You are on page 1of 38

How arbitrary is lexical categorization?

Verbs vs. adjectives

REGINA PUSTET

Abstract

Linguistic surface strucfure often assigns a given concept to more than one lex-
ical category. Verb-adjective pairs such äs to resemble vs. similar are a case
in point. The aim ofthis study is to show that categorization within such min-
imal pairs is not arbitrary, but governed by at leastfour semantic principles.
From the regularities which emerge a scalar model ofthe lexicon can be de-
rived which should enhance the predictability of lexical categorization in any
one language, and which portrays the hitherto monolithic lexical prototypes
noun, verb, and adjective äs interlocking components of a coherent semantic
superstructure.

Keywords: adjective, aspect, iconicity, noun, parts of speech, perfectivity,


process, prototype, time-stability, transitivity, valence,
• ·
verb, word
class

1. Theoretical context
The division of the lexicon into parts-of-speech classes, or lexical categories,
is one of the most fundamental aspects of human language, and consequently
has been one of the most persistently discussed topics throughout the history of
grammar writing and linguistics. The 20th Century has witnessed a new surge
of interest in lexical categorization. Extended research into the structure of
non-Indo-European languages has cast some doubt on the general assumption
that the traditional segmentation of the lexicon into the three major classes
noun, verb, and adjective is an appropriate basis for a universal approach to
parts of speech.
The enormous diversity of formal distinctions within parts-of-speech Sys-
tems has been famously acknowledged by Sapir (1921: 118-119), who con-
cludes that they may not be accessible to theoretical unification at all.

Linguistic Typology 4 (2000), 175-212 1430-0532/2000/004-0175


©Walter de Gruyter

Brought to you by | University of Queensland - UQ Library


Authenticated
Download Date | 7/11/15 10:56 PM
176 Regina Pustet

... the "part of speech" reflects not so much our intuitive analysis of reality äs our
ability to compose that reality into a variety of formal patterns. A part of Speech
outside of the limitations of syntactic form is but a will o' the wisp. For this reason
no logical scheme of the parts of speech - their number, nature, and necessary
confines - is of the slightest interest to the linguist. Each language has its own
scheme. Everything depends on the formal demarcations which it recognizes.
In the same vein, Bloomfield (1933: 266) emphasizes that a description in
formal terms must take precedence over any attempt at defining lexical cate-
gories in terms of meaning: "Class-meanings, like all other meanings, elude the
linguist's power of definition ... To accept definitions of meaning, which are
at best makeshifts, in place of an identification in formal terms, is to abandon'
scientific discourse." In rejecting meaning äs a fruitful component of language
analysis, the protagonists of American structuralism, most notably Bloomfield,
paved the way for the rise of generative grammar, which put a preliminary end
to any dispute over how to cope with formal Variation in lexical categoriza-
tion. In most versions of generative theory, such Variation is explained away
by positing universal deep structures for the categories noun, verb, and adjec-
tive.
However, the wide acceptance of generative grammar notwithstanding, the
possibility that crosslinguistic Variation in the surface structure of lexical cate-
gorization is intrinsically systematic in some sense cannot be ruled out a priori.
Any approach to the parts-of-speech issue (or to any other aspect of language)
that makes crosslinguistic divergences in surface patterns predictable to some
extent should at least be taken into consideration äs an alternative - or supple-
ment - to existing theoretical models. The present study seeks to develop such
a new perspective.
The empirical facts that call for more detailed research in this area can be
summarized äs follows. If the behavior of individual lexemes in morphosyn-
tax is employed äs the sole criterion for establishing lexical categories, it soon
becomes evident that many parts-of-speech Systems diverge from the familiär
Indo-European pattern not only with respect to the number of lexical classes
they comprise, but with respect to the size and, consequently, conceptual con-
tent of these classes äs well (Dixon 1977). The size of the adjectival class, if
it is realized äs a separate class in a given language, ranges between a handful
and an infinite number of lexemes. In the latter case the class can be enlarged
by derivation.
Under the influence of Cognitive Psychology, in particular of research initi-
ated by Rösch (e.g., 1973a, 1973b, 1975a, 1975b, 1978), current non-genera-
tive approaches to parts of speech tend to promote, more or less explicitly,
prototype models of the lexicon which ultimately do not question the suitabil-
ity of the traditional parts of speech noun, verb, and adjective äs generally
applicable descriptive primitives of lexical categorization (for more detailed

