However. ( 2 8 ) .University College London Authenticated Download Date | 5/9/18 6:49 PM . it is the complement of q which follows from p. if they are factual an adversative or concessive reading is called forth. it holds true for all sentences in 2. >P! IP. the following picture emerges. supports as argument the complement of its consequent q. For all types of con- ditionality discussed above it holds true that the antecedent p. . where p contrasts with q on the grounds that. the pro- position which the lexical items in question are in construction with.2. and ( 4 6 ) . The same applies to concessivity as in (46). i. IXie to comple- mentation. The nature of these circumstances determines the adverbial relation holding between p and q: If they are hypothetical the adverbial relation is a conditional one.3 Gcnplementation as underlying mechanism In analyzing the conditional. normally. Or. ( 4 1 ) . for example. including as one member the circumstance denoted by the sentence elements that are in construction with the lexical items investigated in this paper. (47) below). states-of- affairs). 9 (47) conditionality adversativeness concessivity P. it is always a basic set of circumstances (situations. yielded an analysis like (12). in all of these cases it can be argued that the set operation of coiplementation is at work.e. The situation is slightly different for the adversative and concessive readings described in (41): Here antecedent and consequent represent arguments for two opposite conclusions. Just as the investigation of the notional domain of 'exception'. 'ex- ception1 and 'restriction1. 134 2.

and b) that in natural language there may exist con- straints on the nature of the complement to be selected.g. in fact. or G.. but ('exception1/ 'restriction1). acknowledging the limits for the application of set operations to natural language. andermal.University College London Authenticated Download Date | 5/9/18 6:49 PM . This hypothesis gains support from the fact that some ex- pressions may belong to more than one of the notional domains introduced above. Brought to you by | UCL . Remaining problems The major goal of this study was to present material demonstrating that com- plementation can be viewed as a structuring device for the primarily 'functional1 part of the lexicon. anders als ('complementation1/'exception1). and other factors contribute to the constantly pro- gressing determination and delimitation of the universe of discourse? Secondly. the indeterminacy of these two with respect to their reference is greatly restrict- ed: They refer to a day standing in a fixed relationship to its complement. 1O Thank goes to Pieter Seuren for having pointed out this problem to the author. woanders. whereas for anderntags ('the next day 1 ) it is one of determinate posteriority with respect to some day in the past. e. the question arises of where and in what ways natural language behaves differently from what set theory would predict. Both problems converge. A closer analysis of the phenomenon of dual-domain membership was outside the scope of this paper. G.g. otherwise. the presuppositions on the side of addresser and addressee. for example. the relationship is one of limited anteriority with regard to the day including the moment of utterance. In which way exactly do the verbal and situational context. E. What these two examples illustrate is a) that here the universe of discourse is far more restricted than it is for the other expressions mentioned above (thus conventions of different language communities also play a crucial part in its delimitation). e. elsewhere. The first problem has to do with the nature of the basic set including the two complementary subsets. another time. ansonsten ('conplemantation'/'addition'). The same holds true for two other and. anderntags. other than. E. Compared with other expressions belonging to the same domain. sonst. more fundamental problems. however. 135 3.g. the other day or G. E. if one considers some lexemes denoting 'complementation1. For the other day ('recently'). G. e. or E.

1977. 171-198. Perspectives on language in performance.136 REFERENCES Allwood.University College London Authenticated Download Date | 5/9/18 6:49 PM . Osten Dahl. 1977. Cambridge. 1985.. Randolph et al. Cambridge. To honour Werner Hüllen on the occasion of his sixtieth birthday. Cruse. Bernd Kortraann. Lexical semantics.. Oswald Ducrot. Logic in linguistics. A comprehensive grammar of the English language. D. London. Lörscher. Narr. Cambridge University Press.). König. literary criticism. Cambridge. Longman. 1981. Cambridge University Press. 1987.. Deux mais en francais. John. Tübingen. etc. and language teaching and learning. Jean Claude. lars-Gunnar Andersson. Absolute conplementation in the lexi- cal structure of English and German. etc. etc. Ekkehard. Volume 1. Volume 1. Quirk. In: Wolfgang Lärscher.). Wolfgang. 23-40. Lingua 43. Semantics. 1986. Lyons. Cambridge University Press. Anscotrbre. Studies in linguistics. Brought to you by | UCL . Rainer Schulze (eds. 1987. Rainer Schulze (eds. Alan. Jens.

Introduction The main thesis of the present paper can be subsumed under three points: 1) Linguistic expressions can be perceived as cues to set up or identify mental domains (spaces). rela- tivized with respect to a given knowledge frame. central and typical properties of their categories can be identified. 3) The context-free senses of lexical items are underdetermined. functional. and the actual context (Lewandcwska-Tomaszczyk 1987). and the affective. 137 BARBARA LEWANDOWSKA-TCMASZCZYK THE INCREMENT VALUE OF PREDICATES IN THE SEMANTIC LEXICON 1. The properties are attributed according to different types of salience hierarchies. leaving the rest as the background. 2. emotional. the interactional. but the necessary. three levels of meaning are assumed: the cognitive. The affective and the interactional levels of meaning capture the speaker's immediate psychological needs and preferences associated with the linguistic Brought to you by | UCL . elements within these spaces as well as the relations which hold between these elements (Fauconnier 1985. They are assumed to be holistic on the basic cognitive level. etc. axiological. The linguistic form onto which the cognitive-semantic structures are transduced foregrounds the salient semantic information. such as perceptual.University College London Authenticated Download Date | 5/9/18 6:49 PM . assured cognitive per- spective (foregrounding). A concept of lexical meaning In the linguistic model I have developed elsewhere (Lewandowska-Tomaszczyk 1987). Seuren 1985) 2) Lexical meanings are regarded as being dynamic mental entities.

Since this type of cognitive activity is very frequent with language users. (5) The insect was crawling on the table. (4) John wanted to crawl inside. the verbal concept crawl in: (3) John was crawling to his boss all evening. In the case of araul the essential properties seem to be the metaphorical conceptualization of the con- trast evoked by the activation of the image-schematic representation of the upright and the crawling bodily positions. Ihe preferred verbal frame patterning for run is suspended in (2). is equally salient and informative. central!ty. and/or typicality conditioning the conceptual category membership of a given lexical item can be flouted in cases when one cognitive donain is conceived of as another cognitive domain or when one knowledge frame/script is perceived in terms of another frame/script/ e. Brought to you by | UCL . (2) The river ran along the mountain. more concrete one. as well as the negative affective marking associated with typical crawling insects.: (1) John ran along the mountain. Both in its literal and in its figurative sense araal takes an animate subject. the contrast between slowness and persistence. e.g. 138 context and form of an utterance as well as the calculation of the prospective effects of his linguistic choices on the interlocutor. The lexical properties of necessity. the more so as not all metaphors exhibit the same cognitive roots. BAD IS DOWN ('down* is sad. Connected with the evaluation of the differences in bodily positions is the interpretation of the conventional metaphors associated with the physical 'up 1 and 'down1 (lakoff/Johnson 198O): GOOD IS UP ('up 1 is cheerful. evoked by a metaphorical use of a lexical item. on pointing to the simi- larities between domains.University College London Authenticated Download Date | 5/9/18 6:49 PM . victorious). Metaphors do not rely only. more abstract. Also contrasts evoked within the same domain may be equally significant. Bie figurative sense is clearly an instance of a conceptualization of one. though the figurative reading (3) typically involves a human agent. Contrasts between them.g. treating it as an instance of a traditional 'category mistake1 is probably un- justified. humiliating). domain of behaviour in terms of another. as one could believe.

the elements within these domains as well as the relations holding between them. I propose. the increment value of a lexical item must bring in the information coherent with the informa- tion already stored in the discourse domain. 139 Symptomatic here are the Polish and Russian equivalents of metaphorical οτκαύΐ: (6) Pol. etc. logical contradictions. From a differ- ent perspective.g. not present in the Discourse Domain prior to the incrementation and which can be added to X + Υ without any risk of an inconsistency arising. that (7) the satisfiable incrementation of the set X + Υ by the set W -l. i. no matter whether the constructed discourse model is set up by means of strictly linguistic elements (e.Z is the union of χ + Υ with the set which consists of all those members of W and Z which are informative. The inconsistencies mentioned in the above definition are not treated in strict- ly logical terms.g. but are seen as frame-based context-sensitive phenomena con- strained by the principle of relevance (Sperber/Wilson 1986).University College London Authenticated Download Date | 5/9/18 6:49 PM . pacruiacTHBaTbCH (lit. to make oneself down) 3.). changing. tautologies or inconsistencies such as busi- ness is business or α wise fool are linguistically interpretable. From this point of view. Discourse domains Each lexical item builds or evokes one or more conceptual domains. the function of lexical units is to increment the context. i. Since a discourse domain can be assumed to be a set consisting of prepositional (X) and non-propositional image- schematic (Y) information. Each lexi- cal item then functions as a building block in constructing a micro-world of a discourse. a discourse context to which it is added. Discourse incrementation must be consistent with the context. joke. yHHHaTbCfl (lit. modifying Gazdar's (1979) interpretation of the satisfiabl incrementation. fable. ponizac sie Rus. In other words. 'world-creating' verbs) or by an entirely pragmatic means of different language styles (e.e. to make oneself flat (non-perfective)) Pol. pfaszczyc sie Rus. as was mentioned before. context itself influences a lexical item in agreement with the set of default and actualization rules (Lewandowska- Tomaszczyk 1987).e. i.e. Under such con- ditions. Brought to you by | UCL . a lexical unit can be treated as a factor influencing.

(9) Auto nie (lit.e. can be felicitously uttered only in the context where all the information in the preceding context seems to point to the fact that the speaker is a dentist. Brought to you by | UCL . i. An additional consequence of this type of negation. It is interesting to notice that in child lan- guage development negation first appears as an overt sentential operator put either in the initial or in the final position of an utterance. The increment value then of the sentential negative operation not is mainly technical. sentential negation blocks adding to the discourse the textual material mention- ed in the preceding context. as has been noticed in numerous publications. In fact. to put it generally. As a consequence. a negative response to undesired and/or false contextual stimuli. such a role of sentential negation is retained in all discourses. Its primary function is to express rejection: (8) Pol. possibly because they incorporate negation at a certain semantic depth.University College London Authenticated Download Date | 5/9/18 6:49 PM . 14O 4. Blocking the lexical material against its addition to the discourse domain may be either global. car no) "This is not a car'. Dzidzia aa nie (lit. Negative elements in the lexicon Negative elements associated with lexical items are assumed to be stored in different strata of conceptual-semantic hierarchies of a given linguistic system (Lewandowska-Tomaszczyk 1987). (10) I am not a dentist. e. to use Seuren's (1985) terminology. Cnly then is negation used to express a referential truth-value denial. Baby sleep no) Ί don't want to go to bed1. Some negative meanings lexically expressed appear later in language development . A follow-up to (10) in the form of (11): (10) I am not a dentist. (11) ΊΉΘ dentist lives here. Sentential negation from its very first uses by a child is.g. yields an incoherent discourse.they seem both conceptually and linguistically more complex. is the non-referential status of the quantifier a.

the basic domain unfriendly. John in (13). as in (12): (12) John is not friendly. and two other associative spaces: a counterspace friendly. He is unfriendly. He does not exist. He does not exist. or negation de re. The predicate unfriendly in (15) sets up the following discourse spaces (Fig. 1 Brought to you by | UCL . foregrounds the lack of synonymity between the nega- tions expressed sententially and those expressed morphologically: (14) John is not friendly. in contradistinction to John in the de diato reading of (14) expressed in (12). [animate]. though not necessarily. (15) John is unfriendly. (15) cannot be substituted for (14) in (12): (12a) *John is unfriendly. and a space not friendly entailed by unfriendly though not entailing it. typically. / / John I \ \ \ Fig. operating locally: (13) John is not friendly.University College London Authenticated Download Date | 5/9/18 6:49 PM . 141 negation de diato. among others. This fact. blocking the associated existential presupposition. (14) and (15) is used referentially. 1): the address domain.

attached to the main domain.g. A significant function in the increment value of doxastic predicates such as e. which possesses the same technical instruction of barring the increment as the sentential not. are the main 'space-builders' in discourse domains: they set up a new discourse space.University College London Authenticated Download Date | 5/9/18 6:49 PM . So. The technical instruction then in such cases as believe involves the following: (20) suspend the projection of the lexical material from the subdomain to discourse domain and suspend the truth-value judgment of the sub- domain. on the other hand. Intensional predicates such as think. believe. Sentence (17) can be accepted only under the socially valid condition of the positive evaluation of toothache. | Mary might have passed or might have failed the examj Thus. while not is mainly a technical predicate. in the case of believe. is a space builder of positive certainty only in Brought to you by | UCL . the sub-domain con- structed under the predicate know. In the case of lack. Morphological negation increments the discourse by adding the (negative) predicate value to the discourse domain and by setting up a (posi- tive) counterspace and a general negative space. believe is played by the instruction blocking the truth-value judgment of the sub-domain. where the truth-value of the subdomain is strongly entailed: (18) He knows that Mary passed the exam. most cases of lexically incorporated negation prove to represent the mixed type. one of the invited inferences to be made is what Mary might in fact have failed. The predicate know sets up a space of what can be called'positive certainty'. As a consequence we have: (16) He lacks money. say. and (17) He lacks toothache. |Mary passed the exam| (19) He believes that Mary passed the exam. in contradistinction to. 1 covers the properties associated with John as given in (15). 142 The space marked as α in Fig. the frame-sensitive instruction associated with the affective layer of the meaning of lack requires that the nominal object of laak be assigned positive evaluation. etc. know.

they are interactionally used only with reference to the partici- pants other than the speaker him/herself. modal. They set up one subdomain (SD) embracing the address to the agent of the action. Seuren (1985: 419-42O) for the description of the intensional. not the case). they are not fully synonymous with respect to the degree of the speaker's certainty towards the subdomain material ((23) with if is less strongly negative than (22)). while for the addressee the space embraces a possibility continuum between the positive true and the negative false.University College London Authenticated Download Date | 5/9/18 6:49 PM . awkward. ze by^ u mnie wczoraj 'Janek keeps telling me and tries to convince me that he was at my place yesterday (and I consider that false)' Brought to you by | UCL . While the verbs analysed above represent the class of intensional predicates. (24) Janek insynuuje. Due to the language users' negative evaluation associated with the actions expressed in the verbs. Ihe examples above are all interpreted in terms of the negative possibility space. doubt is a space- builder whose experiencer chooses a negative end of the truth scale in the pos- sibility continuum (X believes that Υ is. The predicate doubt is an opposite of believe in this respect. (21) I doubt his honesty. (22) I doubt that he is honest.. etc. which is a meta-space (MS) associated with the addressee and/or connentator of the speech event.) or wmauiac 'keep repeating something which is considered false by the addressee with the goal of convincing the addressee about it1. ze by^am wczoraj w domu 'Janek (falsely) suggests that I was at home yesterday' (25) Janek wmawia mi. while its oppo- site covers a counter-space. most probably. They represent a type of multiple space-builders. As a result. However. which can be differentiated from the positive one in terms of the scope of the implicit or explicit negation. typically about the addressee1 (cf. can be treated as representatives of the predicates of the intensional-phatic scale (cf. 143 the believer's domain. and another one. Cue of the domains built around the continuum corresponds to the believer's space. English insinuate. the weakening of the incrementation of the associated subdonain material to the discourse domain occurs. (23) τ doubt if he is honest. and phatic scales in English). some Polish verbs such as insynuowda 'suggest something unpleasant.

e.g. may be functionally equiva- lent to pistol |i. |I cannot really ask him about it| The complete actual incrementation of the predicates in question then can be established only in terms of the whole text or the ongoing interaction. etc.). |She might have said a ward or not| (28) I can hardly ask him about it. although the perceptual and/or operational properties of the two will be radically different. There are rich lexicalization possibilities in this class of items. differentiating the increment values of particular predicates in this class of item. e. She did leave her room but very infrequently| (27) She could hardly speak for tears. on the other hand.g. barely. e. toy gun. exhibits a strongly interactional character in the sense of one of the 'mention1 properties in Sperber /Wilson (1985).: (29) false hair wig false drug placebo although the analytic units are hardly iscmorphic or synonymous with their lexi- calized equivalents. The discourse function of another group of predicates involving negation is to delete the necessary and/or central properties of the associated nominal con- cepts (Welsh 1983) and to bar the incrementation of these parts of their mean- ings to the discourse domain (e. soaraely set up what can be tented modal-infer- ential spaces and although they cannot be treated as fully synonymous. while dummy pistol. Pseudo. It is the case that placebo is always a false drug. This technical instruction. is com- plemented by a complex lexical material.144 The predicates hardly. but not all false drugs Brought to you by | UCL . with all the functional properties of gun deleted displays the salient functional property of a toy.g. a fake document is not a document. imitation leather is not leather. Neither does the relation of strict implication hold between them. the predicate imitation in imitation leather foregrounds the properties of the perceptual and partly functional properties between leather and its imitation. made to kill). however. paradoxically enough. their scopes move along the true-false (positive-negative) scales and acquire a defi- nite interpretation only in a determinate context: (26) She hardly left her room in those days.University College London Authenticated Download Date | 5/9/18 6:49 PM .

: (31) He is most idiosyncratic in his judgments.g. and false hair is not always a wig. 145 are placebos. they may be poison.subjective .g. can be used to foreground the conceptual perspective negatively evaluated by its user. e. The conceptual perspective equating idiosyncratic and strange is indeed con- tained in the meaning potential of the predicate idiosyncratic |idiosyncratic . e.g. 5. LEwandowska-Tcnaszczyk (1987) for the Polish and English data). Their mixed increment value makes them an especially suitable tool to express different shades of lexical meanings.unusual .: (30) What a hag! Hags bears a negative affective/interactional property even in its context- free use. although its negative value is actualized only in some contexts. This class of predi- cates is very large in many languages (cf. Brought to you by | UCL . He is really strange. Conclusions Table I below presents a typology of discourse incrementation functions of the predicates involving a negative element in Qiglish.strange|. They also exhibit different degrees of con- text-dependence. The last class of items are units which are associated with a negative evalua- tion judgment of varying strength. Similarly not all wigs have false hair (they may have real hair). the predicate idiosyncratic prototypically associated with the sense of subjectivity and uniqueness. while e.University College London Authenticated Download Date | 5/9/18 6:49 PM .personal .

Blackwell. Pragmatics: Implicatures. Seuren. Barbara. George. Discourse semantics. Deirdre Wilson. 1987. London.University College London Authenticated Download Date | 5/9/18 6:49 PM . Dan. Aspects of meaning construction in natural language. Gazdar. Oxford. Relevance. Cynthia K. Brought to you by | UCL . linguistic meaning. New York. Putnam's stereotypes and ccnpositionality.M. Cambridge/Mass. Sperber. Conceptual analysis. 1985. Lakoff. Chicago. The University of Chicago Press. 396-4O7. ji&dz. Gerald. Levrandowska-Tomaszczyk. presuppositions arcl logical form. London. Pieter A. 1980. and verbal interaction.Acta universitatis Lodziensis. 1985. 1983. Gilles. 1979. Academic Press. Mental spaces. Vfelsh. 146 Fauconnier. 1986. Oxford. Papers frcm the 19th regional meeting of the Chicago Linguistic Society. Mark Johnson.. Metaphors we live by. Bladcwsll. The MIT Press.

University College London Authenticated Download Date | 5/9/18 6:49 PM . 147 Table I Discourse incrementation functions of negative predicates 1) barring the increment directly (not lack) 2) blocking the presuppositional increment in a netadonain or in a subdonain (believe doubt) 3) weakening the increment (hardly) 4) adding a (negatively specified) increment (setting up a counter-space and a general negative space) (unfriendly) 5) deleting salient properties of modified terms (different lexical!zation possibilities) (pseudo-scientific imitation leather fake document toy gun _ dunmy pistol false hair wig fals-e drug placebo) 6) foregrounding the negative evaluation component of varying strength and with a different degree of context-dependence (hag hasty idiosyncratic). Brought to you by | UCL .

The considerations presented here are based on the investigation of two corpora: one is a collec- tion of about 350 pairs of opposites as listed in the 1972 edition of Böget's Thesaurus of English words and phrases.: SEMANTIC FEATURES VINDICATED 1. ) . or it aims at de- scribing semantic structures as "units we identify. Mettinger (1988).University College London Authenticated Download Date | 5/9/18 6:49 PM .148 ARTHUR METTINGER PAY CAESAR WHAT IS DUE TO CAESAR . Brought to you by | UCL . assuming that semantic structures should be charac- terized as "conceptual structures that have become conventionalized for pur- poses of linguistic symbolization" (Dirven/Taylor 1986: 6 ) . a corpus of 'op- posites in context1 has been assembled from 43 (predominantly British) novels. of course.. depend on the goals of the analysis: it either aims at investigating "the general character of conceptual structure in human cog- nition" (Talmy 1986: 2 ) .. whose meanings have been checked with Hornby's Advanaed Learner's Dictionary. The choice of a theoretical framework for the description of the opposites established in the corpora does. In addition bo this. and 'semantic feature1 as conceived of by the "Tubingen School of Semantics1 (Coseriu/Gecke- ler 1981: 5} with regard to semantic opposition in English. or postulate as theoretical constructs [that] derive both their essence and their existence from their re- lationships with other units in the same language-system" (Lyons 1977: 231 f . This collection comprises more than 350 pairs of morphologically related and morphologically unrelated lexemes in a variety of contexts. 'semantic dimension'. Cf. also for a comprehensive bibliography of studies on antonymy in English and other languages. In this contribution I want to demonstrate the descriptive and explana- tory adequacy of the notions 'archisememe'.

. yet the meanings of murderess and victim are not the same. 'And she turns out to be a victim.. the 'referent1. Adversativity is. It was for profit and not for pleasure that Lady Tamplin was so anxious for the company of her dear cousin. während die konkreten Bezeichnungsbeziehungen inkonstant (variabel) sind. we know that there are people. it is necessary to observe the important distinction of meaning ('Bedeutung') vs. and that there are persons who are murdered by fellow human beings.] 'Well? demanded Inspector Neele Impatiently [.] Death 1 had come to Adele Fortescue suddenly and swiftly [. You'll have to think again. men and women. and victim are used in text (1) to designate one person. Second. in this case three linguistic expressions have the same designation. The doctor looked at him with slight curiosity..] and saw the face of the woman who had sagged back against the cushions [. "We are made of sterner stuff in the States...] gehört..University College London Authenticated Download Date | 5/9/18 6:49 PM .: (2) "It is as well you are in France. Oppositeness of meaning1.The words Adele Fortescue. A description of "antonymy" (in the broad sense of the term) as an in- stance of systematic paradigmatic relations between the meanings of linguistic signs must be based on the following assumptions: First. Brought to you by | UCL . Hm. . is the term I will use for systemic relations obtaining between the meanings of two linguistic signs." said Van Aldin. and we know that the relation between them is basically an antagonistic one. (Coseriu 1970: 44) This can be illustrated by the following example: (1) [.' said Neele. won't you?" ." (3) She drew out from her handbag the letter she had received that morning from Lady Tamplin. reflected in language use. or label. who deliberately kill other persons. cf. 'meaning1 is understood as pertaining to the language-system. in a (fictitious) extralinguistic world. 'You're taking this hard. She understood the nuances of that letter as well as anybody and the reason of Lady Tampion's show of affection towards a long-forgotten cousin was not lost upon her. Daher sind die Bedeutungsbeziehungen (vom Standpunkt der Synchronie aus) konstant.].. of course. Any special reason?' 'She was cast as a murderess.. designation ('Bezeichnung') and referent ('Bezeichnetes'). 149 2. Thus Ooseriu points out: Im Prinzip sind nur die Bedeutungsbeziehungen strukturierbar/ aber nicht die Bezeichnungsbeziehungen. murderess.Murderess and victim in (1) designate. Katherine was no fool. The term 'adversativ!ty' will henceforth be used for "Gegensätze in Wirk- lichkeit und Denken" (Agricola/Agricola 1979: 16). während die Bedeutung zur 'Sprache' [.. items of extralinguistic reality. on the other hand. Business comes before pleasure there.. Die konkrete Bezeichnung (eines bestimmten Objektes) ist ein Faktum der 'Rede'.

do opposites differ with regard to the language-sys- tem? Ihe answer to this question seems surprisingly simple: systemic opposites are instances of relatively stable. then. be shown that contrast in texts is indicated by the contiguous arrangement of opposites and by the favoured placement of opposites in a syntactically definable environment. She said if the man was a hairdresser.opaque and open . In what way.pleasure I will term 'non-systemic opposites1.words contrast depends for its interpretation on the specific context in which it is used.The reason for introducing a coranon term. Life was composed of things^ literature of words. as I will presently show. Transparent . context-independent meaning-relations constituting minimal lexical fields and can be handled along strictly semantic lines. literature opaque. The experience was an exquisite mixture of pleasure and pain. and criteria for delimiting the two groups will be given below. for both groups is due to the observation that in speech/parole they are used in a similar way. literature a closed system.pleasure and profit . and semantic features. Brought to you by | UCL . in fact.150 Business and pleasure in (2) as well as profit and pleasure in (3) express adversativity. he would have wavy hair. whereas the life . In example (5) both types can be found: (5) In Morris Zapp's view. A similar contextual restriction applies to wavy . It can. the root of all critical error was a naive confusion of literature with life. Life was transparent.University College London Authenticated Download Date | 5/9/18 6:49 PM . semantic dimensions. not straight. pleasure .pain in (4) Philip lay face down on the floor while Melanie walked up and down his back in her bare feet. . whereas the treatment of non-systemic opposites requires contextual in- formation and/or encyclopedic and pragmatic knowledge. Such 'syntactic frames' for English have been established in Mettinger (1988: 45-84). viz. 'opposites'.closed can.pain I will call 'systemic opposites1. 2. For the sake of clarity the term 'contrast1 should be used when semantic op- position on the level of parole is dicussed. but an analysis of the meaning of these lexical items shows that they cannot be regarded as instances of oppositeness of meaning on the level of the language system. Life was an open. The pairs business .literature and things .straight with regard to hair in (6): (6) Caroline dissented. be ana- lysed in terms of archisememes. as is the case with pleasure . All hairdressers did.

1 Systemic semantic opposition (oppositeness of meaning). i. Boy and girl are thus cc—hyponyms with regard to the archisememe QUID (I use the term 'achisememe1 instead of 'archilexeme1 because I am working on an ab- stract sememic basis). 4. Adjectival opposites such as large . A given pair of systemic opposites can thus be regarded as constituting a semantic microfield characterized. ce n'est pas non plus ce qui est commun aux termes d'une opposition (la 'base de comparaison'): c'est ce qui est commun aux d i f f e r e n c e s entre ces terms.mortal in (7) is almost taxononic in nature: (7) 'Seriously. can be adequately handled in terms of a modified version of a subpart of the semantic theory established by Coseriu and further developed by Geckeler and Kastovsky. 'boy' and 'girl1 constitute a paradigmatic semantic opposition along the semantic dimension SEX. intuitively convincing meta- linguistic terms. Kastovsky (1982b: 6 6 f f . where one might take in at a glance the consuranation of man's technological skill and the finest splendours of the natural world. both the archisememe and the seman- tic dimension can be designated by different. the semantic dimension. by a hierarchical relation with regard to the archisemema. between them.female Cf.girl case with the archisememe CHTTD and the semantic dimension SEX. by a non-hierarchical semantic relation obtaining between the members of the pair. this view. on the other.1 Ihe contrast in (8) is an almost classic case of adversativity within the frame of a certain belief-system: (8) It was indeed. 'this venial sin . states with regard to which property/quality the meanings of the lexical items of a pair of opposites are opposed to each other. on the other hand. which regards oppositeness of meaning in the broader context of the lexical field. which in turn is specified by semantic features. as they perform different tasks: the archisememe acts as 'basis of comparison' and thus accounts for the similar- ities in the elements opposed to each other.small or male .mortal sin business is old hat. 151 Ihe pair venial .University College London Authenticated Download Date | 5/9/18 6:49 PM . a perfect marriage of Nature and Civiliza- tion.1 he said. c'est-a-dire ä leurs traits distinctifs. on the other hand. It is very important to keep the notions 'archisememe1 and 'semantic dimension' strictly apart.e. and. (Coseriu 1975: 36) This distinction is quite obvious in the boy . ) for a more detailed presentation. cf. on the one hand. he thought.: La dimension. Brought to you by | UCL .

i. big and small with regard to the semantic dimension SIZE? In view of the relational character of the meanings of big and small it seems most appropriate to assume variable relational features χ and y such that big is specified as 'x SIZE1 and small as 'y SIZE1. or rather.g. (Kastovsky 1982a: 34) 4. c) they are variable.2 In what way. what kind of features must we assume to characterize e. the archisememe for 'male1 and 'female1 is HAVING SEX. 'male' and 'female1 operate over the dimension SEX.3 What characteristics must we assume for the semantic dimensions that relational features operate on? 4. the archisememe for 'large' and 'small' must be labelled HAVING SIZE. as only elements be- longing to the same part of speech can enter into minimal oppositions with each other. can semantic features specify the meanings of two lexemes A and B along a semantic dimension. as semantic dimensions result fron the immediate oppositions betareen the meanings of lexical items and are themselves in turn specified by the semantic features: Semantic features thus characterize the internal semantic structure of an individual lexical item and at the same time specify the meaning relations existing between lexical items. and one might assume that the archisememe should also be SIZE and SEX.e.ratio that determines the type of oppositeness obtaining between the meanings of lexeme A (big) and lexeme Β (small).g. converseness.University College London Authenticated Download Date | 5/9/18 6:49 PM . e. Whether a pair of opposites constitutes a scalar or a digital dimension can be determined if we look at the behaviour of opposite A and Brought to you by | UCL . b) they are relational with regard to each other. Both concepts are strongly interdependent. For a systematic description of oppositeness of meaning in the lexicon of a given natural language it is the semantic dimensions and semantic features that are the decisive factors. 4.1 A distinction must be made between scalar dimensions and non-scalar (= digital) ones. respectively. antonymy. and archisememe and semantic dimension would coincide. analogously.152 seem to cause greater problems: 'large1 and 'small' constitute the dimension SIZE. complementarity. then. unless either of them must meet certain conditions with regard to the range of its value. it is the χ to y . i. they can represent any value along a semantic dimension.3. On the other hand.e. with χ and y exhibiting the following characteristics: a) they are relational with regard to the semantic dimension involved. etc. hyponomy.

in superlative or equative constructions.. and one feature can be represented as the negation of the other.e. 4. good . Along quality scales. 153 4 opposite B with regard to gradability. 'Gradability 1 refers to syntactically observable phenomena such as the insertability of gradable adjectives in syntactic frames of the more . Quality scales thus represent adversativity of the POSmVE/HEGftTIVE or GOOD/BAD type. whereas quality scales. male and female would constitute the digital dimension SEX . than/less .bad along the dimension QUALITY). thus. scales that have a zero. i.g. and bi-directionally open scales. than . 'scalarity 1 . the other feature expresses the appropriate evaluatively 'negative' counterpart (cf.type. scales characterized by a turning point Τ (which is not to be understood as a zero-value but rather as a pivot) from which the scale extends infinitely into opposite directions.or starting-point and extend infinitely into one direction.e. but must be inferred from e. collocability with degree adverbs. i. Which member of a pair of opposites is referentially adequate for describing something usually involves an object-related norm..2 When scalar semantic dimensions are involved. and the combinability with intensifiers. which member of a pair of opposites is regarded as referentially adequate for describing something usually involves both an object-related norm and a speaker-related norm or attitude. 'Scalarity' is thus not directly observable.University College London Authenticated Download Date | 5/9/18 6:49 PM ..3. one feature expresses an evaluatively 'positive' specification of a dimension rather than a degree. Brought to you by | UCL . as a consequence of their scalar!ty. on the other hand. a distinction must be made between 'type of scale1 and 'kind of scale1: 'Type of scale1 refers to the distinction between uni-directionally open scales. Ungradable opposites constitute digit- al dimensions.. denotes the semantic properties accounting for the syntactic behaviour of the respective lexical items. are also inherently degree scales.the semantic features χ and y nave absolute values and exhaust the dimension completely. on the other hand. Degree scales exist on their own. opposites oper- ating over degree scales are thus ultimately reducible to adversativity in terms of MDRE/LESS. 'Kind of scale' refers to the distinction between degree scales and quality scales: along degree scales. the features χ and y express various degrees or different amounts of the properties denoted by the dimension.

slow: The value of slow..shallow. The question arises. the y^value (characterizing weak) moves towards zero.and I'm perfectly certain they wouldn't like me!" Brought to you by | UCL . This type of oppositeness of meaning is characterized by semantic dimensions representing 'uni-directionally open scales with attainable zero-value1 and the feature-relation χ > y * 0. Cruse (1986: 206) points out with regard to the pair fast .strong type holds for pri- mary lexemes only. 5. of whether y can be interpreted as ever equalling zero.3 Yet another type.weak.2.unimportant type consists only of pairs with one prefixed member. frcm the point of view of word-structure the two types just dis- cussed are complementary in distribution: the weak . however. The same applies to pairs like strong . Tommy shook his head.. whereas the important . however.2. etc We can thus single out one type of oppositeness of meaning as being character- ized by a scalar semantic dimension of the type 'uni-directionally open scale with non-attainable zero-value' and the feature-relation χ > y > 0.154 5. clean - dirty. deep .2. safe . This is not a physical fact.unimportant: The semantic dimension IMPORTANCE is scalar..light. never actually reaches it . which is characteristic of uni-directionally open degree scales. 5.1 The pair strong .weak type. although it "tends towards1 zero speed. which goes to show that the feature y (characterizing unimportant) may reach the zero-value of the dimension involved. one feature represents the zero-value of the scale involved: (9) "What about the colonies?" she suggested. In contradistinction to the strong . made up of pairs like certain . Both members are gradable. com- pletely unimportant is perfectly possible. it is a uni-directionally open degree scale. but a linguistic one: we cannot say completely slow when we mean 'stationary1 .1 The interaction of the parameters mentioned above accounts for different sub-types of oppositeness of meaning: 5.weak operates over the scalar dimension STRENGTH.. but vihereas the x-value (characterizing strong) moves to- wards infinite when intensified. moreover.e.2 The situation is different for pairs like important .dangerous is characterized by the fact that the semantic dimen- sions constituting them must be regarded as 'uni-directionally open scales with obligatory zero-value". heavy . i. "I shouldn't like the colonies . Incidentally.University College London Authenticated Download Date | 5/9/18 6:49 PM .uncertain.

on the other hand. the fea- ture χ represents the value zero along the semantic dimensions UNCERTAINTY/ DIRTINESS. of course.. Nevertheless. But it was mighty uncertain whether it would bear my weight . functional one. characterizing uncertain. dangerous. where the dimension is measured in terms of POSITIVE/NEGATIVE evaluation rather than in terms of MORE/ LESS. 6." Thus we can say perfectly /completely /totally certain/clean/safe. and does not postulate any psychological or cognitive reality of the lin- guistic constructs it uses. The feature y. DANGER.. thereby also specify- ing the meanings of linguistic signs with regard to the semantic micro-fields they constitute.2. Systemic oppositeness of meaning can.. Ltd.e. ".. i. Within such a framework semantic features fulfill the task of specifying values along semantic dimensions.pain can be regarded as an instance of a pair oper- ating over a 'bi-directionally open quality scale1. both χ and y do. semantic dimensions. as has been shown. dirty.4 Ihe pair pleasure . 5. however. 155 "Here's to our joint venture/ and nay it prosper!" "The Young Mventurers. also represent degrees along the scalar dimension. but also almost certain/olean/safe. represents any value other than zero along the respective dimensions. be described adequately in terms of archisememes. They put down the cups and laughed father uncertainly. so that we arrive at a feature-specifica- tion 0 = χ < y.University College London Authenticated Download Date | 5/9/18 6:49 PM .!" responded Toory. which indicates that approximation towards a zero-digit is possible.. and semantic feature- relations. a totally intralinguistic.. Such a description is. Brought to you by | UCL . I noticed that there was a long branch running out from the tree in the right direction .

1977. Gruse. Antonyme der deutschen Sprache. 3O-51. The relation of grammar to cognition. Duisburg. Arthur. Mettinger. etc. 197O. von Gunter Narr. 1988. München.Duisburg. Aspects of semantic opposition in English. 1975. Dieter. YEB Bibliographisches Institut. Schwann-Bagel. Francke. Kastovsky. Eugenio. Cambridge.) Cambridge. Bern. Vers une typologie des champs lexicaux. A corpus- based study of binary meaning-relations. Semantics. paper no. Studia Anglica Posnaniensia XIV.. Coseriu.156 REFERENCES Agricola. Wörter und Gegenwörter. 1982b. 165. Eugenio. 1986. 'Privative opposition1 and lexical semantics. The conceptualisation of vertical space in English: The case of tall. 1986. Lyons. Talny. Coseriu. Narr. Erhard Agricola. Kastovsky.University College London Authenticated Download Date | 5/9/18 6:49 PM . lexical semantics. Einführung in die strukturelle Betrachtung des Wort- schatzes. 163. (Studienreihe Englisch 14. etc. Dirven. Tübin- gen. 1981. Wortbildung und Semantik. Wien. paper no. Cambridge University Press. Leipzig. Leonard. Linguistic Agency University of Duis- burg (previously Trier). Cambridge University Press. Series A. Christiane. D. Cahiers de Lexicologie 27. Horst Geckeier. 1979. Dieter. Diss. Brought to you by | UCL . 29-45. John. 1982a. Narr. Series A. John Taylor. Tübingen. Trends in structural semantics.) Düsseldorf.. Alan. Eugenio. Coseriu. In Zusammenarbeit mit Erich Brauch und Gisela Köhler hrsg. Rene. (Cambridge textbooks in linguistics. Linguistic Agency University of Duisburg (previously Trier). 1986.

. the traditional concept of semantic features is accepted as that of the smallest units which constitute lexical meaning. Dictionaries normally seek pragmatic solutions. as this dismisses the problem to some inaccessible metalevel. are still largely unsolved. How- ever. one.University College London Authenticated Download Date | 5/9/18 6:49 PM . the process of disambiguation. however. which creates a number of linguis- tic problems. 2. in that the same form. such as the internal semantic structure of polysemic lexemes and the question of the number of meanings which a word may have. I assume that features are meanings of linguistic forms which function metalinguistically. however. the result of which is that the number of meanings attributed to any particular word by different dictionaries may vary considerably. Rather. 157 EDGAR W. does not convey the same meaning. I do not see very much sense in assuming that features are metalinguistic atomic primes. in a fairly theoretical manner. SCHNEIDER ON POLYSEMY IN ENGLISH. it has to be handled in some way both in communication and in linguistic de- scription. most researchable aspect of natural lan- guages" (Weinreich 1966: 398). Fea- tures are either inherent or 'transfer features' in the sense of Weinreich.. This paper presents both a model of polysemy which takes these points into consideration and a practical application of this model to the description of an English verb. Core problems from a lexicological point of view. Practically. Theoretically. Theory: A model of polysemy As a preliminary assumption. Brought to you by | UCL . Introduction Polysemy is "a characteristic and . Traditional topics in the study of poly- semy have been its delimitation from homcnymy and. CONSIDERING CONSIDER 1. it threatens the Saussurean sign concept based upon the unity of both a sign-if-Cant and a signifio. when occurring in different contexts. they are potentially complex internally but gain their unity through a process of holistic conceptualization (Hypostasierung).

further distinct sub-sets of features com- bine with the common core. the Brought to you by | UCL . The meaning of a polysemic word is organized at three interdependent levels: features. thus creating the 'primary meanings' of a word. some spe- cific context factors present or not. The sememes of a lexeme overlap and are arranged according to the principle of progressive internal differentiation. I shall describe these constitutive elements in turn in greater detail. however. the only real manifestations of the semantic level of a language in a text. its 'compre- hensive potential meaning' or 'semanteme'. sememes are "actual meanings'. the role partners (or 'case frames') which it demands. A certain sub-set of features constitutes the semantic core of all occurrences of a particular word and is always ex- pressed. more precisely. or. and the comprehensive potential meaning of a word. Features constitute the minimum level. polysemy is understood as systematic meaning variation under the influence of context. is the intermediate one: the sememes of a word consist of ordered configura- tions of sub-sets of the semanteme's features. which determines which of these possi- bilities is realized in any particular case of word usage in performance. The maximum level is the total set of features potentially associated with a word form. Variable components (also called optional or inferential components) are typically associated but can be explicitly denied. This procedure can be continued at further levels of sub-specifica- tion. in the following paragraphs. The sememes of a polysemic word have partly overlapping but also partly differing feature configurations. i. The most important level. and it is the linguistic or extralinguistic context. the meanings associated with any particular occurrence of a word in parole. sememes. and they may be either obli- gatory or variable.and its characteristic context factors. They serve to account for sortie of the fuzziness of word meanings and to distinguish pro- totypical from marginal cases of word usage. thus consistently narrowing both the actual meaning and the possibilities of contextual choice.e. or 'contextual meanings'. They are repre- sentatives of superordinate feature dimensions. The number of such rank levels aimed at in a particular analysis is not a fundamental question but rather a pragmatic one. Whereas features and corrprehen- sive lexical meanings are constructs of linguistic analysis. These can be further sub-specified by adding more semantic components on a second level of differentiation.e. At a first-ranking level.University College London Authenticated Download Date | 5/9/18 6:49 PM . A sememe is taken to be fully described by three pieces of information: its semantic feature configuration. they relate to certain role partners required by a word. thus making the meaning expressed even more specific.158 i. the units of description. Thus.

) little needs to be said about the second constitutive factor of a sememe. So. This is why dictionaries differ so drastically in this respect. It is well-known that primarily verbs. more recently. accounts for vagueness and provides the material for possible further sub-specifications. Ihus. the first two of which can be further sub-specified into sememes S1.2. S2 and S3. or S2. Brought to you by | UCL . which includes further. its role partners.2.3. the organization of a semanteme has to be understood as a complex network of partly overlapping and only partly well-defined feature configurations (sememes). and its ill-defined extension. Ihis has to do with a philosophical question of fundamental importance: if we assume . or. and I believe the most reasonable solution to this problem is the pragmatic one of explicitly selecting a degree of resolution regarded as appropriate. and it is admitted that further se- mantic and contextual sub-specifications would be possible but are not consid- ered reasonable. (The degrees of specification and the number of context factors associated in this scheme are. Whether these role partners have to be chosen from a pre-defined set of possible cases or are characteristic of specific lexemes or groups of lexemes is a matter of secondary inportance. usually associated variable features. It seems recommendable that the rank level to be chosen is the one where syntactic and syntagmatic context factors have been fully taken into consideration. 159 number of meanings of a polysemic word can be given for a specified level of precision but not in principle. which con- sists of only the obligatory features accumulated so far and can be regarded as an ideal stereotype. and accept the idea that the meaning of every occurrence of a word in context / parole has the status of a sememe (and both assumptions are reasonable ones).University College London Authenticated Download Date | 5/9/18 6:49 PM . then we have to conclude that theoretically the number of sememes of a word is infinite. a distinction has to be made between the well-defined sememe core. I would lean towards the second of these alternatives. but also other parts of speech. alternative sub-sets of further features re- sult in three major primary divisions S1. by frame theory.that in several occurrences of a word its meaning is always slightly different. at any level of analysis chosen.1 and S1. further sub-specifications would correlate mainly with the text- semantic and extralinguistic environment. as described by FiUmore in his case grammar. Diagram 1 illus- trates this organization principle schematically: features f to f a constitute \ the semantic core of the lexeme. of Bosch (1985) and others do . respectively. S2. chosen arbitrarily. require not only certain syntactic complementation structures but also semantic role partners.1. personally. and S2. The analy- sis is carried out to a chosen rank level.

or frequent collocates). Proba- bilities or frequencies relate to the level of a variety-specific norm in Coseriu's sense. the topic of a discourse). for it seems that "there is a tendency in language to restore to each meaning a form of its own" (Cohen 198O: 44).e. almost always.which is what I shall prefer be- low.e. Syntagmatic collocatability COTES next in importance. . syntagmatic co-occurrence conditions (such as selection restric- tions which determine possible collocation partners. these four types are not of equal standing: syntactic complementation condi- tions are most important and frequently distinguish the primary sememes of a word. usually. What is important to recognize is that the function of context is not primarily subtractive. i. Probabilities can also be indicated verbally by means of adverbs such as occasionally. each of the context factors associated with any particular se- meme is modified by a probability factor p. For example. Individual factors from each of these four levels may correlate with specific meaning configurations. in I remember being there it is the syntactic complementation by a verbal -ing form rather than an infinitive which ΆΛΛ·Ζ meaning components such as "factive1 and 'past1. and when talking about basketball or being on a playground it is likely that the word ball is meant to refer to a 'round object1 rather than to a 'festive event'.e. 16O The factors of context nay be classified into four categories: syntactic com- plementation (e. and extralinguistic context (e. choosing one of a set of alternative sub-specifications. arming further informa- tion components to an existing feature configuration. when think has the dynamic meaning of 'processing Brought to you by | UCL . i. i.g. i. text semantics (e.Categorical conditions cor- respond to what is called 'knock-out constraints' in variation theory. As indicated above.e.g. but rather both selective. For example.g. and additive. but also partly probabilistic. etc. objects or conditions in the immediate environment). Hie connection between specific meaning configurations and specific context factors in a sememe is partly categorical. excluding certain possibilities. while text-semantic and situational elements normally pro- vide finer and less general distinctions.University College London Authenticated Download Date | 5/9/18 6:49 PM . determine that a word is understood in some particular way. which may of course be 1 in the case of categorical conditions but normally lies somewhere between zero and 1. because there is no reliable way as yet of determining their exact mathe- matical value. which concerns what is common and typical in a language and is located between what is structurally possible (the system or langue) and what is actually realized in a speech event (parole). In the model displayed in diagram 1. as is normally assumed. the use of the predicate sleep identifies the respective subject as seman- tically 'animate'. whether or not a verb permits or requires finite clauses as objects).

the set of possible context factors of the chosen senate provides him with structural possibilities and. whereas the alternative sta- tic msaning 'hold an opinion1 categorically requires a simple form and most likely is complemented by a finite object clause. To ease understanding. and S3 and S4. Examples are identified by giving the respective corpus. text number. applications. it is very likely to occur intransitively. understanding) . we have to distinguish system-related (synthetic) probabilities. As far as quantification is con- cerned. leaves a certain degree of choice in generating any concrete utterance. one could try to eliminate some of the cross-classification that occurs. and analytic fre- quencies actually observed in 3. This description is based upon data from a larger study of mine (Schneider 1988). For example.Suffice it to say that all analyses and examples are corpus-based and taken from two machine-readable standard corpora. however. genre. limited by probability conditions. In the text. capitals are used to label semantic roles.University College London Authenticated Download Date | 5/9/18 6:49 PM . I would claim.a point which is important but which I cannot go into here for reasons of space. In synthesis (speech production) . lexico- graphical. In analysis (speech reception. on an intermediate first rank level. this connection between meaning and context factors permits two alternative points of view. Within the framework of a traditional communication model. respectively. e. features are presented simply in the form of brief paraphrases. e. that this analysis of the meaning of the word is much more appropriate than and superior to what is currently available in dictionaries.g. Brought to you by | UCL . the context factors observed by the recipient serve as a probabilistic clue to the actual meaning expressed by a polysemic item in a specific context.g. However. a speaker knows which meaning he intends to express. and line number. and one could rearrange the hierarchy to increase the number of ranks considered. I do not claim that this analysis is exhaustive or could not be improved. Description: An analysis of the verb consider Diagram 2 presents a schematic representation of the meanings and some of the relevant usage conditions of the verb consider within this model. and quite frequently it will occur in the progressive form. which also presents the methods employed in analysis . which are generally relevant figures associated with some form. the so-called Brown corpus and the LOB corpus. by combining S1 and S2. I have attempted to present an arrange- ment which can be grasped more easily and permits practical. 161 information in the mind1 .

the sub- ject is quite commonly realized as a collective body such as Committee. An example of this is: (1) He considered opening a can of beer but vetoed that idea. it approaches an inherent final point which. On rank level 3.1. This proposition has a dynamic predication. city council. while the last two are marginal and infrequent. of course. it is a possible future action which may be performed by the THINKER himself. called INFORMATION (INF). be able to do so. i. a non-factive propo- sition which is currently not real but may perhaps be realized at some future point in time.e. These concern the information thought about. however. In S1. S1 denotes a durative thinking process which is internally terminated. also that little deliberation has been spent on this natter before. Government. i. may be observed in this case: the verb consider frequently occurs in the progressive form. nor- mally in the syntactic function of the subject of the verb consider in an active clause. 138) With sememe 1. Here is an example of this usage: Brought to you by | UCL .2. The THINKER is the one who may cause this proposition to coma true. four of these are common. espe- cially with a subject pronoun Γ and in the present tense. 162 Ihe verb consider" basically refers to a cognitive (mental) activity which re- quires two semantic role partners: a human THINKER (abbreviated as ΤΗ).of the embedded proposition.University College London Authenticated Download Date | 5/9/18 6:49 PM .e. This forthcoming terminating point is the purpose of the process. withdrawal. who must.the affected OBJECTIVE . Club. It is normally assumed that this thinking process is easy and does not require any effort. namely a re- quired decision concerning the selection of one out of a set of at least two alternatives. etc.2. Congress. Two non-categori- cal context factors. it is only one of the complements . full senate. (Brown L18. The implied predication of the embedded proposition is something like 'cause to come true'. etc. a syntactic and a syntagmatic one.1. and the matter thought about. installation. In the case of S1. this sememe is further subspecified. and second. the thinking person will also be the AGENT of this possible future undertaking. is a possi- bility. the OBJECTIVE and syntactic object is an abstract noun which mostly denotes an event and is frequently a deverbal noninalization. which is syntac- tically realized as a verbal -ing clause with the equi-NP subject deleted. forma- tion. addition. without necessarily having to be active himself as the agent. although this is not yet considered very likely . the syntactic complementation of consider is an object noun phrase. no more. It has six primary sememes. such as appointment. has not yet been reached. however.

are procedure. carefully. the implied predication is 'cause to have1. A habitual collocate in this position is application'. situation. it is quite characteristic that the THINKER is not speci- fied at all. it tends to be a plural noun or one with generic meaning.1.2. movement. 78) In the case of S1. 1O8) In S2. If the subject does occur.2. 63) (5) .2. a concrete object: (3) If you are considering a part-time farm.. (LOB G67. question. and the latter is normally. possibility.. also denotes dynamic mental processing.2. . therapist and/or linguist. a durative process of intensively thinking about the INFORMATION in some detail with the purpose of integrating this information into one's own system of knowledge. problem. we frequently find collective bodies in the THINKER role (Council..1 has abstract nouns as objects. the INFORMATION is therefore new to the THINKER. evidence. such as the delegates. because the verb is in the passive or imperative. one which the THINKER is entitled to make. 140) The sememe S2 of consider. fact. the historian. A further step of sub-specification is regarded as reasonable. according to the corpora studied. Commission. and rate. thing. party. and opposed. although there is no association of a great deal of time being spent on it. as in and approve. evaluations. Court. though not always. element. the most frequent one. mostly by buying. Normally. or is the predi- cate of a non-finite subordinate clause. other Brought to you by | UCL . Germans and their allies. the THINKER is at the same time the person to receive the OBJECTIVE. and the process is typically carried out seriously. etc. and with sane attention to detail. factor. (IDE E30. implication. S2. (Brown B26. Again. which is revised in the course of this process with respect to sane of one's opinions. readers. the appeal board]. which is done intentionally and rationally and concerns a piece of INFORMATION which is in some way currently topical. 163 (2) No doubt many readers with new gardens will be considering the planting of hedges. The objects are mostly abstract nouns and denote something complex. The most character- istic variant is S2.. and denied. the University. Syntactically. and we may also note a slight tendency to co-occur in conjunction with speech act verbs which explicitly state this decision.University College London Authenticated Download Date | 5/9/18 6:49 PM . the durative thinking process is internally terminated and aims at an explicit decision at the end. non-Catholics. systematically. the thing which British sociologists need is to consider the implications of Weber's work for their own. collocates in this function. Exanples of this sememe are: (4) Consider the savage wounds that isolationism would inflict. (Brown F13. and others. consequence.

the function of illustrating a principle just mentioned. viz.. Semantically. written information transfer as in a textbook. S2. problem. The sememe permits a fourfold sub-division.2.1 and S2.3.University College London Authenticated Download Date | 5/9/18 6:49 PM . (KB P02. a typical kind of usage situation may be observed.3. S2.e. Brought to you by | UCL . the tendency to delete the subject and use the verb in the imperative (which occurs in about a fifth of all examples of this sememe). S2.2. in a valuable essay. respectively: (6) an informal committee . to the INFORMATION (11)..3 identifies the definite mental object as a detail within a greater context. there appears to be no restriction with respect to the nouns in the INF role. With S2. 164 nouns which have been recorded repeatedly are proposal. and this is manifested in syntagmatically co-occurring adverbials such as in (greater/a little more) detail.1. Finally.2. An optional prepositional complement with for may explicitly state the type of vacancy concerned. yet it emphasizes the concentrated and primarily the inchoative character of this process. to consider the advisability of arranging a conference on .3. aspect and process recur in the corpora. changes.3 denotes a mental process with the aim of the cognitive integration of the INFORMATION into one's own sphere of knowledge.. advice. matter. 97) As is the case with S2. Further. effect.2. Two dominant syn- tactic tendencies affect the TH-role. The following examples illustrate S2. S2. in this/the present paper. question. needs. with the implied predication being 'accept EXPERIENCER to fill a vacancy1. As to context factors. in a later/this section.. case. i. we again encounter the case that only one of the complements of the embedded proposition is realized as the syntactic object. especially the plural pronoun we. as a constituent of a more comprehensive conception (TO).1 denotes just the existence and identification of the specific INFORMATION (example 8). second. the conscious turning of the THINKER'S atten- tion to some aspects of the INFORMATION. the meaning S2. problem.3. correlation. and. namely the choice of a first person pronoun in more than a third of all instances recorded.2. This is obligatorily a human being who has the EXPERI- ENCER role in the proposition. yet seman- tically general nouns are characteristic. or in the inverse order of their presentation.2 initiates a more intensive step-by-step inspection of the matter considered (9). 88) (7) Her impact in the ZING ccranercials led to her being considered for an excellent part in an upcoming TV series (Brown N17. future time reference with shall and introductory and super- ordinate let us are characteristic. and the nouns case.4 attributes the role of an example.

yet infinitive clauses as ob- jects.. (Brown J14.. MjP or PrepP) only. 132) Consider S4 is treated next because it is more general in character although much less frequent and regular than S3. The passive is frequent and the perfect rare . Brought to you by | UCL . the object can be realized as a finite ω/z-clause postponed after the prepositional phrase with as (17). (12) He paused to consider and chatted on. as in The Ministry considers that . 56) In the case of the primary meaning S3.e. (LOB A14. It is implied that the ΙΝΕΌΒΜΑΠΟΝ is in some way important. which is much more frequent and common. and. It is frequently intransitive. Consider.4 is very close to think or ponder in expressing just a general kind of thinking process with the otherwise characteristic aim of mental integration no longer obligatorily associated. to which it is related. 165 (8) The problem we shall consider is the following. Norwich. also transformed into the generic passive. (13) Carvey considers that former Vice President Nixon would be Brown's most formidable foe (Brown B11. 11) S2. which. 50) (11) a nineteenth century urban church . Exceptionally. and the wider context usually makes clear that no decision or inherent terminating point need be envisaged. 25). for example. the degree of subjec- tive certainty is higher. 167) (9) Let us oonsider each of the three types of cost in turn (Brown J50. i.University College London Authenticated Download Date | 5/9/18 6:49 PM . however.167) (10) we consider now the graph of the function f(t) (Brown J20. I propose that an underlying copula predication with the ESSIVE as complement has to be posited. above all. This sememe seems to be somewhat formal and can be used to express an official opinion held by some organization or public institution. and as a complement following the preposition as (16). (LOB P28.otherwise. after some thoughtful de- liberation.. may appear in the surface structure in a number of ways. as a second complement (NP.. and that the opinion has been a- chieved after some deliberation. again. the HUNKER is always a definite person. This is normally assumed to be done rationally and. Syntactically. with the copula deleted (15). It expresses the static existence of a mentally stored serious and rational opinion.. no further context conditions are apparent. the meeting house at Old Meeting. the subjective opinion stored is rated as 'true'. this proposition is semantically 'qualifying' in that it characteristically attributes an inherent quality or category membership (the role ESSTVE) to the subject (OBJECTIVE). following the OBJECTIVE which always holds the position of direct object: as non-finite to be + complement (see example 14). (LOB D02. a non- factive proposition rated subjectively as 'probably true1. are possible as well. It is typi- cally complemented by a finite object clause.

this is easily conceivable within the current framework.University College London Authenticated Download Date | 5/9/18 6:49 PM ... (18) We again consider a fixed point P at . as centers and agents of Westernculture. (Brown D10. occurs re- peatedly in scientific writing but is apparently restricted to that technical variety. 58) S5. 3) (16) Africans and Asians tend to consider not only missions but the local churches . 78) 4. 7) (15) What degree of interference shall be considered excessive (Brown H23. (Brown J08.. Brought to you by | UCL . The object noun is neces- sarily something concrete. and how.. we know our intentions. which. 92) (17) I think we can . sociolinguistic approaches.. (LOB G63. 166 (14) Biological warfare is considered to be primarily a strategic weapon. and the model there- fore permits a connection of both structural and categorical aspects and in- herent variability and. and a variable point Q at . I hope. a certain degree of fuzziness which always remains. as far as I can see. it is congruent with a view of language which has corns to be known in recent years as variation theory.. the problem of counting sememes at least at certain rank levels. (19) . ex- presses a process of visual contemplation in addition to a quiet and somewhat distanced. is not mentioned by dictionaries.. Conclusion The above analysis of the verb consider has demonstrated. 155) Finally. that the model of polysemy suggested earlier can cope with some of the central problems which this phenomenon poses: the internal arrangement of the meanings of a semantically complex word. durative and aimless mental processing.. which is also very infrequent but does occur in the corpora. Although extra- linguistic context has not been taken into consideration in any great detail. (Brown J2O. He considered them with brooding eyes.. and pulled out the six slugs he had taken from the revolver. brows bunched as his brain grappled with the problem (Brown L18. S6. for example. Thus.. consider as an independent question whether. without actually knowing what this value is or might be. The struc- ture is transitive with an appropriate object NP. It means that the value of a mathematical or physical variable is deliberately assumed and fixed for the purpose of further argumentation and analysis. and the corre- lation between meaning configurations and context factors.

Sebeok (ed. Forum Linguisticum 5. 1988. 1966.). Seuren.). 395-477. Volume 3: Theoretical foundations. Foris. M. Gerald Leonard. 1966. Pieter A. (ed. M. Polysemie und Unscharfe der Wortbe- deutung. Brought to you by | UCL . M. Paris. Anton J. J. Dordrecht. Band 2: Studien zur lexikalischen Semantik der mentalen Verben des Englischen. Variabilität. Schneider. Geer A. 251-258. 1985. Band 3: Theoretische und methodische Grundlagen. J. In: Themas A.). Weinreich. M. Hoppen- brouwers. Thomas A.. Pieter A. On semantic differentiation. Meaning and the lexicon. Cohen. In: Geer A. Explorations in semantic theory. Uriel. Anton J. The Hague. Hoppenbrouwers. Current trends in linguistics. Mauton. Lexical meaning contextualized. Weijters (eds. Seuren.). Cinnaminson. Tübingen. 167 REFERENCES Bosch. M. M. 44-52. 1980. Peter. Niemeyer. Weijters (eds. 1985. Edgar W.University College London Authenticated Download Date | 5/9/18 6:49 PM . Sebeok.

168 Meaning configuration context factors core rank 1 rank 2 var.University College London Authenticated Download Date | 5/9/18 6:49 PM . features synthetic (speaker's) perspective analytic (hearer's) perspective f = feature cf = context factor (i) i S = sememe p_ = probability of cf Diagram 1: Schematic representation of a polysemic lexeme Brought to you by | UCL .

2 TH=EXPER(INF) JINF=concrete \ INF=non-factive Lpred(INF)=have Γ object=NP=OBJ(INT) pred ( INF) =dynamiq INF=sub j . form •f s u b j . Ugradually esD. co-occurrence w i t h conj.(INF=propos.2. 3. " of knowledge \\\ [cognitive a c t i v i t y ) tegrate INF INF=definite occ.real in fut Jurative brief duration often no TH (passive TH is able to has purpose: in. no effort] equi-NP deleted objN=abstract .) Sl. f i n i t e clause dynamic determine deliberately _context-mathem. K alternatives .pred(INF)=-copula.=generic !T S2. or non-finite p r e d .3 S2.4 tegrate INF context: no decision \ static INF«proposition ..1 general IMF=topical i S2 f [IKF-def inite 1 subj. 2. 1 *. rational \ 52. pro- bably true S4 f_ . . : progr. un.2 written transfer ftUoften imperative -has purpose: in. emantic s t r u c t u r e of the verb consider Brought to you by | UCL . collocate: application perform pred(INF) S2.2 S2. concrete poss. purpose: further physics name of variable argumentation INF=unknown ^ S5 dynamic process in mind | & observe visually objN-concrete no purpose ^ calm. lINF-important lobi.3.3 \ \< occ.University College London Authenticated Download Date | 5/9/18 6:49 PM . "little previous object=V'ing=pred(INF) : ι 1 process in mind jdurative | i n t e r n . qualifying ^— —_ ' · i passive frequent INF=subi. ΓΗ is able to plINF-abstract ?. terminated . 1 TH=CAUSE of pred. usu. '.INF=proposition P· but ongoing las purpose: pred(INF)=accept L EXPER to fill ΞΞΞΓ frequ.2.3.l — _ thinking 3BJ(INF)=abstract .4 durative INF-new. with inner distance S6 Diagram 2: The .7 pbjN»abstract : f r e q u . . terminated r (INF-propos. S2.) f / . etc.1 *· }objN-abstract.\ subjN often collective iNF=p reposition SI.2.pred(INF)=cause to become real £- SI.— — INF=new. usu. i . .3. we Inchoative i Istep by step | t^ <ith attention S2.2 systematic occ . PrepComp for process in mind exact • intentional — . but ongoing has purpose: se- lect one of sev. core rank 1 rank 2 rank 3 var. TH-collective occ. / + as comp. unknown usu.2 |f- ^ objN=usu. . ) perform pred(INF) tegrate INF known often subj. features context factors 169 Idvnamic JTH=AGENT of pred.=usu.=otten 1st sg. f u t u r e shall intensively INF=detail S2.true S3 some previous "* thinking static have stored serious rational _— lNF=proposition 1INF»non-factive IVF=sub j . TH-definite person S2. superordinate \ let us INF-definite \ L INF-detail characteristic a d v e r b i - INF=example \ als: in detail.1 ι durative ^ intern.Jeasily. OBJ Sl. speech act verb decide a vacancy serious \ intensively EXPER-human careful objN-EXPER-human dynamic S2. relevant INF=not real now.-IINF»definite objN=often a b s t r a c t . syntax: objN + (to be) i comp.deverbal Ν f r e q u . intransitive -INF=theme κ has purpose: in.1 / event.

2 Fodor1s (197O) objections have been answered in Seuren (1985: 2O4-8). has claimed for years now that all lexical meanings are one and indivisible. SEUKEN LEXICAL MEANING AND PRESUPPOSITION1 0. M. assassinate is 'murder1 plus something extra. or can be. 'kill unlawfully. not the prevailing opinion. In fact. going back to Plato. field analysis. Examples like these are eas- ily multiplied.170 PIETER A. probably because of the absence of clear operational criteria for the distinction between presuppositional and assertive content.e. One aim of this paper is to provide sufficient workable criteria in this respect. which gained wide recognition but remained without a follow-up.University College London Authenticated Download Date | 5/9/18 6:49 PM . and various systems have been thought up to implement lexical ana- lytic procedures. of lexical analysis. Introduction Not everyone is agreed that lexical meanings are. forthcoming). And. The lexical aspects of this theory. It is hoped that this study will contribute to further investigations along the lines set out by Fillmore and followed up here. and c. systematic comparison of lexical items irresistibly invites analysis. murder. 'murder an important pub- lic figure on account of his/her public status'. with malice aforethought'. as the legal definition has it. complex and are thus open to analysis. Through the centuries. three main approaches have been developed to carry out lexical analysis: a. murder clearly is 'kill' plus something extra. Compare. i. can be seen as 'cause to 2 die1.. following McCawley. componential analysis b. for example.e. for example. prelexical analysis. the three English verbs kill.. 1 The ideas developed in this paper are a direct consequence of the theory of presupposition as developed in Seuren (1985. which are central in this paper. are directly inspired by Fillmore's seminal paper (1971). however. Whereas kill. there is an old tradition. and assassinate. Fodor. In fact. Brought to you by | UCL . in its turn. i. This is.

a periphery. what we find is two-dimensional space. The meaning of a word is compared with a 'field1. In a still rather blurred way.University College London Authenticated Download Date | 5/9/18 6:49 PM . We find it in Katz/Fodor's (1963) and Katz/Postal's (1964) system of features and markers. (In itself such an exercise would seem to be of great value. or rather a metalexicon. The metaphor could. for the expression of the semantically atomic components. All three types of lexical analysis. Prelexical analysis is historically bound up with the movement in grairmatical theory known as 'generative semantics'. In any case. in Leibniz's famous attempt at a universal analytic metalanguage for the description of all lexical items of a language. For componential analysis it has so far proved impossible to establish a workable metalanguage. to apply any of these three types of analysis to a natural language lexicon as a whole. It is a form of lexical semantic decomposition based on an analogy with 'open' syntax: for certain classes of lexical items. elements of this approach are encountered again in Dixon (1971. or η-dimensional. But the task has so far proved to be beyond the Brought to you by | UCL . flourishing in the late '60s and early '70s. of course be extended to three-dimensional. or downright impossible. of componential analysis of kinship terms. however. it always proved unfeasible. just as geometrical planes are. but then it becomes hard to see what is gained by the notion of space. and the closely related method. but more clearly in the writings by Trier between 1931 and 1934 (Trier's theory of 'sprachliche Felder'). for example. which has a centre. this notion is found in Erdmann (191O). which was followed in classical antiq- uity by a long stream of etymologizing analytic treatises on word forms and word meanings. Word meanings are thus constrained. Field analysis is a 2Oth century innovation. but with an accent on 'nuclear1 versus 'non-nuclear' items (verbs). have so far met with insurmount- able problems: whereas applications to isolated sections of the lexicon seemed feasible. It is found. to a definition in terms of two parameters. since any proposal as to such a metalexicon necessarily amounts to a hypothesis about the cognitive categorization according to which the members of a speech community interpret the world. mainly used in anthro- pology. striking parallels can be pointed out between their syntactic properties and those of their 'open' syntax correlates. space. 171 The method of conponential analysis is the oldest and most widely practised one We find it in Plato's dialogue Cratylus. It is based on an application of the metaphor of geometrical extension. 1972). and topologically definable rela- tions with neighbouring fields. strictly speaking. such as causative verbs. In a less geometrical way.

but without any axiotiatic underpinnings. Given the massive amounts of translations that have to be made in the context of vastly increased international contacts on all levels. a few remarks of a general methodological nature are in order. Three points need to be considered in particular.172 powers of those who acoepted it. it has serious theoretical shortcomings (such as. has so far lacked the generality of application required of a general method of lexical analysis. likewise. it becomes impossible to predict categorically the further course of the process or the Brought to you by | UCL . And in order to make up for deficiencies in the definitions given.) Field analysis is too heavily constrained to stand a chanae of being generally applicable to lexical items (which clear- ly have more than two dimensions of description. therefore. Prelexical analysis. is determined by too many para- meters of too many different kinds must give up the hope of predictive power: when systematic processes or structures can be 'broken into1 at any matEnt or any place by factors whose appearance is largely due to chance. that we still lack a general method of lexical analy- sis. It also suffers fron practical disad-i vantages. Theories dealing with sets of data whose character.if it nakes sense at all to speak of the number of parameters required in this respect). . Word meanings are analysed there on an ad hoa basis. it must be stressed that lexical theory cannot be of the predictive kind. this would seem to be an interest of con- siderable magnitude. But before we can proceed to a discussion of this distinction. to the extent that it is systematic. and that is likely to be of considerable use in systematizing the semantic analysis of lexical items. in partic- ular. lexicog- raphers resort to examples meant to convey the conditions for the correct use of the words in question. Although this practical method has served us well through the ages. This (embarrassing) fact is reflected in the monolingual and bilingual dictionaries that are available today. Definitions are given.University College London Authenticated Download Date | 5/9/18 6:49 PM . in that translation processes could conceivably be speeded up con- siderably if dictionary lemmas were structured in a more systematic way than they are in the dictionaries that are available nowadays. but must of necessity be retrodictive. 1. What this paper is meant to illustrate is that there is in any case one general distinction that can be maintained throughout the lexicon of any language. The upshot is. Ihis distinction is closely linked up with the phenomenon of presupposition. that of definitional circularity). Preliminary points First.


eventual shape and functioning of the structure. In such cases the factors
involved are too many and beyond control. Predictive theories need data that
are sufficiently insulated or encapsulated. If this condition is not fulfilled,
the best that can be achieved is a retrodictive theory, which can locate causal
factors only after a fact has occurred or failed to occur. Its predictive
power is then necessarily restricted, at best, to a range of possible alterna-
tive future phenomena. Typical exanples of retrodictive theories are found in
the historical sciences, most of the medical sciences, meteorology, and lexical
theory. Che should, therefore, not expect lexical theory to be able to predict
whether a given language will have some specific lexical item, or whether items
of a certain form or structure will have some predicted meaning or grammatical
function. The best that can be achieved is, after observation of lexical facts,
indicate the factors that must have been at work, to the extent that they are
identifiable. The first task of lexical theory is, therefore, to detect lexical
regularities and devise a system of categories and possible factors.
The second point is closely connected with the first: lexical items tend to be,
by their very nature, unique. They symbolize or express concepts, and we speak
of a concept when the conceiving subject recognizes phenomena with certain
properties as being of a kind. Conceptualization processes are heavily depend-
ent on idiosyncratic factors of personal or group experience and functionality,
while at the same time being constrained by system-internal structuring prin-
ciples. Lexical items can be said to represent Gestalt knowledge. They arise
typically when new kinds of recognizable phenomena present themselves in the
lives of the speakers. Examples abound, especially in the sciences, but also
in ordinary life. The English word engine underwent a rapid and dramatic spe-
cialization and fixation of meaning when engines became a fact of life. Pop
groups bring out albums, not records, presumably because, in the relevant
circles/ records with pop music on them represent a separate recognizable
category. The word record itself is a recently acquired lexical item with the
specialized meaning of flat disc with engravings that can be transformed into
acoustic signals by means of a gramophone (note that a compact disc is not a
record ). It is, however, not possible to predict that a language will make a
lexical distinction between record, album and compact disc. All that can be
done is say, in hindsight, that, apparently, such and such developments in
public or cultural life have been responsible for the emergence of these lexi-
cal distinctions.
The third point is equally closely connected: to posit internal structure in

Brought to you by | UCL - University College London
Download Date | 5/9/18 6:49 PM

the meanings of (a class of) lexical items does not imply that such structural
analyses should exhaust the meanings in question. In other words/ lexical analy-
ses need not be compositional in the technical sense of the term as it is used
in formal semantics: the total meaning of a lexical item may be more than the
value resulting from the functional calculus of its parts (or, as it is often
put in vulgarizing terms, more than the sum of its parts) .(.Whether it may per-
haps, in some cases, also be less than 'the sum of its parts' is an open ques-
tion that will have to be decided on empirical grounds.) This point applies in
particular to the componential and to the prelexical methods of analysis. It
is quite conceivable that a successful and convincing componential or prelexi-
cal analysis is proposed for one or more lexical items while, at the same time,
it is recognized that the item or items in question have meanings that go well
beyond the analyses given. The reason for this is precisely our second point:
lexical items tend to be semantically idiosyncratic and arise according to
communicative needs. Since new concepts are often specializations of existing,
wider concepts, it is quite natural that speakers should make use of existing
lexical structures associated with the wider concepts and add the specificities
as a small extra ballast, using either a new lexical form or an existing form
which then acquires ambiguity or at least recognizably specific usage. Record
in the sense of 'grammophone record1 is an example of an already existing item,
now equipped with a separate though related specific meaning. The English verbs
murder and assassinate are specializations of the wider item kill, whereas, as
we have seen, assassinate is a further specialization of murder. It is perfectly
possible to maintain that all three of these verbs have the same "skeletal1
structure of 'cause to become dead', while at the same time carrying different
idiosyncratic 'extras'.
The non-compositionality of lexical analyses is mirrored in compound nouns, but
these have the advantage that their 'internal analysis' is there for everyone
to see. Here, too, the resulting meaning is only partly derivable from the com-
ponent parts; a tennis elbow is not like a tennis racket, and the difference
is not reducible to that between an elbow and a racket. Examples are rife: coat-
hanger and cliffhanger, faae towel and face value, city hall and concert hall.
Available background knowledge is clearly essential in determining the correct
final meaning, together with conventionalized linguistic meaning (as with

Chomsky's criticisms in this respect (1972: 72, 142-6) miss the point,
since they are based on the misconception that prelexical analyses should
be compositional.

Brought to you by | UCL - University College London
Download Date | 5/9/18 6:49 PM


ol-iffhanger). Conpound nouns are thus a living proof that structure assignment
need not mean compositionality: sheer memory, i.e., the lexicon, will supply
the missing elements.
The point of having a non-compositional 'skeletal1 structure is obvious: it
cues, or prompts, the associated meaning. Compounds do this overtly, in the
external acoustic material. Single lexical items must be assumed to do the
cueing or prompting internally, thereby facilitating lexical search proce-
dures. Wie main problem is empirical: what kind of evidence will support in-
ternal analyses, distinguishing linguistic from encyclopaedic knowledge? Cnly
partial answers are available as yet, as we have seen. An adequate general
anwer is still lacking, but it should be very high on the list of priorities
in lexical studies.

2. Presuppositions and lexical conditions
Let us pass on now to the main topic of this paper: lexical, analysis in terms
of different kinds of lexical conditions. The distinction links up directly
with presuppositions. We shall speak of a presupposition P of a carrier sen-
tence A when:
a. P is an entailment of the assertive form of A;
b. the sequence 'P and/but A1 forms good discourse, while Ά and/but P 1
does not;
c. a sequence of the form 'possibly not P, but A1 is felt to be
Thus, (1b) is a presupposition of (1a), but (2b) is not a presupposition of
(1)a. Susan has forgotten that today is her birthday,
b. Today is Susan's birthday.
(2)a. Lady Fortune neighs.
b. Lady Fortune is a horse.
(1b) is an entailment of (1a): it is intrinsically and per se impossible for
(1a) to be true while (1b) is not true; then, "Today is Susan's birthday, but
she has forgotten that it is' is a perfectly natural bit of discourse, while
the reverse order produces an awkward sequence; and 'Maybe today is not
Susan's birthday, but she has forgotten it is1 is intuitively contradictory.

For detailed analyses and discussions, see Van der Sandt (1982; 1988)
and Seuren (1985: 21O-3, forthcoming).

Brought to you by | UCL - University College London
Download Date | 5/9/18 6:49 PM


For (2), we notice that (2b) is at least not clearly an entailment of (2a):
is it or is it not possible for someone to neigh while not being a horse? The
sequence 'Lady Fortune is a horse and she neighs' is good discourse, but so is
'Lady Fortune may not be a horse, but she neighs'. According to the definition
given, (2b) is thus not a presupposition of (2a).
It has been observed by a large number of authors that presuppositions have
at least one specific property not shared by non-presuppositional entailments:
when the carrier sentence A is embedded under an entailrnent-cancelling opera-
tor, such as negation, intensional operators of propositional attitude (be-
lieve, hope, wish), or certain modal operators, then the presupposition is,
normally, not simply cancelled or lost, but is retained as what is variously
called an 'invited inference1, "suggested inference1, or 'default assumption'.
Thus, while (3a) presupposes (3b), (4a) does not presuppose (4b) but still
has (4b) as invited inference or default assumption:
(3)a. John's son lives in Kentucky,
b. John has a son.
(4)a. John thinks that his son lives in Kentucky,
b. John has a son.
This property is not shared by Ordinary* entailments, as appears from the
fact that, while (5a) entails (but not presupposes) (5b), (6b) is not retained
as an invited or suggested inference (default assumption) of (6a):
(.5) a. John bought tulips,
b. John bought flowers.
(6)a. John thinks that he bought tulips,
b. John bought flowers.
What interests us here, however, is not so much the phenomenon of presupposi-
tion itself as the question of the structural source of presuppositions. The
(fairly vast) literature fails to give a satisfactory answer to this question:
presuppositions are accepted as 'being there' even though it is not known how
or why. Only in Seuren (1985, forthcoming) is a general answer provided. In
that analysis, presuppositions derive in principle from the semantic proper-
ties of the highest lexical predicate of the carrier sentence. This applies
equally to the so-called 'existential1 presuppositions and to the 'factive1
and 'categorial1 presuppositions.
In the case of existential presuppositions we have to do with entailments of
existence, as in (3) above: for someone to live in Kentucky he has to exist;

Brought to you by | UCL - University College London
Download Date | 5/9/18 6:49 PM


hence the existential presupposition (3b). This is the class of presuppositions
that attracted philosophers: a sentence like 'The present king of France is
bald1 (Russell 1905), presupposes that there is a king of France at the moment
of speaking. In Seuren's analysis this class of presuppositions is derived
from the lexical property of the highest predicate in the sentence in question,
i.e., live or bald, that its subject term must refer to a really existing
entity for that predicate to be applied truthfully. We say that these predi-
cates are extensional with respect to their subject terms. Although predicates
are usually extensional with respect to their terms, this is by no means always
the case. The English predicates look for, talk about or think of, for example,
are not extensional with respect to their object terms: one can perfectly well
talk about something without that something actually existing, and analogously
for the other predicates.

Factive presuppositions have always been associated with sentence predicates,
the so-called f active predicates. These are predicates (verbs, adjectives, or
other predicate expressions) that take an embedded clause as subject or as
object term, with the special property that the truth of the embedded clause
is presupposed, as in (1a) above, with the factive predicate have forgotten.
Examples of factive predicates with factive object clauses are know, realize,
remember, have forgotten, regret, discover, Factive predicates with factive
subject clause are, e.g., be a pity, be a shame, be regrettable, be (un)fortu-
nate, but not, e.g., be probable. Here, the lexical origin of the presupposi-
tion is obvious (Kiparsky/Kiparsky 1971).
The categorial presuppositions are much more idiosyncratic, and thus more
typically lexical. This class is best illustrated by means of a few examples.
Thus, (7a) presupposes (7b), (8a) presupposes (8b), and (9a) presupposes (9b):
(7)a. John has cone back,
b. John was away.
(8)a. John confessed that he had forged the signature.
b. The forging of the signature was bad or criminal.
(9)a. rfy neighbour is buxom.
b. My neighbour is a woman.
Presuppositions such as these can likewise be derived from the semantic condi-
tions associated with the main predicate: oome back requires of its subject
term referent that it was away; confess requires of its object term referent

Brought to you by | UCL - University College London
Download Date | 5/9/18 6:49 PM


that it be bad or criminal; buxom imposes on its subject term referent that
it be a woman.
Notice, however, that some Ordinary1, i.e. non-presuppositional, entailments
are likewise derivable from lexical semantic conditions for truth. In standard
logical systems, the logical entailments are specified on the basis of the
logical structure of sentences (propositions), including certain specifically
logical elements, such as the quantifiers, the truth-functional operators, and
any other sentential operator for which a logic has been devised (in particu-
lar the modal operators of possibility and necessity). But apart from these,
there remain vast classes of non-presuppositional entailments that are purely
and uncontroversially lexical. logicians consider these as falling outside the
realm of logic, but they can hardly be considered as falling outside the realm
of (lexical) semantics. Examples are (10)-(12), where the (b)-sentences are
ordinary entailments of the (a)-sentences:
(10)a. Harold has bought a new car.
b. Harold has either paid money or committed himself to pay money.

(11)a. Harold has died,
b. Harold is dead.
(12)a. Harold is a father.
b. Harold has one or more children.
Entailments such as these can only be derived from semantic conditions asso-
ciated with the main predicate of the sentence in question: buy implies paying
money, die implies an ensuing state of being dead, and be a father implies the
having of children.
Our problem now is that we must distinguish between those lexical conditions
that generate presuppositions and those that generate only ordinary entail-
ments. For this purpose a distinction is made between two categories of lexi-
cal conditions, the preconditions, which generate the presuppositions, and the
satisfaction conditions, which generate the non-presuppositional, Ordinary'
lexical entailments. Formally, we associate with each predicate an extension,
i.e., a set of individuals for unary predicates, a set of pairs for binary

There is a 'remainder' class of presuppositions, mainly those associated
with or generated by words like too, only, even, or cleft and pseudo-
cleft constructions. In Seuren (1985: 295-313) it is proposed to treat
these as, likewise, lexically derived: the words in question are consid-
ered to be (abstract) predicates, and (pseudo)clefts are analysed as
standing under the main predicate of (specifying) be.

Brought to you by | UCL - University College London
Download Date | 5/9/18 6:49 PM


predicates, a set of triples for three-termed predicates, etc., such that the
predicate in question actually holds for the members of the associated set.
These extension sets are, of course, to be defined not by enumeration but by
comprehension, i.e., by stating the conditions that must be fulfilled by indi-
viduals, pairs, triples, etc., to qualify as members. These conditions are
then divided into two categories, the preconditions and the satisfaction con-
ditions. A viable notation is the following, whereby the symbol 'σ 1 denotes
the function that takes the predicate to its extension set, the colon intro-
duces the preconditions, and the vertical stroke introduces the satisfaction
conditions (we take the predicate 'P 1 to be binary: P 2 ):
ι·ο2\) —
ο(r - tι < e f, r > : (preconditions)
. . . . . . . . . . |ι . (satisfaction
. . . . . . . conditions)
. . . . . . . /-.6
Now, a sentence of the form 'P 2 (a,b)' - where 'a 1 and 'b' are the subject term
and the object term, respectively, is true just in case the referent of'a 1
fulfils both the preconditions and the satisfaction conditions of P2 in so
far as they affect the first member of the pair < e,f >, and the referent of
b likewise fulfils both the preconditions and the satisfaction conditions for-
mulated for P2 with respect to the second member of the pair < e,f >. One is
thus entitled to conclude from the truth of P 2 (a,b) to that of any sentence
expressing the fulfilment of the preconditions and satisfaction conditions of
P 2 . Presuppositional entailments are generated by the preconditions; Ordinary'
lexical entailments by the satisfaction conditions.

3. Further aspects
Preconditions, and the associated presuppositions, are highly functional with
regard to processes of actual linguistic communication. Human ccrtmjnication
essentially involves the building up of discourse structures: we would be ut-
terly unable to communicate the way we do if we were restricted to single sen-
tences whose semantic interpretation was entirely self-contained. Philosophers
have, at times, entertained such a view of communication, and devised analyses
that made sentences semantically entirely self-contained, but a problem analy-
sis quickly shows that such a view is untenable if not naive. In speech it is
continually necessary to refer back to an entity or set of entities referred
to earlier. This can very often not be achieved in any other way than by first

6 Informally: "the extension of P 2 is the set of all pairs < e , f > such that,
first, ...(preconditions)..., and, then, ...(satisfaction conditions)...".
7 I am referring in particular to the analysis proposed by Russell in his
'Theory of Descriptions' (19O5), and by Quine in various works, especially
(I960). For a refutation of Geach's (1969; 1972) attempt to save this kind
of analysis, see Seuren (1977; 1985: 319-22).

Brought to you by | UCL - University College London
Download Date | 5/9/18 6:49 PM


fixing reference and then maintaining it, with the help of formal linguistic
means, through a bit of discourse. It is the central function of definite
terms to do precisely that, as in:
(13) Last night I saw a movie. The movie/that movie/it was terrible.
It is now beginning to be accepted that semantic analyses of sentences pre-
suppose (in a non-technical sense) a mental mechanism that sets up so-called
discourse-representations, containing representations of (sets of) entities
that have been set up in the discourse at hand ('addresses'). Definite terms
denote (i.e., select and 'land at') the intended addresses.
The complex of presuppositional phenomena has found a functional interpreta-
tion in terms of discourse representations. It has been established (Van der
Sandt 1982, 1987; Seuren 1985; Fauconnier 1985) that presuppositions (and thus
lexical preconditions) fulfil a dual function: they play a role in the deter-
mination of truth-values with respect to a given model ('the world 1 ), and,
moreover, they contribute in an essential way to coherence and acceptability
of discourses. The general view is that presuppositions must have been incre-
mented to any given discourse domain before the carrier sentence can be deemed
'acceptable1 in that discourse. Tunis incrementation can take two forms: actual
or virtual. 'Actual incrementation1 implies that some sentence expressing the
presupposition in question has been actually pronounced in preceding discourse,
and has thus been incremented in virtue of having been uttered. 'Virtual incre-
mentation1 implies that the sentence has not been actually pronounced, but is
incremented post hoc, when the carrier sentence is uttered. This post hoc, or
'backward1 incrementation (also called 'acccranodation') is made possible by
the fact that presuppositions are structural properties of sentences, derivable
from lexical preconditions. Backward incrementation is, moreover, subject to a
cognitive condition: normally speaking, a presupposition can be incremented
post hoc only if it is compatible with available background knowledge.
Under conditions that are not very clear, this background knowledge check can
be overruled. Typically, in such cases, a metaphor comes about. E.g.:
(14) And the train was waltzing at sunset round the walls of Verona.

8 For a detailed discussion of this aspect of linguistic communication, see
Seuren (1985: 214-8).
9 From: E. M. Forster, Where Angels Fear to Tread.

Brought to you by | UCL - University College London
Download Date | 5/9/18 6:49 PM


This sentence violates a lexical precondition of the predicate waltz, which
requires that its subject term refers to an animate being capable of walking
or dancing. A train, obviously, falls outside that category, and backward
incrementation of the presupposition that the referent of the subject term is
an animate being capable of walking (or dancing) should thus be blocked. Ap-
parently, this veto is nullified on account of a temporary upholding of the
fiction that trains are capable of walking. But we shall not go further into
such questions here.
Sentences that carry presuppositions are thus seen to have a limited usability:
they can be used only in such discourses as either already contain their pre-
suppositions or allow for these presuppositions to be incremented post hoc.
This is a highly functional property of sentences, since (a) it provides for
error-detecting checks on correct understanding of a given discourse, and (b)
it allows for the building up of mental discourse representations with a mini-
mum of actually uttered sentences. This latter property crucially depends on
the mechanism of backward incrementation, with its requirement of compatibility
with available background knowledge.
But let us return to the distinction between 'preconditions1 and 'satisfaction
conditions'. One immediate consequence is that the machinery of preconditions
is crucial for an understanding of the phenomena of polysemy. The term 'poly-
semy1 is used for the at first sight curious fact that identical definite terms
can refer to things of very different categories, depending on their context
of use. Thus, the definite term the school in (15) will refer to a building in
(15a), to an institution in (15b), to a set of people in (15c). And the defi-
nite term the game in (16) will take a temporally limited process as a possible
referent in (16a), but the equipment for an activity in (16b), and the desired
result of an activity in (16c):
(15)a. The school is on fire.
b. The school had excellent results last year.
c. The school has a day off.
(16)a. The game lasted two hours.
b. I bought the game for little money.
c. He had the game in his hands.
The preconditions of the various predicates used provide directives as to the
right category of referent: to be on five requires preconditionally a concrete
object; to have results requires a functioning organism as referent for its
subject term; to have a day off, in its turn, wants a human subject. These

Brought to you by | UCL - University College London
Download Date | 5/9/18 6:49 PM

It should be noted that the distinction between these two kinds of lexical conditions is not limited to words or expressions that function as verbs or adjectives in surface structures. And similarly for (16): given a general (approx- imate) definition of game as "play bound by rules and yielding a winner1. Thus there are conditions for anything to be properly called a house. Likewise. etc. with a systematized descriptive metalanguage and a systematic distinction between preconditions and satis- faction conditions. it is perfect- ly possible to treat even function words as predicates: and.. where a systematic analysis is carried out of German and Dutch verbs of hearing. Such entities can be world individuals. or facts. fortunate. if.University College London Authenticated Download Date | 5/9/18 6:49 PM . A few things strike the eye irttnediately when such analyses are carried out. or properties thereof. there are conditions for anything to be properly called reliable. etc. in terms of preconditions and satisfaction conditions. in principle verbs. So far only incidental probes have been carried out on isolated words to estab- lish their preconditions and their satisfaction conditions. adjectives. i. because. a table. etc. or processes. not. although. irrespective of the surface category of a word. OP. 455). From the point of view of the lexicon. a law. quick. we get "game manifested as desired result1. negation 1O With the exception of Vliegen (1986). And there may be certain advantages in such an analysis. 'school manifested as a group of people1 for (15c). TMs is so because. Formally. any such solution to the theoretical problem of polysemy can only be validated by means of detailed lexicographically adequate semantic descriptions of predicates. But for our present purposes it is sufficient to stipulate that all items with lexical meaning. necessary. possible. we get for (16a) 'game manifested as a process'. and prepositions. Clearly. leads to 'school manifested as concrete object1 for (15a). First.e. For (16b) we get "game manifested as object of purchase'. can be seen as abstract predicates over extensions of (pairs of) clauses (Seuren 1985: 314-46. all items with lexical meaning are predicates. when it has a lexical meaning it imposes lexi- cal conditions on whatever entity it can be applied to. given the predicate have In one's hands. Brought to you by | UCL . And for (16c). 'school manifested as functioning organism1 for (15b). nouns.182 conditions. since that is what the predicate last requires. in association with the general definition of school as 'teaching institution1. are predicates from the point of view of lexical meaning analysis. etc.. adverbs. given the predicate buy.

Let the extension set of bachelor be defined as follows: (bachelor) . content of a sentence. i. in outline. since Strawson (195O). while the differences concentrate in the preconditions. singularly. [-married].University College London Authenticated Download Date | 5/9/18 6:49 PM . In unmarked cases. there is no rationale for the fact that the negation singles out the [-married] feature. male and adult | e is unmarried } Now.3) shows that negation over Negative Polarity Items neces- sarily affects only satisfaction conditions. We can thus say that some- one. compared with German kahl or Dutch kaal. on the principle that unmarked negation affects only the assertive. for example.. in that bald has the extra preconditions that the subject term referent should be (part of) an animal being or textile (with the single idiosyncratic addition of 'rubber tyre 1 ). But we cannot say.e. bald differs from its Eastern counterparts kahl/kaal. either in the same language or across languages. and. that the landscape is bald. for example. or that the wall is bald.{ e : e is human. it is clear why (17) implies that Harold is married. tend to differ little in their satis- faction conditions. leaving the pre- conditions unaffected. normally speaking. [-ttnale]. Ißt us consider. 11 Seuren (1985: ch. Oadult]. In German and Dutch. which says that this predicate applies to an entity which is [+human]. and tends to leave the other features unaffected. to spot presuppositions. All three words have a precondition that the sub- ject term referent be normally wholly or partly covered with a prototypical covering. that my car has bald tyres. However. leaving preconditions alone. Take. However. the case of the English word bald. Brought to you by | UCL . is bald. this fact is systematic. or someone's head or eyebrow. under an analysis as proposed here. that there is a bald patch in the carpet. In these cases the word bare is appropriate. whereas negation over Positive Polarity Items affects them all indiscrim- inately. this will be taken to imply that Harold is a married nan. and not the presuppositional. 183 seems to apply in the first place to satisfaction conditions. Note that the negation test has been widely used. This fact is a central element in his argument for two truth-func- tionally distinct negations and a logical system with three truth-values. or that this tree is bald. the sentence: (17) Harold is not a bachelor. Given such an analysis. A further striking feature of this analysis is that related items. and the satisfaction condition is in all three cases that the cov- ering which is normally there is not there. This does not follow from the standard componential analysis of bachelor.

differs preconditionally from put to sleep or destroy in that the latter two require the object referent to be animal but not human.184 however. Brought to you by | UCL . identical preconditions and different satisfaction conditions. Both have as one of their preconditions that the object term referent did what is mentioned in the prepositional /or-object. and the killing must be related to that importance. the words kohl and kaal. bare. and heads. cross-examine or hear (witnesses). Kill. but now the opinion is positive: it was right that the object referent did it. which have a precondition that the object term should refer to something reprehensible or (especially for admit) to the subject's disadvantage. kill and murder seem to differ in their satisfaction conditions. of course. Cft the other hand. But we can hardly claim full lexico- graphical adequacy here. in its turn. and 12 no separate word corresponding to bare exists. And so forth. are perfectly appropriate. again. Criticize and praise form a pair with. the exception that the typical covering of legs and arms (and their parts). Murder and assassinate clearly have identical satisfaction conditions. kahl. may be of the clothing kind: bare feet. Closer inspection will reveal further conditions. but for the latter three it may be of an ornamental nature. accord- ing to the threefold test for presuppositionhood given above: it is part of the satisfaction conditions of murder that the killing be unlawful and with malice aforethought. Thus bald. Thus we can take together the neutral say or assert with verbs like confess or admit. respectively. which differ precondition- ally as to the category of persons that the object term may properly refer to. Yet. that the subject referent gives his opinion on what the object referent did. 12 This analysis is. or he did it in the right way. again. for bare only there is. kaal require in addition that the typical covering should not be of the clothing kind. We see that the differences are entirely expressible in terms of preconditions. Likewise for the neutral ask (quest-Cons) as opposed to verbs like interrogate. But the satisfaction condition for criticize is that the subject referent gives as his opinion that what the object referent did should not have been done by him. For praise the satisfaction condi- tion is. bare hands·. Within one language.University College London Authenticated Download Date | 5/9/18 6:49 PM . but differ preconditionally in that the object referent must be a person of public importance. bare head. especially preconditions. while the satisfaction condition remains unaltered. apparently. words can often be grouped together under identical satis- faction conditions but different preconditions (reverse groupings seem to be much rarer). incomplete and rough. or that it was badly or wrongly done.

can profit from an integration of the distinction at hand. but the corresponding satis- faction condition of criticize. does not involve any criminal law. and perhaps also practical lexicography. But the other half of the equation is a little less straightforward: accuse carries the precondition that what has been done is criminal or at least morally bad. it does so in a much milder way than accuse. 185 In Fillmore's famous article (1971) the verbs accuse and criticize are com- pared. and FiUmore's main conclusion is that a crucial precondition and a crucial satisfaction condition of the one are satisfaction condition and pre- condition. so to speak (1971: 381). It does indeed seem to be so that the satisfaction condition of accus-e. It is not feasible to give more than a few examples here. Brought to you by | UCL . but it seems not too far removed from what is the case. as we have seen. is a precondition of criticize. respectively. and if it involves morals. that the object referent did what is men- tioned in the <?/-object. in the other: they swop places. as we have just seen (though criticize takes a /bz>-<±>ject). The point of this paper is to encourage further study along these lines.University College London Authenticated Download Date | 5/9/18 6:49 PM . This conclusion nay be a little too simplistic. and to see if theo- retical lexicology.