Brought to you by | University of Queensland - UQ Library


Authenticated
Download Date | 7/11/15 10:56 PM
Ho w arbitrary is lexical categorization ? 177

discussion see Section 5). Thus, a universal model of the lexicon that accounts
for crosslinguistic Variation in lexical categorization, and also offers a criti-
cal evaluation of the Status of the established lexical prototypes, is a veritable
desideratum: "... it is to be hoped that formally-based classes will in any case
have some notional coherence that can one day be defined" (Crystal 1967: 43).
The empirical basis of the present study is a specific section in the lexicon
of seven languages (English, French, Spanish, German, Carinthian Slovene,
Swahili, and Turkish) where apparent verb-adjective synonymies of the type to
resemble vs. similar occur. As it turns out, the semantics of the verb-adjective
distinction in the languages investigated, perhaps in all languages, is of a highly
systematic nature. The forces which emerge here can be taken äs ingredients
of a semantic model of the lexicon whose predictive power for formal category
membership exceeds that of current prototype approaches.
The factor meaning might, after all, contribute a lot to a solution of the parts-
of-speech puzzle.

2. Point of departure
Nouns denote physical objects (i.e., persons, animals, things, and places), ad-
jectives denote properties, and verbs denote events (i.e., actions and processes).
This is, by and large, the traditional semantic formula employed for distin-
guishing the three major parts of Speech. Though intuitively appealing, this
characterization is not very satisfactory; the counterexamples are legion. For
instance, abstract nouns, such äs beauty and peace, do not designate physical
objects. What is more, certain nouns can be interpreted äs denoting events
or properties; examples are motion and whiteness. Many adjectives are more
appropriately described äs expressing states rather than properties. Examples
are adjectives which indicate emotions or bodily sensations, such äs happy or
dizzy. Verbs do not necessarily designate events; like adjectives, they may de-
note both states and properties äs well; examples are to know and to differ.
Croft (1991) replaces the pre-scientific terms "object", "property", and
"event" with bundles of semantic features, and thus arrives at a far more sys-
tematic description of the semantic contrasts between the parts of speech. The
semantic parameters Croft uses are valency, stativity, persistence, and gradabil-
ity (1991: 62-67). The feature profiles he proposes are shown in Table 1.
Croft's model is a selective synthesis of earlier attempts at disclosing seman-
tic distinctions that correlate with lexical class distinctions (e.g., Dixon 1977,
Givon 1979,1984, Langacker 1987, Wierzbicka 1986). However, Croft (1991:
65) points out that his feature profiles still define prototypes only. There are
lexical items that do not match any of the three lexical categories characterized
in Table 1. For instance, the English adjective worth behaves like an adjec-
tive with respect to the parameters stativity and persistence: in Croft's model,

Brought to you by | University of Queensland - UQ Library


Authenticated
Download Date | 7/11/15 10:56 PM
178 Regina Pustet

Table 1. Croft's lexical prototypes

Valency Stativity Persistence Gradability


Noun 0 state persistent nongradable
Adjective 1 state persistent gradable
Verb 1 or more process transitory nongradable

worth would presumably be classed äs a persistent state. With respect to the


Parameters valency and gradability, however, worth deviates from the prototyp-
ical semantic profile of adjectives. For one, the adjective worth has a valency
of 2, since it is furnished with a valency slot both for a subject äs well äs for
an object. Further, worth is not gradable: *worther or *more worth are not
acceptable.
Nonetheless, Croft's feature analysis establishes a connection between the
hitherto monolithic concepts object, property, and event, which implicitly in-
creases the likelihood of the hypothesis that crosslinguistic Variation in lexical
categorization correlates with specific semantic parameters.
The present investigation of verbal vs. adjectival semantics is based on the
well-known observation that the category membership of a concept in lan-
guages that display a verb-adjective distinction which can be defined in terms
of mo hosyntactic criteria may vary. A given concept may be realized äs a
verb in one language, but äs an adjective in another. Thus, the English lex-
emefragrant is an adjective, like its French equivalent odor[ifer]ant, its Span-
ish equivalent fragante, and its Turkish equivalent kokulu. But the semantic
counterpart offragrant in German, duften, is a verb, äs is Carinthian Slovene
disati äs well äs Swahili nukia. This type of categorization shift also mani-
fests itself in the way the concept 'ashamed' is classed in the above languages:
English ashamed and French honteux are adjectives, while Spanish avergon-
zarse, German sich schämen, Carinthian Slovene sramovati se, Swahili aibika
andtahayari[ka], and Turkish utanmak are verbs. What is more, such fluctu-
ations in category membership do not only occur at the crosslinguistic, but at
the language-internal level äs well. Some English examples are: to be asleep
vs. to sleep, to be covetous vs. to covet, to befand of vs. to like, to be sorry vs.
to regret.
These facts suggest that there is a certain degree of arbitrariness in the as-
signment of category membership. As a consequence, one might feel tempted
to accept that lexical categorization äs a whole is arbitrary. However, instead
of abandoning the hypothesis of the semantic determination of lexical catego-
rization, äs the generative approach to language demands, reseafch programs
dealing with parts of speech should treat the phenomenon of double catego-