Davidson. 4 . V. Words and objections. Leon A. Hie Dyirbal language of North Queensland. Essays on the work of W. In: Danny D. O'Brien. Aufsätze aus dem Grenzge- biet der Sprachpsychologie und Logik. D.C. Quine. Gilles. 191O. Steinberg. Cambridge. Report of the 22nd annual round table meeting on linguistics and language studies.. James D. W.).. Carol Kiparsky. 1964. Dordrecht. Logique et Analyse 79 (vol. Bertrand. 1. Cambridge. Oxford. Seuren. M. Noam. Fauconnier. MIT Press. Avenarius. Georgetown University Press. The structure of a semantic theory. Three reasons for not deriving 'kill1 from 'cause to die1. Aspects of meaning construction in natural language. 1971.. Blackwell. Erdmann. M. Jaakko Hintikka (eds. O. 338-347. Mass. 20). Blackwell. Geach. 429-438. MIT Press. Quine. Paul M. (lanua linguarum aeries irtinor 107. 19-33. Richard J. 1972. In: Donald Davidson. Postal. Katz. Brought to you by | UCL . W.186 REFERENCES Chomsky. 1972. 1971. Mind 14. Jakobcvits (eds. Donald.). Leon A. Dixon. In: Richard J. Mass. On denoting. Mental spaces. 19O5. Cambridge University Press. In: Danny D. 1970. Russell. Geach. (Cambridge ötudies in Linguistics 9 f ) Cambridge. 1985.). Charles J. logic natters. Studies on semantics in generative grammar. Fillmore. Jerry A. O'Brien (ed. 479-493. Mass. Robert M. Robert M. Leipzig. Jerry A. Karl Otto.). 345-369.. 1969. Linguistic Inquiry 1 . 1963. Jerrold J. Pieter A. 1971.). Quine1s syntactical insights. 1971. Die Bedeutung des Wortes. Leon A. 1971. Fodor. 1969. (ed. 1960. Jakobovits (eds. Peter T. Willard Van Orman. Dixon. MIT Press.Jerrold J. Language 39. Word and object. Jakobovits (eds. Reidel. 146-157. A method of semantic description.1 The Hague. 1977.). 37O-392. McCawley. Seuren. Cambridge. Fact. Pieter A. Mouton. Forme logique et forms semantique: un argument centre M. 1985. Paul. Washington. Steinberg. 170-210. Jaakko Hintikka (eds.. Peter T. Katz. Fodor.. Geach. Kiparsky. 436-471.University College London Authenticated Download Date | 5/9/18 6:49 PM . In: Danny D. Prelexical syntax.). Steinberg. Types of lexical information. 1972. Discourse semantics. An integrated theory of linguistic descriptions. Oxford.

Verben der auditiven Wahrnehmung im Deutschen und Niederländischen. 1973. Dieter Wunderlich (eds. Ben studie van het projektie-probleem en de presuppositionele eigenschappen van de logische konnektieven. and psychology. M.. Arnim von. Frankfurt/M. Vliegen. Aufsätze und Vorträge zur Wortfeldtheorie. Jakobovits (eds. 1986. Athenäum. 187 Seuren. 195O. Presupposition. Groom Helm. Cambridge. Oskar Reichmann.. Rob A. Pieter A. linguistics. Semantics. Cambridge University Press. In: Arnim von Stechow. Peter F. Steinberg. Maurice. 1971. Dieter Wunderlich (eds.). An inter- disciplinary reader in philosophy. On referring. Nijinegen university. Handbuch der Semantik. Nijinegen University. (lanua linguarum series minor 174. The Hague. Loraaon. Brought to you by | UCL . 1987.). PhD-thesis.). 32O344. Jost. Kdntekst en presuppositie. Rob A. forthcoming. Van der Sandt. Stechow. Context and presupposition. Danny D. Mind 59. PhD-thesis. Van der Sandt. Leon A. (forthcoming).University College London Authenticated Download Date | 5/9/18 6:49 PM . Trier. NIS. Mouton.) Edited by Anthony van der Lee. Eine Beschreibung ihrer semantischen Struktur und syntaktischen Ungebung. Strawson. Nijmegen. 1982.

Brought to you by | UCL .University College London Authenticated Download Date | 5/9/18 6:49 PM .

189 SECTION 3 Brought to you by | UCL .University College London Authenticated Download Date | 5/9/18 6:49 PM .

s. which is more or less equivalent to the definition of 'structure' in the same dictionary: an organized body or combination of mutually connected and dependent parts or elements (ShOED. Kastovsky (1982: ch.v.190 DIETER KASTOVSKY. since the fact that some element is part of a structure provides this element with certain properties that result from its function and position in the structure/ system in question. they are organized into an intricate. 'structure'). 'structure1 involves the notion of analysahility. interlocking system whose underlying principles can be discovered" (Aitchison 1987: 5).University College London Authenticated Download Date | 5/9/18 6:49 PM .. that this framework does not For a more detailed description of the principles of structural seman- tics.. Brought to you by | UCL .and a micro-struc- tural organization of the vocabulary (lexicon) is postulated along the follow- ing lines. It should be pointed out. the definition of 'system' in the Shorter Oxford English Dictionary: an organized or connected group of objects.2 These assumptions. in terms of which it can be analytically described and defined. a whole composed of parts in orderly arrangement according to some scheme or plan (ShOED. As it is defined here. however. s. 'Sys- tem1 involves the notion of structure. where both a macro. STRUCTURAL SEMANTICS OR PROTOTYPE SEMANTICS? THE EVIDENCE OF WORD-IOFMATION 1. 4 ) . And these properties in turn make up the internal structure of this element. 1. ultimately going back to de Saussure.v.1 In one of the most reoent introductions to psycholinguistically oriented studies of the lexicon. cf. cf. constitute the basis of European structural semantics. it is claimed that "words are not just stacked higgledy- piggledy into our minds . 'system').

as Lipka (1985) has shown. the meaning of an individual lexical item is described in terms of some underlying dimension or dimensions and some feature or features specifying these dimensions. both. and the num- ber and kind of opposition on which they are based. Both aspects are thus inter- dependent: a lexical field is constituted by lexemes standing in immediate op- position (s) to each other. and these oppositions at the same time establish one or several semantic dimensions which characterize both the field in question as well as the internal semantic structure of the lexical items it contains. depending on their function.e. OFFSPRING in son : daughter). the vocabulary is organized on the basis of paradigmatic oppositions between the meanings of lexical items into subsystems of varying size called 'lexical fields'. i. both resulting from the oppositions constituting the lexical fields. They are further specified by semantic features which also result from the oppositions just mentioned. In addition to these paradigmatic structures and relations.g. Micrc-structurally. (1) relation lexical number field substance dimension« lexeme property ^·» feature4 function obligatory »inherent contextual·« inferential inferential itocrc— structurally. they may refer to some property (e. subdividing into obligatory and inferential features.g. The dimensions may either have a substantial or a relational character.University College London Authenticated Download Date | 5/9/18 6:49 PM . relation -. Moreover. which are based on semantic implications of the type Brought to you by | UCL . their form in turn de- pends on the substance of the dimension . 191 necessarily have psychological implications. structural seman- tics also recognizes syntagmatic structures or "lexical solidarities' (Ooseriu 1967). the psychological reality of the constructs of structural semantics is as unclear as the reality of any other metalinguistic construct. we have to distinguish between inherent and contextual vs. SEX in man : uoman) or some relation (e.

however. Now.most of the arguments in favour of prototype theory stem from psychological and psycho- Brought to you by | UCL . postulated a strict demarcation between lexical meaning and both prelinguistic concepts and extralinguistic reference. given that prelinguistic concepts and lexical meanings may overlap.University College London Authenticated Download Date | 5/9/18 6:49 PM . Such assumptions. i. Others. This new semantic theory has been embraced by many as a genuine alternative to the older 'checklist1 theory.semantic features are treated as discrete properties. 1978) and has come under severe attack because it is claimed to be deficient in the following respects: . . bark ^ dog Sbj/Ag see. since research into the nature of human categorization (Hosch/Marvis 1975. like Coseriu. And those linguists who. look ^ eye Instr fell D tree Obj kiss Ό lip Instr b. are less radical and rather opt for an 'integrated1 view allowing for both types of semantic analysis in different areas of the vocabulary. e. cf.e. confess 3 [+HUMAN] Sbj elapse o [-tTIME] Sbj 1. will undoubtedly characterize proto- type theory as a theory of reference or of conceptualization. give a false impression of semantic represen- tations in the mind.g. and not of lexi- cal meaning proper. involves a rigid yes/no decision. Geeraerts (1984. . it is claimed. but normally do not coincide except in presupposes clear category boundaries. 1985) and the respective contributions in this volume. like Lipka (1986).what might perhaps be charac- terized as a kind of renaissance of Platonic ideas. . Rosen 1977) and denotational structure (Labov 1978) is said to have shown that lexical meanings as well as denotational categories are basi- cally fuzzy.a given feature is either present or absent.all features have equal weight.3 This kind of framework has recently been dubbed 'Aristotelian1 or 'check- list theory of semantics1 by Fillmore (1975. 192 (2)a. Semantic feature theory should therefore be given up and be re- placed by a non-Aristotelian prototype theory .

BOVINE. reference-based description. i.e. and their meanings can ideally be derived exhaustively from their form..e. or more precisely. 1. in particular so-called 'natural kinds'. Post 1986).1 The most salient property of word-formations is their morphologically analysable and motivated character. at least according to Coseriu. the 'attrib- utes' or 'elements' of prototype theory in these analyses are suspiciously like the semantic features of structural semantics. But there have also been seme attempts to extend prototype theory to abstracts and to verbs. unanalysable. cf. and which would indeed seem to totally invalidate the concept of se- mantic features. i. or at least a complementary. oow. As Lipka quite rightly observes (1986: 92). In- deed. which do not really provide any semantic analysis. at least we need more evidence from areas other than 'natural kinds'. feature analy- sis is only possible with regard to the archilexematic level. 193 linguistic experiments involving simple lexical items with concrete referents. to me. since the meanings of these con- stituents represent distinct parts of the global meaning of a word-formation Brought to you by | UCL . For such lexical items. This is par- ticularly true with regard to the interaction of word-formation with the simplex part of the vocabulary and the conclusions we can draw from it for the kind of semantic structures one should postulate as representations of lexical meanings. from the meanings of their constituents and the underlying morphosemantic patterns. does indeed not exhibit any genuine structural organization. however. Here. so that prototype theory here does not seem to be more than a terminological variant of feature analysis. what features are relevant for distinguishing the meanings of horse. It is certainly no coincidence that it has always been rather difficult if not impossible to establish satisfactory semantic features in the case of such lexical items. the lowest level of organization.. they are syntagmas based on a determinant/ determinatum relationship at the most abstract and general level of interpreta- tion (Marchand 1969: 3). however. sheep.University College London Authenticated Download Date | 5/9/18 6:49 PM . Word-forma- tion is one such area which has so far not figured prominently in the discus- sion of prototype theory (but cf. Ooleman-Kay's study of the word lie (1981). and that. immediately suggests a parallelism with a checklist theory of meaning. FELINE. the species level. items that can be said to con- stitute folk taxonomies closely related to terminologies proper.e. This. etc. prototype semantics does indeed seem to offer a viable alternative.4 Ihe status of prototype theory and its variants as a genuine alternative to traditional (Aristotelian) feature analysis is thus rather problematic. seems to indicate that it would indeed be rash to abandon feature theory completely. arbitrary features such as EQUINE. OVINE. 2. for example. i. cat? Hence the use of global.

g. discern. this implies a constant shift from primary to secondary (complex) lexemes. what is borrowed are foreign syntagmas containing them. unanalysable lexeme to motivated syntagma are . too. unconnected parts of the vocabulary. clear-cut dividing line between opaque (simple) and transparent (complex) lexical items.reinterpretation of the direction of the derivation with back forma- tions. if simple and complex lexical items belonged to two completely separate. but this is regained once the base is borrowed. 2. e. which in turn suggests a parallelism of semantic structure in the items involved. First of all. receive. This.adoption of foreign word-formation patterns. —ive. . If these are borrowed before their bases. While the first two phenomena concern individual lexical items. hammook ->· G Hängematte. inert. dis-3 in-.University College London Authenticated Download Date | 5/9/18 6:49 PM . there is no neat. But a word like laudable. rather. as in the cases of (4) laudable. -ize. (3) pedlar f to peddle/0 'act as pedlar1 (back formation) to peddle > peddl/er "someone who peddles1: pedlar (monomorphemic) ->· peddl/er (bi-morphemic) . -ify. Moreover. 194 syntagma. inane which have remained monomorphemic in English. magnify. pensive. This then results in a synchronic cline of analysability/motivation rather than an all-or-none dichot- omy.folk etymology. especially in the earlier stages of the borrowing process. -able. cf. there are transitions in both directions in the history of English and presumably also in most other languages.2 Phenomena that exemplify the shift from monomorphemic. In periods of large-scale borrowing. there will also be cases where the basis of the original derivative has not been borrowed. is only con- ceivable if the semantic structures of both types of lexical item are suffi- ciently similar if not identical. the last affects larger parts of the vocabulary and even its overall structure. Affixes are not normally borrowed directly. de-. re-. they lose their syntagmatic status. Such a parallelism would not necessarily be of interest. in the same way as semantic features do in a checklist theory of meaning. -ation. as in Middle and Early Modern English. derived from the verb Brought to you by | UCL . the English affixes co-. however.a process parallel to back formation. or created by analogy with other similar pairs . But this is not the case. cf. due to its origin as the derivative laudabilis 'praiseworthy1.

i.1 The shifts from simple lexeme to syntagma just discussed are instanta- neous/ while the opposite phenomenon is a gradual process/ due to the operation of lexicalization and concomitant progressive idiomatization/demotivation. it seems justifiable to assume that the first type of development dees not involve any change of lexical meaning proper. and the meaning of ear is contained in hear and listen. lord (< OE hläfweard). only the shift from associating meaning components with the whole lexical item to association with its parts. additionally. examples such as (5) blackboard. i. atomize vb the semantic implication is accompanied by a formal- morphological implication. lady (< OE hlaefdije). 195 laudare 'to praise1/ will necessarily have a semantic structure which is par- allel to the semantic structure of a syntagma such as acceptable 'worthy to be accepted1. On the other hand. Thus. or both. the question of compositional!ty. cupboard.e. only in that with hammer vb. holiday.3. They differ from pairs like miaow : cat. 2. 2.University College London Authenticated Download Date | 5/9/18 6:49 PM . But such pairs belong to word- formation. probably the first to draw attention to this phenomenon. As a re- sult. the meaning of oat is contained in miaow.2 Similar affinities exist between lexical solidarities such as those ex- emplified in (2)/ and word-formation syntagmas. and. cf. at least on the Brought to you by | UCL . such implications obviously go beyond the domain of word-formation and characterize primary lexemes as well. where it functions as a semantic component. etc. Similar arguments apply to the other examples in (4) and to countless other similar loans.3. or the pattern meaning is obscured. forecastle. which he termed 'wesenhafte Bedeutungsbeziehungen' notes that pairs like härrmern : Hammer exhibit the same type of relationship. Moreover. dees involve some change of meaning. And it is precisely this kind of implication which is the essential property of motivation characterizing the field of word- fornation. But these semantic changes do not affect the basic semantic structure.e. Lexical solidarities are syn- tagmatic implications holding between lexemes such that the meaning of one lexeme or of a whole lexical field is contained in the meaning of some other lexeme. the formal analysability may also be impaired. Porzig (1939). because either the constituents lose their original meaning. The opposite process. synchronically we are confronted with a scale of various degrees of moti- vation/ and diachronically this implies that a word-formation syntagma may move along this scale from one extreme to the other. however. from complete motivation to complete arbitrariness.

erase : eras/er. sun : sol/ar (8) consume : consump/tion. since analysabil- ity is only possible on a foreign. and there is a limited possibility of new formations. and involves a great deal of idiosyncratic morphophonemic alternations which gave rise to SPE-type phonology. I would once more like to re- turn to the assumption that purely semantic implication on the one hand and semantic accompaniment by formal implication on the other are endpoints of a scale and not an all-or-none phenomenon. die : kill (7) father : patern/al. as we shall see presently. exemplifies the other endpoint of the scale. deceive : decep/tion (9) science : scient/ist. give : gif/t. Conversely. mouth : or/al. tooth : bite. illustrates purely semantic implication. sing . break itr : break tr. i. Neo-Latin basis. song. lexical solidarities proper. legal : legal/ise. This is corroborated by the following examples: (6) steal : thief/theft.e. it has (10) as a native parallel. (7) contains items that are in principle analysable. however. Brought to you by | UCL . on the other hand. 2. haimer sb : hammer/0 vb. but the pattern is at a completely different level from (11). (6).4 Before turning to this aspect/ however. the semantic relations existing in the primary vocabulary are very often taken up in word-formation and are made explicit by corresponding formal relations. The remaining groups fall in between these extremes. shoot : shoot/ing.University College London Authenticated Download Date | 5/9/18 6:49 PM . 196 semantic level. Type (8) is also basi- cally foreign.-caus. (11). wide : wid/th (11) rob : robb/er. fully motivated word-formation syntagmas produced by productive word-formation rules.e. i. high : heigh/t. pirate : pira/cy (1O) fly : fligh/t. look vb : look/0 sb. viz. size : big/small. thick : thick/ness. father : father/ly. it is com- pletely unproductive. but in terms of analysability it is closer to (11). mother : matern/al. even on a foreign basis.

This results from a general tendency "to see a thing identical with another already existing and at the same time different from it" (Marchand 1969: 11). because the derivation as such takes place on a Neo-Latin basis/ and the result is then borrowed into English without its formal basis existing there.1 Another important aspect. The status of (7) is unclear. finally. sailing ship or steamer. freighter. 2 Cf. antonymy. also the relationship between say /tell and unanalysable predict. which is parallel to tell and foretell.5. already briefly mentioned. will somehow have to be handled in the lexicon and thus again suggest some kind of compositionality of meaning parallel to the semantic structures associated with word-formation syntagmas. ao-author. Therefore. Thus SHIP and steamship. 197 Type ( 9 ) . But it differs fron (8) in that the suffixes are productive.University College London Authenticated Download Date | 5/9/18 6:49 PM . and consequently are indicators that feature theory in semantics might be more relevant than the adherents of prototype semantics are inclined to admit. This specification may assume quite different forms. but at the same time it is differentiated fron it. the mean- ing relations exemplified in ( 6 ) . 2. although such rules will have to be lexically gov- erned. but it largely seems to follow the sense relations pos- tulated for the simplex vocabulary by structural semanticists. Word-formation syntagmas are based on a determinant-determinatum relation. which are completely par- allel to those in (9) and (11). author. because the identified object differs from the gen- eral category by an additional specification. in contradistinction to (8). resembles type (8) in so far as it is foreign. relations such as hyponymy. which. The identification-specification pattern can manifest itself in terms of an archilexeme-hyponym relation.e. etc. only (9) and (11) are candidates for English word-forma- tion rules proper. tell. and the de- rivative can be analysed as containing an allomorph of the basis. as we have seen. Nevertheless. (8) and (10). schooner and others they belong to the lexical field SHIP. tanker. A certain phenomenon is categorized as something already known. multiple taxonomy. perhaps it is best classed with (6). are related to each other as archilexene and hyponyms. etc. i. complementarity. itorphophonemic alternations between bases and deriv- atives could and probably should be handled by appropriate morphophonemic rules. The same kind or relationship characterizes prefix formations such as rewrite. which occurs as an independent word in English. In other words. is the observation that word-formation syntagmas partake in the sense relations characterizing the simplex dictionary. (7). frigate. Together with barque. is a purely semantic phenomenon. Brought to you by | UCL . foretell with regard to their bases write.

wise : unwise. unfortunately. in analogy to to open : to close. this archilexeme-hyponym relation is typical of the relationship between the determinatum of a compound and the compound as a whole. other possibilities exist and are Brought to you by | UCL . Directional opposition. white : non-^white. smack. transformable : untransformable. a systematic enpirical study does not yet exist. as Lipka has shown (1985: 340). slap and beat do not by any means exhaust the dimension Instrument. pound. punch. Another dimension. militarize : demilitarize. Manner. "with a stick1. This is only a small selection of such parallels. etc. occurs with reversa- tive verbs of the type (14) tie : untie. (13) kind : unkind. cf. These. The opposition between kick : punch : slap constitutes the dimension of Instrument by obligatorily implying the meanings of foot. join : disjoin. HIT. the relationship between base and word-formation syntagma is one of complementarity.e. But kick. but may be cancelled by the con- text. viz. Beat. word-formations typically occur in those instances where there is no primary lexeme to represent the respective sense-relation. which characterize the internal structure of the lexical fields. despite their importance for the overall structure of the vocabulary. 2. and the semantic dimensions. knock. Elsewhere (Kastovsky 1981: 441f. 198 Obviously. i. the archilexeme. are examples of lexical solidarities. is unspeci- fied with regard to the instrument or manner involved.University College London Authenticated Download Date | 5/9/18 6:49 PM . viz. results from the opposition between bash. etc. are of the utmost importance for this parallelism and also for the over- all organization of the vocabulary. in- volves an inferential feature along this dimension.) I have illustrated this with the lexical field HIT. etc. In other cases. for which. fist and flat object. incidentally.5. beat. steward : stewardess. and it also applies to numerous prefixations and suffixations. widow : widower Antonymy is also possible. as in (12) edible : inedible.2 Obviously. this feature is often present by implicature.

The other verbs are based on the dimension 'removed object1.. Miss Pride . But there is another function. We thus get the following simplified picture. Dimension kick bash punch smack slap pound hammer beat club Another example is scrape. by means of which certain syntactic material is converted into verbs. etc. do we assume that the stone-chucker.. hammer. 199 realized by syntactic paraphrases such as hit with the knee. dumper of green lady and telephonist are one and the sane person and that this person is also the murderer of Miss Cost? . identical properties. scrape) and verbs like peel. is to find out more about the semantic properties underlying word-forma- tion syntagmas.. bark. in fact. tablespoon. Scrape does not imply a specific object. which. has to be envisaged as a scale. the examples in (16)a. etc.. the creation of new labels for sane segment of extralinguistic reality salient enough to re- quire a name instead of being described by a syntactic construction. elbow. where the possible overlap of the di- mensions Instrument and Manner is disregarded: (15) HIT (Archilexeite) INSTRUMENT MANNER .1 All the examples just discussed have demonstrated that the semantic structures underlying word-formation syntagmas and simple lexical items must have similar. debark. there- fore. to cudgel. is convinced that the ringer-up was Miss Cost. to alub. but are neutral with regard to manner. But this problem can only be approached in connection with the function of word-formation. skin. syntactic re- categorization.. cudgel. . defined as 'to remove (unwanted material) from a surface by pulling or pushing an edge firmly across it repeatedly1 (Longman's Dictionary of Contemporary English. This func- tion is typically served by many N + N compounds of the type ice bear. of labelling. And if a fixed expression is needed. wire-stretcher.v. alub. but contains a specification of the manner in which the removal is done. 3. etc. if not.. word-fanration inter- venes and produces verbs like to hammer.e. cf. again. adjectives or nouns. teaspoon.. to cosh.University College London Authenticated Download Date | 5/9/18 6:49 PM . s. composite letter-writer. eggplant. i. viz.. At one end we have the function of nomination. The crucial problem. Brought to you by | UCL .

as in (17b).. h. She remained female for life. Solarians did not bud. d. and why in the more recent lexicalist frameworks the theory of inheritance of syntactic properties such as thematic relations. may either function as re- categorization... depending on the context. Manner. also figure in word-formation. Agent. He defisted to gesture.and polymorphemic lexical items. i. and the female was always the birther. Object. as in (17a). bricklayer. so that one and the same lexical item. And he knew no one was going to take him offstage and beat him. g. 200 b. Prototype theory thus is not an absolute alternative to the analysis of the Brought to you by | UCL . f. .. This is why syntactic-semantic categories such as Instrument. and up to now Rodget had been silently determined that it was going to be a split. opera singer. carpenter is probably not radically different frcm that of song writer. the semantic structure of doctor. If that's not civil. architect. and he goes and washes them. the beating of prisoners was not authorized. no matter how many times she birthed. b. butcher. It's the pettiness. etc. civilise it and tell me. Once or twice he ahuakled. they birthed. baker. plays an important role... Thus it seems to me that the analysis of the simplex vocabulary also cannot do without them.. . that's so awful. It's blood on his hands. He wouldn't give his name and didn't mention bloody hands. His hands get covered with blood. . and whether our own conversation doesn't sound a little potty. . not visible to anybody else. And indeed. Such formations overtly incorporating syntactic relations are of course also integrated into the vocabulary and can become labels as well. either they would split up or decide to make it permanent. it has to be assumed that such syntactic properties must have their counterpart in simple lexical items. . "Don't you know a single person who ought to be murdered?" He wondered why his host should appear to set so much store by his acquaintance with potential murderees. But the latter contain elements that function in the same way as the semantic features of a 'checklist theory1 of meaning.. e. etc. a larger male not only stole his peanut but gave him a beating. At the end of it.University College London Authenticated Download Date | 5/9/18 6:49 PM . In view of what I have said about the shift between mono. c. or as label. (17) a... you know. He made fists. etc. but hardly liked to ask. It was following one of those chuckles that Paul Drake drawled a question.

as opposed to a earving knife. concepts. and meaning. But with the complex lexical items in (19). tea spoon (Tee-/Kaffeeloffel) The 'natural kind1 terms in (18) might conceivably be defined in terms of proto- typical shapes. And the same seems to hold true for much.2 Perhaps one final remark should be added concerning the relationship between prototypes. of the simplex vocabulary. Or. In this case. carving knife b.University College London Authenticated Download Date | 5/9/18 6:49 PM . 'knife intended for carving1. which will also have to be of a certain size and shape. 201 semantic aspect of lexical items in terms of semantic features. a fact which had been recognized even before Frege. pitch fork. meaning is usually not identical with a concept or a referent. prototypical concepts and meanings are definitely not the same. which will probably have a toothed blade. which clearly involve a purpose relationship: "knife intended for cutting bread1. put differently. pastry fork c. 3. this cannot be done so easily. bread knife. though perhaps not for all. Brought to you by | UCL . For this purpose. spoon (19)a. I can of course envisage a prototypical breadknife. and their meanings might well be identified with this kind of definition. But these prototypical representations do not in any way specify the meanings of these lexical items. table spoon. at least as long as it hasn't found a way to incorporate this syntactic aspect into its framework. fork. compare the fol- lowing sets of lexical items: (18) knife. kitchen knife. pocket knife.

Prototype theory and diachronic semantics. 1981. 1978. Aleksander Szwedek (eds.). Jean. Probleme der strukturellen Semantik. Fillmore. Narr. Language 57. 22O-26O. Geeraerts. Bern. Pädagogischer Verlag Schwärm. 1987. Coseriu. Labov. Karol W. 239-253. Coleman. 202 REFERENCES Aitchison. 1978. In: Horst Geckeier et al.University College London Authenticated Download Date | 5/9/18 6:49 PM . 1981. Jacobsen. In: Donka Farkas. 429-445. 1978. Historical semantics. (eds. 26-44. Berlin. Bagel. Kastovsky. 1985. Dieter. Credos. 1985. Proceedings of the first annual meeting of the Berkeley Linguistics Society. Farkas. In honour of Jacek Fisiak on the occasion of his fiftieth birthday. Chicago. Amsterdam. Aleksander Szwedek (eds. 339-354. Lexical fields and word-formation. Mouton de Gruyter. 1967. 148-173. Linda. Volume 1. On the organization of semantic information in the lexicon.). Kastovsky. Tübingen. Lipka. 1986. Denotational structure. In: Dieter Kastovsky. Prototype semantics: The English word LIE. Donka. Lexikalische Solidaritäten.). Wasley M. An alternative to checklist theories of meaning. Semantic features and prototype theory in English lexi- cography. A case study.). Eugenio. Düsseldorf. Wortbildung und Semantik. Charles J. 1-32. Berlin. Words in the mind. (eds. Jacobsen.). Brought to you by | UCL . 1973. New York. April 14-15. Geeraerts. Geckeier. ed. Berlin. 1984. Eugenio. Todrys (eds. Karol W. In: Jacek Fisiak (ed.). Poetica 1. Papers from the parasession on the lexicon. Francke. 85-94.). Jacobsen. Indogermanische Forschungen 89. 1981. Jacek (ed. Lipka. In: Donka Farkas. Volume 3.Studia linguistica in honorem Eugenio Coseriu 1921-1981. Dieter. Karol W. Coseriu. 1985. Cognitive restrictions on the structure of semantic change. by Dieter Kastovsky. de Gruyter. New York. Todrys (eds. Historical word-formation. 1986. Wesley M. Dirk. Paul Kay. 111. William. 123-131. 1982. Fillmore.. Logos semantikos. 127-153. Todrys (eds. Dieter. Charles J. Kastovsky.). Horst et al. 1975. Wesley M. Madrid. Blackwell. New York. Mouton. Chicago Linguistic Society. Mönchen. Inferential features in historical semantics. Dirk. Leonhard. 1978. In: Jacek Fisiak (ed. Chicago Linguistic Society. Linguistics across his- torical and geographical boundaries.). Leonhard. Fisiak. Amsterdam.). An introduction to the mental lexicon Oxford.

Carolyn B. Family resemblances: Studies in the internal structure of categories. Rösch. 1986. Walter. Volume 1. Eleanor. Neil (ed. Denominal adjectivalization in Polish and English. Wroclaw. Rainer Schulze (eds. Rosch. Prototype semantics or feature semantics: An alternative. Marchand. Narr.).University College London Authenticated Download Date | 5/9/18 6:49 PM . 1987. 1-49. Volume 1. 1969. In: Neil Warren (ed. The categories and types of present-day English vrord- formation. Lörscher. Minchen. Hans. Human categorization. etc. Studies in cross-cultural psychology. 573-6O5. 1977.. Cognitive Psychology 7. Wolfgang. 1987. 1934. Academic Press. 1975.).). Wydawnictwo Uhiversytetu Wroc^awskiego. literary criticism. 282-298. Perspectives on language in performance. Post. Rainer Schulze (eds.). Studies in linguistics. Leonhard. Eleanor. Brought to you by | UCL . To honour Werner Hüllen on the occasion of his sixtieth birthday. and language teaching and learning. In: Wolfgang lörscher. Beck. Mervis. 203 Lipka. Tübingen. MLcha^. 196O Porzig. Warren. 7O-97. Wesenhafte Bedeutungsbeziehungen. London. Beiträge zur Geschichte der deutschen Sprache und Literatur 58.

is termed an endocentric compound. is termed an appositional compound.. the compound is not a hyponym of the grammatical head: a redskin is not a type of skin.. Yet even Bauer's analysis leaves many questions unanswered.204 GUNTER ROHDENBURG SEMANTIC FRINGE PHENOMENA INVOLVING NCMINAL CCMPOUNDS IN ENGLISH 1. an armchair is a kind of chair. Bauer's subdivision may be summarized as follows: (1) a. a number of hitherto neglected oonpounds can be accommodated only with great difficulty. Bauer's (1983) classification of nominal compounds Traditionally. possessive. copula compounds. etc. This type .. AB^ B. AB A appositional d....). is a hyponym of some unexpressed semantic head ... is normally given the Sanskrit name of dvandva . AB«t=B. AB«=B endocentric b. AB = A + B dvandva In contrast to comparable classifications this proposal gives the impression of being a relatively comprehensive and well-defined system...c) recognizes at most two semantic relations. This type . Brought to you by | UCL . termed an exocentric compound .University College London Authenticated Download Date | 5/9/18 6:49 PM .... maidservant is a hyponym of both maid and servant· a maidservant is a type of maid and also a type of servant. the compound is a hyponym of the grammatical head: a beehive is a kind of hive. AB«PA.. if at all.. the compound is not a hyponym of either element. AB«t B exocentric = c. which distinguishes four basic semantic types: In the first two examples. . Thirdly. In the second two examples. This type . (Naumann 1986: 6O ff. compound nouns have been subdivided into general semantic cate- gories such as endocentric.. The final division of compound nouns is exemplified by Alsace-Lorraine . This is mainly due to the fact that the classi- fication in (1 a .. The point of departure for the present study is provided by Bauer's (1983: 3O-31) analysis. In particular. However. Here .. but the elements name separate entities which combine to form the entity denoted by the compound. Schematically. This type . in terms of Bauer's framework...

(2) a. In the past. the distinction may also be found in the case of endocentric compounds and their second constituents. the claim has often been made that all relations of hyponymy may be de- fined in terms of formula (3) (cf. the only relations which Bauer uses to analyze the relationship holding between the compound and its second constituent are those of hyponymy and incompatibility. incompatibility (disjunction) c. compatibility (overlap) d.for instance. This is due to an altogether different reason: the compound as a whole is gen- erally interpreted as a synonym rather than a hyponym of its second constituent.. is regularly rejected by English informants. (4) ? A paper boy is a kind/type of boy. paper boy (unlike armchair) is not a taxonym of boy. (cognitive) synonymy (identity) Apparently. (5) ? A beehive is a kind of hive. Brought to you by | UCL . but only an ordinary hyponym. one of Bauer's illustrative examples itself.1 Taxonyms and other hyponyms Vfe turn first to a problem concerning the hyponymy relation itself. sentence (5). Naturally enough. Cruse 1986: 86-87). of course.University College London Authenticated Download Date | 5/9/18 6:49 PM . In Cruse's work the distinction between taxonyms and ordinary hyponyms is il- lustrated by reference to pairs of lexemes which happen to be morphologically unrelated. hyponymy (proper inclusion) b.) has shown that formula (3) is only diagnostic of a special case of hyponymy. 2. that exccentric compounds invariably stand in a relationship of incompatibility (or are incompatible) with their respective second constituents. Lyons 1977: 292-293)i (3) An X is a kind/type of Y. However. as is evi- denced by the deviance of (4).. 2.2 Synonymy between compounds and their second constituents Furthermore. This presupposes. which is referred to as taxonymy. Cruse (1986: 137ff. For instance. 205 according to the logic of classes we can distinguish four basic relations as set out in (2) (Lyons 1977: 154ff. The examples quoted above strongly suggest that the traditional assumption is shared by Bauer. Endocentric compounds and related phenomena 2.

such as motor-cay. A relevant example is given in (8). Given this situation the analysis of boxing-glove proves to be problematical. synonymy is most closely related to hyponymy (Lyons 1968: 455-456. Rohdenburg 1985a. all of these compounds repre- sent the same general type. examples like beehive are largely similar in their distributional be- haviour to endocentric compounds. both types tend to occur in structures like (7). As is sometimes acknowledged by dictionaries. motor-car. Rohdenburg 1988). As is well known. which serves to clarify the relevant interpretation of the second constituent (Bauer 1983: 95). however. any of the four types recognized by Bauer. In the case of glove and mitten. By contrast.206 Sane additional English and German examples of this type include the follow- ing: (6) pigsty. Fahrrad. In nrj earlier papers I have shown that the competing relations may differ greatly in strength. For example. which is clearly different from. Brennessel. it was found that the general sense of glove 1 is relatively weak while the specific sense of glove 2 is very strong. Brought to you by | UCL . b. Vogelbauer Conpounds which are synonymous with their second constituents have originated in at least two different ways. They centre round a phenomenon which has received scant attention in linguistic research: two items may function as a hyponym and superordinate as well as incompatibles (Lyons 1977: 305-311. boxing-gloves belong to the cate- gory of mittens. bone marrow.3 Other readings alongside endocentric interpretations? The ideas presented in this section are somewhat speculative in nature. (7) An AB is a B. In some cases. From the synchronic point of view. Accordingly. (8) glove 1 glove 2 mitten Glove has two distinct senses. Not surprisingly. 2. statements such as the following are regarded as perfectly acceptable by native speakers. a general one (glove 1) functioning as a super- ordinate of mitten and a more specific one (glove 2) which is incompatible with mitten and also a (theoretical) hyponym of glove 1. the second constituent has been "specialized to the point that the modifier became re- dundant" (Lyons 1977: 547).University College London Authenticated Download Date | 5/9/18 6:49 PM . in cases like bone marrow (contrast- ing with vegetable mavrow) the first element is a later addition. for instance. there- fore. Lyons 1977: 292).

Thus boxing-glove. These facts are summarized in (12). respectively. we should have to classify them as appositional. gardening glove. As was suggested above. Brought to you by | UCL . In the productive type represented by the examples in (10). In terms of glove 2. kid glove. 207 (9) Boxing-gloves are not really gloves.University College London Authenticated Download Date | 5/9/18 6:49 PM . Similarly. This seems to distinguish boxing-glove from all other compounds using glove as their second constituent. the compound would have to be interpreted as exocentric.this type has an obvious parallel in examples like beehive (cf. In the present case. which is surprisingly often reflected in dictionary definitions. the only prominent sense of glove. In addition. If these compounds were also hyponyms of their first constituents. AB <= B b. beech tree. But then there is also a weak relationship of hy- ponymy between mitten and glove 1. the first and second terms involve a hyponymy relation. boy child. A <= B With regard to the synonymy relation in (12b). (10) rubber glove. (12) a. puppy dog Judging by his discussion of these examples. driving glove. On the periphery of appositional compounds 3.1 Synonymy between compounds and their first constituents This section is primarily concerned with compounds such as the following: (11) a. Clearly. AB = A c. which is unambiguous itself. cod fish b. Bauer (1983: 94-95) must regard them as endocentric. however. But this recognizes only the fact that the compounds in (11a) and (11b) are taxonyms and ordinary hyponyms. palm tree. they are mittens. the type under consider- ation must be regarded as a close relation of the appositional compound. such an analysis does not do justice to all of the relevant relations involved. the type be&hive is closely related to endocentric compounds. glove is associated with the specific sense.2). outside glove So boxing-glove gives the iitpression of being a lexicalized ccmpound. compound and first constituent are even synonymous. 2. 3. of their second constituents. would seem to be compatible with two competing analyses: there is a strongly favoured exocentric interpretation and also a weak endocentric one.

. a girl friend is both a friend and a girl .despite its enormous productivity in casual speech the type exemplified in (14) has been virtually neglected. the compound is generally regarded as being compatible with rather than hyponymous to its first constituent.. this Atkinson chap (Amis).208 In respect of this type.. the Callaghan woman (Amis).Faiss distin- guishes clearly between 'subsuraptive copula compounds' like oak tree and 'at- tributive copula compounds' like woman writer.University College London Authenticated Download Date | 5/9/18 6:49 PM .. Faiss assigns this item along with woman writer to the category of 'at- tributive compounds'. Following Marchand (1969: 41-43). By contrast. but only compatible with it. Therefore. girl-friend is not (any longer) hyponymous to the first term. (13) Geiger counter. this Bunge girl (BFBS) c. Bertrand's blonde and busty Callaghan piece (Amis) Brought to you by | UCL . 3.Bauer's treatment is obviously inferior to earlier analyses such as Faiss (1978: 41-42). girl-friend and woman writer should not be lumped together.. research has concen- trated on examples like those in (13) (cf. the Loosemore girl (Amis) d. Consider a case like girl-friend. Faiss 1978: 6O. for instance. By contrast. the analysis adopted by Faiss is found to be no less inadequate. Following Marchand (1969).. Chomsky adjunction. exanples of the second type do not involve the last two relations. We can be more precise than that: in cases like woman writer A and B stand invariably in a relationship of compatibility (or are compatible with each other)." In present-day English. The first type is defined in terms of all of the relations listed in (12). Incidentally. Here. Bauer (1983: 2O3) commits the same type of error by assigning boy-friend to the class of appositional compounds. distinctions of this kind are necessarily obscured. In other respects/ however. too. that Catchpole fellow (Amis) b. which is based on unadorned copula sentences like the one quoted above. however. (14) a. 1O5- 106 and Bauer 1983: 204). In Marchand's and Faiss' analysis.2 Compounds containing personal names as first constituents As far as compounds involving proper names are concerned. London train These are ordinary endocentric compounds. In Marchand (1969: 40) this analysis is supported by statements such as the following: ". that Sean boy (BFBS).

retained: in the type Sean boy the first constituent is coreferential with the whole expression. Roughly speaking. the English type has not been investigated. dog-dog is neither a taxonym of dog (or any other lexeme) nor has it any taxonyms itself.3 Reduplicative compounds While reduplicative compounds like dog-dog are still unusual in German (Günther 1981). the thing about shipbuilding. however. for instance: (15) "Have you sent it off to anyone else?" "Yes. they may have gained some slight degree of currency in the colloquial language used in the United States and England. and the first and second constituents involve the relationship of class-membership. it would seem inappropriate to postulate these kinds of lexical relation for proper names. Ihe parallel is. Admittedly. the first and second constitu- ents are identical. however.. As was shown above. (Amis) Despite their close similarities to strict apposition.. In addition. the first constituent of boy child is synonymous with the compound as a whole and also hyponymous to the second constituent (cf. that Caton chap who advertised in the T. 209 Like proper names. (1972: 638). Obviously. examples like these occur only in referring expressions. So far. In the first case. the term expresses a positive evaluation Brought to you by | UCL . In examples like Sean boy the compound may also be regarded as an ordinary hyponym rather than a taxonym of the second constituent. dog-dog is used to designate a real or prototypical dog. the examples under con- sideration do not belong to Bauer's category of appositional compounds. By contrast." (Amis) (16) That chap Caton's taken my article.. Ch the second interpretation. ( 1 2 ) ) . the distinction drawn by Bauer between endccentric and appositional compounds has become completely pointless. the English informants who are at all familiar with such compounds tend to give them a fairly uniform treatment. Interestingly. Since.L. an example like dog-dog is a hyponym of both its first and second constituents. Cf. There appear to be two general interpretations. In an experiment conducted with German informants such reduplicative compounds were associated with very different interpretations (Günther 1981: 27O-271). which will not be further considered in this paper.University College London Authenticated Download Date | 5/9/18 6:49 PM . 3. they are largely replaceable by coreferential exanples of strict ap- position in the sense of Quirk et al. reduplicative compounds are synonymous with their corresponding simple lexemes: they are used to clarify what is re- garded as the fundamental meaning associated with a potentially ambiguous item. a couple of months ago . In fact. of the types mentioned they appear to be most similar to that represented by boy child in (11b).S.

the examples in (18): (18) a.b). c. abstract nouns seem to be totally inadmissible. Definite determiners as in (18d) are excluded. Oarrespondingly. Evidently. Brought to you by | UCL . * It's a kind of [K*«»v3 dog-dog.are subject to severe syntactic and collocational restrictions. definite. That's a real (genuine. an example like oat-oat does not normally make a lot of sense. Moreover. Examples like those in (18a) might have been expected to be pleonastic.. such compounds have a close affinity with contexts which explicitly establish a referent as a member of a particular category. In other cases like. * That dog-dog you mention . MDreover. oar-oar the role played by size may be rela- tively unimportant. For example: (19) a. That's quite a dog-dog. 3. Instead we find an intensification of the pro- totypical notion. In addition. prototypes and marginal mem- bers of the category involved must be easily distinguishable and a certain amount of individual variation is also required. d. old) dog-dog. any at- tributive adjectives must belong to (a subset of) the class of emphasizers in the sense of Quirk et al. respectively. Consider in this respect the following exartple given by an informant: (17) It's a dog-dog. the criteria assigning a referent to the relevant core of the category involved are largely determined by pragmatic factors. They concern the items girl-friend and oak tree.1 seem to warrant two conclusions which are relevant to this section. proper) dog-dog. brown.210 of the referent involved. I'm looking for a real dog-dog. For a reduplicative compound to be fully acceptable. b. as is suggested by (18 a . Something similar occurs in (18e). it was found that reduplicative compounds . Accordingly.quite apart from any phonological constraints . Cf. * It's something like a dog-dog. (1972: 260).University College London Authenticated Download Date | 5/9/18 6:49 PM .. In fact. for instance. In this particular case the most important criterion seems to be that of size. e. expres- sions which weaken the validity of a given category assignment are rejected as deviant. b. * That's a nice (big. not just a poodle.4 Categorial indeterminacy involving the first constituent The data presented in 3. and the indefinite article in (18c) seems to have only an unspecific reading.

So someone using t^ingr-compounds involving this kind of interpretation ex- plicitly reserves the right to be wrong in their categorial assignment.-friend are generally regarded as being only conpatible with their first constituents. In this respect. In examples like box thing. however. has sort of flipper things instead of proper feet. the number of categorial alternatives is left indeterminate. (Conversation) (23) It looks like a bowl with a kind of stixrer thing. It might be thought. 45) holds the view that they may be defined in terms of (12a) and (12b) alone. the speaker is assigning the object involved to the periphery rather than the centre of the category box. This seems to suggest that compounds of this kind cannot be formed spontaneously by means of any productive rules. In terms of Bauer's classification we should have to treat examples like box thing as endocentric.. such compounds containing thing as their second constituent are only appropriate to the extent that the relevant categories themselves involve Brought to you by | UCL . the compound as a whole is only compatible with its first constituent. In using box thing rather than box in (24). are disproved by a type of compound which ... (Mis) (22) Depot have a kind of fridge thing where . (Conversation) (21) A turtle . This assumes that (12c) is automatically associated with (12a) and (12b). Both assumptions. to a big box thing.University College London Authenticated Download Date | 5/9/18 6:49 PM . The referent may be a box but it might just as well belong to some other cate- gory. Well it looked like a box. Yet surely this would not do justice to their transitional status between endocentric compounds like armchair and the type oak tree. that these items were similar in their general semantics to that of the type oak tree.. it's a sort of small leopard thing. there is an important difference between the type oak tree or boy child and the compounds exemplified in (20) .as in (24). (Sinclair 1987) (24) He limped away . Some relevant examples include the following: (20) A: Was ist ein Ocelot? B: Well. therefore. 211 Girl-friend and also bay. How- ever.has so far been overlooked. Besides.. Concerning so-called 'sub- sumptive cotpounds1 like oak tree/Faiss (1978: 42. (Sinclair 1987) The compound in these examples is a hyponym (though not a taxonym) of the sec- ond constituent thing.. the types dog-dog and box thing are almost diametrically opposed.despite its enormous productivity in casual speech .(24). It is true that we find comparable compounds which have to be analyzed in this way. and in all likelihood they constitute lexicali- zations. Apparently.

. thing is primarily (though not exclusively) asso- ciated with concrete nouns.? a dog thing Par obvious reasons.? a table thing b. Der Mansch ist ein Säugetier. However/ in cases like (2O) . a bush thing . Ein Säugetier ist ein Tier.(23). for slicing the cobs and a paintbrush affair for spreading the melted butter on them were passed from hand to hand. (29) a. Der Mensch ist ein Tier. Let me begin by considering a German example which first drew my attention to the phenomenon.. b. For example: (26) A razor-blade . Other general nouns which may show essentially the same behaviour as thing include affair. also (19a)).so far - also gone unrecognized. As is to be expected. such compounds have no hyponyms. In compounds of this type.? a car thing c.212 a reasonable amount of indeterminacy. Rather than producing pleonasm in these com- binations. The area between endocentric and exocentric compounds Compounds which are compatible with their second constituents have .? a tree thing d. stuff and effort. Consider in this respect the following acceptability ratings supplied by one informant: (25) a. a desk thing . Yet they frequently occur in conjunction with a sort of / a kind of. nor can they be taxony- mous to any lexemes. examples like (20) - (23) bear a striking resemblance to those in (18a).University College London Authenticated Download Date | 5/9/18 6:49 PM . Brought to you by | UCL . In this respect. (28) Where did you get hold of that jacket effort? 4. where it is only the addition of a kind of which provides the necessary degree of categorial fuzziness. a sort of /a kind of serve to reinforce the category-blurring func- tion of the relevant t/it?^-compounds. a van thing . c. It is presented here in the form of a syllogism. emphasizers such as those in (18a) do not occur as modifiers of examples such as box thing. a kind of dog thing . (Amis) (27) We were given some kind of stew stuff. these phrases have a similar function to thing-cctnpaanas themselves (Cruse 1986: 138): they are explicitly used to provide only a provisional or approximate categorization of the refer- ents involved (cf. This is particularly noticeable in the case of (25d).

bellboy. Säugetier has more in cannon with endocentric com- pounds than with exocentric ones. which is also compatible with (though not hyponymous to) its second constituent. This similarity is further reflected in con- structions of type (30). 213 It is generally felt that (29c) is patently absurd in ordinary non-technical German. (34) foothill. However. In this respect. which is simply misleading. Säugetier is to be classed as a para-hyponym of Tier. football boot These items function also as para-hyponyms of their respective second constit- uents. (30) X's and other Y's. this distinguishes Säugetier fron Geburtstagskind. Comparing examples like box thing and Säugetier or Geburtstagskind we find that they involve very different kinds of compatibility. (31) dogs and other pets This special case of compatibility has been referred to as para-hyponymy (Cruse 1986: 99). As is pointed out in Cruse (1986: 99) this kind of phrase is not necessarily diagnostic of the hyponymy relation. where this would be impossible. C32) Säugetiere und andere Tiere For many people. (33) ? Geburtstagskinder und andere Kinder Some possible English examples of this transitional category include those in (34). This being so. Thus Säugetier has to be assigned to the no-man's land between endo- centric and exooentric compounds. Wiile in the case of box Brought to you by | UCL . As is revealed by the normality of (32). Säugetier and Tier only involve a relationship of compatibility. In sore cases like (31). Geburtstags- kind would seem to be an exception.University College London Authenticated Download Date | 5/9/18 6:49 PM . In (29b) the distinction between hyponymy and compatibility happens to be obscured by the predicate involved: unlike (29b). as is suggested by the potential confusion surrounding (29b). Unlike morphologically unrelated items this may be typical of compatibles involving compounds and their second constituents. most students I have questioned do not realize imne- diately or not at all that the erroneous conclusion is to be blamed on (29b). (29a) permits the addition of notwendigerweise 'necessar- ily' . the lexemes oc- curring in such constructions involve only the compatibility relation. While Mensoh certainly is a hyponym of Säugetier in the colloquial language.

which is known to both speaker and hearer. Furthermore.he or she must be an adult.2 and parts of least in part - been analyzed before.4. 3. a particular Säugetier may well be a prototypical specimen of the class of animals. the type under discussion seems to involve only one categorial alternative on the same hier- archical level. 2. In addition.4). So if a particular person referred to as a Geburtstagskind is not a child.University College London Authenticated Download Date | 5/9/18 6:49 PM . Apparently the statement in (35) was gen- erally held to be true only 3O years ago: (35) A football boot is necessarily a kind of boot. While the examples associated with the first group have . 5. 131) af- fecting the compound though not its second constituent. 3.3).2. and a bellboy may be a boy or a man (Sinclair 1987). In some cases at least the change of category must have been brought about by a referential change in the sense of Lipka (1981: 124.3) and the distinction between Ordinary hyponyms1 and 'taxonyms1 (cf. It might be supposed that compounds of this type cannot arise according to specifiable productive rules of word-formation. a foothill is either a hill or a mountain.1). However. Brought to you by | UCL . 127. 3. In view of the large number of compounds using boy as their second constituent we cannot rule out the possibuty that further examples of this type will be created by analogy with the existing ones.3 and 3. 2. In the case of football boot this development is very recent.214 thing the referent is characterized as a fringe phenomenon in relation to the category of box. 2) They are (only) compatible with their first or second constituents (cf. there are two general aspects which were found to be relevant to the present study: the phenomenon of 'autohyponymy' (cf.2. The majority of these exam- ples may be divided into two groups: 1) They are synonymous or (»referential with their first or second con- stituents (cf. this assumption may well turn out to be inaccurate. and parts of 3. those belonging to the second group have so far been completely ignored. 3. unlike the type box thing. In the meantime the typical football boot has changed into an object that would more properly be referred to as a shoe.1. The examples discussed may be assumed to have originated as endocentric com- pounds. 2. Similarly. 4.1 and parts of 3. Summary A number of types of compound have been presented which defy analysis in terms of the general scheme put forward by Bauer (1983).

In: Geer A. Dogs.University College London Authenticated Download Date | 5/9/18 6:49 PM . Introduction to theoretical linguistics. M. J. John (ed. LLpka. 1978. John. 63-71. Marchand. Beck. Rohdenburg. Leonhard. Naumann. Seuren. Weijters (eds. etc. Randolph et al. Dietrich Nehls (eds. Brought to you by | UCL . Tübin- gen. Hartmut Günther (eds. 1981. Wortbildung. 1981. 215 REFERENCES Bauer. 1986.). Geer A. M.). Problems of hierarchical organization and lexical specificity involving compounds and their equivalents in English and German. Cambridge Uni- versity Press. Cambridge. Günther.). D. Cambridge University Press.. Anton J. 1988. 258-28O. 1972. Tübingen. Lyons. Dordrecht. etc. Cambridge.). Journal of Seman- tics 4. 196O. Günter.etc. Cambridge. Lipka. John. English word-formation. English Language Dictionary. M.). Klaus. Collins CQBUILD. 1969. Einführung in die Wortbildungslehre des Deutschen. Longman. Günter. Günter. bitches and other creatures. Anton J. PLeter A. 1977. Groos. lexical semantics. 1985. 1988. M. Hartmut Günther (eds. 1981.). Weijters (eds. Semantics. Hoppenbrouwers.) Sinclair. Rohdenburg. 1985b.. Cambridge. Pieter A. Klegraf. Darmstadt. Narr. 1983. 1968. Cambridge University Press. 119-132. 1972. Alan. Collins..Ein Beitrag zur Theorie und Praxis der Wortbildung.). Cruse. Meaning and the lexicon. Glasgow. The categories and types of present-day English word- formation. 1985a. J. 117-135. M. Dietrich Nehls (eds. Niemeyer.. Hoppenbrouwers. Quirk. Rohdenburg. A synchronic^^iLachronic approach. 1987. In: Josef Klegraf. Poris. 2 vols. Verdunkelte Compounds im Englischen. Bernd. Cambridge University Press. London. Faiss. Unmarked and marked terms in English. Hans. Seuren. 1986. London. Hartmut. Essays on the English language and applied linguistics on the occasion of Gerhard Nickel's 6Oth birthday. etc. Laurie. Lyons. Josef. M.. Hartmut Günther (eds. Wissen- schaftliche Buchgesellschaft. Leonhard. Zur Lexika i sierung im Deutschen und Englischen. In: Leonhard Lipka. A granmar of contemporary English. Heidelberg. In: Leonhard Lipka. München. N+N: Untersuchungen zur Produktivität eines deutschen Wortbildungstyps.

by means of their referential function. denote extra-linguistic things and events. it is the intended purpose attributed to a certain piece of furniture by the 1 Cf. and extra-lin- guistic knowledge resulting directly fron the characteristic features of the objects or events referred to by the lexical units. Obviously Pottier believed himself able to establish the differential features of meaning by taking into account the objective features observable in the respective denoted classes of objects themselves. Since lexical units are considered to be represented by those linguistic forms which. any semantic analysis of these units runs the risk of con- founding linguistic facts in the sense of functional meanings. if we have to ana- lyse linguistic units referring to concrete objects. also Gipper (1978: 42-44). is symptomatic for the phe- nomenon in question. so that a pretended lin- guistic description of such lexical units often turns out to consist of a de- tailed objective description of the extra-linguistic objects themselves.University College London Authenticated Download Date | 5/9/18 6:49 PM . Consequent- ly.216 BRUNO STAUB EXTRA-LINGUISTIC KNOWLEDGE AND SEMANTIC ANALYSIS 1. Gipper (1959) had already examined the semantic relationship between the German lexemes Sessel 'arm-chair' and Stuhl 'chair 1 and pointed out that the classification of a concrete piece of furni- ture under one of the toro classes of referents depends on other factors than solely its being equipped with arm-rests and/or padding. Brought to you by | UCL . It is obvious that we espe- cially run the risk of confounding the two sorts of features. According to Gipper. 1. he considered the featues 'with arm-rest' and 'with padding1 to represent the differential semantic features of the two lexical units chaise 'chair1 and fauteuiI 'arm-chair'.1 The structural semantic analysis of the lexical field of the words for 1 'seat in French. proposed by Pottier (1963: 11-18). Four years before Pottier's analysis.

and on the other hand that they can only be analysed differentially. If we oppose these generic suffixes to other nominal suffixes like -merit in changement "change" and -£έ in fierta 'pride'. According to him. 'instrumental nouns'.University College London Authenticated Download Date | 5/9/18 6:49 PM . generic compounds belong to the formal type of suffixal derivatives with the suffixes -eux. -oir.2 Regarding the methodology of structural semantics. 2. Cf. in the terminology proposed by Coseriu. that is fay opposing them one to another. -ier. 1981) and Coseriu/Geckeler (1981: 61-63) Brought to you by | UCL . can be called 4 "generic composition'. 1973. Cf. iMs theoretical conception and the methodological postulates that de- rive from it are still far from being generally accepted in the semantic analy- sis of lexical units. the distinction made by Coseriu (1968) between 'primary 1 and 'sec- ondary paradigmatic structures' which is equivalent to the distinction between 'primary lexemes' and 'derived words' (including composites). though being supplied with padding as well as with arm-rests. Ihis criticism is even more valid if we leave the domain of the so-called primary lexical units and take into account one of the sectors of word formation which. it is highly in- structive to consider Gipper's approach to borderline cases. generic composition represents a special type of word forma- tion comprising those secondary lexical units which in more traditional termi- nology are called "agent nouns'. 1977. In Coseriu's view. and 'place nouns'. is nevertheless classified as a Stuhl because of its not being intended for rest and recreation. which in Goseriu's terminology belong to the word-formation The differentiation between 'signification' and "designation1 in the sense of two fundamental levels of linguistic meaning is made here in accordance with the respective distinction postulated by Coseriu (Coseriu/Geckeler 1981: 54-55). 217 respective view of the speaker rather than the objectively observable charac- teristics of the given concrete objects which has consolidated in the lexical meaning. a dentist's chair. Coseriu (1968. and -iste. 1. to mention only the most prominent French generic suffixes. With regard to their material aspect. As a result we can conclude on the one hand that linguistic meanings represent generalizations of designative relations.

Brought to you by | UCL . do not possess this transpositional function. As opposed to primary lexemes. way.similar to primary lexemes . which generic suffixes exhibit in a constant. This procedure not only postulates Cf. Gauger (1969). who examines in a critical way determinative rela- tions in suffixal derivatives in general. which enables them to refer to extra-linguistic objects. the lexi- cal content of generic suffixes is more general. In accordance with their prc—lexematic content they have. it is precisely this global or generic mean- ing which represents the main difficulty. an inherent cate- gorical function . and therefore they can only refer to extra-linguistic objects in a global way. Be- cause of this faculty. Therefore the subsumption of the two categories of suffixal derivatives under forehand's (1969) type of derivation is only possible from a mere material point of view. 2.218 type of 'development'.1 Cn the one hand. in the practice of semantic analysis.2 The characteristics of generic suffixes can therefore be seen in the fact that they possess a pro-lexematic content with an inherent categorical function 'noun'. we gain two differential semantic features. Generic suf- fixes. However. since the content of the lexical units is most often identified with the referential class of the extra-linguistic objects designated by the generic compounds. the suffixes of development possess a function of categorial transposition so that any derivation by means of these suffixes implies a change of the lexical category of the basis.which is restricted to the category "noun1. With regard to content. that is to say functional. however. however. the two categories of suffixes mentioned do not coincide as to their word-formative function. the only difference between nominal and generic compounds can be seen in the more global reference effected by generic determinata.University College London Authenticated Download Date | 5/9/18 6:49 PM . 2. On the other hand. Thus. on the contrary. which make it obvious that suffixations with generic elements represent a special type of word-formation. derivatives with these suffixes represent combinations of two lexical contents in a determinant-determinatum relation. generic suffixes show a content which corresponds to the content of primary lexemes in so far as it enables these suffixes to refer to extra-linguistic objects. This inherent categorical function 'noun1 is responsible for the fact that derivatives by means of generic suffixes are nouns regardless of the category of the determinant.

We can conclude from this observation that the generic suffixes in French form two parallel paradigms so that the respective members oppose each other only in the frame of their paradigm. Apart from the theoretical and methodological problems of this procedure we can observe further negative consequences concerning the influence of extra- linguistic knowledge on the semantic analysis of the two categories of generic compounds. This affinity is due to the fact that the basis of the formation is a verb. by the designation of a substance (person or thing). 219 or tolerates an extensive synonymy or polymorphy of the respective suffixes but also neglects the fact that at least in the Romance languages generic com- pounds are strictly separated into the groups of deverbal and denominal de- rivatives.). including the suffixes used in than. Generic compounds with verbal determinants superficially show a pro- nounced affinity to nominal developments. we cannot neglect the fact that the suffix in the sense of the determinatum is in one case determined by the designation of an action.University College London Authenticated Download Date | 5/9/18 6:49 PM . the substance designated by the suffix of a deverbal generic com- pound is represented as taking part in an action. 2. and here espe- cially under the aspect of topicalization stressed by Marchand (1969: 32ff. they can only be analysed differentially within their paradigmatic relation. two important facts: Firstly. Consequently. kept strictly apart. so that a formation like ahanteicc 'singer* is considered as a topicalization of the subject because of the iden- Brought to you by | UCL . Secondly. which is nominalized by combining it with a suffix. precisely in French. the determinatum of a suffixal derivative is regard- ed as representing a syntactic function. are not described in a differential and analytic way. This neglect implies that the respective formations. deverbal generic compounds seem to be particularly suit- able for an analysis of word-formations under syntactic aspects. According to this concept. but only referential- ly. Denominal generic compounds. However. the respective suffixes used in the different derivational relations are. Furthermore. Therefore. however.3 The subsumption of these two groups under classes of designation such as "agent nouns'. and 'place nouns' largely disregards. 'instrumental nouns'. in my opinion. in the other case. nearly all current works on French word-formation neglect precisely this semantic difference between deverbal and dencminal generic compounds. 3. do not imply any participation in an action so that the ele- ments of the compounds imply a mere relation between two substances. on the contrary.

this attempt represents a kind of link between the syntactic approach under semantic aspects and the traditional onomasiological and semasiological approach to word-formation. If this conclusion is generalized. Therefore it is considered to represent the instrument of the action. Thus a deverbal generic compound like arrosoir 'watering-can' is said to refer to an extra-linguistic object in connection with the verbal basis arroser 'to water1. in this respect the different underlying syntactic structures of and ar>iOSoip proposed by Dubois (1969: 145).220 tification of the suffix -ewe with the subject of the underlying sentence -LI ohante. 'locative1. which obviously does not coincide with those formations in -air that designate the place of the verbal action. The weakness of this referentially orientated approach appears most clearly if the classification of the extra-linguistic things into pretended semantic cate- gories is less evident. resulting in the postulation of either a polysemic suffix or even two or more hcrnonymous suffixes. so that the suffix desig- nating this thing is said to express the syntactic function of the instrumental ccmplement in an underlying sentence.1 My objection to this approach is based on the fact that what is de- scribed here is not the linguistic content of a deverbal generic compound. In the first version of Fillmore's case grammar (1968: 24f. all formations containing the suffix -oir will appear to be topicalizations of the syntactic function "instrumental complement'. What is characteristic of this approach is the direct influence of world knowl- edge on semantic categorization. 'instrumental1. This leads to the consequence that the variation in the referential function of the suffix is integrated in the semantic analysis. 3. The knowledge about the extra-linguistic thing leads to the conclusion that this thing cannot be regarded as the agent of the action expressed by the verb. so that the intrinsic semantic content of the suffixes in the sense of a generalization of their designation properties is ignored.University College London Authenticated Download Date | 5/9/18 6:49 PM . etc. these syntactic functions are inter- preted semantically as 'agentive'. Brought to you by | UCL .). Regarding the analysis of suffixal derivatives. In the sane way arrosoir 'watering-can1 from a&rosev 'to water' is said to be the topicalization of the adverbial complement. but an extra-linguistic object designated by the formation. Cf. In the present case this occurs namely with those suf- fixal derivatives in -oir and -oive that designate a vessel or a container.

drinking trough1 can bei interpreted either as places or as instruments. these con- stant semantic values in the sense of functional meanings must be found some- where else. which by the way is necessary for a differen- tial analysis.University College London Authenticated Download Date | 5/9/18 6:49 PM . since especially the feminine generic compounds in -euse or -trice show a marked tendency towards the designation of machines such as perforatrice "rock-drill (for tunneling)1. not even the differentiation between agentive and instru- mental functions is matched by corresponding linguistic distinctions in French. ohargeuse 'conveyor belt1. 221 The classification of these things as 'places' or 'instruments' is left en- tirely to the arbitrary decision of the analyst. This is why we have to accept the fact that generic compounds in -eur can nei- ther be regarded as agentive nor as instrumental formations.3 We thus have to postulate a new approach in semantic analysis independent of existing "results' but starting nevertheless from designations. 3. 3. If we add to these examples those formations in -eur that designate idiosyncratically either a person or a corresponding machine wa realize that this suffix is indifferent to the classematical distinction between person and thing and that the corresponding formations are generally not predictable in this respect. What is meant by these observations is that the identification of real world categories such as "acting person" and "instrument" with linguistic meanings must be regarded as misleading. so that their func- tional meaning must be looked for elsewhere. It is true that formations in -oir/e cannot be regarded as agentives because of their inherent classematical feature 'inanimate'. kettle1 baignoire 'bath-tub'. and abreuvoir 'horse-pond. in- strument. If agent. if only be- cause two suffixes cannot correspond to three semantic categories. If we pursue this idea further and include the competitive suffixes. crib'. Thus the extra-linguistic things designated by such generic compounds as bouilloire 'boiler. and calculatriae 'cal- culator1 . we come to the conclusion that the formations in -oir/e do not have a functional meaning 'instrument1 or "place" either. not even the oppo- site suffix -eur in French possesses a constant functional feature 'animate1 or 'person'. In a treatise on generic compounds in French and Spanish I made this attempt Cf.2 What is more. mangeoire 'manger . for further details Staib (1988). However. and place do not constitute the constant semantic values. watering-place. Brought to you by | UCL .

generic compounds of direct participation are classematically indifferent. towel rack'. I termed this semantic differentiation 'direct' vs.fondoir 'melting house (for tallow)'. While the classenatic feature 'thing' or 'inanimate' is inherent in generic compounds pertaining to indirect participation in the action. Their function- al opposites are represented by the generic compounds in -oir. we can observe that the suffix -eia· always designates persons and things which participate directly in the action by performing it themselves.s'eahoir 'drying room.4 Taking into consideration this semantic opposition. the distinction between instrument and place does not represent a functional distinction either.plantoir 'dibble'. the question raised by Kastovsky (1982: 190f. or supposed. conveying. On the other hand. However.University College London Authenticated Download Date | 5/9/18 6:49 PM . fondeuse 'melting pot1 .) concerning the linguistic level of the instrumental and the personal-agentive meaning of dish-washer can be answered in the case of -eur by affirming that we are here confronted with a differentiation on the level of designation. which can only be Brought to you by | UCL . the content feature 'agentive' is only one of a bundle of three functional features pertaining to -ewe. etc.arrosoir "watering-can1. be- cause in French there exists no corresponding distinction on the level of ex- pression. Depending on the designated extra-linguistic en- tity. soudeuse 'welding unit1 . or is able to do so.222 and cane to the following conclusions concerning the suffixes mentioned above: Hie functional difference between formations in -ewe and those in -oil· has pri- marily nothing to do with the distinction between acting person and instrument for the performance of the action. which designate entities whose involvement in the action is only indirect in so far as they merely enable the performance of the action by another entity. The distinction between direct and indirect participation in the action is particularly evident if we take into consideration word-formation doublets containing the opposite suf- fixes and both designating inanimate things. "indirect participation in the action1 in the sense of functional content features. planteuse 'planter' (in the sense of 'machine for planting1) . seaheuse '(electric) clothes-dryer1 (machine) . As examples from French we can mention: arroseur 'sprinkler1 . Instead the generic suffixes -eur and -oiv exhibit two opposite ways in which the entity designated by these suffixes takes part in the action des- ignated by the verbal basis. the distinctive meaning that the entity performs the action on its own or is either meant. 3. If we analyse the existing formations of French from this point of view.sou- doiv 'soldering-iron1. they can function as names for persons as well as for things. however.

The detailed analysis of these manifold oppositions yields a highly differentiated structure that I g can only sketch here. used mostly in the plural. im- plying a virtual (potential) agentive participation in the action. which is in perfect correspondence with their going back to present participles. This functional domain is differentiated further in dependence on the suffixes functioning here. -eau in tratneau 'sleigh1 frcm tratner 'to trail. does not imply any intention to perform the action. What is highly interesting in this respect is the fact that French does not possess special generic suffixes for the structuring of this domain. and forma- tions in -ant with the implication of an actual agentive participation. and the direct object in the sense of a patient. 8 For more details. which from a syntactic point of view can be re- garded as tcpicalizations of the direct object. Staib (1988. represented in French mainly by the suffix -oir/e. -ette in sucette 'dumny. chapter I I I . Thus the generic compound voyeur des- ignates a person who wants to see something. the subject in the sense of an agent. whose central function is not a generic but a modifying one. which means that he is character- ized rather by his intending to do so than by his performing the action con- stantly or perpetually.University College London Authenticated Download Date | 5/9/18 6:49 PM . 223 identified if we include in the analysis the other formal means existing for the designation of the participant in the action. l ) . where the substantivation of the past participle represents the most important procedure to fill the gap existing in the domain of generic suffixes. Instead we find several suffixes like -on in noiartsson 'baby1 from nourrir 'to nourish' . On the contrary. Brought to you by | UCL . is opposed by the meaning of direct par- ticipation in the action. which represent a functional opposition. haul1. the formation les voyants. This dif- ference corresponds to the semantic opposition between formations in -ewe. direct participation in the action corresponds to the two main verbal actants. The functional meaning of indirect participation in the action. c f . First of all. Since the designations of the respective functions correspond to different procedures we have to distinguish between agentive and patientive formations. This means that the corresponding generic compounds can only be idiosyncratically identified as such. The same applies for the domain of actual patientive participation in the action. The domain of agentive formations is mainly divided into formations in -eur and in -ant. lollipop1 from sucer 'to suck1. This functional opposition between a virtual and an actual direct agentive par- ticipation in the action is matched by an analogous opposition in the domain of patientive generic compounds.

the formations including a functional classeme 'person1 can be regarded as ncminalizations of the subject including an implicit verb. in which these semantic features occur only occasionally. According to Laca. Motsch (1970) postulates the implication of general verb classes under the condition that the relation betwsen the constituents of the compound depends on a positively specified verb. Spanish derivatives in —ista can only designate persons because of their functional classematic con- tent.University College London Authenticated Download Date | 5/9/18 6:49 PM . Consequently. we are immediately confronted with the problem of the implicit verb raised by Lses (1963) and Marchand (1965). In a transformational-genera- tive approach. and 1987: 166). 4. and suffixal derivatives without such a classematic content. The latest work on the so-called 'subject nominalization' done by Laca (1986 and 1987) generally assumes an implicit verb in denominal generic compounds as well. In both cases. and other formations such as those in -era. Brought to you by | UCL . Ill. analogously to French formations in —ier.1 While deverbal generic compounds allow a syntactic interpretation of their constituents because of their derivation fron a verbal basis. denominal formations do not lend themselves to such an approach. it is neither necessary nor possible to postulate an implicit 9 An exception to this tendency is represented by Polenz (1980). 10 Cf. Laca wants to distinguish between such formations whose suffix contains the classematic feature 'person1. with the constant feature 'person' together with the invariant 'agent1.2 In this approach we cannot overlook the fact that Laca's differentia- tions are based on elements of idiosyncratic designation. while 9 denominal formations have scarcely been analysed from this point of view. If we want to describe dencminal generic compounds under the aspect of top- icalization. therefore. However. which thus influence directly the semantic description of the formations. ch. while derivatives in -era. show a classematic indifference so that they can refer to persons and things. while the other formations show a different word-formation meaning. In fact most demonstra- tions of the topicalization approach are based on deverbal formations. especially Laca (1986.224 4. she differentiates between 'real1 denominal agent nouns such as Spanish derivations in -is to.

This implies that linguistic content as a constant and functional entity pre- cedes referential designation. 225 verb on the functional level of linguistic content. if dencminal generic com- pounds do not imply any verbal meaning. the postulation of a strict separation betareen linguistic facts and extra-linguistic knowledge of the things of the external world has to be maintained. Brought to you by | UCL . since the use of words and forms concerning their reference to extra-linguistic reality is able to provide most useful and decisive indications as to their linguistic content. In fact. since this verb could only emanate from our extra-linguistic knowledge. which must be regarded as a variable. This does not imply. 5.University College London Authenticated Download Date | 5/9/18 6:49 PM . what is meant by linguistic content is nothing other than speakers' intuitive knowledge of the meaning of words and forms. However. their categorization as agent nouns is inadequate and tends to efface the fundamental distinction between deverbal and dencminal formations. that designative relations can be neglected in semantic analysis. With reference to the theory of semantic analysis. however.

In: Peter Cole. Juli 1976. Sadcck (eds. Theodor Elwert (ed. Charles J. 297. 1980. Paris.3. Leonhard Lipka (eds. New York. Geburtstags von Hans Marchand am 1. Coseriu. 1968. Nieneyer. Beiträge zum Wuppertaler Wortbildungskolloquium von 9. Coseriu.). Perspektiven der Wort- bildungsforschung. Hans-Martin Gauger. Autorisierte und be- arbeitete Nachschrift besorgt von Dieter Kastovsky. Karl E. Inhaltliche Wortbildungslehre (am Beispiel des Typs 'coupe-papier'). La phrase et les trans- formations. Festschrift für Hans Marchand. 1968. Herbert Ernst. Les precedes semantiques dans la formation des mots. Brekle. Heidolph (eds. Universale in linguistic theory. Eugenio. Sadock (eds. Determinatum und Determinans im abgeleiteten Wort? In: Herbert Ernst Brekle. Hans-Martin. New York. W. Eugenio. Holt. Bierwisch. Brekle. Wolf. Probleme der Semantik.).. Dieter Kastovsky (eds.). 1968. Harms (eds. Horst Geckeier (eds. Lexikon der Germanistischen Linguistik. Wortbildung. Progress in linguistics. In: W. Jerrold M. A collection of papers. 30. Enron.). Vorlesung gehalten im Wintersemester 1965/66 an der Universität Tübingen. 1981. Granmatik und Wortbildung romanischer Sprachen. Herbert Ernst Wiegand (eds. The Hague. 1987.). Grammaire structurale du francais.) Tübingen. Peter. 59-81. Robert T. Rinehart and Winston. (ZFSL Beiheft Neue Folge 1. Cole. Herbert Ernst.). The case for case. Narr. 1968. Mouton. (Tübin- ger Beiträge zur Linguistik 158. 1981. 3-16.-1O. In: Etnmon Bach. San Francisco. Robert T. Eugenio. Bouvier. Brought to you by | UCL . Dieter Kastovsky (eds. Steiner.1985.10.). 1977. 226 REFERENCES Althaus. Jean. Anläßlich des 70.) Tübin- gen. 93-1O8. Leonhard Lipka (eds. Gauger. 1969.) Tübingen.9. 197O. (Tübinger Beiträge zur Linguistik. Bach. Volume 8: Grammatical relations. Manfred. 1977. etc. Harms (eds. Narr. London. Hans Peter/ Helmut Henne. Charles J. Fillmore. Academic Press. In: Herbert Ernst Brekle. Eugenio.) Wiesbaden. Syntax and semantics.). 1977. Oktober 1977. Paris. Fillmore. Syntax und Morphologie. Probleme der strukturellen Semantik. Larousse. 3-16.). (Tübinger Beiträge zur Linguistik 40. Coseriu. Dietrich. Jerrold M. The Hague. Horst Geckeier.). 1968. The case for case reopened.). Beiträge zum Deutschen Romanistentag in Siegen. 1973. . Mouton. Bonn.). 48-61. Narr. 1-88. Cahiers Ferdinand de Saussure 35. Dubois.University College London Authenticated Download Date | 5/9/18 6:49 PM . 1968.). Elwert. Eugenio. Tübingen. Coseriu. Trends in structural semantics. Les structures lexematiques. 1977. Theodor (ed. Coseriu.

Analyse von Komposita mit zwei nominalen Elementen. Indogermanische Forschungen 70. 169-180. Gipper. 196O. Düsseldorf. (Tübinger Beiträge zur Lin- guistik 286. 1987. Beck. Wortbildung und Semantik. Sprache . München. 151-169.). 117-145. Marchand. Hueber.) Tübingen. Peter von. Heidolph (eds. In: Hans Peter Althaus. The categories and types of present-day English word- formation. Bagel. Pottier. Pädagogischer Verlag Schwann. (Serie A: Linguistique appliquee et traduction automatique 2. 1986. 1963. Hans-Martin Gauger. Helmut (ed. Herbert Ernst Wiegand (eds. Brenda. Band 1. Staib.) Nancy. 1980.) München. A synchronic-diachronic approach. 208-223. Gipper. Dieter.) Tübingen.). 1959. 1965. Indiana university Press. In: Helmut Gipper (ed. Bernard. Hans. Brought to you by | UCL . Bruno. Sessel oder Stuhl? Ein Beitrag zur Bestürmung von Wbrt- inhalten im Bereich der Sachkultur. Generische Komposita. 1969. Laca. Probleme der semantischen Beschreibung denoninaler Nomina agentis. Untersu- chungen zur spanischen Subjektnominalisierung. 271-292. Laca. (Lshrgebiet Sprache. The grarrmar of English nominalizations. Marchand. Narr. Kastovsky. Wortbildung. Wolfgang. Qrientierungshilfen für Lehrende und Lernende. Recherches sur l'analyse semantique en linguistique et en traduction mecanique. 1982. In: Manfred Bierwisch. Niemeyer. 1963. Helmut Henne. Robert B. Funktionelle Untersuchungen zum Französischen und Spanischen.). Bloomington. Bern. Motsch.). 1988. Festschrift für Lso Weisgerber. Karl E. Francke. Düsseldorf. 197O. Mönchen. Lees.Schlüssel zur Welt. Polenz. In: Wolf Dietrich. Publications linguistiques de la Faculte des Lettres et Sciences humaines de Nancy. Schwann. 227 Gipper. Brenda. 1978.). On the analysis of substantive conpounds and suffixal derivatives not containing a verbal element. (Beihefte zur Zeitschrift für Romanische Philologie 221. Helmut. Sprachwissenschaftliche Grundbegriffe und Forschungs- richtungen. Hans. Helmut. Die Wortbildung als Grammatik des Wortschatzes. 1959.University College London Authenticated Download Date | 5/9/18 6:49 PM . Horst Geckeier (eds.

University College London Authenticated Download Date | 5/9/18 6:49 PM .Brought to you by | UCL .

229 SECTION 4 Brought to you by | UCL .University College London Authenticated Download Date | 5/9/18 6:49 PM .

Essen- tially. Vfe may perhaps begin with the apparently simplest type of non-congruent struc- ture. naive native speakers tend to be quite unaware of this and characteristically assume. that 'things' exist to the extent that there are 'words' for them in their own language (Leisi 1952: 25-7). such structures being established on the basis of sense relations hold- ing between individual lexical items. The resultant inter- ference patterns are not easy to counter because of the difficulty of making significant. this is a notorious area of neglect in second-language instruction at all levels. as it were. 230 MARTIN DUKRELL SOME PROBLEMS OF CONTRACTIVE LEXICAL SEMANTICS My aim in presenting this paper is to identify and explain systematically seme instances of non-congruence between selected lexical structures of English and German.University College London Authenticated Download Date | 5/9/18 6:49 PM . Only in this way may we analyse the systematic differences between the lexis of dif- ferent languages in a way which may provide a basis for explaining the charac- teristic interference phenomena occurring in second language acquisition. if not consciously. Such structures are in my view largely language-specific. in different places" (cf.e. which by contrast has one lexeme broadly covering the same range of applicabili- Brought to you by | UCL . However. that the structures of their own language are canonic and universal and that distinctions drawn there will be generally valid. i.e. if any. can be drawn from this. i. easily applicable generalizations about lexical structure. the structure of sense-relations holding in two (or more) languages may be non-congruent. The underlying motive for this exercise was initially the practical one of attempting to pinpoint reasons for inter- ference at the lexical level in foreign-language acquisition/ and ultimately to see what implications of a more general kind. also Lakoff 1987 ). who puts it as follows: "It is a characteristic of languages that they impose a particular lexical 'characterization' upon the world and draw the boundaries. where one language draws a distinction not present in a second. following Lyons (1968: 426).

only has the specific setzen. in English this possibility is ex- cluded. though more seriously because more lexemes are involved. cushion and pillow. for Uhi>. but not in all. in that each language has terms at one level in the hierarchy but not at the other.but because in German one is forced into being specific in contexts where it is unnecessary in English. more to follow). on the other hand.they are demonstrably present .on the other hand.). etc. by using a compound (Armbanduhr^ Sonnenuhr. etc. German. but rather more complex case of non-congruent hierarchical struc- tures is provided by the terms for die in English and German.University College London Authenticated Download Date | 5/9/18 6:49 PM . and thus lacks the possibility given in English of being less specific under appropriate pragmatic conditions: (1) set stand hang lay setzen stellen hängen legen English speakers find this set difficult to handle in German not because equiv- alents are lacking for the German lexemes . In many of these cases. stand (tr. the inoongruence nay be seen in terms of differences in taxonomic hier- archies. count) and Obst (mass.e. nur (limiting) and erst (non-limiting. One is forced to be more specific irrespective of pragmatic consider- ations. whilst English has the 'specific1 terms watch and clock (and. Here we have a relatively straightforward hierarchy. for f i f e . possesses as a simple lexeme only the generic superordinate Uhr. with the generic put functioning as a superordinate to the hyponyms set.). timepiece being rath- er archaic (and not covering sundials). in the case of the lexical set headed by put in English. legen. even though it is conceptually clear. for English fruit. Examples between English and German are well-known and manifold: for German Schnecke. The consequence of this is worth making explicit: whilst in German one may be more specific if the situa- tion demands through appropriate modification of the simplex Uhr. A similar. Conversely. Brand (conflagration) and Feuer (general) and notoriously. which by contrast is the only one represented in the second lan- guage. it does not have a superordinate term for the class. German has Frucht (general. edible). This set has been extensively discussed for German by a number of scholars since Baumgartner (1967) Brought to you by | UCL . stellen. sundial) (Cruse 1986: 145-7). for only. for Kissen. 231 ty as the two in the first. i. German. English has slug and snail. etc. watch and clock. for that matter. hang (tr. e. lay. The same is true.g. Thus. hängen.).

my impression. .. A result is that English learners of German are slow to acquire umkommen as an active item and they tend to use sterben more frequently than a native speaker would. is that. no intermediate level such as is given by umkommen in German (except perhaps for the stylistically highly marked perish. for the lower level. and. which is at the moment a very tentative one based on my experience of the German of English-speaking learners. which is uncommon in actual usage).e.University College London Authenticated Download Date | 5/9/18 6:49 PM .with one exception. the unspecific die will often be used in cases where German usage would prefer the most specific terms. here too. Now. i. I suspect that this reflects a general tendency if one language has paraphrases covering roughly the same area of reference as simple lexemes in another. However. Brought to you by | UCL . as in the cases discussed earlier. In English there is really no such com- parable hierarchy. English does not have the possibility of more specific reference given by German umkommen and will employ in its place the generic term die. as Lipka (198O) has pointed out. and indeed one which will in practice prove very difficult to test rigorously because of the problems involved in eliciting controlled and strictly compa- rable data. for the most specific German. but I am aware that this can at the moment be no more than a subjectively based initial supposition. and the relationship between the various verbs my be repre- sented by the following diagram: (2) sterben ankommen verdursten ersticken erfrieren verbrennen ertrinken. it is not unusual for one language to "use a syntagm where another language employs a single lexeme with roughly the same meaning" and English indeed lacks the derivational posslblity for forming 'terminative1 verbs through prefixes as is given by er. as it entails it and is more specific. etc. indeed taxonyms of wrikormen. .or ver. the English phrasal equivalents may in practice be less frequently employed than the simple lexemes of German and. only a series of phrasal equivalents such as freeze to death.232 and IXorrell (1981). as Lyons (1977: 262) points out. But. de- noting a way of dying occasioned by external causes. We have a generic term die. the lack of such a hierarchy of single lexemes in English has a number of interesting consequences. The third level terms are hyponyms. drown. Umkommen is a hyponym of sterben. First. die of thirst.

In its most current sense it denotes a single place setting. the case of Besteek with its parts would seem to constitute an incontrovertible instance of a purely linguistic meronymy specific to German. i. but where in fact the parallelism is only ap- parent. However. Such cases of apparent and rather misleading parallelism between structures are frequent and by no means restricted to branching hierarchies of the kind just outlined. A Bestedk is constituted by the set of implements necessary for one person to eat a meal. in his terms. 233 We may new pass to cases where the two languages appear to be congruent in the structure of such hierarchies. German Besteek. we have in English the lexemes hill and mountain which appear to correspond exactly to German Hügel and Berg in that in both languages they form part of similar catenary sets which are inherently ordered in terms of size. fork and spoon corresponding to German Messer. e. i. a 'labelled part-whole hierarchy'. although cutlery seems to have the 'right1 meaning to be considered a superordinate to the other three terms. We are clearly not dealing here with a hyponymic or taxonymic relationship.): Brought to you by | UCL .University College London Authenticated Download Date | 5/9/18 6:49 PM . for English (Cruse 1986: I87f. thus rather a 'quasi-superordinate1. on the other hand. is a collective count noun. that in neither case are we dealing with an exact hyponymic rela- tionship: a knife is no more a 'kind1 of 'cutlery1 any more than a Messer is a 'kind' of 'Besteck'. the relationships in the two cases are rather different. Thus. it is rather the case that this set satisfies Cruse's (1986: 161) test-frame for what he terms a 'meronymy'. on the grounds that they reflect extra-linguistic organization and are of no relevance to the semantic structuring of the lexis.g. one each of Messer.e.e. The linguistic nature of meronymies or part-whole hierarchies has been challenged (Cruse 1986: 180).g. as Cruse (1986: 97) points out.: (3) cutlery Besteck knife fork spoon Messer Gabel löffel However. Thus. first. it belongs to a different syntactic category as a mass noun denoting eating implements in general and is. even of a less exact kind. We must recog- nize. Gabel and Löffel and they appear to be hyponyms of the ge- neric term outlery just as the German lexemes are of Besteck. e. In English. Gabel and Löffel. it nay be used with the indefinite article and has a usual plural Bestecke. and each of these nay be explicitly referred to as ein Teil des Bestecks. English has specific terms knife.

it would not seem a satisfactory account of the lack of congruence between the terms in the two languages to say that the average hill is larger than the average Hügel. There is evidently no one-to-one correspondence be- tween the lexemes in the two languages. and the semantic relationship between the lexemes in the two languages is not given fully in terms of their relative position within a size hierarchy. Although. hillock). and for German: (5) {. it may be seen as a subordinate or contingent aspect. clearly.. larger lexical sets exhibiting noteworthy incongruence between different languages which can only be described. the classification is being drawn in a different place.. mound.. for which Luther has "ich hebe meine Augen auf zu den Bergen". hill may be distinguished from mountain in everyday usage.. We may compare the psalmist's "I lift up mine eyes unto the hills". is uncommon in actual usage.University College London Authenticated Download Date | 5/9/18 6:49 PM . if not in the specialist termi- nology of geographers. They are very familiar from second language instruction. It is rather that we are dealing here with oppositions dependent on different idiosyncratic and lexeme-specific relationships in the two languages. This would be confirmed by the fact that the plural of Hügel. typically refers to an isolated entity (DUDEN (1970: 343) defines it as a "leicht anstei- gende Erhebung in einer sonst ebenen Landschaft"). as in the case of the set just discussed. Indeed. Anhöhe). Berg However. if in practice there is a certain degree of referential overlap such that Cruse (1986: 288) regards them as 'plesionyms1. if not ungrammatical. As a typical instance of such Brought to you by | UCL .. it may not even depend primarily on an opposition in terms of relative size within an inherently ordered set. e. hill. hills being smoother. mountain. if we consider expected contextual equivalences. as an originally North German word. in terms of prototypical shape and configuration. mountains rather more jagged.g. Hügel. if they tend to be neglected in many accounts because their potential contribution to general semantic theory is clearly more limited than that of more abstract recurrent relations. In English. whilst Berge typically occur in ranges. are extremely frequent. In German. with English setting the boundaries rather higher than German. In practice.234 (4) (. those of central Germany. size is a rele- vant aspect of the semantic opposition in both languages. in terms of different idiosyncratic relations between the in- dividual items. we find that although a mountain is usually a Berg and a Hügel is usually a hill. many Berge will nat- urally be hills. Hügel..

den Boden bearbeiten 2) ground er liegt auf dem Boden 3) floor der Boden des Zimmers (also = loft. soil.) Schematically. Ihe complex nature of the interlingual incongruence between these may be seen initially by looking at common translation equivalents: (6) Erde = 1) soil er ruht in fremder Erde. bottom. Erde. foundation. and without the restriction placed on the contextual occurrence of earth by the presence of the more spe- cific soil and floor. Much of the detail has still to be worked out. etc.e. although it is obviously close to it and there is a certain degree of referential overlap. 235 a messy set we may consider the lexemes earth. (and may be equated in certain cases with modifi- cations of these. we thus may observe the following relationships of equivalence: (7) earth soil floor ground Erde Boden Grund This is a particularly troublesome set. provisionally it would seem that the contrast between Boden and Erde in German is not dissimilar to that between ground and earth in English. floor and ground in English and their usual equivalents in German.g. in that Brought to you by | UCL . It should be noted that although these are more specific than any of the German terms. as in no case in either language do we seem to be dealing clearly with any relationships of a more abstract kind. but with the proviso that ground as a point of reference will normally be Erde but in other cases Boden. Boden and Grund. they may be partial synonyms in the sense of Lyons (1981: 5O-55). trockene Erde 2) earth die Erde wird im Frühjahr warm 3) ground die Saat in die Erde bringen Boden = 1) soil fruchtbarer Boden. Erdboden for soil or Fußboden for floor) they are not inoontrovertibly hyponyms of either of the other two English words since they do not fit satisfactorily into any of the standard test frames. e. soil is not unequivocally a 'type1 of earth.University College London Authenticated Download Date | 5/9/18 6:49 PM . i. Ihe same is true of floor and ground. Thus. etc.) Grund = 1) ground bis auf den Grund zerstört (also = bottom. as would be confirmed by the fact that Boden dees not translate earth.