Brought to you by | University of Queensland - UQ Library


Authenticated
Download Date | 7/11/15 10:56 PM
Ho w arbilrary is lexical categorization ? 179

rization äs a potential source of valuable Information on the subtleties of lex-


ical class distinctions, which should be explored in greater detail. If there are
any observable semantic differences between verbs and adjectives, they should
surface within verb-adjective minimal pairs.

3. The data
3. L The sample
3.1.1. The languages investigated. This study is based on seven languages
which display an acknowledged adjective-verb distinction that can be defined
in terms of structural criteria: English, French, Spanish, German, Carinthian
Slovene, Swahili, and Turkish. The main factor responsible for the specific
choice of languages was the availability of native Speakers. The formal corre-
lates of the adjective-verb distinction in these languages include features such
äs copularization vs. non-copularization in predicate position, agreement with
nouns with respect to categories such äs case, number, and gender/noun class
vs. absence of such agreement, gradability vs. non-gradability, etc.
For instance, in all the Indo-European languages listed above, adjectives
require the presence of a copula in predicate position (la), while verbs do not
combine with copulas (Ib). The following English examples illustrate this:

(1) a. The car was white.


b. / bought the car.
• ·

In Swahili, the same basic predicate structures are found äs in Indo-European


languages: adjectival predicates combine with the copula ni (2a), while verbal
predicates are not compatible with copulas (2b):

(2) a. kitabu ni ghali


book COP expensive
The book is expensive.'
b. Ali a-na-m-tazama simba
Ali 3SG.SBJ-PRS-3SG.OBJ-watch Hon
'Ali is watching the Hon.'

In Turkish, similar copularizing structures are used, although the Turkish cop-
ula is on its way towards getting eroded (Kornfilt 1997: 77-88). Nevertheless,
copulas are still present at least in some morphosyntactic contexts in Turkish,
for instance, in infinitive formation. The infinitive marker -mak/-mek can be
directly attached to verbs (3a), but not to adjectives. When adjectives are com-
bined with the infinitive marker, the latter must be affixed to the copular root

Brought to you by | University of Queensland - UQ Library


Authenticated
Download Date | 7/11/15 10:56 PM
180 Regina Pustet
- i
(3) a. ye-mek
eat-lNF
'to eat'
b. yeni ol-mak
new COP-INF
'tobenew'

The lexeme lists compiled should represent the verb-adjective minimal pairs
found in the respective languages fairly exhaustively. The samples established
for the five Indo-European languages are reproduced in füll length in the Ap-
pendix. Swahili and Turkish are even richer in verb-adjective minimal pairs
than Indo-European languages. However, in order to avoid an over-extension
of the Appendix, not all of the Swahili and Turkish minimal pairs investigated
have been included.

3.1.2. Verb-adjective minimal pairs. A verb-adjective minimal pair is de-


fined äs follows:
v
(4) A verb and an adjective constitute a semantic minimal pair if and only
if there is at least one linguistic context, i.e., one complete clause, in
which the respective verb and adjective can be substituted for each
other, according to native Speaker intuitions, without affecting the
communicative content of the clause.