236 (8) He fell to the floor. but worse than this one. but it is by no means evident that floor is a "type1 of ground. This may be illustrated using the type of test-frames outlined in Cruse (1986: 207ff. that native speakers of German vary on this point. nay clearly express the same proposition. the German antonyms gut and schlecht differ from their apparent English counter- parts good and bad in that they are of the 'polar' rather than the 'overlapping1 type. This is a not untypical example of the quite disparate structuring in terms of idiosyncratic. like English good and bad. presupposing bad) The German adjectives gut and schlecht. as is indubitably the case in English (Cruse 1986: 219). Thus. however.University College London Authenticated Download Date | 5/9/18 6:49 PM . but better than this one. as Cruse (1986: 219) points out. as Cruse maintains. or that (7) entails or is entailed by (8). (11) TThat essay was good. Within the area of antonymic relations we may find rather simpler patterns of incongruence which may be explained in a rela- tively straightforward manner.e. My own tests with native informants. 1987: 23O-7) suggest that they belong to the 'overlapping'type. Brought to you by | UCL . with both yielding pseudo- comparatives: It should be noted that the analyses of German gut and schlecht by Bierwisch (1967. would tend to support the view that the German pair is rather of the 'polar' type. non-recurrent. like most measure antonyms. In many syntagmatic environments they are mutually exclusive and must be seen as in- compatible. on the other hand. Cn the other hand. English good (in one sense) operates over a scale of MERIT and its com- parative better is a 'pseudo-comparative'. It is possible. i. bad operates over a scale of BADNESS (which overlaps with that of MERIT). (9) He fell to the ground. but whilst that with good is 'impar- tial1 . operate over a single scale of MERIT. expressing not 'good to a greater degree1 but Of greater merit1. in particular where they may be correlated with differences in antonymic type. nevertheless.) or Lehrer (1985): (10) That essay was bad. that with bad is 'committed'. language-specific sense relations with which a contrastive semantics must cope. so that things that are bad may become better. it presupposes that the entity in question may appropriately be designated as bad per se: (12) How g&od was the essay? ('impartial' enquiry) (13) How bad was the essay? ('committed1. Both adjectives yield normal howquestions. but what is good cannot normally be qualified as worse (Cruse 1986: 213).

However. droughts. Much attention has been paid since the classic paper by Bierwisch (1967) to the se- mantic properties of polar adjectives. which is then impartial: (16) Wie gut war der Aufsatz? (17) ?Wie schlecht war der Aufsatz? However.. aber schlechter als der von letzten Jahr. but even here we are rarely dealing with neat pairs of opposites. (15) Der Aufsatz war gut. In the terms used by Cruse (1986: 241) we have 'hypo-antonyms' in opposition to a 'super-antonym'. it has to be said that. particularly the adjectives of measure- ment. but with overlap. even within such a well documented and fully researched area such as antonymy.e. It cannot be identified as an antonym of gut and is used characteristically in collocation with nouns denoting things which are inherently bad. etc. 237 (14) Der Aufsatz war schlecht.University College London Authenticated Download Date | 5/9/18 6:49 PM . asymetry and a marked degree of interlingual incongruence. it is precisely the English equivalents of such words which do not normally collocate with better and cannot be questioned with How good is . An open question for the moment is whether the fact that schlimm refers to inherent badness and thus shares similar properties with English bad is responsible for the fact that English learners of German characteristically overuse schlimm and apply it in contexts where only schlecht would be appropriate. i. schlirrrn.g. famines. as German has a further equivalent to English bad. e. With measure adjectives Brought to you by | UCL . as the recent work by Lang (1987) demonstrates quite conclusively.. Only the 'positive1 member of the pair yields a how-question. A number of English dimension adjectives exemplify what appears to be a general tendency in this set for a distinction made in the 'positive' adjectives to be neutralized in a 'negative' one. aber besser als der von letzten Jahr. accidents. which is distinguished from schlecht precisely in that it operates solely along a scale of BADNESS (Leisi 1952: 44). languages may differ quite radically in their relational structuring to an extent which is not always fully realized.: (18) ein schlimmer (?schlechter) Unfall As Cruse (1986: 214) points out. major illnesses.? It would thus appear that the distinction between the antonymic types in the two languages which we have seen is linked to the pres- ence of the additional term schlimm in German. this is still an inconplete account of this lexical set in the two languages.

Dunn is a super-antonym to dick and dicht. dick and dicht are clearly semantically distinct. All these sets correspond to rather different structures in German. Lyons 1977: 276). with the former clearly referring to a transverse measurement and the latter to relative density .again an iso- lated and language-specific relation which has no real counterpart in English: (20) dicke Wolken i dichte Wolken Finally. but the distinction between them is quite different to that between thiak and fat in English. However. Whilst fat and thiak may be seen as cognitive synonyms (Cruse 1986: 241). kurz correlates only with long.University College London Authenticated Download Date | 5/9/18 6:49 PM . thick and fat the antonym thin and broad and wide the antonym narrow. toll and long share the antonym short. the relation between breit and weit cannot be correlated in any way with that between broad and wide. Thus in English. the former collocating primarily with animates and the latter with non-animates (Leisi 1952: 44). although they share Brought to you by | UCL . we see that German breit and weit both form distinct antonymic pairs with schmal and eng respectively whilst English has the super-antonym narrow to both broad and wide. as the following table shows: (19) short lang · kurz dick- thin ^>dünn thick'' dicht- wide ^ weit eng narrow broad'' breit -schmal We see first that German lacks a lexical contrast corresponding to that made in English between tall and long (Durrell 1988). and a distinction nay be vulnerable as percep- tually less significant (Lehrer 1974: 27.238 this may evidently be seen as motivated inasfar as the negative will approach sane lower limit or zero point.

this is not true of its antonym breit: (21) Die Frau ist schmal geworden. valuable though this is. apparently arbitrary collocational restrictions. Ihus although sahrvzl may be predicated of humans in certain contexts. "daß sich die Dimen- sionsadjektive zu einem sehr unregelmäßig strukturierten Feld zusammenfügen" appears even more justified in a contrastive context. although weit may be used naturally of expanses. this was not always possible. 239 approxiiiately the sane range of application.): (23) ein weiter Platz (24) ein kleiner (?enger) Platz Qiglish broad and narrow exhibit comparable. even within an apparently straightforward group like polar antonyms in the field of measurement. I have attempted in this paper to draw attention to a number of kinds of inter- lingual incongruenoa at the level of the lexis. Lafrenz's (1983: 156) observation on the basis of an exhaustive study of the Genien adjectives in this group.University College London Authenticated Download Date | 5/9/18 6:49 PM . it is rare to find instances of precise antonymic pairs. Durrell 1988). Cn occasion a study of the abstract relations provided only an initial or partial explana- tion. Although in a few cases we were able to correlate the incongruence with relatively straightforward differences in terms of abstract sense relations. Brought to you by | UCL . eng cannot. Similarly. but in essence we are dealing in a contrastive semantics with the comparison of specific and idiosyncratic sense relations holding between individual lexemes which are largely non-recurrent and quite internal to the particular languages. Essentially. (22) ?Die Frau ist breit geworden. and its antonym in such contexts is klein (Lafrenz 1983: 149f. and in both languages we may ob- serve certain awkward and apparently unmotivated asymetries in usage between the positive and negative members of these pairs (lafrenz 1983.

Band 1O. 1984. 397-429. Durrell. Zu einigen deutschen und englischen Dimensionsadjektiven. 198O. Heidelberg. 1985. Lang. Heidelberg. Lehrer. 1987. Dezember 1977. Methodology and representation in the study of lexical fields. Martin. 1987. Zu den semantischen Strukturen der Dimensionsadjektive in der deutschen Gegenwartssprache. 93-115. J. Acta Universitatis Gothoburgensis. Manfred. Zeitschrift für Phonetik. 1975. Perspektiven der lexikalischen Semantik. What categories reveal about the mind. Adrienne. Klaus. In: Manfred Bierwisch. Munske et al. 287-458. Durrell. Cambridge. Alan. Semantik der Dimensionsauszeichnung räumlicher Objekte. Lipka. Lafrenz. Bierwisch. Manfred et al. Some semantic universals of German adjectivals. Markedness and antonymy. Dieter (ed. In: Horst H. Bierwisch. 1987. 1985.University College London Authenticated Download Date | 5/9/18 6:49 PM . 1967.240 REFERENCES Baumgärtner. Ernst. and dangerous things. Mannheim. Winter. 165-97. Bibliogra- phisches Institut. 91-286. In: Manfred Bierwisch. 1952. Semantic fields and lexical structures. 1983. Lakoff. 198O. 490-512. Ewald Lang.). Bedeutungswörterbuch. Grammatische und konzeptuelle Aspekte von Dimensionsadjektiven. 2 Lsisi. Ewald. Chicago. Praxis der englischen Semantik. Beiträ- ge zum Wuppertaler Semantikkolloquium vom 2. Brought to you by | UCL .).). Lexical semantics. Sprache der Gegen- wart 1. London. Contrasting the lexis of English and German. (eds. 1986. Peter G. Martin. Wien. Manfred. 1970. North Holland. 1974. 35-54. Eine vergleichende Analyse.). Amsterdam. Bierwischf Manfred. 1987. Semantik der Graduierung. Quelle & Meyer. Ewald Lang (eds. 664-686. JL 21. Leonhard.). Sprachwissenschaft und Korittunikationsforschung 37. Kastovsky. Foundations of Language 3. (studia gramnatica 26/27.) Berlin. Gruse. Cambridge University Press. The University of Chicago Press. Ernst.) Göteborg. Ewald Lang (eds. In: Dieter Kastovsky (ed. 1-36. leisi. 1981.-3. Bouvier Verlag Herbert Grundmann. Lehrer.. Adrienne. In: Charles V. 1973. Bonn. etc. (Göteborger Germanistische Forschungen 26. 1967. Zürich. Akademie Verlag. Der Wbrtinhalt. 93-114. D. fire. RUSS (ed.). George. Dimensionsadjektive: Semantische Struktur und begriffliche Interpretation. 1988. DUDEN. Bierwisch. Die Struktur des Bedeutungsfeldes. Women. Seine Struktur im Deutschen und Englischen.

1977. John. Metzlersche Verlagsbuchhandlung. B. Lyons. Band 219. Groos. Brought to you by | UCL . Horst H.) Stuttgart. 1985. Mouton de Gruyter.). Introduction to theoretical linguistics. John. Geburtstag von seinen ^terburger Schülern.. 2 vols. Heidelberg. Lexikalische Semantik. 241 Lutzeier. Laxikologische Studien. Ludwig Erich Schmitt zum 80. John. etc. (Sammlung Mstzler. Peter Rolf.University College London Authenticated Download Date | 5/9/18 6:49 PM . New York. Lyons. RUSS. 1988. 1968. etc. Berlin. Cambridge university Press. Cambridge. meaning and context. Charles V. Munske. (ed. Language. Contrastive aspects of Qiglish and German. J. London. Cambridge University Press. Semantics. (eds. Lyons. London.). 1981. et al. 1981. J. Deutscher Wortschatz.. Pontana.

Irrespective of its specific nature. cross-cultural lexicology enjoying a reputable tradition in both linguistics and anthropological studies. Hence the term 'socio-semiotic1 as "indicating a general ideology or intellectual stance. An earlier version of this approach was delivered at the Vlllth World Congress of the Association Internationale de Linguistique Appliquee (AILA) at Sydney in August 1987. Brought to you by | UCL . a conceptual angle on the subject" (Halliday/Hasan 1985: 3). The legitimacy of the object of research of this study will not be questioned. and there- fore open to debate. it should reveal the determining impact exerted by socio-semiotic factors on the ways in which different speech- connunities cope with reality (here: that part of reality which we shall choose under III). The way in which we . varies across cultures. in other words along social systems. is our way of looking at our research object: we take a cross-language socio-semiotic view. 242 WOLFGANG KÜHEWEIN A sccio-siMionc WAY OF LOOKING AT CROSS-CULTURAL LEJO:COLOGYI this. What is less traditional. One such way of coping with reality is by language as one among many semiotic systems with which humans master their environment. WHY? .University College London Authenticated Download Date | 5/9/18 6:49 PM . exerts a general threefold determining power. the way of looking at one's object of research.Raison d'etre/aims 1.and languages . this 'conceptual angle1.

for example.2 A normative philosophy of science. within which internally legitimized theory-formation is encompassed by (at least largely) externally legitimized constitution and fi- nalization. the teaching of the use of these lexemes with. for example. It has a twofold bearing. For a detailed treatment of the integral view as advocated here.2 It affects the finalizing phase of research.1 It affects the constitutive phase of research. admits external. for the study of language. for example social determination of research aims. Which aims nay be considered legitimate depends on one's philosophy of science. of its methodological impact.3.2 It affects one's research object via the constitutive power which it exerts on one's concept of reality or even on reality itself.University College London Authenticated Download Date | 5/9/18 6:49 PM . the causa effiaiens. the ultimate aim of achieving better cross-cultural international understanding. affects. cf. 3 general threefold determination as exerted by one's 'conceptual angle' calls for a concept of linguistics which grants the scope necessary for an interplay between external and internal legitimization. 2.2. or even determines possible aims of research. including the possible range of finalizations.3. 1981). in particular for lin- guistics as part of both natural and social science. From the realm of natural sciences rich evidence is provided for this position by Kühn (1962. Kühlwein (1985: 133ff. cf. 14-21). As to the specific nature of the 'conceptual angle'. 243 1. Kühlwein (forthcoming b: section 3.1. 1. cf. here.1).a precarious position. here the interest in the social determination of the use of a set of lexemes as pre- sented in (III). 1. the oausa finalis. Brought to you by | UCL . 1. 1.3. Kühlwein (1987a: 6 O f f .3. ) . to which one's results can be put. of the role of external determination and finalization. Kühlwein (1986).1 It must affect research methodology to ensure adequacy.3 It influences. cf. cf. in other words a concept which replaces the (pseudo-) opposition between theory and application by an integral view. of constitutive elements. a socio- semiotic one seems to us to be well in accordance with such an integral con- cept of linguistics. 1.1 A descriptive philosophy of science legitimizes research aims internally. by 'pure' science exclusively . Kühl- wein (1987c: 11-14 resp. on the other hand. also Kühlwein (forthcoming a ) . cf. of some aspects of the internally legiti- mized theory-dynamic phas-e between constitution and finalization.2. 1.).

cf. Kühlwein (1987b: 4-7). Brought to you by | UCL . frequently make use of pragmatic evidence for the mere purpose of accounting for structures of other. let alone for cross-cultural. it is only via their mani- festation in performance that cognitive strategies become accessible. . it would . contrastive purposes would necessitate a pragmatic deep structure to serve as the tertium aomparationis.lack sufficient explana- tory power.Despite valuable lexical field studies. Why? For a more detailed presentation of the implication of the philosophy of science view adopted here and for cross-cultural lexicology.The angle(s) 4 1.University College London Authenticated Download Date | 5/9/18 6:49 PM . 1. study. 1. for a contrastive. . cf. for a preliminary model. cf.2 The knowledge-oriented view. Kühlwein (1987b: 7-9). For a greater variety of examples. . cognitive psychology will not provide an ultimate reference point for cross-cultural lexicology. as the terttum comparati'onis.The deficiencies of the system-oriented and of the knowledge-oriented views call for contrastive interactive competence studies . non-pragmatic language levels. 244 II. 1. systemic contrastive lexicological studies have remained scarce for three rea- sons: the problems (a) of semantic primes and universals. (b) of tevtiwn aomparationis and equivalence. Kühlwein (forthcoming b ) . As a con- sequence. in this case for a 'cross-mental'. Furthermore. despite recognizing its basic significance for the question of how different speech-communities conceptualize the world.3 The behaviour-oriented view. and (c) of linguistic models and procedures suitable for contrastive. however. Furthermore.despite a high descriptive power . To see hew the socio-semiotic approach works across cultures we briefly set it into relief against two others.To relate different structures in differ- ent languages to corresponding differences in the mind and to differences among the cognitive strategies which cause different types of conceptualization in different speech comnunities will without doubt have more explanatory power. but we lack the universal grid of cognitions which would be required to serve as the reference point. These.1 The system-oriented view. HOW? .as indicated by a growing number of pragmatic contrastive studies. for example pragmatics as providing the illustration of behavioural and/or communi- cative consequences if one gets one's semantics wrong. studies. Even if such a deep structure could be developed to a sufficient extent.

as their major variations are not across individuals but across societies. but carries en: '"Hie distinction between them is rather obvious and simple" (1978: 13). But which one should be considered to be the more encompassing? Psycho-sociolinguistically. He no- tices that what makes it complicated is "the fact that it is possible to embed one perspective inside the other" (1978: 13). ethnolinguistic level. for our purpose of trying a socio-semiotic 'intellectual stance' (cf.University College London Authenticated Download Date | 5/9/18 6:49 PM . not a com- petence. after all. Section (1. are reflected/manifested accordingly on the lin- guistic plane.a potential. is no less exclusive. . has a potential of latent semiotic properties.and what is actu- alized from latency differs accordingly. of semiotization. we Brought to you by | UCL . or event. As seen on an individual basis they can be described as psycho- semiotic processes of cognition. members of the same society understand each other's ways of semioticizing reality . Members of different societies/speech-comnunities make differing choices from this latent semiotic potential . which is subjective" (1978: 38). despite recognizing the 'ccftplementarity1.a view to which Halliday. would reply that it would mean "taking the intra-organism ticket to what is actually an inter-organism destination" (1978: 38). language behaviour has to be considered as an emanation of language knowledge . opposing what the speaker knows to what he does — there is no need to bring in the question of what the speaker knows.Thus. these processes of semiotic profiling. pragmatically described. which is objective. it can be adopted. Each person. Halliday (1978: 13) observes: "These are two ccnplementary orientations". the background to what he does is what he could do .for example the different semioticity of a woman's bulky stature in a culture in which women are supposed to do most of the physical labour versus that in a highly industrialized culture. at least).socio-psychosemiotics? Nevertheless. reflections of deeper contrasts on a sociocultural. object. However. above) when looking at lexicology across cultures.3) indicates that the psycho-semiotic and the socio-semiotic angles should not be seen as diametrically opposed to each other. anthropo-. are. A property that is semiotically strik- ing to members of one speech-<xxTinunity may well be marked differently by mentoers of a different speech-community or perhaps even fail to catch their attention at all . however. "So in an inter- organism perspective there is no place for the dichotomy of competence and performance.more or less. Obviously. however. but fairly rigid and unilinear as well: 'We can only say what we can mean and we can only mean what we can do' . actually. about whom/which we communicate. the nature of these processes is of a socio-semiotic kind (after all. 2. 245 Cross-cultural differences. Halliday's own 'ticket1. provided.

But: Cf.Data. that are used to attribute B to somebody. Brought to you by | UCL . Nies (1978. results 1. languages: English (E) and French (F) .University College London Authenticated Download Date | 5/9/18 6:49 PM . to assess the degree of its semiotic relevance: the macroscopic view (cf. the gen- eral parts (I) and (II) call for this specific part (III) as a corpus-based illustration of what cross-cultural lexicology can do under socio-semiotic auspices. The underlying procedural and numerical data are derived fron the semasiological- onomasiological thesis of Nies (1978). language is both nature and nurture.). 2 below). I: 4ff.e.2 In the interest of the above-mentioned 'ultimate1 aim of coming to better cross-cultural international understanding we should be able to put each actual utterance referring to B into relief against the overall inclination/ disinclination to attribute or express B at all in the culture in question. 2. The second prerequisite for a proper understanding of an actual utter- ance is the knowledge about the specific socio-semiotic conditions that must be met to assign B to somebody in a certain culture. WHAT? . vol. the sets of adjectives in E and F. 1. III. 1. linguistic realizations. m order to get beyond the theorizing and postulating stage. After all. 3 below). 246 regard the activity of socio-semiotic profiling as part of doing.1 We choose: . Macroscopic results a cultural sphere: beauty (B) of human beings . about the specific ways of socio-semiotic profiling: the microscopic view (cf.1 The overall inclination of members of the F speech ccmtunity towards attributing B is far greater than that of the E one.2 In both speech comnunities B is attributed much more frequently to women than to men.

aesthetic judgment (con- stitution. whereas in the Erench speech conmunity it equals 4 to 1 . 3.1 P1 [+ well dressed] vs. Basically the socio-semiotic thrust of B seems to be greater in F.University College London Authenticated Download Date | 5/9/18 6:49 PM . 247 2. 3. Microscopic results Ihe socio-semiotic parameters that turned out to have the highest discriminating power were: semantic reference (to clothing or tody). however. Furthermore in E there is a stronger tendency towards + P2- adjectives despite coarse features. these differences in socio-seroi- otic signification will certainly matter beyond simple comprehension. P2 in F. 3.3 The relation between explicitly and positively attributing [+ B] on the one hand and stating the absence of B equals 2 to 1 for wonen in the E speech community. less easily in E. + P1 triggers the association of + P2 just a little bit more easily in F than in E.4 When it comes to 'true understanding1. .As for men. in particular for wonen. Conclusion: sex-specificity seems to play a considerably more important role for the contrast P1 vs. non-aesthetic judgment as to physiological con- ditions (age. and psychosomatic conditions (naturalness. coarse features] In F [+ refined features] easily trigger + P2-adjectives.the E nan in the E speech community being characterized by an overall dominance of [. 2. harmony).B]. psychological conditions (vanity.2 P3 constitution [refined vs.+ P2 triggers + P1 even more easily for F women.+ P1 easily triggers + P2 in Γ. sex-typicity). for this very reason. warmheartedness). In E + P1 -adjectives go with both refined and coarse features with equal frequency . coarse features seems to be stronger in F. Conclusion: the semiotic thrust of the contrast refined vs. a higher degree of markedness is due to each actual utterance of B in E. seriousness). Brought to you by | UCL . perfection. and F obviously has a stronger disinclination than E against associating + P2 with + P1 ^adjectives. [+ coarse features] easily trigger + P1-adjectives.obviously a lower socio-semiotic impact of this contrast. . P2 [+ good physical appearance] Ihere are mutual dependencies: .

who in turn is treated more leniently (i.University College London Authenticated Download Date | 5/9/18 6:49 PM . in the case of .e. this is not so in E.P6 they are treated more harshly than the respective F woman. On the other hand.248 3. 3. but more leniently than the respec- tive E man. 3. In both F and E there seems to be more leniency with the man than with the wonan. but the respective correlations are weaker than the above- mentioned one for F women.P1O easily evoke P2-adjectives. in particular for the F woman. cultivated] For the F woman both + P1O and . Conclusion: F being rather generous to the woman.5 P6 age [+ looking one's age (middle-age)] For the F woman + P6 can cause P1-adjectives much more frequently than in E - the E speech community/ obviously. the E man.especially for the E man. In F there is a significant correlation between . specifically evoking P2-adjectives. and the E woman P2-adjectives are more easily compatible with . it is fairly relevant for the F woman. For the F man. E is somewhat more lenient with the man. 3. unassuming vs. 3.3 P4 perfection [+ consuntnate outer appearance] In F + P4 is more important for attributing P1-adjectives than in E.P10 and P1 -adjectives.7 P8 vanity [+ the intention to impress by a good appearance] Biis property primarily affects Pi-adjectives in both E and F. As for F man. 3.P1O than with + P10. almost irrelevant for both E men and women. warmhearted.P5 is more striking for the French speech community. Conclusion: . well-groaned. Brought to you by | UCL . is attributed B-adjec- tives more frequently) than the E wonan.4 P5 harmony [+ harmony of appearance] While in F .9 P1O naturalness [+ natural. in E P4 is more important for attributing P2-adjectives than in F .6 P7 sexually typical appearance [+ feminine/masculine looking] Both E and F show a high correlation of this property with the attribution of B-adjectives. being quite harsh towards women who try to look younger than they are.8 P9 warmheartedness [+ amiable. friendly] P9 causes a somewhat stronger reaction in the F speech conrounity. In addition for the F woman + P7 easily causes the attribution of P1-adj ectives. 3.P5 more or less excludes P1-adjectives completely.

whose attribution seems to be more strongly oriented according to the specific person concerned. The same holds true for aesthetic properties in the case of men (con- stitution. earnest vs. a cross-lan- guage comparison reveals: 4. evoking P2-adjectives for F women fairly easily. What might be a socio-semiotic norm in one culture/language might well be socio-semiotically marked or deviant in another despite various kinds of formal.1) and a group which is (highly) incompatible with either male or female (4. Obviously. 249 3. cheerful. cf. 4. the occasional admittance of absolutely contrary properties for the evocation of these adjectives. In Ε the respec- tive spectrum is even somewhat broader for men than for women. there is no such thing as complete socio-semiotic equivalence as a reference point for cross-cultural lexicological studies. serene] This property is of quite lew significance for the attribution of B-adjectives in both speech communities. This is the point where our results should be passed on to both the lexicographer and the foreign language teacher. however.2 In F psychosomatic properties obviously matter much more for wonen than they do in E (naturalness.2). 4.University College London Authenticated Download Date | 5/9/18 6:49 PM .aesthetic and . harmony). functional or other equivalences. 5.extra-aesthetic judgments in F when women are concerned. Obviously in Ε the attribution of these B-adjectives seems to follow a more clearly profiled image of the woman as opposed to the corresponding adjec- tives in French.1Ο Ρ11 seriousness [+ serious. seriousness/cheerfulness) when it comes to assigning B or not.1 This group of adjectives can be evoked by a far wider scope of . + P11. Brought to you by | UCL . perfection. Cross-language comparison According to the distinction of the B-adjectives investigated into a group which is ccnpatible with both male and female on the one hand (4.

Klegraf. 1987. The interdisciplinary framework of the theory-dynamic phase in finalized linguistics. Externality and finalization in linguistics. Tübingen. Wolfgang. Groos. Vic. Soziosemiotische Detenninanz im kulturübergreifenden sprachlichen Erfassen der Wirklichkeit. New York.). The structure of scientific revolutions. 127-140. A sociosemiotic view of the 1grasp of lan- guages at reality: The lexical field "aesthetic judgement . Berlin. Univer - sity of Chicago Press. Michael Alexander Kirkwood.University College London Authenticated Download Date | 5/9/18 6:49 PM . Aleksander Szwedek (eds. In: Kari Sajavaara (ed. Narr.)./ Australia. Wolfgang. I. Groos. 1978. Kühn. Diss. Chicago. Warner-Heisenberg- Vorlesung. 25O REFERENCES Halliday. Deakin University. Nies. Ms. Kühn. 1-22. research goals. Wolfgang. Roger W. Halliday.). Kühlwein. forthcoming a. Kühlwein. The social interpretation of language and meaning.Aleksander Szwedek (eds.L. Edward Arnold. Linguistic integrality: Research methodology. Language. Heidelberg. Language as social semiotic. In: Eis Oksaar (ed. forthccming. Michael Alexander Kirkwood. Wolfgang. Wolfgang. Onomasiologisch-semasiologische Analyse ausgewählter franzö- sischer und englischer Schönheitsadjektiva. Eis (ed. and text: Aspects of language in a social-semiotic perspective. Nehls. Josef Klegraf (eds. Linguistics across historical and geographical boundaries.). In: Josef Klegraf. Themen XXXIV. Ruqaiya Hasan. Dietrich. Kühlwein. Dietrich Nehls (eds. 1985. Brought to you by | UCL . 1986. Amsterdam. Oksaar. The need for integration of applied and theoretical linguistics: Research objects.). Mouton de Gruyter. Josef.). In: Olga Mis'eska Tonic. 9-24. Dietrich Nehls (eds. 51-74. In: Dieter Kastovsky. Soziokulturelle Perspektiven von Mehrsprachigkeit und Spracherwerb. Kastovsky. 1311-1319.). 1987c. Was sind wissenschaftliche Revolutionen? (Schriften der Carl-Friedrich von Siemens-Stiftung. context. forthcoming b. 24. 1O. Dieter. Kühlwein.T. In honour of Jacek Fisiak on the occasion of his 5Oth birthday. 1986. Universität Trier.) Ms. London. Wolfgang. Thomas S« 1981. 1985. Volume 2: Descriptive. 3 vols. Kühlwein. Essays on the English language and applied linguistics on the occasion of Gerhard Nickel's 60th birthday. forthcoming. 1978. 1987a. Februar 1981. Guy. 1962. contrastive and applied linguistics. 'Let your science be human1: Linguistics as applied linguistics. Wolfgang. Kühlwein. 1987b. Thomas S. Lin- guistics as Applied Linguistics (AIIA-Review IV). Review of Applied Linguistics 67/68.). Contributions to English linguistics presented to Gerhard Nickel on the occasion of his 6Oth birth- day. Heidelberg.). Kühlwein. Shuy (eds.

251 Sajavaara. Kari (ed.University College London Authenticated Download Date | 5/9/18 6:49 PM . forthcoming. Roger W. Shuy (eds. The relation of theoretical and applied linguistics. New York.). Olga Miseska. Brought to you by | UCL . TOmic. University of Jyväskylä. London/ Plenum. Jyväskylä. 1987.). Proceedings of the 1986 synposium of the Scandinavian Association of Applied Linguistics.

or. A recent collection of false friends between English and German gives line drawings for some entries. but we would nevertheless have similar differences with sets of 'representative1 examples. (Neuhaus 1982a. House and Haus would probably never be classified as false friends. a Warenhaus and a Lagerhaus (Pascoe/Pascoe 1985: 176-177). for that matter. Initial computer data preparation for the research reported here was made possible in part by support extended by the Minister für Wissen- schaft und Forschung des Landes Nordrhein-Westfalen (Az. FREGE'S SENSE. most people would be satis- fied with a collection of some 'representative1 examples. between English. however. and Speicher (am Hafen).University College London Authenticated Download Date | 5/9/18 6:49 PM . Instead. Brought to you by | UCL . Of course. IIB7-FA75O7 and FA7977). There is a note on American English usage as well: store instead of department store.Warenhaus has two such drawings showing a department store and a warehouse. and Australian cases. Certainly. 1982b). The entry warehouse . American. and looks more like a shopping centre than a department store. or. there are both warehouses and Warenhäuser which look very different. Besides Lagerhaus two other German words are also mentioned: Lagerhalle. as do similar drawings in dictionaries and encyclopedias. But the Warenhaus drawing seems to represent a rather marginal case. The drawings themselves pose serious problems. more 'representative' examples would show further differences be- tween English and German cases. JOACHIM NEUHAUS FALSE FRIENDS. It features parking spaces and trolleys. This is a more general characteristic of natural kind terms and does not depend on the false friend relation. to use the German words.252 H. and it will be im- possible to draw a 'typical1 Warenhaus. AND WORD-FORMATION 1.

This conment is a bit odd at first sight since Warenhaus is still current today.Warenhaus pair belongs to the traditional canon of false friends." Historically. loanwords may develop a different meaning in the receiving language. warum in der einen Sprache sich gerade diese bestimmte Bedeutung durchgesetzt hat. nächst verwandten. Jobber.and argues for mere historical ac- cident: Die unterschiedliche Verwendung der gleichen Formen von einer Sprache zur anderen. department stores were not well received in Germany. Politically.University College London Authenticated Download Date | 5/9/18 6:49 PM . unmittelbar benachbarten Sprache gerade diese andere Bedeutung. 253 The warehouse . the new word lost considerable appeal. There was even a special tax. Boycott. It was repealed in 1919. The institution of department stores is known in 18th century France and England. mit anderen Worten der ge- schichtliche Zufall. of the Warenhaus für deutsche Beamte . which appeared in 1860. or Trust. Later on. in 1889. Wandruszka. There is a citation from the New York Times (November 1917): "Our country is the warehouse and the bank of the Entente. the founding. 2. and the nowadays old-fashioned Warenniederlage. Krüger (1928: 162) octrroented on the pair using the sane translation equiva- lencies: Lagerhaus. It has been said that the mere existence of so many false friends is a serious problem for current semantic theories. German Warenhäuser were warehouses as well. For Germany it is a late 19th century phenomenon. for example. There even is an explicit date. but this is unlikely for Warenhaus. The word is alreaiy well attested in 18th century sources. uses Lagerhaus and Warenlager in its lexical explanations (186O: 712). which builds on the original warehouse semantics. The special tax was called Warenhaus Steuer . which was introduced to protect small business. das alles zwingt uns zu dem Schluss. sees an immense number of heterogeneous factors. Speicher. in der anderen. Daniel Sanders' Wörterbuch der Deutschen Sprache. The semantic shift can be studied in Grinm/Grirm (1922: 2008). unsere Unfähigkeit zu erkennen. ja auch nur zu vermuten. die souveräne Inkonsequenz und Asystematik so vieler Abweichungen und Verschiebungen. In retro- spect the word chosen in 1889 seems to have been a rather successful public re- lations and advertizing idea. It can be understood with some background information. Thereafter there seems to have been no further reason to avoid the word. In Grimm it is simply stated that by the time of World War I the older and well-established Kaufhaus is replacing Warenhaus. Mackensen (1977: 2452) wants to group the word as an English loan together with other 19th century loans such as Banknote. The current semantics of the German word Warenhaus is a rather recent innovation. (Wandruszka 1978: 223) Brought to you by | UCL . and suggests that even ordinary public servants may now have their own 'ware- house1. dass für das Zustandekommen so vieler falscher Freunde letztlich eine Unzahl heterogener Faktoren verantwortlich sein muss.

but to another factor of linguistic change" which he later on identifies as 'borrowing1 (1933: 361). die Polymorphie (verschiedene Formen für die gleiche Funktion) ein entscheidendes Merkmal unserer natürlichen Sprachen. and the conventional expression. But the slogan has both a phonetic and a semantic interpretation. Both concepts are taken as essential properties of natural languages with an interesting twist in their respective definitions towards a 'functional'. For Warenhaus there could be made out a good case for 'internal1 borrowing.. pragmatic understanding. 73) Finally. So wie die Polysemie (verschiedene Punktionen für die gleiche Form) ist auch ihr Gegenstück.we both have to know something about the reference. Brought to you by | UCL . the socially and culturally determined mode of presentation. Interestingly enough. and it seems to have most often been used against the classical neogrammarian position of exceptionless sound laws. (Wandruszka 1977. 254 In Wandruszka's argument for mere historical accident the concepts of 'polysemy* and 'polymorphy' are used with explanatory intentions. which both Malkiel (1967) and Christmann (1971) used as a title of their special studies. The obsolescence of the former sense is quite important: there is no polysemy of two distinct senses. Conventionally. 3. the slogan was integrated into structural theories. Blocmfield formulated a common defence of the structural position in explaining phonetic irregularities as "due not to sound change. Blocmfield's way out is obviously more difficult to substantiate. against which it seems still to be used now and then.University College London Authenticated Download Date | 5/9/18 6:49 PM . the slogan is taken as a summary of achievements in modern linguistic geography. such as in Wandruszka's aside against "die strukturalistische Phonologie mit ihrem apriorischen System" (1977: 76). The 1889 sense is the accepted sense. Wandruszka's insistence on historical accident is reminiscent of the slogan 'Each Word has a History of Its Own'. and the former sense is now obsolete. Later on. we mentioned the set of 'representative' cases. . Sociolinguistic investigations show that this solution is an over- simplification on the phonetic level as well. this free play of polysemy and polymorphy ("Das Spiel der Bolymorphie und Polysemie" (1977: 75)) is offered as a 'solution1 to the problem of false friends. On the semantic 2 level. or Art des Gegehenseins to use a Fregean expression (1892: 26).. In discussing the semantics of Warenhaus we are talking about the con- ventionality of word senses.

such as the expression 'the least-rapidly converging series' (1892: 28). Frege mentions a funda- mental metalinguistic attitude: "We presuppose reference1 (1892: 31). and Schneckenhaus ('snail shell1) are even further removed from 'representative' core members. would probably have a good chance to specify sense and reference authentically. though. or dog-house could be commented upon in a similar way. In his essay there are examples of sense without reference. In Frege's Morgenstern example there is another proper name interpretation: the planet Venus. Frege himself did not discuss exarrples of this type. The most extreme interpretation is always a proper name interpretation. Word-forma- tions such as bird-house. A corresponding case could be made for lighthouse. But this is not the normal case. Perhaps. The name is said to be 'motivated' in traditional terminology. There is the arbitrary proper name interpretation: Morgenstern as a family name. But examples such as the unmarked Valfisoh show that there is a certain semantic inertia and historical continuity in spite of advancement of learning about external references. The very existence and semantics of the English word chicken. blocks interpretations which are available although strictly unconventional for the German word-for- mation Hühnerhaus. for example. The modern star planet distinction may account for the fact that Morgenstern is now rather archaic in this sense. The German formations Kernhaus ('apple core'). and there is Morgenstern as a star-shaped war- club still used in the 17th century. where it can be proved matheratically that there is no reference. Constructing senses for single word-formations in isolation is no routine task. 255 in this example the compound Warenhaus. although for quite different reasons. in constructing various senses for word-formations presented to them in isolation. Brought to you by | UCL . Again. especially in eating contexts (roasted chicken versus *roasted hen). whose head is set with spikes. Speakers have no difficulty. though. this is why they are unknown in English. for the sense to be constituted. Such an inter- pretation disregards all structural hints of the word-formation itself. We could make up word-formations in such a way that they are either contradic- tory or plainly absurd. But word- formations as such are much too general and abstract semantically to specify their sense autonomously. all three examples would never be admitted as core members into the set of 'representative' cases for house.University College London Authenticated Download Date | 5/9/18 6:49 PM . namely 'a restaurant specializing in chicken'. A speaker who never encountered the formation hen-house. Both in English and German there are two further senses.

and bottles. which lists English word-forma- tions with ware in chronological order. decanters. glasses. (The New Yorker Nov 23. This is even more striking if we compare the English list in Table 1 with the German lists in Table 2 and 3 which show a much more balanced distribution. with the requisite magnify- ing glasses. In Table 1 there are earlier senses for hardware and software. whereas the former sense of soft wares is probably completely lost.) 3 things used in operating a computer: software (= programs). As expected. there seems to be no alternative to the conclusion that the English word ware is more restrict- ed semantically than its German cognate. is on an English kick. MDst prominent in this list is the formation warehouse. etc. usu. constructions with ware as their first element are surprisingly rare. Whereas the 16th century has many formations with corresponding German Brought to you by | UCL . of course. esp. and some are reproductions. At the same time. computer-oriented "sense" is. e t c . knives. It even escaped the editors of the supplement to the OED (IV: 332). glasses. and more recent formations. There are quite a few undated word-forma- tions/ which were not included. There are 12 entries with warehouse as a constituent. The dates of first occurrence are taken from the Oxford English Dictionary. there is some obsolescence and the third. 1987: 105) Instead of arguments based on single citations it is quite profitable to study sense conventions historically as in Table 1. The time perspective is quite reveal- ing. too recent to be documented in the OED. for use in the home: glassware (= glass bowls. etc. ) / silverware (= silver dishes. (page B14) The entry for the single word wares is very specific in a similar way: "small articles for sale.University College London Authenticated Download Date | 5/9/18 6:49 PM . In the case of hardware there seems to be a certain semantic continuity. tend to overspecify the sense of word- formations.256 4. too. not in a shop: a pedlar's wares" (1987: 1185). knives. Dictionaries. Even a cursory glance shows that there are many instances of word-formations not covered by the three 'dictionary senses'. Thus the list is not comprehensive. some of its wares are old. It is easy to find citations with expensive wares sold in stores: (A new store) It. tableware (= plates. Although it would be misleading to compare compounding in English and German on the same scale. (infml) liveware (= people who operate computers). The 1987 edition of the Dictionary of Contemporary English gives three dictionary senses for word-formations with -*ware as their second element: 1 articles made of the stated material. even modern ones.) 2 articles used in the stated place for the preparation or serving of food: ovenware (= dishes for use in the oven).

Brought to you by | UCL . black. The use of colour wards as a first element is restricted to Weißjare. and there are some of and for constructions instead of compounds. 257 formations this is less so later on.University College London Authenticated Download Date | 5/9/18 6:49 PM . brown. The underlying conventions can be studied in a system- atic way by comparing the 'morphological families' of each word. A partly explanation for the restricted use of ware in English can be seen when translation correspondences are analysed. as first element. English has yellow. Weißwaren in German. Goods is most often used for German Waren. or red. Table 4 lists sane of these.

(eds. Bonn. 1933. Dümmler. Nobel Symposium 39. Ann Arbor. Grirtm Jacob. Leipzig. Malkiel. Lautgesetze1 und Wortgeschichte. Joachim. Ge- burtstag von Hans Eggers. Henriette Pascce. Bertil Malmberg (eds. 1971. 1978. Leonard. Glossa 1. 1978. München. Yakov. Zu dem Satz 'Jedes Wort hat seine eigene Geschichte . Holt. EDW: A reference work of various types of formal correspondences between English and German words. 9. Lohnes. Deutsches Wörterbuch in 3 Bänden. In: Walter F. Gustav Korlen.University College London Authenticated Download Date | 5/9/18 6:49 PM . W.). Deutsches Wörterbuch. Wigand. Steiner. Wandruszka. Gustav. Karoma.258 REFERENCES Bloanfield. Goseriu. Festschrift zum 75. Laitenberger. Stockholm. Wiesbaden. München. 37-48. Hirzel. Wandruszka. 1892. Theory and practice of translation. Each word has a history of its own. W. 1971. Schwierigkeiten des Englischen. Eugenio.}. Fix. Mich. 1977. Language. 137-149. Helmut. Über Sinn und Bedeutung.). 1976. 1860. In: Liliebill Grähs. 1985. Grähs. anfassende Darstellung des lebenden Englisch. Mario.). Christmann. Bern. New York. 13.) Berlin. Wolf-Dieter Stempel (eds. Geburtstag.). In: Eugenio Ooseriu. 1982b. Band. Neuhaus.). Südwest. The contrastive grammar of English and German. 1982a.). Edwin A. Hueber. Pascoe. Lang. Wilhelm Grimm. Daniel.). Edwin A. Hopkins (eds. (eds. Sept. Joachim. Fink. 25-5O. Walter F. Brought to you by | UCL . 213-234. Festgabe für Julius Wilhelm zum 8O. Lohnes. Mario. Hugo. München. NF 100.) f 111-124. AQ-Verlag. 2 Bände.. Krüger. Die 'falschen Freunde' des Übersetzers. H. Zeitschrift für Philosophie und philosophische Kritik. In: Hans Fix et al. 53-77. (ed. Fehler und Mißverständnisse bei Gebrauch und Übertragung. 1977. In: Hugo Laitenberger (ed.). Mackensen. 1922. Sanders. Hopkins (eds. Englisch-Deutsche Wortentsprechungen: Plan für ein systematisches Wörterbuch. Neuhaus. 6-1O. Liliebill. by Martin löpelmann. Wörterbuch der deutschen Sprache. Hans et al. Leipzig. 1982. H. Gustav Korlen. Lutz. Sprachfallen im Englischen. 145-151. Bertil Malmberg (eds.(ed. 1967. Sprache und Geschichte. Gottlob. Festschrift für Harri Meier zum 65. 1982. 1977. Graham. Dudweiler. Frege.. Geburtstag. 1928. Wolf-Dieter Stempel (eds. 'Falsche Freunde': Ein linguistisches Problem und seine Lösung. Sprachen und Computer. Juli 1982.

1851 software 1612 halfpenny ware 1857 agate ware 1612 penny ware 1859 warehouseful 1615 warehouse-room 1859 woodware 1617 Nürnberg ware 1859 woodware 1617 small-ware 1860 silverware 1634 chinaware 1860 tinware 1635 warehouseman 1863 jasperware 1638 porcelein ware 1866 pottery ware 1655 ware trash 1872 Satsuma ware 1659 ware-man 1881 crackle-ware 1673 earthenware 1885 willowware 1682 hollow ware 1888 warehouse-knife 1683 warehouse-keeper 1889 flatware 1683 Brazen ware 1894 bisque ware 1683 stoneware 1894 crockery ware 1698 ware-barge 1894 Parian ware 1699 Japanware 1895 granite-ware 1707 garden-ware 1915 warehousage 1713 home-spun ware 1930 stemware Brought to you by | UCL . obs. 1799 warehoused 1508 warestall. obs. obs. obs. 259 Table 1 English word-formations with ware in chronological order 1000 ware 1714 Delftware 1300 watresware. 1836 brownware 1554 grocery-ware 1837 Italian warehouse 1562 Field-ware 1838 warehouse-door 1563 hale ware 1838 Tunbridge ware 1577 whiteware 1842 dye-ware 1591 whole ware 1849 ware-basin 1592 lady ware. obs. obs. 1723 hardware 1349 warehouse 1727 wcodenware 1398 ironware 1745 glassware 1399 fell-ware 1748 Millinery ware 1400 wormesware.University College London Authenticated Download Date | 5/9/18 6:49 PM . obs. 1732 1832 redware 1533 ware chamber 1832 table-ware 1550 fed ware. 1761 Dresden ware 1408 Courseware 1768 pebbleware 1429 mercery-ware 1782 queen's-ware 1435 lentrinware 1784 Staffordshire ware 1436 peltry-ware 1795 warehousing 1465 plough-ware. obs. 1797 mottled ware 1477 haberdash-ware 1797 yellow ware 1488 codware 1798 Wedgwood ware 1489 ware almery 1799 warehouse-rent 1499 ware-doth. 1572 1811 ware-room 1515 haberdasher-ware 1825 lustreware 1515 hardwareman 1832 black ware 1531 poultry-ware. obs. 1722 kitchenware 1300 windesware.

26O Table 2 German word-formations with Waren as a first element Ware Warendurchzug Warenkorb Warenstand Warenabgabe [Zoll] Wareneinfuhr Warenkredit Warenstempel Warenabkommen Wareneingang Warenkunde Warensteuer Warenabsatz Wareneinheit Warenladung Warenstube Warenabteilung Wareneinkauf Warenlager Warenstück Warenakkreditiv Warenerzeugnis Warenlagerbuch Warentausch Warenakzept Warenerzeugung Warenlieferung Waren tisch Warenannahme Warenfälscher Warenliste Warentermingeschäft Warenanschlag WarenfaB Warenmagazin Warentest Warenartikel Warenforderungen Warenmakler Warentransport Warenaufmachung Warenfracht Warenmarke Warenumsatz Warenaufnahme (Inventur) Warenführer Warenmarkt Warenumschlag WarenauFzug Warengattung Warenmasse Warenverbot Warenausfuhr Warengeschäft Warenmesse Warenverbrauch Warenausgangsbuch Warengesellschaft Warenmuster Warenverkauf Warenaustausch Warengewölbe Warenniederlage Warenverkehr Warenballen Warenhalle Warenpack Warenvermenger Warenbaum (Webstuhl) Warenhandel Warenpapier Warenversand Waren bazar Warenhändler Warenpfandschein Warenversandgeschäft varenbeladeo Warenhandlung Warenposten Warenverschluß Warenbericht Warenhaus Warenpreis Warenversender Warenbeschau Warenhausdieb Warenpresse Warenvertrieb Warenbestand Warenhausdiebin Warenprobe Warenverzeichnis Warenbestätter (Spediteur) Warenhausdiebstahl Warenraum Warenvorrat Warenbesteck (Behälter) Warenhauskette Warenrechnung Warenwage Warenbestellung Warenhaus Steuer Warenrechnungsbuch Warenwechsel Warenbezeichnung Warenkammer Warenrest Warenzahlung Warenbezug Warenkasten Warenschau Warenzeichen Warenbörse Warenkenntnis Warenschein Warenzeichenschutz Warenbreite Warenkiste Warenschiff Warenzettel Warenbrett Warenknappheit Warenschild Warenzoll Warenbuch Warenkontingent Warensendung Warenzug Warendiener Warenkonto Warensortiment Warendurchfuhr (Transit) Warenkontrolle Warenspeicher Brought to you by | UCL .University College London Authenticated Download Date | 5/9/18 6:49 PM .

or Waren as second element Agrarware Geldware Marketenderware Selchwaren Apothekerwaren Gemischtwaren Marktware Siderolithware Ausfuhrware Gewürzwaren Massenware Silberware Ausschußware Glasware Materialware Simplexwtre Austausch wäre Grünwaren Messerschmiedware Spezereiware Backware Gummiware Messingware Spezereiwaren Ballenware Gürtlerware Metallware Spielwaren Bandware Gu&eisenware Meterware Sport ware Bannware Gufiware Metzerware Stahlware Batzenware Hafnerware Mischware Standardware Baumwollware Halbware Mittelware Stapelware Bückware Hamsterware Modeware Stückware Blechware Handelsware Musterware Stein ware Bleiware Haushaltwaren nasse Ware Strickware Bronzeware Holzware Nürnberger Waren Strickwaren Chinaware Importware Papierwaren Strumpf wäre Dauerbackware Industrieware Partieware Stuhlware Dauerware Interlockware Pastaware Softwaren Delftware Irdenware Pelzware Tabakwaren Delikateftware Irdenwaren Pfennigware Talmiware Doppelware italienische Waren Portefeuilleware Teigware Dutzendware Jasperware Porzellan wäre Teigwaren Eierteigware Juwelierwaren Primaware Textilware Einfuhrware Kettenware Putzware Tonware Eingangsware Kettware Qualitätsware Töpferware Eisenware Kleinverkaufswaren Ramschware Transitware Eisenwaren Kolonialwaren Ränderware Trikotware Ellenware Kommissionsware Raucher wäre Trödelware Eftwaren Konfektionsware Rauchwaren trockene Ware Fabrikware Konsignationsware rauhe Ware Tuchwaren Fangware Korbware Rauhware türkische Waren Farbware Kramware Retourware Walkware Feinrippware Kulierware Schamotteware Webwaren Fertigware Kunstware Scharnierware Wedgwoodware Fertigungsware Kupferware Schleichware Weißware Fettware Kürschnerware Schleuderware Weiftwaren Feuertonware Kurz wäre Schmuckware Wirkware Filetware Kurz waren Schmuggelware Wirkwaren Filzware Lack wäre Schneidware wollene Ware Fischwaren Lederware Schnittware Wollenware Fleischware Lederwaren Schofelware Wollware Fleischwaren Leinen wäre Schreibware Wurstware Flintware Lokoware Schreibwaren Wurstwaren Galanterieware Luxusware Schuhwaren Zuckerware Galanteriewaren Mangelware Schundware Zuckerwaren Gebrauchsware Manufakturware Seidenware Zündwaren Gefrierware Markenware Seidenwaren Brought to you by | UCL . 261 Table 3 German word-formations with Ware.University College London Authenticated Download Date | 5/9/18 6:49 PM .

pattern Warensendung consignment of goods Warensortiment assortment of goods Warentest goods test Warenumsatz goods turnover Warenumschlag movement of goods Warenverzeichnis goods.University College London Authenticated Download Date | 5/9/18 6:49 PM . credit on goods Warentermingeschäft commodity futures trading Warenmarkt commodity market Warenkenntnis merchandise knowledge Warenkunde merchandise knowledge Warenverkehr merchandise traffic Teigwaren farinaceous products. knitwear. specimen. stationary Apothekerwaren pharmaceutical goods Gewürzwaren spice goods. knitwear Wirkwaren knit goods. textiles Warenaufzug goods lift. doths. spices Web waren woven goods. drapery Galanteriewaren fancy goods Gemischtwaren general goods. list of Warenakkreditiv commercial letter of credit Warenkredit commercial credit. pasta Fleischwaren meat products Rauchwaren tobacco products Tabakwaren tobacco products Warenbörse produce exchange Warenhausdieb shoplifter Warenhausdiebin shoplifter Warenhausdiebstahl shoplifting Warenhaus department store Warenhauskette chain of department stores Warenlager storehouse. groceries Strick waren knit goods. produce Tuchwaren dry goods AB. depot. stock Warenakzept trade acceptance Warenabkommen trade agreement Warenwechsel trade bill Warenforderungen trade debtors Warenbezeichnung trade description Warenzeichen trademark Warenzeichenschutz trademark protection Brought to you by | UCL .262 Table 4 Cannon translations of German compounds with Waren as first or second element Kolonialwaren colonial goods. freight elevator Warenaustausch exchange of goods Warenbestellung orders for goods Wareneingang goods received Wareneinheit units of goods Warenknappheit shortage of goods Warenkontingent goods quota Warenkonto goods account Warenkontrolle goods inspection Warenlieferung goods delivery Warenmuster sample of goods Warenprobe sample of goods. hosiery Leder waren leather goods Papierwaren paper goods.

University College London Authenticated Download Date | 5/9/18 6:49 PM . 263 SECTION 5 Brought to you by | UCL .

The traditional definition of an idiom as "the way of expression typical of a person or a people in their use of language" (DXE). and that their opacity will increase with the number of constituents which cease to maintain their inde- pendent meaning in a complex phrase. because he does not measure up to their imposed standards" (ODCIE 2). Their division according to subject areas.. idioms and phrases are an integral part of every language. has given rise to the vague idea that idioms are a clue to the 'national character1 or even to the 'genius' of a foreign language. black market is more trans- parent and less idiomatic than black sheep (of the family). Generations of foreign language teachers have tried to rake the pupil use the right idiom in the right situation. Introduction As a natter of fact. For advanced pupils. whereas black sheep refers to a member of a family or other group "who is thought to be a disgrace to other members . and in particular an idiosyncratic feature of its wordstock. A subtler distinction can be reached by making the learner aware that sore idioms are more opague than others.. The same holds for to spill the beans ('to Brought to you by | UCL . word idioms like a skeleton in the cupboard ('a shameful family secret1) or born with a silver spoon in one's mouth ('born in affluent circumstances') in fact give no clue to their complex meaning by the meanings of the individual words. or at least to its stylistic peculiarities. speech acts or on lexico- logical grounds according to parts of speech will provide a possible method in the initial stage. At first sight. the foreign language teacher will feel the need for a classification of idioms and phrases. Thus. because blaakmarket refers to an illegal transaction or some inofficial trade or 'market'.264 ROSEMARIE GLfiSER THE GRADING OF IDICMATICITY AS A PRESUPPOSITION FOR A TAXONOMY OF IDIOMS 1.University College London Authenticated Download Date | 5/9/18 6:49 PM . and for university students in particular.

specific character of this. however. as it is often used to describe Brought to you by | UCL . but differ in placing priority and explicitness. in the IDCE (1978) and the CDEL (1985) agree in the scope of meanings of this poly- semous lexeme. and also for idiomatic sentences such as Make hay while the sun shines. the linguistic dictionary by Crystal (1983)). whereas the QAIDCE gives priority to "language of a people or country.g. or to one individual". period. 3. the definitions of the word 'idiom1 listed in the QMJXE (1974). 4. Fools rush in where angels fear to tread. one peculiar to a country. the IDCE and the CDEL quote a 'phrase' or 'word group1 which means something different than the meanings of its constituents. 265 tell a secret. linguistic usage that is grammatical and natural to native speakers of a language. the third meaning refers to specialist/technical vocabulary. line and sinker ('completely'). The term 'idiomatology'. but have a different constituent structure and also vary in their degree of idionaticity. however. The fourth meaning is nearly synonymous to that of style. i. group of people. 2. roughly corresponds to what Soviet and Eastern Euro- pean linguists have called 'phraseology1. hook. usu. All these phrases are idioms. as for example (It was raining) oats and dogs. Under the key word 'idiom'. on the other hand (Hartmann 1981). which. It's six of one and half a dozen of the other. Cftly the CDEL quotes the abstract noun'idiomaticness'. It is the term 'idiomaticity* that is by now an estab- lished term for the semantic property of an idiom (Fernando/Flavell 1981). the linguistic description of set expressions whose meaning cannot be derived from the meanings of their parts. The word 'phraseology'. the characteristic artistic style of an individual. the characteristic vocabulary or usage of a specific human group or subject. is not a cur- rent linguistic term.University College London Authenticated Download Date | 5/9/18 6:49 PM . etc. district. unintentionally'). a group of words whose meaning cannot be predicted from the meanings of the constituent words.e. is itself ambiguous. in the first place. Ihe word 'idiomatic!ty1 is absent in the three dictionaries altogether. school. Definition of idiom In general. 2. whereas the term 'idiomaticness' is sometimes used for the same phenomenon (cf. The definition given by the CDEL covers the whole range of meanings (sememes) of the semanteme 'idiom' and clearly exceeds the above men- tioned in detail: 1. e. The first and second meanings pertain to'idiom1as a phraseological unit.

semantic markers or semes) carried by the constituents that form the word group.266 1. This phenomenon of idiomaticity has often been described as 'transferred mean- ing' (Klappenbach 1968). but with a literal meaning. Obviously the meanings of the constituent words wet and blanket contain not a single component which might serve as a clue to the trans- ferred meaning of this word in the idicm under analysis. which could result in overcoming zero equivalence in linguistic terminology.) for further details). reproducible word group which has semantic and syntactic stability and whose meaning cannot be derived from the meanings of its constituents. Gläser (1986: 31ff. The denotational meaning of an idiom is characterized by a specific choice and combination of semantic components (i. or 'exosememic meaning' (Pilz 1978). If the term1 idicm1 may pass as a generic term and a functional equivalent of the term 'phraseological unit1. whereas its sty- listic connotation is that of the informal. Its expressive connotation implies a negative value judgment as derogatory. The unpredictability of the meaning of an idiom from the meanings of its constituents may be illustrated by the example wet blanket 1 (informal) a person whose low spirits or lack of enthusiasm have a depressing effect on the others'. and also 2.e. Quite a number of idioms have homonyms as counterparts.University College London Authenticated Download Date | 5/9/18 6:49 PM . These are phrases with the same constituent structure. the linguistic subdiscipline of lexicology which studies and classifies set expressions(phraseological units in the broadest sense). Thus the idicm to Brought to you by | UCL . It is worth noting that in addition to its denotational idiomatic meaning. the term 'idiom1 may be defined and explained as follows: An idiom is a lexicalized. the phrase wet blanket has expressive and stylistic connotations. The idiomaticity of a word group is the result of a diachronic process of idicmatization which can only be recon- structed by an extensive sample collection and interpretation of context mean- ings and of the cxranunicative functions of a set expression compared to its possible variants. which is rooted in British and American linguistics. 'anomalous mean- ing'. Against the general and rather sketchy background of the phraseological unit. the term 'idiomatology'. colloquial level (cf. but have been added 1 from outside'. An idiom may comprise such semantic components as have no representation in the semantic components of the word group at all. 'isolated meaning' (Schippan 1975). may be regarded as a functional equivalent of the continental term "phraseology1. the inventory of phrases or set expressions/ and not only idions.

Mstaphor and metonymy may be understood as archetypes of idiomaticity. or an essential part may stand for the whole (e. however.'livelihood'). each criterion representing a single property of this complex semantic phenom- enon. as plain as a pike- staff "very easy to see or understand' (pikestaff was originally paokstaff "a stick on which a traveller supported the bag he carried with him when he stopped to rest. idiomaticity may have its roots in language history and be interpreted as an archaic. which has a literal meaning and may refer to an everyday action. In a similar sense. or marring an effect in order to avoid spending a trifling amount1. ordinary syntactic structure (or syntagm). of failing to achieve one's purpose by trying to save in a small detail. reminiscent of the way of punishing disobedient soldiers). Another salient feature is that most idioms are based on the semantic relation of metaphor or metonymy.g. In their critical study on idiom and idiomaticity. to run the gaunt- let . the word ship in this context was originally sheep (pronounced ship in many parts of Qigland) and hints to the habit in rural areas of protecting a sheep's wound against infec- tion with a patch of tar (Ridout/Witting 1969: 99). and not yet dry piece of thick cloth. Thus a gesture may stand for the whole action as in the idiom to voll out the red carpet. Fernando/Flavell (1981: 19) have suggested that idicmaticity can be best defined by a bundle of criteria. a wet blanket could as well be a newly cleaned. Fleischer (1982) describes such examples as 'phrase-internal meaning' ('wendungsinterne Bedeutung1).University College London Authenticated Download Date | 5/9/18 6:49 PM . such as cleaning the floor or applying fitted carpeting. and postulate five properties of idiom: Brought to you by | UCL . Such idioms are e. whereas in the case of metonymy it is that of contiguity. In the remaining cases. 267 voll out the red carpet 'to give an enthusiastic and carefully prepared welcome1 has a coexisting parallel structure. criticism1. They hold the idea of "varying degrees of idiomaticity correlating with different types or categories of idiom". anger.'to risk danger. an instrument may stand for the action (e. The semantic relationship underlying a metaphor is that of identity.g. It would be worn plain (smooth) from continual use1 (IDEE)) and It is no use spoiling the ship for a ha'p'orth of tar 'the folly of false economy. This parallel morpho-syntactic structure. opaque meaning which is inconsistent with the se- mantic system of present-day English. is no longer a set phrase but a loose. bread and butter . The very fact that sate idioms have literal counterparts is an impor- tant criterion for grading idiomaticity.g.

unless the construction has an intentional literal meaning. to make s. Kunin 1986).to make s. to spill the beans . Exceptions are.University College London Authenticated Download Date | 5/9/18 6:49 PM . as thick as a cable/hail- stones /blackberries. Such tests may lead to the addition.o. (2) an idiom is a unit that either has a homonymous literal counterpart or at least individual constituents that are literal. (Fernando/FlaveU. E. fear. the resulting construction will be grammatically cor- rect and empirically sensible. Permutation test Especially irreversible binomials prove their stability as idioms in that a Brought to you by | UCL . chilled/frozen to the bone/marrow. 's blood run sold ('to cause s.hook and sinker (only part of fishing tackle). e. (4) idioms constitute set expressions in a given language.o. (3) idioms are transformationally deficient in one way or another. line and sinker .g. lexicalized variants like a holy/sacred cow. 3. One transformational defectiveness of idioms The semantic and syntactic stability of an idiom can be convincingly proved by applying a number of syntagmatic and paradigmatic tests which reveal restric- tions in the syntactic behaviour of idioms as opposed to non-phraseological lexical units.'s blood run.o. E. (5) idioms are institutionalised. 1981: 17) These criteria corroborate the general characteristics of an idiom which have also been pointed out by other lexicologists (Makkai 1972. though the expres- sion as a whole would not be interpreted literally. but it will cease to be an idiom. elimination. however. Augmentation test This test may include the adding of an adverbial of degree or other lexical constituents. Elimination test The deletion of a necessary constituent of an idiom will result in a mere syn- tactic structure.268 (1) the meaning of an idiom is not the result of the compositional function of its constituents. the idiom wet blanket does not allow the change a very wet blanket. spill the peas (in the kitchen). or terror1) . substitution or reshuffling of constituents of set expressions.g. As soon as these practical procedures are followed. to feel distress. Substitution test By replacing a constituent of an idiom by an arbitrarily chosen word the mean- ing of the idiom will be entirely changed and lead to a syntactic structure.

This property of idioms has been described as their 'transformational deficiency' (Fernando/Flavell 1981) and elaborated by the following criteria: a) Blocking of predication α sleeping partner.. to spill the beans .Reconstitution: to keep one's uord.a wetter blanket (unless referring to drying one's washing) - the wettest blanket at our party (unless this is a deliberate play with idio- matic meaning).g.His kicking (of) the bucket was quite unexpected. hook. line and hook. 4. His scale of idionaticity combines semantic and syntactic criteria and starts in a 'bottom-up direction' with free combinations of words (L stands for ' Lexeme'). to let the cat ouz of tie Brought to you by | UCL .the bucket was kicked (unless in the kitchen). b) Blocking of the formation of comparative and superlative a wet blanket .('a person who provides a share of capital of a business but does not share in the management1) . Eraser's 'frozenness hierarchy1 (197O) has been derived from the transformational behaviour of idioms which ranges from completely frozen idioms whose meaning appears opaque and unmoti- vated. The grading of idiomaticity The syntactic and paradigmatic properties of idioms have been the basis of var- ious attempts at a gradation of idiomaticity. e. L5 .the/some beans were spilled at a party. since boyhood1) - boy and man 0 two males'). to unrestricted. old-fash. c) Blocking of nominalization transformation to play a waiting game ('informal. man and boy ('informal. d) Blocking of passive transformation or passivization to kick the bucket . It covers the following types: L6 .the playing of a waiting game. to delay one's action until the conditions are favourable') . door. to do homage. In addition to these restrictions to syntactic changes there exist a nunber of paradigmatic restrictions which in fact block syntactic transformations of idioms as opposed to other lexical units where they can easily operate. free collocations.the/a partner is sleeping.University College London Authenticated Download Date | 5/9/18 6:49 PM . asleep (unless this transformation is deliberate for making a scurrilous joke).Unrestricted: to close a. the same holds for to kick the bucket . line and sinker - sinker. this level contains no idioms. 269 pemutation of their constituents is blocked.

transparent expressions: to out wood.27O These semi-fixed expressions allow the passive and nominalization transformations. These phrases permit the formation of the gerund and nominalization. semi-transparent phrases: to skate on thin iae. which the authors have tested with questionnaires given to various target groups of informants.Permutation: to bring down the house. to break eggs. to give birth to.Insertion: to give ground to. on the other hand. to put on weight. L3 . These phrases permit passivization and the insertion of an adverbial of place. Their scale of decreasing idicmaticity is a "top-down approach1 as compared to that suggested by Fr ser (1970) and Fernando/Flavell (1981). to add fuel to the fire^ etc.University College London Authenticated Download Date | 5/9/18 6:49 PM . These are no idioms at all because their meaning can be derived from the constituents. as metaphorization - along with metonymy . L2 . to give the benefit of doubt to. Here constituents may be added. to teaah new tricks to an old dog.Extraction: to hit the nail on the a general property of idicmaticity. opaque phrases: to pull s. semi-opaque phrases: to burn one's boats. tarred with the same brush. Fernando/Flavell (1981). but are not completely unintelli- gible. to pass the buck. 2.Completely frozen: to bite off one's tongue. the authors of the ODCIE 2 (1983: ΧΕΙ f f . It is obvious that the 'scale of idiomaticity1 suggested by Fernando/Flavell (1981: 28) is rather vague in its main categories. because it has recourse to language feeling. These are metaphors and have a counterpart with a literal meaning. From a more practical angle of lexicography.Adjunction: to burn the oandle at both ends. A reshuffling of the constituents is possible. They distinguish between: 1. Ll . These metaphors are in fact idioms. Brought to you by | UCL . These phrases are full idioms whose meaning cannot be derived from the meanings of the individual constituent words. 4. These phrases are idioms in the sense of prototype and do not permit any syntagmatic or paradigmatic changes. LO . to sit on pins and nee- dles . 3. Mareover. They use the term 'idiom1 in the British tradition as an umbrella term for all kinds of set expressions. L4 . off the top of one 's head. offer a classification of idioms which is based on the degree of motivation and thus of semantic intelligibility. f s leg. in their scale metaphors can hardly be separated from idioms (as in class 2 and 3). and may be described as free collocations with literal meanings. a pink shirt. ) suggest a taxonomy of idicms which is based on both semantic properties and the syntactic behaviour of idioms. to take interest in.

271 1. 's memory.g. so that it has been occasionally interpreted as an 'inter- level1 of the linguistic system. It is the heterogeneous linguistic material under analysis which places this linguistic discipline in a mediating position between word- formation and syntax.g. virtue. i. Weinreich's concept of uni- lateral and bilateral idioms has proved incomplete so that it calls for the category of 'multilateral idioms' which covers such nominations and propositions as a bird in the hand.g. phrases of a considerable length. Ihe place of idioms in the phraseological system Phraseology as a linguistic subdiscipline of lexicology describes and classifies phraseological units. to burn one's boats. With regard to the fact that idioms may cover multiple-word con- structions. to rub noses with sb. Pure idioms: These are 'petrified' or 'congealed 1 phras.) in which both constituents have a transferred meaning so that the meaning of the whole word group cannot be derived from that of its and thus form the end of the process of idiomatization of a fixed word group. grace.University College London Authenticated Download Date | 5/9/18 6:49 PM . They have no homonym in normal use with a literal meaning. 1983: XII). e. to beat one's breast. 'idiomatic1 meaning. Vfeinreich (1972) distinguishes between "unilateral idioms' (e. and those which have a literal meaning. 2. and Beggars cannot be choosers. to charge an account) where one constituent is self-explanatory. The authors hold the convincing opinion that "idioms are not divided as a small water-tight category from non-idioms but are related to them along a scale of continuum" (ODCIE 2. idic— maticity will be graded according to the semantic relationship of those constit- uents of a word group which have a transferred. to blow the gaff.g. sin. black market.e. Noun idioms of this type may produce variants within a certain collocational range: a cardinal error. Figurative idioms: These phrases hardly allow any variation and are in fact on the border of pure idioms: e. 3. a blind alley. The fact that phraseological units may have the character of words (nominations) and also of sentences (propositions) has re- Brought to you by | UCL . and 'bilateral idioms' (e. Each constituent of an open collocation is used "in a common literal sense". to jog one's/sb. white lie. 5. e. cold war. Restricted collocations or 'semi-idioms': They combine one constituent with a transferred. early bird. For the purpose of the taxonomy of phraseological units which will be elaborated in the following chapter in the context of a more comprehensive approach. before you can say Jack Robinson. Open collocations: These are "free" or "loose" syntactic structures which are in accordance with semantic stability of words and thus are no fixed expressions at all. once in a blue moon. Thus.g. 4. and one with a literal meaning.

Non-idiomatized phraseological units are a very important part of set expressions. however such idioms as by virtue of. Mast phraseological units among the nominations are in fact idioms. that is to say. hotly contested. and Földes (1987) include sentence-like phraseological units such as proverbs. malignant tumour so that the distinction between common language and language for special purposes becomes rather blurred. as nouns. Fleischer (1982). routine (or conversational) formulas. Non-idiomatic nouns among phraseological units include onymic units such as the New opposed to phrasal verbs which are all idioms. 1972) includes here specific technical terms such as reinforced concrete. Brought to you by | UCL . and they all share the following features: + lexicalization + common usage + reproducibility + semantic and syntactic stability ± idiomaticity ± connotations ± an emphatic. 272 suited in the ocnoept of phraseology in a narrower or in a broader sense. to put to the test . Non-idiomatic verbal set expressions also include a number of cliches such as to attempt the impossible. 1986). slogans. in order to). by dint of. Kunin (1970. quota- tions. the Middle East. itensifying. adjectives. beyond belief. on the spot. Adjectival non-idiomatic phrases include cliches like approximately correct. peaceful coexistence. and thus support the broader concept of phraseology. the idiom may be described as the prototype of a set ex- pression or phrase. to stand on ceremony. Gläser (1981. opinions differ on the non-idiomatic sector of phraseological units in nominative function. Phraseological units which are not idioms may occur among all parts of speech which have a denotational meaning. over and over again. which include 'idiomaticity' as an optional feature. There are also a few non-idiomatic adverbial phrases like beyond compare. and terminological word groups like displaced person. e. 1972. Ifcn-idio- matic verbal phraseological units are the so-called paraphrasal verbs. verbs and adverbs. but not as function words (cf. 15iis observation does not stand in contradiction to the general charac- teristics of the phraseological unit. On the whole. or expressive function in the text.g.University College London Authenticated Download Date | 5/9/18 6:49 PM . etc. to take a walk. The models suggested fay Kunin (1970. In view of its variety of types and frequency of occurrence among all sorts of phraseological units. 1986). commonplaces.

the Middle West. over and over again. although they are composed of appellative constituents (e. by dint of) and conjunctions (in order to.g. Idiomatized prepositions (by virtue of. which modem linguistics owes to the Prague School of Lin- guistics of the 1930s. They may occur as reductions or ellipses of sentence-like idioms. before you can say Jack Robinson non-idioms: beyond compare. to rack one's brains non-idioms: to take into consideration. dressed to kill non-idioms: approximately correct. processes. displaced person adjectives: bred in the bone. The area adjacent to the centre of the phraseological system is the transition area in which we find phraseological units which are nominations. born under a lucky star. the Golden Twenties. The centre of the phraseological system of modern English comprises phraseo- logical units in nominative function. adjectives. states. which again shows that idioms are proto- typical. word-like units which designate phenomena. This area comprises the following types: Brought to you by | UCL . but his system follows a slightly different approach. bread and butter non-idioms: running water.University College London Authenticated Download Date | 5/9/18 6:49 PM . in the outside world. and which has proved its longevity and applicability to a number of fields of the linguistic system. The overwhelming majority of word groups or set expressions in the centre of the phraseological system are idioms. relations. 273 Similar overlaps are obvious between phraseological units which are in fact proper nouns. seen in the perspective of synchrony and diachrony. The linguistic model underlying the following taxonomy of English idioms and phrases in the function of nominations and propositions is that of "centre1 and 'periphery'. on the spot. to have a swim. and a few propositions and con- junctions. events. the Glorious Revolution). and a few word groups which only function as opera- tors in that they designate relations between phenomena or objects. i. This model was also used by Fleischer (1982) for classifying the phraseological inventory of modem German. these are also known as function words. but at the same time parts of propositions. etc. qualities. verbs and adverbs. nouns: wet blanket. red tape. gainfully employed verbs: to go the way of all flesh. on condition that) may serve as examples. actions. to put to the test adverbs: by leaps and bounds.e. Alongside of a much smaller proportion of non-idiomatic phrases they cover nouns. objects.

the tertium comparationis can only be retrieved Sron the referential mean- ing of the verb and the connotations that accompany it. hit or miss. a rolling stone (gathers no moss).alive and kicking.University College London Authenticated Download Date | 5/9/18 6:49 PM . to hum and haw. because the tertium comparationis is present in the surface structure: as proud as a pea- cock. as the Outer circle1 of the phraseological system. 5. two or more constituents of the same part of speech are con- nected. Allusions of this kind presuppose a considerable amount of background knowledge on the part of the speaker or listener. The majority of them are non-idioms. 4. on the other hand. to breed like rabbits. Fragments or reductions of proverbs such as a new broom (sweeps clean). or states of affairs. covers phrase- ological units which are chiefly propositions. The stereotyped comparisons functioning as adjectives illus- trate the intensifying function of the simile. which depend on the given subtype of phrases. In the stereotyped comparisons carried by verbs. Irreversible binomial idioms: These so-called word-pairs or 'stereo- typed (unchangeable) set expressions' (Arnol'd 1986) or 'irreversible binomials' (Makkai 1972) have a fixed order of constituents which does not allow permutation. Allu- sions to English literature are: a Jekyll and Hyde. The periphery.274 1. e. Embedded in a sentence or a stretch of text. to throw out the baby with the bath-uater. adjectives. Stereotyped comparisons or similes: They compare two qualities or two actions. a Mrs. The alliteration may support the internal phraseological binding of the words compared. 2. verbs and adverbs. Proverbial sayings: These idioms likewise have a nominative function. adjectives: shipshape and Bristol-fashion. Examples: nouns: bread and butter. Literary allusions and fragments of quotations: These nominations are also parts of whole sentences and may associate them in the text. 3. a Catch-22 situation. by means of a figurative ex- pression.g. to put the cart before the horse. processes. The phraseological material allocated here are pro- Brought to you by | UCL . They may occur as nouns. to and fros sooner or later. these idioms associate the whole proverb. verbs: wait and see. which is often done in the appropriate situational context. Grundy. although their idiomatic charac- ter varies greatly. Examples: to fight like aat and dog. adverbs: up and down. the ups and downs. but can be easily turned into proverbs.

1986. are not idioms.A stitch in time saves nine. they are hardly ever idioms. Proverbs: These popular phrases express folk-wisdom of various gener- ations and from different regions of the language conntunity.You never can tell. We live and learn. Christmas comes but once a year. Makkai 1972. Their idiomaticity is not based on the semantic quality of metaphor or metonymy. 3. As slogans must be self- explanatory. which is often an indicator that they date from past centuries. quotations. cf.g.. Slogans: These phrases have an imperative or at least hortative charac- ter. e. internal rhymes and assonances. Gläser 1981. (ellipsis). Brought to you by | UCL . idiomatic meaning (and therefore cannot be taken literally). . -It's a small world. Most of them express a self-evident fact. Commonplaces or truisms: These phrases differ from proverbs in that they have no didactic content.Any port in a storm. Examples: An apple a day keeps the doctor away. Ask for Haig. slogans. maxims. 2. 1. but are rather entities of a socio-cultural system. Their memorability is supported by such linguistic means as allit- eration and rhyme patterns (internal and final rhyme). and are chiefly used in oral ccmnunication. Not all of them belong to the lexioon of the living language.Practice makes perfect. they are idioms. 275 verbs. comnonplaces. Value for money. Another group are truisms in a more limited sense and express a self-evident fact. . Examples: Don't be vague. Some commonplaces are even tautologies: Boys will be boys.. Others may refer to a general experience.University College London Authenticated Download Date | 5/9/18 6:49 PM . (inter- nal rhyme) . Since proverbs always have a trans- ferred. Routine formulas: These set expressions comprise idioms and non-idioms.He that is down need fear no fall. Qiite a few proverbs contain alliterations. but results from a particular fusion of the meanings of the individual constituents. Examples: How do you do? ('form of greeting when being introduced on first meeting1) and Come again? ('coll. (allit- eration) . please say what you just said again because I did not hear you the first time'). winged words. in a short. memorable and mostly figurative way.lie only live once.All roads lead to Rome. Fleischer 1982) such propositions are discussed in great detail. . 4. (assonance) . In linguistic descriptions of the phraseological system in a broader sense (Kunin 197O. commandments and routine formulas.

6. Ihe internal structure of the phraseological system is illustrated by the follow- ing diagram. Commandments and maxims: These propositions express rules of conduct in private and public life. the origin of quotations is known in most cases. Centre: Nominations II. but on em- pirical values gained in working with a corpus of idioms and phrases (Gläser 1986): The internal structure of the phraseological system of Modern English Legend: Subdivision of the phraseological system of Modern English I. Roosevelt). which is only a sketchy representation of the types of idioms and phrases discussed above. D. Transition area: Parts (reductions) of propositions III. Example: Speak softly and earvy a big stick.Thou shalt not steal. (F.University College London Authenticated Download Date | 5/9/18 6:49 PM . Ihe proportion of idiomatized and non-idiomatized phraseological units pertaining to the different types of nominations and propositionsin the whole system is not based on frequency counts. Examples: Thou shalt not kill·. Periphery: Propositions non-idioms idioms Brought to you by | UCL .276 5. You will get far. Quotations: In contrast to proverbs.

The learner's sensibility to stylistic differences (i. Last but not least. Conclusion Ihe phraseological system of present-day English as illustrated in the diagram of 'centre'. expressive and stylistic connotations) between phraseological units in the foreign language and in the mother tongue. The wealth of idioms and phrases can be brought to mind. idioms and phraseological units in the broadest sense against their socio-cultural background will provide a rich source of general education and increase the pleasure in foreign language teaching and learning. Brought to you by | UCL . the bilateral and multi- lateral idioms in the transition area. the sys- tem of word-like idioms and phrases in the centre.University College London Authenticated Download Date | 5/9/18 6:49 PM . with their varying degree of idiomatic!ty as unilateral and bilateral idioms. On the whole. and the sentence-like idioms and phrases in the periphery.e. to a sys- tematic study of equivalence relationships between idioms and/or their para- phrase. idioms and phrases lend themselves to a comparison with the mother tongue. Moreover. embedded in an analysis of their stylistic function in the text. may be developed on different proficiency levels. Cn closer inspection it consists of further transition areas and a num- ber of borderline cases. 'transition area1 and 'periphery1 is not divided into neat pigeon- holes. which can be effectively checked by the elimination or substitution tests. 277 6. These resist a strict allocation to a particular type of idiom but rather enjoy hospitality in various types. have proved teachable to students of English philology and even to non-philologists in courses in English for Special Purposes at the University of Leipzig.

Idioms within a transformational grammar. (masch. K. Aleksandr Vladimirovii?. Frazeologija sovremennogo anglijskogo jazyka. Leipzig. Exeter linguistic studies. 1986. On idiom. London. Aleksandr Vladiinirpvic'. Földes.und Sprachwissenschaftliche Reihe 17. Gesellschafts. Moskva.) MDskvai. Aleksandr VladimiroviX. Collins. Frankfurt/M. 1982.). Applied Science Publishers Ltd. (ed. Leksikologija sovremennogo anglijskogo jazyka. 1980. VEB Ver- lag Enzyklopädie. A first dictionary of linguistics and phonetics.Csaba. Izdatel'stvo 'VysSaja skola1. Diss. Bruce. Critical views and perspectives. Reinhard R. Editorial director: Lawrence Urdang. Stork. Ruth. 1981. Kunin. 1976. Hartmann.) Fräser. Andre Deutsch. 1972. Exeter. Roger Flavell. Lehrmaterial zur Ausbildung von Diplomlehrern Englisch. 22-42. K. London. Moskva. David. Brought to you by | UCL . Rosemarie.278 REFERENCES Arnol'd. Gläser. Foundations of Language. Exeter Linguistic Studies. 5. PH Potsdam. (repr. University of Exeter. Hartmann.. Crystal. 1981. 1973. 1972. Opyt sistematizirovannogo opisanija. 1986. Leipzig. 1981. Vol. Irina V. A Jena. K. 197O. Moskva. Phraseologie der deutschen Gegenwartssprache. London. Aspekte phraseologischer Äquivalenz in der ungarischen. Athenäum. 1983. Kiefer. Izdatel'stvo 'Mezdunarodnye otnosenija1. General editor: Reinhard R. 5. Ferenc (ed.). CDEL = Collins Dictionary of the English language. Gläser. 1970. Exeter. 1985. Dictionary of language and linguistics. Kunin. Fernando. Lizenzausgabe Tübingen. Wissenschaftliche Zeitschrift der Karl-Marx-Universität Leipzig. Probleme der Phraseologie. Kurs frazeologii sovremennogo anglijskogo jazyka. Niemeyer. Hartmann. Phraseologie der englischen Sprache. Phraseologie der englischen Sprache. University of Exeter. On idiom. Critical views and perspectives. (Teoreticeskij kurs. Wissenschaftlich-Tech- nisches Zentrum. 1968. Chitra. VEB Bibliographisches Institut. Glasgow. Wolfgang. Klappenbach. TJie English word.6. Kunin. Francis C. (repr. Semantik und generative Grammatik. "Vyssaja skola1. 1986.). Reinhard R.. 1987. Roger Flavell. Rosemarie. deutschen und russischen Gegenwartssprache. 'Vyssaja ¥kola'. 5: Chitra Fernando.University College London Authenticated Download Date | 5/9/18 6:49 PM . Fleischer. Anglljskaja frazeologija. 221-227.). Vol.

Kümmerle. 279 LDCE = Longman Dictionary of Contemporary English. 1984. VEB Bibliographisches Institut. Paris. QALDCE = Oxford Advanced Learner's Dictionary of Current English. Leipzig. Albert S. 1972. Vol. 1974. Harlow. LDEI = Longman Dictionary of English Idioms. Anthony Paul Cowie. English proverbs explained. Brought to you by | UCL . Oxford University Press. Lexikologie der deutschen Gegenwartssprache. Einfuhrung in die Semasiologie. Göppingen. etc. London. 1979. Editor-in-chief: Paul Procter. Uriel.) The Hague. Schippan. VEB Biblio- graphisches Institut. Longman. Ronald. London.University College London Authenticated Download Date | 5/9/18 6:49 PM . 1967. 48. (Janua linguarum. Oxford University Press. Schippan. Series maior. Begriffsbestimmung und Systematisierung unter besonderer Berück- sichtigung der deutschen Gegenwartssprache. Klaus Dieter. Editorial director: Thomas Hill Long. Leipzig. Probleme bei der Analyse von Idioms. Hornby. Thea.). 1978. 1972. 1978. 2: Phrase. Harlow. Isabel R. Landen. Ronald Mackin. London. Ridout.. Gimson. 1972. In: Ferenc Kiefer (ed. Phraseologie.. 1969. Mam. Clifford Witting. Idiom structure in English. London. Longman. 1975. Makkai. . ODCIE = Oxford Dictionary of Current Idiomatic English. 1984. Pilz. Pan Books Ltd. Thea. Versuch einer interdisziplinären Ab- grenzung. 1983. Weinreich. Clause and Sentence Idioms. etc. Alfred C. Anthony »Paul Cowie. Mouton. 415-475.

i. Since some linguistic expressions must be learned and remembered separately while others can be reconstructed on the model of some previously acquired structure (Coulmas 1985: 49). signs that primarily refer to the outside world in some way or other (Rothkegel 1973: 159 f f . paradigms of modifications with an intra- textually related meaning potential (in a rather broad sense of the concept of relational meaning). and vocabulary on the other hand.g.1 Contrary to what generative granmarians tried to postulate. memorization. the problem I intend to discuss in this paper is how to categorize idiomatic chains as distinct and distinguishable from free syntagms. 280 KARL SORNIG IDIOMS IN LANGUAGE TEACHING Ο.2 Which of the conceivable and palpable data of description should cate- gorization and organization start from? The lexicographer may make his choice Brought to you by | UCL . As a matter of experience. As a matter of fact. i.e. what teaching is all about is just this possibility and the feasibility of generalization. it has not too recently become obvious that the effective implementation of linguistic competence consists of far more than just the application of rules of grammar as it is commonly understood (Coulmas 1985: 47). O.e.University College London Authenticated Download Date | 5/9/18 6:49 PM . fluc- tuating relationships and an unstable balance obtain between what we have cone to call grammar on the one hand. linguistic)forms and strategies once he/she has acquired the pattern to which they pertain. i. for didactic purposes of storage. ) . and re- trieval. To teach somebody how to do things means demonstrating to the learner that he/she will be able to handle a considerable number of similar (e.e. the lexicon contains various kinds of syntagrratic concatenations. While descriptive linguistics may find it necessary or at least sufficient to restrict itself to rules of morphological variation and sentential corobinatory patterns and feel justified in calling the application of these the grammar of a given language.

University College London Authenticated Download Date | 5/9/18 6:49 PM . Furthermore.e. Brought to you by | UCL . surface structures. i. Lipka in this volume). lexical kind of motivation. To group words and word groups any kind of simi- larity or contiguity will do (cf. i. at least in word-formation. For one thing. set on fire. seemingly redundant and near-synonymic. regardless of the fact that they may not be so for the 'stranger1. or else intelligibility and availa- bility of a given expression. are nevertheless merely systemoid. there is a fair. or else deep structure. see a doctor. you see. predictability and instability may be misleading on various levels and to variable degrees. see a joke.3 I feel inclined to venture a few remarks here on the concept of moti- vation (Gauger 1971: 12). I see.. there is a third category of motivation or transparency. though restricted amount of innovatory creativity and productivity at work. We presume that sate similarities will serve certain kinds of purposes more effectively.e. the native tends to confound familiarity with systemic. whenever a language is used in everyday speech or for an artistic (poetic) purpose. one that would arise from the familiarity with certain preformulated tex- tual specimens and/or patterns: quotational and proverbial language use. Provided motivation is understood as the predictability and transparency of the relationship between the surface and content sides of semiotic structures. 0. viz. i. since idiomatic meaning is non-compositional: a dark horse. either re- trievability in a lexicographical sense. 281 frcm between two kinds of phenomena: features of surface structure. idioms cannot be categorized by means of one single lexical component. As a matter of course. Ihe native speaker tends to experience and regard things he is familiar with as motivated. and their transparency. Derivational patterns. Besides. fire sb. a. semantic. apparently motivated and systematic. there is no easy way out of this. or even both. fire away! Although there are always lacunae in the lexical or the derivational repertoires of any natural language system. at least one can and must not at one's own discretion try to fill up what seems to be missing. To group expressions that contain certain catchwords under the same heading would mean to ignore essential differences in meaning (Rothkegel 1973: 163). one might recognize various levels of motivation: Apart from the well-known primary motivation (onomatopoetic and sound-symbolic) and a secondary. transparency. surface resemblance of word stems is not too useful for the acquisition of communicative devices at all. Anything the native is used to because he has been using it successfully and repeatedly in communication will invariably seem self-explanatory (Wsinreich 1969). white elephant. re- spectively. need not be so in semantic or pragmatic terms.e.

Note that Fräser (197O) utilized their idiomaticity. since these are useless for any imaginable didactic·purpose Brought to you by | UCL .University College London Authenticated Download Date | 5/9/18 6:49 PM . 1. Koller 1985: 28). go west.similarly to idioms .like anything we say . like son. 282 1. The idiomaticity of so-called 'dead metaphors' is high- lighted .2 Another characteristic of idioms.may be regarded as originating in meta- phorical language use. spick and span. as right as rain. aheap as dirt. compositional meaning. Idioms may or may not have been free syn- tagmatic configurations before they got attached to a particular (contextual) meaning. either: easy come. 2. the degree of their syntactic and morphological inflexibility ( c f .3 Idioms .by juxtaposing them against the literal.1 The most prominent characteristic of idiomatic expressions (= idioms) is their holistic semantic representation. which has been extensively discussed in linguistic literature.e. viz.2 One side-product of this tendency towards preservation of formal structure are fossilized remnants or freezes (Cooper/Ross 1975). of obsolete or obsolescent surface structures: dree one's weird. is their (variable) inflexibility as to modifications and transformations (Chafe 197O. 'varying com- mutability1 in Rothkegel (1973) or 'syntactic anomalies' in Coulmas (1982: 2 6 f . i..0 Among the attempts at a categorization of idioms that start from surface structure are the following: There is a considerable number of cases where the separable parts have no meaning of their own whatsoever. 1.e. pins and needles. like father. easy go. burn the aandle at both ends. despite the fact that their surface can be segmented into constituent parts that usually can have a sepa- rate (but not the same) meaning and function in the language. Goulmas 1982: 25. still the density of these holistic semantic structures is variable (Coulmas 1985: 52). clean as a whistle. Piischel 1978: 156f. No mention will be made of alphabetically ordered lists in workaday phrase-books. which they seem to lose inside the idiomatic unit. ) ) as a point of departure for the categorization and hierarchization of idiomaticity. kith and kin. i. Specific a-grammatical phenomena are not too rare. some idicras are not amenable to any kind of modification at all. which on its part would have to be triggered off by a specific reason or occasion and would therefore necessitate an additional de- coding effort: a big noise.

4 Collocations. part and parcel. fool-proof. are learned. i. I will mention but a few: Worth mentioning is the fact that bi-nominals do not just serve various kinds of intensifying intention.e. Russ. i. a dare-devil. but very often impart . foxglove.e. moreover. in a specific order. they impart undisputed moral and pseudo-sociological knowledge and insight: Puss. from those concepts and entities in extralinguistic reality that serve as 'Bild- spender1 in a Weinrich'ian sense. least said.University College London Authenticated Download Date | 5/9/18 6:49 PM . the early bird catches the worm..2 More or less fixed attributional or stereotyped adverbial collocations very often serve evaluative functions and intents: a cheap-Jack.some stable ideological content. Gläser 1985: 70): with might and main. Obviously.e. i. It is not surprising that fields of exis- tential significance and momentum should figure foremost here. at sixes and sevens.e. rough and ready. Ihey are widely known and accepted idiologemes. 3.3 Ccnpounds are conbinations of words between which varified syntactic. come a cropper. 2. stored and processed as more or less fixed syntagms: woolgathering.more or less clandestinely . as plain as a pikestaff. Their meaning is dependent on their occurrence in combination and. soonest mended. 2. Brought to you by | UCL . 2. case relations. categorization may start from semantic content. as cold as charity. i. na posloviau i suda net ('one cannot argue against a proverb1). sometimes for the refutation of such behavioremes. t>H-V+(N)-groups. 283 2. 2. As an alternative. rack and ruin. which may not be 4 reversed (Burger 1973: 46ff. Sometimes this combination is realized contrary to sore of the commonly accepted grammatical and morphological rules: topsy- turvy.1 Bi-noninals (sometimes tautologic). kl'ua silneje zamka ('the key is stronger than the lock'). bag and baggage. null and void. such as quotations (the green-eyed monster] and anonymous sayings. obtain.5 Of course. calf-love. breath-taking. in a number of cases supported by such memorizable devices as rhyme or alliteration. the highest rank as far as inflexibility is concerned is due to proverbial text-fragments (Coulnras 1982: 33). proverbs have a clear-cut interactional function as stimuli for reactive and active social behaviour. heart-breaking.

tooth and nail.e. put a spoke in sb's wheel. know the ropes. hit below the belt. rest on one's oars. dar gli otto giorni 'sack sb.1 Cosmic phenomena (the sun and the moon. as black as thunder. keep abreast of sth. 3. pick up the threads. the thin end of the wedge.. at the dead of night.5 Obviously the most interesting examples of metaphoric and idiomatic usage are provided by fossilized elements from stages of cultural development and civilization as mentioned above. keep sth. from the frying pan into the fire (cf. hen-pecked. look daggers. vehicles. hare- brained. such as tools.g. for a rainy day. 3. be under a aloud. bagnato come un pulcino 'soaking wet1. and hence provide idiomatic expressions: e. throw in the sponge. the weather. colours. get sth. vividly represent the environmental elements that are of existential relevance for everyday human subsistence.University College London Authenticated Download Date | 5/9/18 6:49 PM . etc. very probably and plausibly various of the numerous absurd semantic elements in idioms can be explained by relating them to an origin in taboo-governed restrictions and superstitious customs and traditions. It. pay through the nose. crafts and sciences: you can't make bricks without straw.284 3. Numerous are the vestiges of by-gone stages of civilization. It. take the wind out of sb's sails. treat sb. sometimes Brought to you by | UCL .) nay serve as sources of metaphorical comparison. white-livered (cf. a fair-weather friend. i. weapons. see how the cat jumps. an iron in the fire. run the gauntlet. animals and esp. Very often they represent aspects and techniques of jurisdictional and magic institutions. expressions borrowed from the sportsman's or the warrior's jargons: e.4 Allusions to technical elements. It. avere del fegato 'courage'). fan the flames.'.2 Considerable numbers of idioms take their origin from competitive and aggressive language use. blow off steam. not to care a fig (It. keep the pot boiling.. parts of the human body. an axe to grind. by rule of thumb. non valere un fiaao secco). 3. stick to one's guns. 3. Thus. to teach one's granny to suck eggs/grope ducks. viands. high-handedly. off one's chest. and human activities) are by far the roost frequently used lexical elements constitutive for metaphoric and idiomatic expressions: e. all is grist that comes to his mill.g. artifacts. etc. of the same kidney. get one's teeth into sth.3 Animate entities (plants. go at it hammer and tongs.g. as busy as a bee. Elements with a social and cultural background: be born on the wrong side of the blanket. mettere troppa carne al fuoco). in clover. double-edged. It.

one would have to admit that. ) . even among native speakers. Consequently. moon-struck. Uncertainties. svegliare il cane ehe donne (cf. was glänzt). It. Russ. no rule can be derived from unique cases. donkey's years ago. 285 with a story behind it: hat trick. essere oome i ladri di Pisa 'lead a cat-and-dog life1. i. a little bird told me. demon- strate that it is legitimate to use notions like 'competence1 and 'acquisition1 in the area of idiomatic usage. But are we entitled to speak of 'teaching' in this realm of a language system and its implementation? Since teaching must be regarded as leading the learner towards the recurrent application of generalized rules. mistakes and errors. tocca ferro! fare la corna). Chafe (1968) is one of the few authors who recognizes the familiarity (and frequency) of a linguistic item as part of its (connotative) meaning (cf. like a bull in a china shop (G. If the competent handling of idioms is how a native proves and utilizes his/her command of the language. As a matter of fact. Es ist nicht alles Gold. absurdities. whilst anything can be learned. Span. Span. not everything can be taught. dressed up to the nines. more precisely. show the cloven hoof. let sleeping dogs lie). ne vs jo zoloto s"to blestit (G. if Apart from Coulmas' approach. It. the fact that its meaning is neither literal nor composi- tional? Still another consideration is this: What actually should induce the native to use a newly coined expression instead of another. how does he/she find out about the idiomaticity of an expression.e. es tan aomo perros y gatos. castillos en el aire).University College London Authenticated Download Date | 5/9/18 6:49 PM . Brought to you by | UCL . crocodile tears (same in G . cf. bring back the ashes. raise Cain. It. one ought to ask oneself how it is that the native him/herself acquires the knowledge and use of newly coined expressions (Olshtain 1987). 4. also 'semantischer Mehrwert' (Kühn 1985)). Elefant). play gooseberry. the crucial concern and didactic consideration concerning idiomatic elements (Coulmas 1985: 55) must be the problem of how idioms are stored in the mental lexicon and how they can be listed in the dictionary in a way that would facilitate retrieval. Kirchen- maus) . more familiar one. so that they rather belong into the category 'faux amis'. as poor as a church mouse (G. parallelisms between source and target languages are very often misleading (Mrazovifc 1985). castles in Spain (G. semantic and gramratical: by a long chalk. in the jaws of death. touch wood (It. non capire un cavolo. Luftschlösser. For one thing. rragic traditions and beliefs: black sheep.