At this point a more explicit discussion of the relationship between linguistic


context, more specifically, syntactic context, and the parts of speech noun, verb,
and adjective is necessary.
Since antiquity, grammar writing has linked certain parts of speech with
specific syntactic functions so persistently that the syntactic functions in ques-
tion have, at least in common doctrine, become part of the definition of these
parts of speech. Thus, nouns are generally conceived of äs fulfilling argument
function in syntax; verbs are above all thought of äs functioning äs predicates;
and attribute function is considered essential for adjectives. There is, however,
some evidence particularly in the newer typological literature suggesting that
it might be advantageous to view the notions of lexical category and syntactic
function independently (e.g., Croft 1991, Pustet 1989). It is certainly true that
the prototypical functions of nouns, verbs, and adjectives are those of argu-
ment, predicate, and attribute, respectively - but all three parts of speech may
also figure äs the nucleus of syntactic constituents they are not prototypically
associated with. There are predicatively used nouns and predicatiyely used ad-
jectives; there are attributively used verbs (äs in the case of relative clauses)
and attributively used nouns (äs in the case of appositional nouns); and verbs

Brought to you by | University of Queensland - UQ Library


Authenticated
Download Date | 7/11/15 10:56 PM
How arbitrary is lexical categorization ? 181

äs well äs adjectives can be used äs the nucleus of arguments, äs in the case of


headless relative clauses and "nominalized" adjectives.
The empirical basis of the present survey, which is summarized in the Ap-
pendix, are verb-adjective minimal pairs which are contrasted in predicate
function. Thus, native Speakers were asked to form pairs of clauses äs in ex-
amples (5a) vs. (5b), which are based on the English verb-adjective minimal
pair to be asleep vs. to sleep:
(5) a. The cat is asleep.
b. The cat is sleeping.
Theoretically, the syntactic slots of attribute or argument would have been
equally suitable environments for investigating verb-adjective minimal pairs.
However, a difficulty that must be dealt with in the attempt to create syntactic
environments in which the members of a given miminal pair can be substituted
for each other is that individual lexemes may not be admissible in all three of
the syntactic functions listed above. For instance, the English adjective asleep
cannot be used äs an attribute; the English adjective main, on the other hand,
is not acceptable äs a predicate (for additional examples of these two adjective
types in English, cf. Bolinger 1967). However, all the verbs and adjectives
included in the Appendix, in all the languages investigated, lend themselves
to usage äs predicates. This is the main reason why the syntactic function of
predicate has been chosen äs the empirical backdrop for data compilation in
this case.
A major objection to the methodology developed here might arise from the
tacit assumption nurtured by over two millennia of traditional grammar writing
which has been addressed above: attributive function is so essential for defin-
ing the notion of adjective, and predicate function so essential for defining the
notion of verb, that comparing potential members of a verb-adjective minimal
pair in a single syntactic function is unthinkable, even if both members of the
pair qualify for the syntactic function in question. Such views are, however,
in need of critical reconsideration. For instance, there is nothing "unnatural"
about adjectives functioning äs predicates. An adjectival predicate is äs good
a predicate äs a verbal predicate, and therefore äs worthy of empirical investi-
gation äs a verbal predicate. The observation that adjectival predicates are less
prototypical than verbal predicates boils down to a mere quantitative issue, that
of statistical discourse frequency. Verbal predicates are far more common than
adjectival predicates across languages, äs Croft's (1991: 87) discourse counts
reveal.
The sample established by means of definition (4) comprises a total of 164
verb-adjective minimal pairs. In the majority of cases, the members of the
minimal pairs cannot be used interchangeably in all contexts in which they oc-
cur. Such partial synonymies yield the däta this study is concerned with. The

Brought to you by | University of Queensland - UQ Library


Authenticated
Download Date | 7/11/15 10:56 PM
182 Regina Pustet

meaning differences that emerge here are systematic ifi nature. They can, by
and large, be described in terms of four semantic parameters: TRANSITIVITY,
ENERGY RELEASE, PERFECTiviTY, and PROCESSUALITY. If synonymy re-
lations within minimal pairs are determined solely on the basis of these four
parameters, only 56, i.e. 34.1 %, of the minimal pairs contained in the sample
are fully synonymous, while the remaining 108 minimal pairs, i.e. 65.9 %, are
partially synonymous.
Lexemes which are members of verb-adjective minimal pairs are marked by
gray shaded boxes in the Appendix; in addition, lexemes which are involved
in füll synonymy relations are underlined. Lexemes which lack an intralin- -
guistic counterpiece in the opposite lexical class, but which are semantically -
equivalent to a lexeme which is a member of a verb-adjective minimal pair
in another one of the sampled languages, are reproduced in unshaded boxes.
Such lexemes are not directly relevant for this study, but they demonstrate the
frequency of crosslinguistic categorization shifts, and have, therefore, not been
eliminated from the Appendix. A blank box simply indicates that a given lan-
guage is not furnished with a monolexemic representation of the concept in
question within the category in question.
Historically, adjectives often originate in participials. Within this study,
which deals with "genuine" verbs and adjectives only, a participial form counts Ji
äs an adjective only if diachronic change has brought about the erosion of the
verb stem it is derived from. If the verbal base is still in use, the respective
participle is considered part of the verbal paradigm. On these grounds, forms
such äs English astonished, scared, and worried, which coexist synchronically
with the verbal bases to astonish, to scare, and to worry (tr.), have been ex-
cluded from the sample. Ashamed, on the other hand, exemplifies the type of
participial whose verbal base has been lost, and which can therefore be classed
äs an adjective.