Necessary though the description of the in- tricacies and peculiarities of idiomatic surface structure nay be. by keeping one's ground and carrying one's point in the presence of one or several interlocutors. this time with respect to their communicative functions. Generally speaking... Brought to you by | UCL . cf. they are not remembered because of these peculiarities (due to the fact that these function contrary to the system.convince him / seduce her. labelling in collections of proverbs and common sayings is usually quite obviously (and plausibly for that matter) oriented along topics like the following: love. the tactics of impressing people and depicting things in a memorable way must be learned and are learned from the management of situations (Glaap 1985: 96f. cf. how to get along with your neighbour. Learning a language takes place in and by successfully handling communicative situations.e. we shall venture a few suggestions towards a categorization of idiomatic expressions. etc. truth and falsehood. In ccnniunicative interaction. friendship and animosity. not just to express an opinion.University College London Authenticated Download Date | 5/9/18 6:49 PM .hope- fully ..286 it were not for its higher expressive intensity. chatting up sb. they are remembered and memorized by virtue of their situational impressiveness and ccnmunicative effectiveness or ". etc. Metaphorical as well as quotational language constitutes and requires compe- tence. a need that invariably arises is this: not just to call a spade a spade but to do it in a colourful way. Interestingly. or delicate situations as asking for help. the implementation of words beyond their 'truth value1. albeit at the price of its transformational inflexibility. but to give one's interlocutor a piece of one's mind and . etc. i. i. an aspect still badly in need of further research and blatantly missing from most of the collections of idiomatic usage (for a laudable exception.). Seme of the means serving these purposes are expressive and metaphorical language elements. also Cruse in this volume). making excuses.e. 5. the examples given in phrasebooks and the situations described are most frequently concerned with such quandaries and embarrassing. possibly on a higher and more intricate semiotic level. Friederich (1976). etc.0 With didactic feasibility and effectiveness in mind. all 'errors' in this domain of language com- petence tending towards systemic adequacy). wsil ihre Bedeutung primär sozialer Art ist" (Koller 1985: 30). Moreover..

monologically . witticisms and the like (any of the examples mentioned above might just as well be listed here). in un lampo = 'at once1. the tail wagging the dog. flog a dead horse. talk 19 to the dozen. as treated in Coulmas (1981. acqua in bocca (keep one's silence). etc. make one's hair stand on end. characterization as well as evaluation Generally speaking. idioms have intensifying functions here: It. where angels fear to tread .very roughly .distinguish categories like the following: 5. It. lead sb. It. make one's flesh areep/one's mouth water. not care a/tao hoot(s)/a pin. 5. It. greetings. freddo da diavolo/da lupi/da cani ("extremely cold1). have a slate missing/a screw loose..4 Quotational and proverbial elements that may serve various expressive ends. get one's goat/monkey up.represent attitudes and/or signal the emotions of the speaker him/herself (of course. It. have one's heart in one's mouth/boots. have a bee in one's bonnet. be in a brown study. 287 Ihis perlocutionary aim is most effectively realized by idioms. fare I'indiano (pretend not to know). as merry as a cricket. 5. formulas. It. be in a blue funk. talk the hind legs off a donkey/horse/cow/dog/bird/brass pan. from a pragmatic. up to the eyes. dead as a door-nail. point of view we could .3 Affective signals are partner-oriented dialogical ccttinunicative moves. lose one's hair. enamorado hasta los ojos = up to the ears in love. 1985: 64f.0. make a clean breast of sth. even drastic..). 5.e. frequently directive. no room to swing a cat. "seven shirts'). heart and soul. Span. i.2 Patterns of utterances that .. 5. inamorato cotto (G. even objectionable and invective language: It. spill the beans.. una buona lana (a sly dog). second Brought to you by | UCL . turn a deaf ear to sb. They signal the assessment of a situation and role-relationship and try to impress the interlocutor by the use of strong. It. break a butterfly upon a wheel. tongue in cheek.University College London Authenticated Download Date | 5/9/18 6:49 PM . preferably ideological ones.). sudare sette camiaie ('sweating hard1.1: Descriptive lexical elements that refer to extra-linguistic reality and aim at its expressive. sometimes with an argumentative aim (Koller 1985: 30ff. Bius. proverbs. green with envy.. such as stereotypes. functional. 1982. up the garden path. bis über beide Ohren).1 Phatic signals. swear words and any four-letter-expression will serve the purpose admirably well): make one's blood boil. toacare il aielo col ditto (be in high spirits).

Russ. Yet. sow one's wild oats. Beg ne "cesten.. the breaking up and re- activating of fixed syntagms (Glaap 1985: 103) leads towards renewed trans- parency and insight. the darkest place is under the aandle stick. da zdorov (flight may not be honourable3 but it aan save you). Ironic play upon idions has become very popular in advertising and subcultural literature (''to spray or not to spray .. I am not sure whether this realm of creative modification of fixed expressions should not be reserved to the native speaker exclusively.").288 childhood. Brought to you by | UCL . Thus. apple of discord.5 Beside the fact that the native's competence is restricted and not free fron errors and despite the inflexibility mentioned above there is pro- ductivity and innovation on all levels and all the time. 5.University College London Authenticated Download Date | 5/9/18 6:49 PM .

Koller. Sprache und Literatur 56.). München. Ein Stein des Anstoßes. Gläser. Papers from the parasession on functionalism. Niemeyer. 11 Oct. 95-1O4. Athenäum. Ausgewählte Probleme der Idio- matik. Skating on thin ice: Literal meaning and under- standing idioms in connections. Raymond W. Grossman.. Idioms within a transformational granner. 17-36. Kühn. Hockett. Gibbs.Besseres Englisch? Zu einem vernachlässigten Bereich des fremdsprachlichen Unterrichts. L. Timothy J. 1973. Wallace. 1970. 1956. 1985. The Hague. 1956. Coulmas. 1976. L. Glaap.Hueber. Werner. 1981. 1975. Gauger. Sprache und Literatur 56. 1966. Florian. Moderne deutsche Idiomatik. Tübin- gen. Wolf. Idiom formation. Discourse Processes 9. 1985. Phraseologismen und ihr semantischer Mehrwert. 47-66. Idiomatik und Sprachvergleich. 67-73. 1985. Robin E. Studium Linguistik 13. Halle. In: Robin E. ("jemandem auf die Finger gucken" in einer Bundestagsrede).University College London Authenticated Download Date | 5/9/18 6:49 PM . Durchsichtige Wörter. Idiomatisches Englisch . 289 REFERENCES Burger. Peter. Chicago. April 17. 1982. 1985. 111. Rosemarie. 17-30.). Foundations of Language^ 109-127. Florian. 1968. Horace G. Zur Theorie der Wortbildung. Timothy J. Albert-Reiner. Foundations of language 6. Harald. (Germ. 26-36. Friederich. Diskursive Routine im Fremdsprachenerwerb. 63-111. James San. A^habetisches Wörterbuch mit Definitionen und Beispielen. Sprache und Literatur 56. Wallace. Idiomaticity as an anomaly in the Chonskyan paradigm. 37-46. Hugh McLean (eds. Vance (eds. Sprache und Literatur 56. Grossman. Horace G. Coulmas. 1985. Meaning and the structure of language. Heidelberg. Chicago Linguistic Society. Vance (eds. London. 1986.. Wiesbaden. Die einfachen Wahrheiten der Redensarten. 222-229. For Roman Jakobson. 1970. Sprache und Literatur 56. Arbeitshefte 16). Bruce. 1975.). Idiomatik des Deutschen. Winter. Florian. Zur pragmatischen Fundierung der Idiomatik. Chafe. In: Morris Halle. World order.). Hans-Martin. Hugh McLean (eds. Essays on the occasion of his birthday. James San. Brought to you by | UCL . Mouton.. Lunt. Chafe. Chicago. Cooper. 1971. Charles F. Fräser. 22-42. jr. Lunt. 1956. William E. 1975. Morris. Coulmas. The University of Chicago Press. John Robert Ross. Routine im Gespräch.

Grundlagen. 1985. Weinreich. König- stein/Ts. Olshtain. Piischel. Christoph Schwarze. Niemeyer. 1985. 1973. Wortbildung und Idiomatik. 1985. Zeitschrift für Germani- stische Linguistik 6. Handbuch der Lexikologie. Jaan (ed. The acquisition of new word-formation processes in second language acquisition. 1969. Walter. Puhvel. Zum literarischen Spiel mit der wörtlichen Bedeutung von Idiomen.University College London Authenticated Download Date | 5/9/18 6:49 PM . 151-167. Problems in the analysis of idioms. Berkeley. 88-94. Wunderlich.290 Mrazovic. Strukturbeschreibungen und automatische Analyse. 1987.). Studies in Second Language Acquisition 9. Athenäum. Tübingen.). 1969. Uriel. Pape. Ulrich. 2-13. Brought to you by | UCL . Los Angeles. Elite. In: JaanPuhvel (ed. 23-81. University of California Press. Substance and structure of language. 1978. 221-232. Rothkegel. Dieter. Gleichartige Phraseologismen im Deutschen und Serbo- kroatischen. Zwischen Sprachspiel und Sprachkritik.. Parica. Sprache und Literatur 56. Annely. Sprache und Literatur 56. Feste Syntagmen.

291 SECTION 6 Brought to you by | UCL .University College London Authenticated Download Date | 5/9/18 6:49 PM .

This paper is based on part of a chapter in Dunbar (forthcoming). Let us examine each of these deficiencies in turn. MYERS CONCEPT COMBINATION AND THE CHARACTERIZATION OF LEXICAL CONCEPTS1 0. In sum. the lexical concept. (3) prototypes are not coherent in isolation from cognitive models. Here. prototypes do not display those properties that would enable lexical concepts to play a computable role in a natural language understanding system.University College London Authenticated Download Date | 5/9/18 6:49 PM .292 L. This view has gained currency because of its advantages over the so-called classical view of concepts in relation to certain well-known psychological results. should be characterized in such terms.e. This paper examines the view that the meaning of a word can be charac- 2 terized as a conceptual prototype. and (4) prototypes are not constant (i. there are some good reasons for doubting whether the semantic part of a lexical entry. their inferential relations vary). Brought to you by | UCL . Four arguments will be provided in support of this contention: (1) prototypes cannot be combined. (2) some properties inherent in prototypes are not inherent in lexical items. it is argued that although gcodness-of-exemplar ratings indirectly reflect the meaning of a word. 'Prototype' is defined here as in Dunbar (1987: 2 ) : "a (possibly struc- tured) set of default values along certain dimensions to which exemplars can be compared and rated for typicality in terms of some similarity metric (examples departing least from the defaults being the most proto- typical) ". DUNBAR TERRY F.

For example. degree of membership can be formalized in terms of the relative truth of the predi- cation. for example robins are birds and beds are furniture. He systematically varied the degree of rated truth of the indi- vidual statements in such pairs and attempted to predict the truth of their conjunction. Translation rules function in a logic to indicate how. Although the multiplicative rule is a better match to subjects' intuitions about complex statements. This could be better modelled by the multiplicative rule than by the minimum rule. which they take to be intersective. the 'middle' is excluded). and so this translation rule is not adequate. it too has serious short- comings. argue Osherson/Smith (1981). in a standard predicate logic the set of objects which are red squares is defined by a translation rule as being the intersection of the set of objects which are red with the set of objects which are square. which he attrib- utes to Goguen (1969). Syntactic non-compositionality of prototypes One way of formalising the concept of prototype that has been adopted in the literature is Zadeh's 'fuzzy logic1 (1965). an object cannot have a higher rating with respect to Brought to you by | UCL . Osherson/ Smith (1981) examine fuzzy intersection in relation to the concept striped apple. Oden (1977) proposed as an alternative a multiplicative rule. the interpretation of some construction may be determined. whereby the GOE rating of a complex concept will be the product of the GOE ratings of its constituents. with char- acteristic functions which assign values anywhere on a continuum between 0 and 1. For example. 293 1.cannot be defined. the predicate is either true or false of any particular entity.University College London Authenticated Download Date | 5/9/18 6:49 PM . One contention of critics is that the corresponding function in fuzzy logic telling you how relatively true it is that some object is a red square from information about how red and how square it is. The key attraction of fuzzy logic is that it replaces the characteristic functions of set theory. it has been argued. Such functions offer an appealing analogy with distributions of 'goodness of exemplar' (GOE) ratings (Roth/Msrvis 1983). a striped apple will be a better exemplar of the cate- gory striped apple than of apple. Thus. The obvious translation rule in fuzzy logic (the 'min1 rule) entails that the 'goodness1 of a complex cannot exceed its goodness as an example of either constituent concept. But. given the interpretations of its constituents. What he found was an interaction between the ratings of the con- stituent statements. which are binary (an entity is either in the set or out of it. To test this he asked subjects to judge the truth of pairs of sentences.

The crucial difficulty with only takes account of one half of the conjunct. they will be equally good red squares (so long as both are better 'squares' than 'reds'). For example. Indeed its GOE rating must be lower. like the minimum rule. no simple function of the constituents can adequately characterize their conjunction. Hence this rule. and stipulating that that member's rating for goodness should be 1. At the half-way point. while deriving that relative goodness from goodness in constituent categories. Both work (to oversimplify) by assuming that in any category there will be a 'best1 member. the object should be equally as good a ball as a cube.a conclusion that violates one's intuitions. two red squares will be equally good as red squares so long as their values for the poorer of these properties are equal. Thus. Zadeh 0982) discusses the 'normalization' of complex concepts (raising the maximum value of the set to 1 by dividing all values by the pre-nounalization maximum). if it is based on the 'rain' rule. This is done by imagining a cube and a sphere which metamorphose in such a way that each grad- ually approaches the other's shape. then this entity should be a pound ball to the same degree that it is a round cube . By way of response to these arguments. unless both of its constituent ratings are 1. then if the best red flower were a perfect flower it could not be a perfect red thing.294 how good a member of the conjunction of A and B it is than its rating for ex- emplariness in either of the constituents. This makes it technically feasible for a member of this set to have a higher degree of membership of it than of either of the constituent sets. no simple function of constituent goodnesses can predict correctly the goodness of an exemplar in a complex category. This is counter-intuitive. because its goodness as a red flower would have to be less than 1 in Brought to you by | UCL . even if one is a worse square than the other. This is said to be brought about through focussing on an intersection "by giving it a label" (1982: 291). is that. cannot allow any striped apple to be a better striped apple than it is an apple. if they are equally poor shades of red. Osherson/anith (1982) also demonstrate that if a multiplicative rule underpinned normalization. pointed out by Osherson/Stoiith (1982). If the goodness of something's membership in a conjunction is a simple function of its goodnesses in the constituent categories. Osherson/Smith (1982) sketch a proof that. Ihus. Like ranking (Jones 1982) this result is achieved by considering the 'absolute' goodness of members of a complex category to be a function of their relative goodness with respect to one another independently of their goodness in the constituent categories.University College London Authenticated Download Date | 5/9/18 6:49 PM . in fact.

They emphasise that the process is "asymmetric [and] knowledge-dependent" (1984: 51). the effect of baldness on a man's typicality depends on his age. Effects of correlational structure such as this would. They argue that domain knowledge also modifies. knowl- edge of the world influences the typicality structure of complex concepts which consequently. Cohen/Murphy (1984: 51) describe concept combination as: a process of combining representations according to certain generative rules within a domain of knowledge. then. there are instances of concept conjunction that Osherson/Shiith (1982) assert would defeat the scheme. and hence the goodness of the complex concept. For example. having slots called 'roles'). these authors ask. Jones' rule could not permit an entity which is prima in two cate- gories to be a better exemplar of their conjunction. have to be considered in evaluating typicality with respect to complex concepts. even if ranks could be assigned. Paragons. he [was] more highly rated as a handsome actor. it would seem. what contingency is there among properties that impels no entity to be perfect in more than a single respect? Jones' (1982) attempt to tackle these issues founders on technical problems associated with the need to rank entities before determining their typicality (Osherson/Shiith 1982). different contexts promoting some roles for modification more strongly than others. 295 order to prevent the product of the constituent goodnesses of any red flower equalling their normalized product. it is not the case either that Brought to you by | UCL .University College London Authenticated Download Date | 5/9/18 6:49 PM . for ex- ample. for example. will not be predictable on the basis of the con- stituent concepts alone. Thus. that which role gets modified will de- pend on context. where by'knowl- edge-dependent1 they mean. some of the role values of pet when it is used in pet fish. But. rise above the canons of logic. although Paul Newman [was] both the best actor and the most handsome entity. Although the 'min' rule is undoubtedly wrong to consider only one of the constituents in deter- mining exemplariness in the complex concept. A crucial problem facing any approach to conceptual combination is that the two terms in the conjunction may not be interchanged without altering the typicality rating of an exemplar in the complex concept. but this would entail further use of domain knowledge. They also note that the effect of a role value speci- fication on typicality can depend on the setting of other roles. They take from the knowledge-representation literature in AI the notion that sub-concepts may specialize a super-rconcept by role modification (nouns are treated as frame-like. But. However. they continue.

Thus there appear to be two difficulties confronting attempts to model the combination of concepts by means of fuzzy logic. We will consider two sources of evidence. obvious.e.University College London Authenticated Download Date | 5/9/18 6:49 PM . and a pancake is pottable but not a pot. whenever there is a 'head-modifier1 relationship between the terms. Such non- ccnimitativity confounds the attempt to model the combination of prototypes by means of fuzzy logic. and second its asymmetry. He proposes instead to list only a few senses from which other uses may be derived by general pragmatic principles.296 each contributes equally. for example. a cow oat eher is a type of catcher. first. point is that many words can be used to refer to more than one sort of referent. a milk bottle is a kind of bottle. the other from Jackendoff (1983). a constituent contributes more to the determination of exemplariness in a complex concept when it is playing the role of modifier than when it is playing the role of the head. Brought to you by | UCL . the knowledge-laden nature of combination. 2. and not a dairy product. as would be predicted by the multiplicative rule. 'Sports which are games'). Thus. The principles may be summarised in the fol- lowing way: the designatum. His study used expressions of the type: noun plus relative clause modifier. As Hampton (1988) has recently shown. and not a sort of cow. 'Games which are sports'. can be used to refer to Hampton (1988) has. the head is typically the one that determines the superordinate category. Lexical indeterminacy of types Our second reason to doubt that lexical concepts have a prototype structure is motivated by the finding that some properties inherent in prototypes are not inherent in lexical items. He argues that this is a very general process that is not adequately handled by trying to list all the different senses. Nunberg's first. produced tentative evidence that entities may be treated subjectively as members of a complex category even if they are regarded by the same subjects as non-members of the constituent cate- gories. He found that a given instance (for example chess) might be rated more highly in relation to this form of the putative conjunction than the other (i. one de- riving from the work of Nunberg (1979). for example. Moreover. for example. or basic sense of a word.

an exam script. there are cases in which cue va- lidity appears to be equally high in both directions. we must determine the cue validity of the referential function from singer to song and the inverse referential function from song to singer. cue validity is the same for the referential function from trees to species of trees as it is for the inverse function from species to trees. more than one marker may be involved with a single script. Consequently. It concerns the choice between a token or a type reading of a phrase. and indeed. he will refer to the candidate. This can be used to refer either to a singer. That is.g. That is. Bob Dylan. Now candidates pro- duce only one script while markers handle more than one. it is related to the set of examination candidates. Which of the two is more basic? In order to compute this. while pointing to the exam script. 'Cue validity1 is the exactness with which cues (in this case. In this example. The more basic sense of Dylan is taken to be the one in the domain of the referential function with higher cue validity. and secondly it is related to a set of markers.University College London Authenticated Download Date | 5/9/18 6:49 PM . Dylan des- ignates the man.for example. the songs of Bob Dylan. Brought to you by | UCL . Consider the word form Dylan. each tree is of just one species. (Given certain assumptions about the background knowledge of speaker and hearer. cases in which it is not clear which of the two referential functions has greater cue validity. Since the cue validity of the referential function from Dylan to his work is greater than the cue validity of its inverse. This object may be considered with respect to two relations: firstly. Take. if a speaker says. Either this refers to tokens (the trees themselves) or it refers to types (different species of trees). or to his work. three trees. such as the one we have just described. the referential function from the exam script to the candidate will have a higher cue validity than the referential function fron the exam script to the marker. He has a good grasp of the subject. Nunberg next considers the cases in which we are most interested for the pur- poses of this paper. Of his several examples we will consider just one. e. Therefore. and each species picks out just some trees and not others. 297 an entity if the cue validity of the referential function (RF) mapping fron the designatum to the entity is sufficiently high to clearly and unambiguously pick out the entity in question from any alternatives. referential functions) pick out a given referent.) Nunberg's argument continues as follows: Just as we can talk about relative cue validity of referential functions in cases of deferred ostension. we can also use relative cue validity to determine which senses of a word are most basic.

resemble and become) .g. tense and aspect present a further source of constraint. A policeman has a truncheon with A policeman has the truncheon). Jackendoff (1983) provides some evidence that the choice of a type or a token reading for an NP is determined through the satisfaction of a number of gram- matical constraints. choosing prototypes as the characterization of lexical con- cepts would imply that the type sense is more basic. However. in a talk to the PHTiTNG Conference. Donnellan's (1966) referential-attributive distinction). But if that is the case. in particular. However. Moreover. all predicate NPs express tokens. What Jackendoff's evidence suggests is that the type/token distinction is not determined by a lexical feature of the head noun of a noun phrase. if a verb is stative all its NP arguments must be indefinite for a type reading (cf. since both tokens and types incorporate features of the type in which they are a member. there is a reciprocal relation- ship between types and tokens: just as tokens presuppose types of which they are instances.298 If Nunberg's arguments about cue validity are correct. then his conclusion with respect to the type-token distinction is of great interest and relevance to our own argument. the choice of verb. a type reading results. Thus. this is precisely the asymmetry that Nunberg argues we cannot empirically justify. The definite article. However/ a verb that expresses an event must appear in simple present rather than present progressive for a type reading (cf. His ambiguous instances. so types presuppose tokens that exemplify them. A spider spins a web with A spider is spinning a web). Rumelhart (1987. gives rise to an ambiguity with respect to a token or a type reading (Margaret Thatcher is the woman drinking a martini) (cf. A type reading does not derive from information inherent in a lexical entry. Thus. Another way to Brought to you by | UCL . When the verb is buy. In these cases the distinction is both lexically and syntactically under-determined and would need to be resolved at the level of discourse representation. There is a sense in which a type classification may be seen to be more basic in Jackendoff's analysis. regardless of definiteness. Edinburgh) has recently observed that prototypes could as easily be termed'types'. with the token sense being derived from it. suggest that neither sense is the more basic. If the verb is be and the article indefinite (Margaret Thatcher is a politician). and definiteness of the article. on the other hand. for example. Others do not. A number of verbs behave similarly (e. though we arrive at it by a different route.University College London Authenticated Download Date | 5/9/18 6:49 PM . choice of a type or a token reading for the predicate NP (in English) depends upon its syntactic position. We reach the same conclusion from a consideration of some work by Jackendoff (1983).

and in terms of their current goals. Consequently. what others have called variously 'images'. 299 interpret the sense in which types might be more basic than tokens would take us into genetic epistemology. Things belong to- gether in a category if they are sufficiently similar to the prototype. and therefore. well beyond the scope of this paper. Instability of prototypes Another doubt as to whether lexical concepts are prototypes arises from recent studies which have provided evidence that typicality structure is not a fixed property of a word.University College London Authenticated Download Date | 5/9/18 6:49 PM . but in terms of "interactional properties' related to our exchanges with that world. We "structure and make sense of our experience" in terms of cognitive models (roughly. not the lexical concepts that cue them. The prototype reflects the central tendency of instances in relation to the dimen- sions along which similarity is computed. 'Cognitive models' are defined "relative to idealized circumstances". Brought to you by | UCL . Coherence of prototypes Prototype structure has usually been described with respect to some similarity metric. According to Morphy/ifedin (1985) GCE ratings are 'heavily con- strained1 by knowledge of the world. Miller/Johnson-Laird (1976) had argued that semantic fields have a conceputal core that derives from the implicit theories shared by members of a linguistic community. Lakoff (1987: 68) observes that our categories are not determined with respect to properties of the real world. and that its focus can be shifted by context. Earlier. 4.. this position has been chal- lenged by those who believe conceptual coherence is determined by the theories people have about their physical and social worlds. As Miller (1978: 1O4) once wrote: lexical knowledge . However.. is not isolated from [the] general conceptual system. where similarity has been taken as both the criterion of category mem- bership and the principle underlying conceptual coherence. the lexicon has a cognitive structure only because it is an Integral part of everything a person knows and believes. And this is so even in the case of basic level categories. For the present. 'frames' and 'scripts'). In a similar vein. 3. we take Jackendoff's evidence as further reason to doubt that lexical concepts are prototypes. and conceptual coherence is derived from a theory that explains why things are the way they are. ide- alized cognitive models are what give rise to the prototype structure of our categorizations.

however. These results held for both taxonomic and goal-derived categories. too specific and not independently coherent.g. 'African'. as if membership involved category membership verification (when the antecedent is more general than the anaphora). Problems arise. so for the moment we rest con- tent to record the observation that prototypes as defined appear to be unstable. Roth/Shoben examined whether reading times were affected by a single GOE structure in differing contexts. people can flexibly generate different concepts in different contexts. What then. Summary Four arguments have been presented indicating that lexical concepts are not idealized summary representations corresponding to prototypes. Brought to you by | UCL . Thus. In comparison. Their subjects rated exemplars (e. Thus robin is a poor exemplar in the context The bird walked across the barn-yard. but assigned them different degrees of centrality according to the cultural perspective ('Anerican1. ostriah. and it is not yet clear whether this approach will be viable. however. What various findings illustrate is that prototypes are unstable. It is akin to some recent ideas proposed by Barsalou (1986) in its emphasis on the dynamic nature of conceptual structure. or whether context could affect speed of resolution by changing GOE structure. Barsalou/Sewell (1984. hard to combine. it is more radically'minimalist1(Cruse 1988).300 Roth/Shoben (1933) base a study on a finding of Garrod/Sanford (1977) that the speed of anaphora resolution depends on the semantic relatedness of anaphora and antecedent. are lexical concepts? Dunbar (forthcoming) has put forward a view of the relation- ship between lexicon and cognition that attempts to solve these difficulties. 5. swan) for typicality as birds. they vary with context. The finding was that resolution for typical exemplars (typical for the cate- gory in isolation) was slower when the context was biased to make that exemplar a bad fit than when the context was neutral.University College London Authenticated Download Date | 5/9/18 6:49 PM . reported in Barsalou 1985) found that GOE structure can change dramatically when people take different perspectives on a concept. robin. or 'Chinese') they had been asked to assume.

central tendency and frequency of instan- tiation as determinants of graded structure in categories. Brought to you by | UCL . George A. Interpreting anaphoric relations: The integration of semantic information while reading. 1983. What categories reveal about the mind. Jackendoff. Memory and Cognition) 14. Diss. Overextension of conjunctive concepts: Evidence for a unitary model of typicality and class inclusion. Murphy. George A. Sanford. Models of concepts. 1966. 1977. Cruse. Semantics and cognition. 1979. Mass. London. Lakoff.). Goguen. Philosophical Review 75. ΜΓΓ Press. Gregory L. Reference and definite descriptions. Edinburgh. The non-uniqueness of semantic solutions: Polysemy. Dunbar. Cambridge. 629-654. Journal of Ex- perimental Psychology (Learning. 281-290. Psychological Review 92. Jones.. London. Philip N. 1986. Lawrence W. Linguistics and Philosophy 3. George A. The Belknap of Harvard University Press. Johnson-Laird. Anthony J. London. Miller (eds. Joseph A. Dunbar. Geoffrey. 60-118. Chicago. 1969. Word meaning and encyclopedic knowledge. Cognitive Science 8. Miller. Ideals. Cohen. forthcoming.. 1987. 651-652. Edingburgh Ms. Alan. 27-58. Hampton. Simon. ΜΓΓ Press. Gregory L. James A.). 1982. George L. (In this volume) Donnellan. George A. Barsalou. D. Journal of Verbal Learn- ing and Verbal Behavior 16. fire. The role of theories in conceptual coherence. Language and perception. Stacks not fuzzy sets: An ordinal basis for a proto- type theory of concepts. Ray. and dangerous things. 301 REFERENCES Barsalou. Joan Bresnan. Cambridge. Gregory V.. Miller (eds. 1987. Cambridge.. The University of Chicago Press. Journal of Experimental Psychology (Learning. 77-90. Keith S. George. Medin. 1985. Garrod. Morris.University College London Authenticated Download Date | 5/9/18 6:49 PM . 289-316. In: Morris Halle. Cognition 12. 281-3O4. Mass. Prolegomenon to a cognitive science of lexical meaning. 325-373. Women. Are there static category representations in long- term memory? Behavioural and Brain Sciences 9. 1985. 1984. Nunberg. 1988. Halle. Mass. Linguistic theory and psychological reality. Manory and Cognition) 11. The logic of inexact concepts. 1978. Douglas L. Murphy. Ihe cognitive lexicon.. George L. 1978. Miller. 143-184. Lawrence W. Synthese 19. Benjamin. 1976. 1988. Semantic relations among words. Joan Bresnan.

1983. Emilie M. Integration of fuzzy logical inforrration. Edward J. 1983. On the adequacy of prototype theory as a theory of concepts.. Gradedness and conceptual con- junction. Shoben. Information and Control 8. Osherson. 509-525.. Cognition 12. 1965. Journal of Verbal Learning and Verbal Behavior 22. Lotfi A. Roth. A note on prototype theory and fuzzy sets. Brought to you by | UCL . 346-378. Fuzzy set theory and class inclusion relations in semantic categories. Smith. Edward E.302 Oden. The effect of context on the struc- ture of categories.. Gregg C. Carolyn B. Cognition 12. 1982. 1981. Daniel Ν. Emilie M. Smith. Osherson. Edward E. 35-58. Roth. 1982. Journal of Ex- perimental Psychology (Human Perception and Performance) 3. 291-297. 299-318. Cognition 9. Cognitive Psychology 15. 565-575. Mervis. Daniel N. Zadeh..University College London Authenticated Download Date | 5/9/18 6:49 PM . Lotfi A. 338-353. Fuzzy sets. 1977. Zadeh.

Mental representation refers to the meaning of a single word. One goal of this paper is to con- tribute to an answer. I share the position of Jdinson-Laird/Herrmann/Chaffin (1984) who argue that an adequate theory of the representation and organization of word meaning in memo- ry has to take into account sense as well as reference. Until now. This criticism holds true for semantic network theories. Zimmer/Mohr 1986). for semantic feature theories as well as for meaning postulate theories. there has been no clear answer as to why this is the case.University College London Authenticated Download Date | 5/9/18 6:49 PM . Both are necessary parts Brought to you by | UCL . Why not use these theories? The answer is straightforward: Because almost all of them consider the representation of sense but ignore the repre- sentation of reference although this distinction has been kncwn at least since Frege (1892) who introduced it at the end of the last century. The main goal. 2. is to show that nouns and verbs differ in their mental representation and in the organization of their mean- ings in memory. several theories of word meaning have been proposed. Two experiments will serve to test various assumptions as to the representation and organization of both nouns and verbs. Each of them can readily be translated into the others as Hollan (1975) has shown. these assumptions suggest reasons why nouns are better retained in memory than verbs. Theoretical considerations As can be seen from the literature. organization refers to the connections among the mental representations. What is to be expected It is well known that nouns are better retained than verbs in memory experi- ments (Horowitz/Prytulak 1969. 303 JOHANNES ENGELKAMP NOUNS AND VERBS IN ΊΗΕ MENTAL LEXICON 1. however.

They postulate one representational system. Unfortunately. The value of the theoretical Brought to you by | UCL . Paivio does not distinguish between sense and reference.University College London Authenticated Download Date | 5/9/18 6:49 PM . Almost all of the existing theories assume or imply that the process of translating natu- ral language into a sense representation amounts to comprehension. part of the representation of words conserves information about their sound. these theories postulate a more abstract. do take into account the distinction between sense and refer- ence. Others. anomaly and instantiation . however. conceptual system which contains information that comes close to the sense of words (Nelson 1979. Motor representations. Similarly. They contain information that is used to identify referents in the ex- ternal world. we distinguish object and motor representations which contain appearance information of objects and movement pattern information of actions. Wippich 1981). as we postulate them. As Johnson-Laird/Herntiann/Chaffin (1984) have shown. Thus. there are phenomena . in which words as well as objects are represented. Similarly.which cannot be dealt with adequately without resorting to reference. understanding language cannot normally be reduced to translating sentences into sense. Each can refer to the other in memory. There are. for instance. 304 of meaning representation. respectively. preserve the movement pattern of simple motor actions such as to knock or to wipe (Engelkanp/Zinner 1985. It is assumed that part of the representation of objects conserves information about their appearance. Hubert Zimmer and I have extended this latter type of model by adding motor representations to visual object representations.such as ambiguity. Besides this representational system. These representations have a structure which closely corresponds to the perceived structure of words and objects in the external world. Reference to the external world is almost always involved in under- standing language and consequently is referential representation. They deal less with the meaning of single words than with the general structure of word meaning. of course. besides concepts which represent the sense of words. Paivio (1986). Engelkamp 1987a). They over- look the possibility that sense representation remains meaningless unless it is connected to a more direct representation of the external world (Engelkamp 1985). The nonverbal has largely been re- stricted to visual imagery. distinguishes between a verbal and a nonverbal coding system. some theories which take into account at least some as- pects of reference and which refer to word meaning although usually such theo- ries operate at a different level of theoretical analysis from those mentioned above.

It is evident that objects are usually denoted by nouns and actions by verbs. It is well known that objects can be categorized on different levels of abstraction . Engelkamp/Zimmer 1987a). that a ball is round or a tower is high. Oily specific aspects of action representations are also contained in the verb con- cepts. Therefore the mental representation of nouns and verbs should differ at least with regard to their referential components because the basis of perceiving objects differs from that of experiencing actions. there is a considerable overlap of sense and reference for nouns. The theoretical problem is more complicated with respect to verbs than to nouns. There are also differences in the way that representations of nouns and verbs are organized. or more precisely. retrains the same concept. the overlap of sense and reference representations is less strong for verbs than for nouns. Zimmer (1988) has shown experimentally that the time needed to deal with perceptual properties of objects depends on whether they are accessed on the sense or reference level. Because of this quality. on both representational levels. Thus. of concrete nouns and action verbs. so to speak. Collins/Quillian 1969). a plant - accordingly. 305 extension we have proposed is most clearly seen in the representation of nouns and verbs. Per- ceiving objects as units seems to be an inherent quality of the operation of the perceiving system. And they may also differ in their sense components if we assume that some aspects of the referential representations are also normally represented at the level of the sense repre- sentations. For the sake of simplicity I will use the terms nouns and verbs only. There are aspects of the movement pattern of actions which are not normally part of the sense representation of action verbs. This organizational principle. is far less important with respect to verbs Brought to you by | UCL .University College London Authenticated Download Date | 5/9/18 6:49 PM . The referential representations of actions appear to be more complex than those of objects.a tulip may be seen as a tulip. Thus actions have objects as elements. which I will call taxononical. on the sense level as well as on the reference level. whether a cherry or a sandwich is eaten and whether a child or an adult eats (Engelkamp 1987b. objects form representations which preserve information about the appearance of objects. Normally the type of object as well as the type of agent is ignored in the representation of action concepts. class inclusion is a central aspect of noun organization in memo- ry (Bousfield 1953. Furthermore. We perceive objects as natural figures against a background. Actions basically comprise both changes of states of objects and a causing agent. Some of this in- formation is often also part of the corresponding sense representation. for instance. We 'know1. Eating. a flower.

Experimental tests I will now turn to two experiments which aim at testing some of the assumptions sketched above. 306 (Huttenlocher/Lui 1979). The evidence suggests that nouns as well as verbs are organized in memory according to such episodes. Here. Taxonomy is however not the only organizational prin- ciple in memory. verb meanings are more complex than noun meanings. relational en- coding refers to the activation of connections between representations. Recall effects of word lists are widely agreed to be a function of item-specific and relational encoding (Hunt/Einstein 1981). 3. noun as well as verb meanings are represented on two levels in memory: On a referential level which is relatively close to actual experience of the external world and on a sense level which is more abstract. Generally. Vfe also store the typical arrangement of objects in the ex- ternal world (Hoffmann/Klein 1987) as well as of typical sequences of actions (Schank/Abelson 1977). Some com- ponents of meaning are represented on both levels. It is assumed that the item-specific information consists of concept and object representation in the case of nouns and of concept and action representation in the case of verbs. Both taxonomical and epi- sodical organization are more typical for nouns than for verbs. Nouns and verbs differ in their sense as well as in their referential representations. If nouns and verbs are listened to solely with the inten- tion of retaining the words. A brief examination of the way in which words are processed and of the consequences of this processing on memory performance will provide a frame-work for assessing the results. This type of organization I will call episodical. And it can be assumed that nouns are organized episodically somewhat more strongly than verbs because their referents are experienced both in (pure) spatial contiguity (seeing a bathroom) and in spatial-temporal contexts (taking a bath). but the probability that the referential representation is also activated should Brought to you by | UCL . sense and reference are more closely connected in memory for nouns than for verbs. the experience of spatial and/or temporal contiguity of objects and actions determines how their representations are organized in memory. Furthermore. Summarizing. concepts should be activated in both cases. but the differ- ence is more marked in the former case. whereas referents of action verbs are experienced only in the latter situation (Engelkamp 1987c). Item-specific encoding refers to the activation of the internal structure of a representation.University College London Authenticated Download Date | 5/9/18 6:49 PM . Noun and verb meanings also differ in their mental organization.

If a list is structured episodically. Experiment 1 In the first experiment. either for nouns or for verbs. Four lists were constructed. the relational encoding of nouns should be much better than that of verbs. Because noun meanings are more strongly organized along taxonomies than verbs. apple. relational encoding of nouns and verbs depends on the connections which exist between their representations. to seize. the terms 'modality-free1 and 'modality-specific1 encoding will be used. there should also be better relational encoding for nouns than for verbs. As recall performance is assumed to depend on relational encoding. The same finding is expected with episodi- cally structured lists. peach (fruit) Verbs: to handle. A taxonomically-organized list of nouns and of verbs and an episodically-organized list of nouns and of verbs. to tap (touching) Brought to you by | UCL .University College London Authenticated Download Date | 5/9/18 6:49 PM . A corollary of what is said above is that modality-specific encoding has little influence on relational encoding. Because recall is a function of relational encoding a similar pattern of results is expected for free recall scores. to stroke. Nouns: banana. If a list contains words from cannon taxonomic categories. These connections can be manipulated by varying the structure of word lists. but the difference between both word classes should be smaller than with taxonomically structured lists. Thus a further hypothesis is that relational encoding which is reflected in organizational scores should not change from a modality-free to a modality-specific encoding instruction. but the difference should be less pronounced. instructing subjects ex- plicitly to form images of objects should enhance item-specific encoding of nouns less than instructing them to enact the actions in the case of verbs. Cue example is given of each type. This should be reflected in organizational scores. the organizational score of nouns should be higher than that of verbs with taxonomically structured lists. To identify the condition when subjects only listen to the to-be-remembered words and the condition when they either have to form images or to perform actions. pear. By implication. the two hypotheses about relational encoding were tested. Each list con- sisted of 24 words which belonged to six categories of four items each. By definition. Taxonomically organized lists. the same pattern of re- sults is expected for free recall scores. 307 be greater for nouns than for verbs.

91 .01).62 . to rake.30) = 9. p < .30) = 2. Table 1: Clustering effect (ARC-scores) as a function of list structure and word class List structure taxonomic episodical nouns . the other half was given the verbs first.30) = 20.University College London Authenticated Download Date | 5/9/18 6:49 PM .4.41.60 Before the results are discussed. Each list was presented twice and a free recall followed each presentation. wine. The results show that there is on the average more clustering with nouns (x = . Brought to you by | UCL .85 . I will report on a second experiment. There is also a main effect of word class in free recall (F(1.79 verbs . menu (restaurant) Verbs: to water.61 verbs .30) = 167.54 As expected. Within each block the items were randomized. Half of the subjects of each group received the nouns first.01).14). the difference in ARC-scores for nouns and verbs is smaller in epi- sodical than in taxonomic lists.33.42.53) (F(1. Furthermore. the expected interaction for clustering could be observed (F(1.01). to weed. p < .76) than with verbs (x = .82 versus x = . Table 2: Relative free recall performance as a function of list structure and word class List structure taxoncniic episodical nouns . tip. Analyses were conducted for free recall and for the adjusted ratio of cluster- ing (ARC score according to Roenker/Ihompson/Brown 1971). p < . The inter- action effect between list structure and word class is not significant in free recall (F(1.61). 308 Episodically organized lists Nouns: waiter. Nouns are better recalled than verbs (x = . to sow (garden work) The taxonomic and episodical lists were presented to two different groups of subjects with the verbs and nouns in blocks. p < .52 .

Each list con- sisted again of 24 words. and second. One group received a modality-free. p < . Nouns and verbs were presented in blocks with the words within the blocks being randomized.77) than do nouns (. p < .01).77 This interaction effect is.45.85). as expected. nouns enjoyed a clear recall advantage (5? = . Each list was presented twice and a free recall followed each presentation. Conclusion This discussion has sought to show that it should be assumed that noun as well as verb meanings are represented on two levels in memory: on a referential level Brought to you by | UCL .60 versus . Table 3: Relative free recall as a function of encoding instruction and word class Encoding instruction modality-free modality-specific nouns .68) (F(1. the other a modality-specific encoding instruction. The order of noun and verb presentation was bal- anced between subjects. that item-specif- ic encoding of verbs benefits more from a modality-specific encoding instruc- tion than does encoding of nouns. 309 Experiment 2 The goal of this experiment was to test two hypotheses: First.79 . As in Experiment 1. The lists were presented to two different groups.60 .5.University College London Authenticated Download Date | 5/9/18 6:49 PM .32) = 10. The noun superiority effect almost disappears under the modality-specific encoding instruction.79 versus . In this experiment only episodically structured lists of nouns and verbs were used because with these lists the noun-verb difference in relational encoding under a modality-free encoding instruction was less pronounced. Analyses were conducted for free recall and ARC scores as in Experiment 1.32) = 97. there is no significant interaction between word class and encoding instruction. that relational encoding is not influenced by a modality-specific encoding instruction. This interaction shows that verbs benefit substantially more from modality-specific encoding in- struction (. Although ARC-scores are significantly higher for nouns (x = .67) than for verbs (x = .01) and there is a clear interaction between word class and encoding instruction (F(1. divided into six categories of four items each.51). 4.82) over verbs (x = .85 verbs . Both groups learned nouns as well as verbs. independent of relational encoding.

a bi-level representational structure proves to be more adequate (for a more general discussion of this point see Engelkamp 1987d. although the fact that item-specific encoding of nouns is also better than that of verbs cannot be excluded. Zimmer/Mohr 1986. The noun superiority effect which has been consistently observed under a mo- dality-free encoding instruction (Horowitz/Prytulak 1969. This latter assumption is also supported by the fact that the noun superiority effect could be modified by manipulating item-specific encoding. It was shown that the use of a modality-specific encoding instruction reduces the noun superiority effect. The results fit nicely into this conception insofar as in- structing people to use visual imagery and motor processes to encode word mean- ing influences recall performance. Since it was also observed con- sistently that the ARC-scores of nouns were higher than those of verbs.University College London Authenticated Download Date | 5/9/18 6:49 PM . Zimmer/Mohr 1986). has also been observed in these experiments. In the same way as it was demonstrated that the noun superiority effect can be influenced by item-specific encoding independently of relational encoding. we may conclude that better relational encoding of nouns is one important cause of the noun superiority effect. The effect of modality-specific encoding is obviously independent of changes in relational encoding since it does not influence the ARC-scores either in Experiment 2 or in the experiment reported by Engelkamp (1987b) or Zimmer/ifohr (1986). EngeUcamp/Zimmer 1985). EngeUcamp/Zirtmer 1987b and Zimmer 1987). Engelkamp 1987b). similarities between visual perception and visual imagery (Finke 198O). Engelkamp 1987b). The interaction effect is in agreement with results observed in experi- ments using unrelated lists of nouns and verbs (Zimmer/Mohr 1986. The pattern of results also fits the generally stronger effect of doing as compared to forming images on free recall performance repeatedly observed for action phrases (Engelkamp/Krumnacker 198O. The result was expected because it was assumed that motor encoding by doing enriches the activation of the mental represen- tation of verbs more than imaginal encoding does the mental representation of nouns. Although in Experiment 2 modality-specific encoding improved recall performance only of verbs/ it has been shown repeated- ly that it can also improve recall performance of nouns (Bower 1970. when other phenomena such as the differential access characteristics (Zimmer 1988). I Brought to you by | UCL . and selective interference effects (Zimmer/Engelkamp 1985) are taken into account. 310 which is close to perceiving objects and experiencing actions and on a more ab- stract sense level. Although the results could also be explained without pos- tulating two levels of meaning representation.

the noun superiority effect should became greater if a taxonoraic list structure is presented (as opposed to an episodical). be excluded. Brought to you by | UCL . the interaction between list structure and word class can clearly be observed with ARC scores. With free recall scores. In this case. This interpretation is supported by the fact that the interaction could be observed by Engelkamp (1987c). As can be seen from Table 2. Thus one can conclude that the generally better relational encoding of nouns as compared with verbs contributes largely to the noun superiority effect. the fact that the interaction is not statistically significant might be due to the high recall performance of nouns which indicates a ceiling effect for nouns. however. In any case.University College London Authenticated Download Date | 5/9/18 6:49 PM . If relational encoding is influenced by varying the list structure. the interaction does not yield a significant effect. it was pos- sible to demonstrate that enhancing differentially the item-specific encoding of verbs reduces the noun superiority effect. taxonomic structure is more typical for noun than for verb organization. Better item-specific encoding cannot. As postulated above. Acknowledgment: This research was supported by a grant from the Deutsche Forschungsgemeinschaft (En 124/7). 311 tried to show that the effect can be influenced by relational encoding largely independently of item-specific encoding. however.

). Zimmer.1985. in press. Johannes. Collins. Johannes. 1987b.). 1987. Lawrence Erlbaum. 1987a. Hubert D. Wissensrepräsen- tation und Wissensaustausch. Motor programs and their relation to semantic memory. EngeUcanp. Mental organization of concrete nouns and action verbs. Bericht über den 35. Engelkamp.). Kuno Lorenz. In: Michael Gruneberg. 1979. Engelkamp. Modalitätsspezifische Gedächtnissysteme im Kontext sprachlicher Informationsverarbeitung. Johannes. 1987a. 198O. Dieter Wunderlich (eds. Ronald A. Engelkamp. 511-33. Hampson. 240-247. Cermak.)· 1987. Richardson (eds. Ingbert. Göttingen. Hubert D. Allan M. Horst Krumnacker. ) . Levels of processing in human memory. Peter Morris. 73-100. Johannes. Johannes. 1980. Die Repräsentation der Wortbedeutung. In: Peter J. Engelkamp. 1969. 1-28. The occurrence of clustering in the recall of random- ly arranged associates. Fergus I. 292-313. 229-24O. Weston A. Bower. 1953. Finke.University College London Authenticated Download Date | 5/9/18 6:49 PM . Journal of Verbal Learning and Verbal Behavior 8.N. 415-42O. Johannes. 312 REFERENCES Altelang/ Manfred (ed.). Zeitschrift für Philosophie und philosophische Kritik 10O. Journal of Genetic Psychology 49.). Gordon H. The German Journal of Psychology 9. in press. Kuno Lorenz. Johannes.. Frege. Modalitätsspezifische Gedächtnis- systeme: überflüssige Gebilde oder nützliche Fiktionen? In: Manfred Amelang (ed. Engelkamp. Cognitive Psychology 1. 1987c. Zimmer. Retrieval time from semantic memory. 1987b. Qigelkamp. Barbara Sandig (eds. Modality-specific encoding and word class in verbal learning. 113-32. 277-288. R. 18-46. Johannes.Sykes (eds. In: Christoph Schwarze. Hogrefe. 1987d. M. Brought to you by | UCL . David F. Craik (eds. Marks. M. Zeitschrift für experimentelle und ange- wandte Psychologie 198O. 239-54. Gottlob. Zeitschrift für Psychologie.. Johannes. Engelkamp. Imaginale und motorische Prozesse beim Behalten verbalen Materials. Levels of equivalence in imagery and perception. 1970. Imagery and action: Differential encoding of verbs and nouns. Bousfield. Kongreß der Deutschen Ge- sellschaft für Psychologie in Heidelberg 1986. Zimmer. 1892. Röhrig. Organizational factors in memory. St. Laird S. Engelkamp. Hubert D. 25-5O. John T. New York. über Sinn und Bedeutung. Psycho- logical Review 87. 1985. Ross Quillian. Zeitschrift für Psychologie 195. Johannes. Arguments for a visual memory system. In: Johannes Engelkamp. Barbara Sandig (eds. Engelkamp.).

Leonard. Gilles O. 295-310. 141-162. Zimmer. Brown. M. John T. Nr. Rosemarie Klein. Douglas L. 1971. 'modalitäts- spezifischen1 Lernweise.University College London Authenticated Download Date | 5/9/18 6:49 PM . 1977. 1984. Zwei verschiedene Repräsen- tationen für visuell-sensorische Merkmale? Sprache & Kognition 7. J. Luby S. In: Johannes Engelkamp. Zimmer. Acta Psychologica 58.). Douglas J. Johannes Engelkamp.. Psychological Bulletin 76. Saarbrücken.).). Philip . New York. 497-514. Richardson (eds. Janellen. Fergus I. 1981. Roger J. 1986. Schwarze.. Semiotik 3. Dieter Wunderlich (eds. Hampson. Joachim. Abelson. Peter J. 45-76. Wippich.)· 1987. James D. Horowitz. 1981. Hubert D. Die duale Kode-Theorie und die Konzeption der Analyse- stufen. 1987. 45-48. Einstein. Frankfurt/M. Margit Mohr. Routledge & Kegan Paul. Thompson. Chaffin. Hubert D. Ihe semantic organization of sate simple nouns and verbs. 1975. Arbeiten der Fachrichtung Psychologie an der Uni- versität des Saarlandes. Wiley. Scripts. Prytulak. 1985. Heed. 1969. 154-155. Hillsdale. 25-39. Marks. Craik (eds. Journal of Verbal Learning and Verbal Behavior 20. 1988. Schank. 1979. Allan. 1985. Athenäum. Organisation und Organisierbarkeit von Verben und Substantiven bei einer verbal-semantischen bzw. Journal of Verbal Learning and Verbal Behavior 18. Lawrence Erlbaum. Herrmann. David F.. Hollan. Mental representations. Kuno Lorenz. The sensory-semantic model. 4O-5O.. Features and semantic memory: Set-theoretic or network model? Psychological Review 82. Psychological Bulletin 2. Sykes (eds. Argumente für ein motorisches Gedächtnissystem.. 49-72. Daniel L. Cermak. Imagery: Current developments. Cnly connections: A critique of semantic networks. Felicia Lui. Christoph. Johnson-Laird.. Practical aspects of memory: Current research and issues. Robert P. 313 Gruneberg.N. Universität des Saarlandes. Zimmer. 292-315. Hoenker. 1980. New York. 1979. Roger C.. Brought to you by | UCL . Peter Morris. Oxford University Press. 1987. goals and under- standing. Huttenlocher. Barbara Sandig (eds. Paivio.. Hubert D. Hunt. Kontexteffekte bei der Benennung und Entdeckung von Objekten. Relational and item-specific information in memory. 519-31. Reintegrative memory. Nelson. Charles P. Sprache & Kognition 7. R. Hoffmann. Psychological Review 76. N. Formkonzepte und Bildmarken. 1986. Zimmer. Handbuch der Lexikologie. 10O. Michael.). Hubert D. Werner. Comparison of measures for the estimation of clustering in free recall. In: Laird S. plans. Sam C. London. 81-1O6. An attempt to distinguish between kinematic and motor memory components.

Brought to you by | UCL .University College London Authenticated Download Date | 5/9/18 6:49 PM .

315 SECTION 7 Brought to you by | UCL .University College London Authenticated Download Date | 5/9/18 6:49 PM .

The cognitive semantic approach was triggered off by psychological studies on categorization (cf. summarized in Rosen (1978)). and later on the European variant of componential analysis) and (ii) the lexical-semantic tradition originating with Katz/Fodor's seminal paper (1963). viz. since prepositions constitute a (rather restricted) field of lexical items. two important theoretical frameworks immediately stand out. Introduction If we consider the past fifty years in the history of lexical semantics. also a number of im- portant studies on prepositional semantics have appeared. and was taken up from there by linguists such as Lakoff. Rosen's results on prototypes and basic-level terms.and Langacker. 316 HUBERT CÜYCKENS SPATIAL PREPOSITIONS IN COGNITIVE SEMANTICS1 0. (I should add. Fillmore. that studies on prepositional semantics do not exactly abound.) This title differs from the title of the oral presentation of my paper at the colloquium: 'Prototypes and spatial prepositions'. It is not really surprising that each time a particular lexical-semantic frame- work is at the height of its popularity. I am grate- ful to the people who offered suggestions after the presentation of my paper. As such. Brought to you by | UCL .University College London Authenticated Download Date | 5/9/18 6:49 PM . the classical and the cognitive semantic approach to lexical semantics. Classical lexical seman- tics encompasses (i) the structuralist insights into word meaning (with Trier's field theory. though. we can witness an outflow of detailed studies and applications within that framework.

Moreover. no vague boundaries. Brought to you by | UCL . If there is a lexical item that flags or labels that concept. Classical approaches to SpP-semantics What characterizes the classical approach to word meaning in general? At the heart of the classical view lies the assumption that the mental represen- tation of a concept (that is associated with/labelled by a lexical item) is a summary description of an entire class/category of entities or relations in the world. I will confine myself to the analysis of spatial prepositions (henceforth. All the entities that fit the conditions in the concept constitute the category corresponding to or repre- sented by the concept. This summary description consists of a number of defining or necessary and sufficient features that stipulate the conditions that an entity in the world has to meet in order to fit that concept. concepts have clear boundaries. SpPs) . of the spatial meaning of (a number of) prepo- sitions. There is no degree of membership. 1. I will briefly say something about classical approaches to SpP- semantics. the conceptual conditions can be seen as the meaning of that item. no fuzziness. to put things in their proper histori- cal context. I will be mainly concerned with their analysis within a cognitive semantic framework. Schematically: WORLD 1: unstructured LANGUAGE MEND DDDD -*· categories WORLD 2 This classical view of word meaning and concept structure can also be found in the company of other characteristics: (i) Since concepts/categories (labelled by lexical items) have defining features. Everything is either a member of a concept or not.University College London Authenticated Download Date | 5/9/18 6:49 PM . or rather. 317 In this paper. However.

they should only consist of those semantic elements that mark off lexical items from each other (structuralists) or that have a systematic role in semantic structures (Katz/Fodor 1963). (Katz/Fodor's distinguishers do comprise .encyclopedic information. Let us now turn to some classical SpP-analyses and see how the classical ideas are exemplified in those studies. As we shall see. The SpP in is analyzed as follows in these three studies: Brought to you by | UCL . They are the result of our mind registering observable common properties in entities and relations as defining conceptual features. Leeoh (1969). Actually. They do constitute world knowledge. the semantic struc- ture of a word should not be equated with the conceptual structuring of enti- ties or relations in extralinguistic reality. each member has the same status. Three well-known analyses within the classi- cal paradigm are Oooper (1968). the classical conceptual structure is very valuable. in turn.fron the classical point of view . then the only possible conceptual structure is the classical one. but it is not the only one. Where do the features come from that define the classical conceptual structure corresponding to a lexical item? Historically speaking . The defining features do not represent world/encyclopedic knowledge. Now. This only starts with the advent of cognitive science (and cognitive semantics). if the activity of the mind is confined to registering common observable prop- erties. make up the classi- cal conceptual structure corresponding to those entities or relations.University College London Authenticated Download Date | 5/9/18 6:49 PM . incorrect. The fact that it will not do in a number of cases proves that our mind cannot confine itself to simply registering. Lakoff (1987) goes to great lengths in arguing that these classical concepts are part of an epistemological climate in which conceptual structures are not (yet) the consequence of "the nature of human biological capacities and of the experience of functioning in a physical and social environment" (1987: 12).they have to admit . these intra- linguistically motivated defining features making up a classical concept don't just come out of the blue. that is. and Bennett (1975).this is the wrong question to ask. Why? Classical semanticists view semantic structure as autonomous.318 (ii) Since each member or instance of a concept shares the same defining features. Defining featues are intralinguisti- cally motivated. historically speaking. my entire expose on the classical approach is. which.) Now. Seme members are not more representative than others.

are some of these difficulties with classical SpP-analyses? 2. In other words. volume). Herskovits (1986). Brought to you by | UCL . then. and Lakoff (1987).1 What. but also with conceptual struc- tures that are a consequence of how we function in our environment.1 As Herskovits (1986) points out. Hawkins (1985). These geometric descriptions arise through (i) geometric conceptualization or (ii) metonymy. and interacting with reality. but some apply to schematic geometric figures/descriptions that are associated with these objects (e.g. y) χ is located internal to y with the constraint that χ is smaller than y Leech: in: χ is enclosed or contained in a two-dimensional or three-dimensional place y Beoinett: in: locative (interior ( y ) ) : χ is located in the interior of y. This is not to say that our proposal is going to differ widely from recent studies by Brugman (1983). If we take a closer lock at these classical definitions of in.2 The classical analysis of SpP-semantics presents us with a number of difficul- ties.University College London Authenticated Download Date | 5/9/18 6:49 PM . 2. It goes without saying that in these circumstances language can no longer be seen as autonomous. By systematically reviewing some of them. 319 Cooper: in: 1) L (x. surface. 2. The cognitive approach to SpP-semantics In cognitive semantics. These conditions are reflections of proper- ties observed to be cannon in a number of spatial relations in reality (with as little involvement as possible from the experiencing subject). point. geometric relations do not always apply to objects. that are a consequence of our mind processing. i (y)) 2) SR: smaller (χ. outline. we notice that the concept in is defined in terms of necessary and sufficient conditions that are couched in geometrical terms. the meaning of lexical items corresponds not only with classical conceptual structures that are the result of our mind registering common properties in extralinguistic entities. ordering.1. I think that a selec- tion from their ideas can lead to a satisfying semantic study of SpPs. we will build up our cognitive semantic approach to SpP-semantics. Vandeloise (1986).

Sentence (1) is taken from Herskovits (1986: 16). 2. (ii) Synecdoche (metonymy): In the children in the back of the car. the mind is already present in process- ing reality. Brought to you by | UCL . Consider the following situations: (1) The potato under the bowl /*==*^ /CD Λ (2) The smoke in the bowl (3) She stretched her legs under the table. One way of interacting with the world is that we use some objects as containers for other objects. the village on the road to London. in is not allowed in (1) and (3). where village is conceptualized as a point. and where a contiguity relation is asserted between both arguments. lines. The cog- nitive semantic approach can solve this problem because it also takes into account factors that result from our functioning with entities in the world. while it is allowed in (2). here. In (1). the classical Spl^account does not work.1. this relation should be lexical!zed as in. The conceptual structure that is labelled by in can be de- scribed in terms of the property ΟΟΜΓΑΙΝΜΕΜΓ. (2) from Vandeloise (1986: 233) and (3) from Herskovits (1984: 2 5 ) . How- ever. because a container normally controls the position of the contained (control should be understood as physical control).2 Even if we allow for the fact that the geometric relations proposed above do not apply to objects but to geometric descriptions. points. They are descriptions imposed by the mind. trees.32O (i) Geonetric conceptualization: Consider. In lexicalizes only those spatial relations where y functions as a container for x. for instance. Now. section 1).(3). In the clas- sical account (cf. road as a line. So. *in In (1) . y is a container. respectively. but it does not function as one for x. are not observable prop- erties of villages. outlines. the car refers to the passenger compartment.University College London Authenticated Download Date | 5/9/18 6:49 PM . roads. we are dealing with a geometric INCLUSION-relation. etc.

It appears. In (3). One could try to save the classical account by turning these experiential factors into necessary and sufficient features. the table is normally not considered as a functional container. This is also in contradistinction to the classical theory.University College London Authenticated Download Date | 5/9/18 6:49 PM . 2.3 Consider the following examples: (4) The apple in the bowl (5) The house to the right of X X Π (6) The plane is flying over the hill. which outlaws the use of in. as sections 2. that a full account of SpP-semantics will require the in- clusion of experiential factors. for instance.3 and 2. Brought to you by | UCL . This means that sane instances of the category or concept have a property to a cer- tain degree and hence are members of that category to that degree. so in can be allowed here. It is hard to say. These exaitples show that a lot of SpPs show a fuzzy category structure. where the border is between a aup and a mug. 4 Examples (4) and (5) are taken from Herskovits (1986: 14-15). ftoreover.4 will show. This goes against the classical tenet that concepts have clear boundaries. en the other hand. (6) is taken from Lakoff (1987: 4 5 4 ) . this 'rescue operation1 will not get us very far either. They are more peripheral than more central members. y).1. which claims that contextual features do not change category boundaries. it often depends on con- textual information to find out whether a particular spatial relation is an instance of an SpP or not. 6 Context does play a role. the bowl does control the position of the smoke (it cannot escape). 5 Artefacts can also show a fuzzy concept structure.1. then.y) instead of INCLUSION (x. In would then be defined in a necessary and sufficient way by CONIAUSMEOT (x.1. 321 In (2). However.

Schematically: (i) (i) (iii) — (iii) Brought to you by | UCL . because mothers can also be: . with a central subcategory. According to the classical theory. These extensions are understood in terms of the central subcategory. . but that the mind itself must set up structures that reflect how we experience and interact with reality.the closest female ancestor (ii). Mother shows a radial structure.the person who contributes the genetic material (iv). all of them).the nurturing and raising mother (iii). There are different kinds of games that are all related. often.the wife of the father (i). there simply are none). An example is mother. but one cannot draw up any necessary and suffi- cient conditions to delineate the concept. the different instances of the concept are all related to each other through the mutual similarity of their features. often. it is particularly clear that concept structure cannot exclusively be the result of our mind reg- istering common observable properties (because. . type of family resemblance concept is the one that dis- plays a radial structure (Lakoff 1987). but carmen.e. a family resemblance structure has been proposed. The subcategories/instances are related to each other because they share a number of features. In these cases. .1. But 'a woman who has given birth to a child1 is not a defining feature. A special. in which a number of definitions converge (ideally.University College London Authenticated Download Date | 5/9/18 6:49 PM . because the members of the category or concept all resemble each other. we simply cannot find any defining or necessary and sufficient features for a concept.4 Ihe main problem with the classical approach is that. we should be able to give necessary and sufficient conditions for mother that fit all cases. A typical example is game (Wittgenstein 1953). i. 322 2. and extensions. For the concept game.

If over is a radial concept. as in (8) (8) He flew over the hill. and '+ laying eggs'. there is a link between (7) and (8). For radial concepts such as mother. So. but that cannot be described in terms of a (set of) defining feature(s). SpPs are not analyzed in terms of features but in terms of image-schemas. how. I wonder whether we go through all these links when we try to comprehend an instance of over. A large number of SpPs also show a wide variety of meanings that are all related to each other. does this concept fulfil its categorizing function? How does one decide whether an object is a member of a particular category? How do we comprehend instances of a category? For classical categories. then. Ihis suggests (cf. We comprehend members through the defining features of the concept. as is shown in Appendix 1. I doubt whether an example such as (7) (7) Ώιβ veil over the woman's face is comprehended/is assigned category membership by comparing this instance with a central case. I am afraid Lakoff's analysis presents us with a problem. then owl also belongs to the category bird because it possesses these three criteria! features. However. but the link is very indirect and goes through quite a number of image-schema transformations. then all instances of over should be understood in terms of the central subcategory. the different realizations of over are all transformations of that central image schema. Notorious examples are around (Hawkins 1985) and over (Brugman 1983. Of course. If Appendix 1 repre- sents the concept structure of over. If bird is characterized as '+ wings'. 323 Let us now return to SpPs. In his proposal. categorization occurs by checking whether an entity meets the conditions stipulated in the classical conceptual structure. category membership is defined in terms of the central subcategory. '+ feathers'. 2. game) that (a number of) SpPs at least show a family resemblance structure and maybe also a radial concept structure. with a central subcategory from which various extensions depart.2 SpPs as radial concepts Lakoff (1987) treats over as a radial concept. A foster-mother is understood in terms of the central mother.University College London Authenticated Download Date | 5/9/18 6:49 PM . Brought to you by | UCL . Lakoff 1987).

Lakoff (1987) hints at this when he suggests that over has three distinct. I really think there is a need for this type of abstractness in the conceptual structure of SpPs.University College London Authenticated Download Date | 5/9/18 6:49 PM . I want to suggest that we should give up the primacy of a central subcategory in terms of which all the concept's instances are understood. The higher level is essential Brought to you by | UCL . . Also the central subcategory in a radial concept such as mother shows abstractness. For instance. This means a return from SpPs as a radial concept to SpPs as a general family resemblance concept. Indeed.covering. but I don't think this entire structure is set in motion when it comes to conprehending/categorizing in- stances of a concept. because they are realized in a necessary (but not sufficient) way in the set of instances.above.above and across. The cover terms that we propose are abstract enough. in other words. 324 tfy solution to this problem is twofold. detailed senses. but it shows up as a necessary condition in all of them. Let us therefore have a look at other types of concepts. SpPs such as over have two levels. (i) For SpPs such as over (and I think this holds for most SpPs). but related broad senses: . 'above and across1 has different realizations. because large chunks of semantic information in the central subcategory are present in the other instances. In short. . which is exemplified in a theoretical way in (9) ABC BCD CDE DBF EPG (ii) But proposing a family resemblance structure for SpPs is not enough. Classical concepts show abstractness in that the same defining conditions cover all instances. serves that purpose much better: instances are under- stood in terms of one of these covering senses. which consists of a limited number of broad covering senses. because the semantic properties in the central subcategory (or any combination of them) occur in a large number of the in- stances. They constitute an inherent part of over's concept struc- ture because we relate instances of over to them when trying to understand those instances. the conceptual structure of over needs to be supplemented with these three cover terms. The higher level. The lower level shows all the links between the various.

3. detailed senses of the SpP in a family resemblance fashion. namely CQNTAINMEM'. what about in? for in. this higher level is represented in the following figure: covering above New. Note that this cover term should not be understood as the necessary and sufficient condition for in. which in this case consists of one abstract cover term. For over. this variety is best captured by a two-level structure: (i) the lower level stipulates the different. but related senses. The lower-level part of the struc- ture of in is represented in Appendix 2. I also propose two levels. Brought to you by | UCL . Conclusion A large number of SpPs show a wide variety of different. To my mind. 325 because its level of abstraction is essential to categorization. struc- tured as a family resemblance with a number of different. and (ii) the higher level consists of a limited number of broad covering senses that serve a categorizing function. but related meanings. A lower level.University College London Authenticated Download Date | 5/9/18 6:49 PM . and a higher level.

A semantic analysis of English locative prepositions. 142. 1984. Fodor. Longman. 1587. 1963. Series A. Berkeley Cognitive Science Report no. 20. Cooper. David C. Lakoff. 1986. Herskovits. Cambridge University Press.University College London Authenticated Download Date | 5/9/18 6:49 PM . The structure of a semantic theory. Leech. Linguistic Agency University of Trier. Annette. paper no. George. Chicago University Press. Claudia M. Brought to you by | UCL . Trier. Editions du Seuil. Beranek and Newman report no. Hawkins. London. Ludwig. Oxford. Philosophische Untersuchungen. Wittgenstein. Cambridge. Longman. The semantics of English spatial prepositions. 1O2. Jerrold J. Linguistic Agency university of Trier. Annette. 19O-210.. Series A. Blackwell. London. Claude. An interdisciplinary study of the prepositions in English. Katz. Herskovits. 1983. 1958. Bolt. L'espace en fran<?ais. language and spatial cognition. Spatial and temporal uses of English prepositions: An essay in stratificational semantics. Geoffrey N. 1986. What categories reveal about the mind. Towards a semantic description of English. Chicago. 1975. Comprehension and production of locatives. 326 REFERENCES Bennett. paper no. Bruce W. Gloria S. Brugman. 1968. Trier. 1953. and dangerous things. etc. Vandeloise. London. Semantique des prepositions spatiales.. Paris. Jerry A. fixe. 1987. Women. Language 39. Story of over. 1985. 1969.

University College London Authenticated Download Date | 5/9/18 6:49 PM . 327 Appendix 1 The spider crawled all over the ceiling There were flies all I walked all over over the ceiling the hill The guards were posted all over the hill There was a veil The ice spread over her face all over the windshield 3. The painting over the fireplace The power line stretches over the yard TR 1.3 The bird flew oysr the wall *->· Sam climbed over the wall Brought to you by | UCL . Clouds came over the city The board is over the hole 2.2 The plane flew over the hill +-*· Sam walked over the hill «-»· The town is t over the hill 1.1 The bird flew over the yard -*->· Sam drove over the bridge ·«-+· Sam lives t over the bridge 1. 1.

._ -v v 1/2/3/4 I ' ' 1 1 4 3 1. the bump in the surface y is a 2DIM container + part-whole relation 7. the bird in the tree 4. The diamond in the box the milk in the jar 2. Brought to you by | UCL . the muscles in his legs y is a 3DIM container " 6.University College London Authenticated Download Date | 5/9/18 6:49 PM . the fish in the water the nail in the beard y is a 3DIM container 3. the man in jail containment relation + added the children are in school element: a person (x) is in the the women in church "institution (y) in a typical function associated with it" (Herskovits 1986: 154). ·» 6 «- -* . he is standing in the circle y is a 2DIM container 9._.328 Appendix 2 5 <. the bird in the air 5. the curve in the road y is a 1DIM container between χ the points in the line and y 8.

and 5) essive verbs. it is possible to propose a unified semantic classification based on case relations. i. I found that a similar division (con- taining two additional classes. 329 RENfi DIRVEN A CCQirnVE APPROACH TO CONVERSION 1. Within each major class. especially when seen from the viewpoint of a linguistic theory that claims an autonomous status for linguistic categories as isolated from cog- nitive categories. Brought to you by | UCL . The result (cf. 3) locative verbs. When the study was finished.e. and/or on the basis of more par- ticular features (such as using something as a utensil (7) or fulfilling a social role ( 2 6 ) ) . random or hocus pccus list of fea- tures. Classes of converted verbs based on case relations In a former paper on conversion (Dirven 1979) I was irainly interested in show- ing that instead of the heterogeneous criteria underlying Zandvoort's (1961) syntactic classification of these conversions..University College London Authenticated Download Date | 5/9/18 6:49 PM . temporal and measure) had been proposed in Leitner (1974. 2) in- strumental verbs. 4) manner verbs. Table 1) was a division of converted verbs into five classes: 1) object verbs. various subclasses can be distinguished on the basis of more general semantic-syntactic criteria such as (a) transitivity and (b) animate/inaniinateness of the underlying noun. 1977). The set of more particular features used to characterise the various subclasses may rather look like an extremely ad hoc.

ranging from 1 to 5. pro- posed and analysed in Radden (forthcoming) (cf. conversely.University College London Authenticated Download Date | 5/9/18 6:49 PM . I will mainly limit the discussion to three questions: (i) What is the relationship of the case relations in conversions to the total set of possible case relations? (ii) What cognitive principles turn out to underlie the formation. each of the major groups in the total set of case relations is remarkably well represented in conversion. indicated as 'animate cases' differs from the other categories in that it does not invoke a cognitive domain belonging to the set of domains involved in the rendition of a state of affairs described by a verb and its various semantic cases. almost all converted verbs are dynamic verbs. In this survey there are 5 cognitive groups of case relations.) 2. the sixth category.e. But it seems to be preferable for the present analysis to leave the division as it is. 330 The present paper locks at that classification in a more principled way and tries to shew that we can use. Consequently. but the so-called 'animate1 cases are not. I will stick to the survey. or for that matter a cause. or. Now. only makes sense in the context of a dynamic verb. The relation of case relations in conversion to the total set of case relations 2. the agentive Brought to you by | UCL . The category 'animate cases' is rather based on a different principle. Thus. build. 2.2 This fact raises the question: why should this be so? An agent. which can be seen as the interactants in our conceptualisation of the state of affairs. it would be more logical to allocate the case 'agent1 to group 2 'causal cases'. Table 2). 'experiencer1 and 'benefactive' into one separate category. interpret and classify conversions and offer explanations for doing all this only by invoking our cognitive or en- cyclopedic knowledge. Or to put it the other way round. use and interpretation of converted verbs? (iii) Is there any explanation for the varying productivity of conversion in the single subclasses? (This question will make the division into 27 subclasses and the features used for it somewhat more plausible. but which offers no rationale for putting 'agent'.1 For the total set of case relations. an anthropocentric criterion which may help to explain the large number of agents. since the group "animate cases' does not typically underlie conversion. a dynamic verb requires an agent or cause as its primary interactant. i.

There is. to be an author. a cook or to write. 'instrument1. that the possibility as such exists. 331 notion would be represented twice in the conceptual structure of an 'agentive1 converted verb. 'manner' and 'class membership' in general reflect the cognitively more salient aspects of our experience of states of affairs? Limit- ing oneself to the area of conversion. What about conversions such as (25) to author then? As suggested in my earlier paper. the derivation of agentivity goes in the opposite direction. In fact. Wfe can here only state this fact for the present corpus of 26O converted verbs. from verbs to derived nouns as in the pairs to cook.4 This functional principle is also the guideline to decide what case relation is to be claimed as the underlying case for a single converted verb or a set of converted verbs. So within the area of conversion these subclasses ('patient'. case relations such as 'patient'. but the question has a more general validity. 'goal'. this is not really a derivation from an agent but from an essive. pamphlets or anything else. to 'instrument' in the caus- al cases. which goes far beyond the scope of the present investigation.e. 2. i. one can see that two facts go hand in hand: each of the 5 major groups of the case relations is so well represented. 'manner' and 'class membership1) are clearly prominent and may justify two hypotheses: (i) the weaker hypothesis (to be checked in further research) is that some case relation- ships are more prototypical of their group than others. Brought to you by | UCL . 2. to 'manner1 in the circumstantial cases and to 'class membership' in the essive cases. apart from agent or cause. and (ii) the stronger hypothesis that these 5 case relationships are the more fundamental anchors in the conceptualisation of our experience of reality.e. however. 'instrument'. in other words. Iliis question is: do. a conversion from a stative to a dynamic situation. Although there is no a-priori reason to exclude such a con- struct/ it does not seem to be the normal word-formation pattern. allowing one to introduce an agent into the description of the state of affairs: the process of converting to be an author into to author thus mainly results in creating the possibility of creating an agent. because one subclass in each group is well represented. This applies to the case relation 'patient' in the object cases.University College London Authenticated Download Date | 5/9/18 6:49 PM . i. a writer or to manage. The process of back-formation as in to stage-manage shows. a manager.. to 'goal1 in the locative cases. 'goal'. and per- haps also of giving up the contextual constraint 'books': one can author books.3 Sane case relations seem to be more essential for the interplay of inter- actants in the states of affairs we want to describe than others.

The notion of purpose is even lacking in the paraphrase for the verb to box. Still. In the former group one can paraphrase by means of 'to provide with 1 . In fact this iirplicit use of case relations is but one aspect of their more general. the notion of putting things into a container is more salient and primary than the purpose for which this is done.1 Inspection of some five English dictionaries reveals that dictionary makers intuitively make use of case relations in their paraphrase of converted verbs. food and other artefacts are preserved. to anger (4) is analysed as an object verb since it allows the paraphrase 'to instill anger1. you provide men to a certain institution. COD (Concise Oxford Dictionary) is an older dictionary based on the Oxford English Dictionary and serves as the input for ALD (Advanced Learners' Dictionary).University College London Authenticated Download Date | 5/9/18 6:49 PM . which in Collins is defined as 'to put into boxes' (whereas in COBUILD it is not mentioned any longer with this sense). As the general characterisation 'pre- serve1 for this verb suggests. Similarly. i. but has only an American English written corpus as its basis of description. whereas its successor COBUILD (Collins-Birmingham University Integrated Linguistic Database) is a fully corpus-based approach. Table 3 summarises the paraphrases of to board in 5 dictionaries. the underlying noun could have been considered as an instrument by means of which drinks. Brought to you by | UCL . Collins is a non-corpus based dictionary.e. but in the case of to police the paraphrase is rather that one gets control over an area by means of posting police there. The same functional principle also prevails in the analysis of to bottle (14) as goal verb rather than as instrumental verb. a fact which dictionary makers have been aware of all the time. These examples may also show that the analysis of the meaning of linguistic forms cannot be severed from our encyclopedic knowledge. 332 To give a few examples. one could raise the question why to man (2) is analysed as object verb and to police (11) as an instrumental verb? The answer is based on a functional criterion. Cognitive and encyclopedic principles underlying conversion 3. 3. encyclopedic or cognitive approach to the analysis of the meanings of items. Longman's Dictionary of Contemporary English (LDOCE) claims to be a corpus-based description of English.

to push with the shoulder (= instrument) In the goal meaning of to shoulder the case configuration or case frame is: [. get on to.2 As the various paraphrases of the 5 dictionaries show there is a re- markable congruence between them. to place (as a load) on the shoulders (= goal) 2.g. get on or into.3 Many converted verbs have double or even triple valencies. By 'goal meaning' or 'instrumental meaning 1 of a converted verb is always meant the case relation of the noun (phrase) that served as the input for the converted verb. supply with or give. The sense of providing meals of to board is a relationship which is typically denoted by provide with. in LDOCE we find the following description: to shoulder: 1. e. to lift or carry on the shoulders (= place or instrument) 3. 3. but in fact it also has a goal meaning.. (ii) to shoulder the responsibility/the entire burden. that they belong to the total encyclopedic knowledge upon which every language user or dictionary maker draws. In the list of Table 1 only the instrumental meaning of to shoulder is mentioned. Locative]: (iii) I shouldered the guard aside and went in. Agent. Agent. where the phrasal verb to board up is taken as the dictionary entry.University College London Authenticated Download Date | 5/9/18 6:49 PM . Patient] as illustrated in the following examples fron COBUILD: (i) The soldiers shouldered their kit and moved on. which in most of the 5 dictionaries is even mentioned first. Patient. A typical instance is to shoulder (10). Brought to you by | UCL . Thus to board in the sense of 'go aboard1 is paraphrased by means of goal prepositions such as embark on. the case configuration of to shoulder is [-. The sense of using boards to cover something is typically paraphrased with the instrument preposition with. the actual case frames given for the converted verb describe its present valency or case frame. The fact that dictionaries use paraphrases which are so clearly related to case relations shows that these case relations are part and parcel of the lin- guistic intuitions of the language user. In the instrumental meaning. (iv) They shouldered their way through the crowd. which is based on an intuitive use of case relations. 333 3. except in COBUILD. or more specifically.

This is especially clear in the case of to board. according to the different senses of the input meaning of the noun board. thirdly. which is the input for the goal meaning. i. Thus to board is. Fourthly. extra-linguistic knowledge that dictionary makers tap and that are most convincingly illustrated in the paraphrases and 4 examples given by COBUILD. etc. con- textual domain and the profiles that accounts for the various senses of a lexi- cal item and for the possible conversion processes leading to different valen- cies and to two different domains for the converted verbs. bus. in other words. (ii) travelling. train. The cannon domain of board is the artefact "sawn timber1 or any other material such as cardboard. and (iv) eating or living. 'to come alongside (a vessel) before attacking or going aboard1 (Collins). i. Brought to you by | UCL .4 Ihe double or triple valency of converted verbs may also be due to the lexical polysemy of the item that serves as the input for the conversion.University College London Authenticated Download Date | 5/9/18 6:49 PM . 'to cover with boards'. board can be profiled as 'a smaller piece of rigid material1 as in the compounds breadboard. aheeseboard. The relationship between this quality of COBUILD and its corpus—based nature cannot be a coincidence. it can also be profiled as the side of a ship. Secondly..e. plastic. which is the input for the instrument meaning of the converted verb to board. It is this encyclopedic. boat. 'to get or supply meals'. But it may be interesting to note that. In other words. that these linguistic case frames are the reflection of extra-linguistic scenes in which to shoulder is either a way of carrying something or else a way of using (literal or figurative) power to move oneself or others.e. etc. (iii) attacking. which is the input for a more specific goal meaning of to board. i. 334 Vfe see.e. aeroplane. it can also be profiled as the whole platform (found in on board) of a ship. which may metonymically stand for the food itself and thus serve as input for the object meaning of the verb to board. 3. It is thus the relationship between the common. the various uses of an item in actual language use may easily and quickly lead to a recognition of the encyclopedic dimensions in the item's total meaning. The meaning of a lexical item can be seen as an interplay betoreen a domain (or conceptual context) and the way this is profiled (Langacker 1987: 183). This can first be profiled as 'long wide flat re- latively thin sawn timber1 (Collins). pro- filed against the domains of (i) covering.

An instance of metaphorisation in and during the conversion process is found in the manner sense of to bolt (24) in the context of doing things suddenly (a. i. These fig- urative uses may be due to one or several of the processes of metaphorisation. Ihus to board as an object verb is based on the metonymic use of the noun board for food and is an instance of metaphorisation before the conversion process.) Amongst these. the metaphorisation can take place e. which differs from the goal verb to board (15). but not with its technical meaning of securing a lock. It is hardly imaginable that this meaning should be derived from the literal meaning of to bolt (7). or you eat your food so quickly that you hardly chew it or tasteit) (COBUTLD). metonymy.University College London Authenticated Download Date | 5/9/18 6:49 PM . i. a horse suddenly starts to run fast.e.e.g. also the precise stage in the conversion process must be spot- ted.5 As the analysis of to board and other converted verbs shows. straight movement of pushing a bolt. vegetables such as lettuce or onions grow too quickly and produce flowers and seed. Or in other words. but now only means 'to give shelter to1 as in to harbour a criminal (Collins). the underlying concept is rather 'to move (suddenly and quickly) like a bolt 1 rather than to 'secure a lock'. before. 'to secure a lock with or as with a bolt or bolts' (Collins). i. this verb implies the sense of getting people into an enclosure which also functions as shelter. 335 3.g. things are done too suddenly or too quickly. In addition to the type of metaphorisation. A common aspect in all these senses is the negative connotation that goes with to bolt. i. synecdoche.e. Brought to you by | UCL . An instance of metaphorisation after the process of conversion is found in to harbour.e. synaesthesia and metaphor proper (Dirven 1985: 95ff. metonymy and metaphor proper are the most likely candidates for meaning extension in the case of converted verbs. during or after the conversion process. But to harbour no longer has this literal sense. this lexi- cal category abounds with a mixture of literal and figurative uses. This is of course based on the sudden.

sauce with butter1 or in LDOCE.. I have opted for the 'objective1 case relation. it is in fact the extra-linguistic. 3.University College London Authenticated Download Date | 5/9/18 6:49 PM . i.7 The dictionary draws upon a great deal of extra-linguistic or encyclope- dic knowledge for the characterisation of linguistic items. Consequently. But others are much vaguer and constitute instances of borderline cases. [ . This encyclopedic knowledge may (or must). to board with someone. In this respect to butter is different from the clear instances of instrumental (food preparation) verbs such as (9) to ioe or to pepper.e. Accompaniment] has been transferred to the converted verb.6 The above instances of conversion are rather prototypical cases. to salt. encyclopedic information (which is reflected in good dictionaries) that can solve the fuzziness of border- line cases.. be noted that the transitive causative meaning Brought to you by | UCL .e. i. 'to spread butter on (bread) ' or 'to cook with butter1 as ALD suggests. 'to spread. In the three more 'traditional1 dictionaries COD3 ALO and LDOCE. so that the butter melts and spreads over them1. etc. in which two main types of food.e. i. toast. Its lexical base-profile configuration offers no problem. but not yet as 'lodging1. This can be illustrated for to board again. i. you spread butter on top of them. Collins combines both aspects as 'to give or receive meals or meals and lodging in return for money and work'.e. where the main case relations stand out clearly and unequivocally. but the valency resulting from the conversion does: does it rather have two meanings. Agent. since COBUILD says:'When you butter bread. cook. 336 3. i. usually in return for payment'.e. It should. i. A typical case is to butter (4). this verb is characterised as 'to get or supply meals'. which could imply an objective and an instrumental case frame? Or is there just one instru- mental case frame as is suggested by the paraphrases for to butter in COD. moreover. moreover. This is typically shown in COBUILD's description of the two senses of to butter. bread and vegetables.e. 'to spread with or as with butter1? In the analysis in Table 1. change according to the changing socio-cultural context. It is interesting to note that the case frame of live with someone. which refer to ingredients in the preparation process but which are not garnishing supplements like butter. are used as the starting-point to characterise the two senses given. but only says 'If you board with someone. when they have been cooked. you live in their home for a period of time. And COBUILD goes even one step further in that it does not mention meals any more.

a high productivitiy. one in the manner class. i.e. Table 1). (24) to bolt (26 verbs) and one in the essive class. 337 'to board someone1 is no longer retained in the COBUILD paraphrase. These figures may well represent sane general priori- ties in cognitive interest: objects as the top group. (7) to angle (24 verbs). (25) to author (23 verbs). manner verbs 33 and essive verbs 36. listed in Table 3. it may be supposed to reflect a large part of the socio-cultural contexts determining the present-day uses of the various lexical items. Productivity of the subclasses of converted verbs Finally. i. Table 1 repeats the classification of Dirven (1979) with its 27 subclasses and adds the total number of verbs for each subclass plus a productivity label (high. one in the instrumental class. There are six subclasses of the high productivity type: three in the class of object verbs. But especially the high productivity subclasses may reveal seme deeper insight into the cognitive needs of speakers to develop these classes.e. together they represent about half of the verbs in the corpus.e. instrumental verbs number 55. instnments and loca- tions as the middle group and manner and ways of being as the bottom group. a cognitive or encyclopedic approach is also necessary for the expla- nation of the question of why some subclasses of converted verbs are far more productive than others. This may finally be re- flected in the order of the various senses listed in the dictionary: the sense of travelling of to board may have become more prominent than the sense of living with people and thus be given priority in the dictionary arrangement of the various senses. (1) to butter (26 verbs). Sine» COBUILD is based on a very large corpus of about 20 million words. and a low pro- ductivity for a subclass of 5 to 2 items. limited and low) and draws attention to the cognitive/encyclopedic na- ture of the entities denoted by the nouns that serve as input for the con- version.University College London Authenticated Download Date | 5/9/18 6:49 PM . i. Brought to you by | UCL . a limited productivity for the subclasses between 13 and 6 items. (4) to anger (20 verbs) and (6) to duel (18 verbs). when the subclass comprises about 20 verbs. locative verbs 6O.e. The figures in Table 1 show that object verbs constitute about one third (83 in 266 verbs). i. 4. For this purpose the small corpus of about 26O con- verted verbs has been divided into three scales of varying productivity (cf ·.

Thus the 'remove1 verbs as in (3) to bark seem to be limited on the ground that there is only a limited number of entities from which we want to remove a part. as shapes as in (18) to bundle. but on the contrary. or course. Likewise. the language here only takes stock of very prototypical cultural events. The verbs in the group of subclasses with limited productivity practically al- ways derive from nouns denoting entities which are intrinsically limited in occurrence. This anthropocentric interest underlying the high productivity group of converted verbs is then a cognitive principle that helps to explain what aspects of experience tend to get categorised in a given language. water. Although the class of locative verbs has an overall high productivity (60 verbs). or better limited in relevance to human experience. interactive performance (to duel). surprisingly. It is not unlikely that a great number of explanations This hypothesis needs further research. seem to be limited in productivity for conversion. restricted by the surface (land. social activity. The predominantly anthropocentric nature of converted verbs thus stands out clearly: what English seems to be most interested in when creating converted verbs is the designation of all kinds of processes related to man-made artefacts or to interpersonal.338 The most striking aspect of these six subclasses of converted verbs is that they are based on nouns denoting artefacts (to butterf to angle) or human behaviour such as psychological states (to anger). air) and the kind of power used. It is based on the observation that in food preparation verbs only very prototypical in- stances get lexicalised as converted verbs.University College London Authenticated Download Date | 5/9/18 6:49 PM . or as dwelling places as in (22) to camp. the body parts (10) are strictly limited. as- sociations of manner (to bolt) and social roles (to author). as containers as in (14) to bottle. none of the subclasses themselves are highly productive: this may be explained by the limited number of relevant natural entities that can serve as landmarks with 'arrive1 verbs such as in (13) to beach. Brought to you by | UCL . The same holds for the vehicle verbs such as (8) to boat: the number of various vehicles is not unendingly large. Also abstract cultural artefacts as in (12) to artiale and (19) to bill. although one would rather expect a different result here: possibly. It would take us too far to try to account for each of the subclasses of the low productivity group.

for producing young animals as in (5) to calve. On the one hand these linguistic forms can often only be inter- preted if conceptual or encyclopedic knowledge is seized upon. to spice. milk. Although we use tens (or even hundreds) of food preparation additives. only a few have led to converted verbs such as (9) to iae. etc. the area of conversion (and probably of word-formation in general) proves to be an area where the unity between so-called linguistic categorisation and conceptual categorisation stands out very clearly. for paths as in (2O) to channel or sources as in (21) to mine. we see that entities in the world are only categorised in the language if they are relevant for human interests and activities: the process of conversion may thus be a measuring instrument for those aspects of our experience of reality that have obtained prominent relevance. fat. the principle of relevance to human action and interaction seems to be the ultimate criterion. which is an indication that the former are felt to be far more prototypical food prepa- ration additives/ though far from being the only ones.University College London Authenticated Download Date | 5/9/18 6:49 PM . Brought to you by | UCL . Oanclusion Perhaps more than any other area of the language. to salt. to pepper. But conversion in the area of food preparation has not worked with vinegar. All in all. to sugar. oil.