3.2. Thefacts
\.

3.2.1. Semantic parameters. Although the parameters transitivity, energy


release, perfectivity, and processuality account for the bulk of semantic diver-
gences within the verb-adjective minimal pairs assembled in the sample, there
are occasionally additional factors, such äs style. Examples from English are
the adjectives stinky and aglow: the former is a colloquial expression, while
the latter tends to be used in poetic contexts.

3.2.1.1. Transitivity. The most obvious manifestation of transitivity is the


grammatical phenomenon'of valency. The valency of a prototypical adjective
is l, while that of a prototypical verb is either l or 2. Lexemes whose valency
value is 3 are rare; an example is to give. However, a valency value of l is

Brought to you by | University of Queensland - UQ Library


Authenticated
Download Date | 7/11/15 10:56 PM
How arbitrary is lexical categorization ? 183

not a reliable diagnostic of adjectives. If grammatical transitivity is taken äs


simply referring to the obligatoriness of an argument in addition to the sub-
ject, transitive adjectives do indeed occur, äs examples like worth [+OBJ] and
devoid of show. The pair worth [+OBJ] / devoid of further points to a differ-
entiation which allows a refinement of the traditional concept of transitivity: in
contradistinction to the common characterization of transitivity äs "the ability
to take an object", the Interpretation proposed above is not sensitive to the cod-
ing type of the secondary argument. With devoid of, the genitive phrase is äs
obligatory äs the object phrase is with worth.
On the other hand, oblique coding of obligatory secondary arguments is by
no means confined to adjectives, but occurs with verbs äs well. Examples are
the German verb gieren nach 'to be avid for' or the English verbs to hunger
for and to thirstfor. Of course, the normal coding device for obligatory argu-
ments are "grammatical" and not oblique case markers; for this reason, the fact
that the valency of a given lexeme is not determined by the marking type of
its argument(s) is easily overlooked. Object-marking of secondary arguments
is neither a necessary nor a sufficient condition for bivalency; even objects are
dispensable, äs pairings like to jwnp a fence vs. to jump demonstrate. Thus,
within the framework developed for this study, a lexeme is classed äs gram-
matically transitive if its use in discourse1 obligatorily requires the presence of
two arguments, no matter how these arguments are coded.
Detaching the obligatoriness criterion from the coding type criterion in defin-
ing transitivity allows us to capture the somewhat less conspicuous, yet regu-
larly occurring transitivity distinctions within verb-adjective minimal pairs. A
semantic criterion for defining transitivity is compatibility with a target phrase.
A TARGET is defined äs the entity or set of entities that a predication is di-
rected at in a more or less abstract sense: a target is, typically, a patient (to hit
SOMEONE), but it can also be the result of an activity (to make SOMEONE), or
an entity emotions are projected at (to love SOMEONE), or a basis for compar-
isons (to resemble SOMEONE), etc. Thus, TARGET serves äs a loose cover term
for a variety of semantic roles which do not necessarily have much in com-
mon. The purpose of this deliberately vague concept does not lie in semantic
unification, but rather in exposing certain facts about the semantic potential
of members of verb-adjective minimal pairs: lexemes which are synonymous
at first glance may differ considerably with respect to their compatibility with
the same semantic type of non-subject argument. The German lexeme triplet
fürchten [+OBJ] (verb) 'to fear' vrs. sich fürchten (vor) (reflexive verb) 'to be
afraid (of)' vs. ängstlich sein (adjective) 'to be afraid' illustrates this. Ac-
cording to traditional analyses, fürchten is transitive, while sich fürchten (vor)
and ängstlich sein are both intransitive.
»
This analysis glosses over the fact
that sich fürchten (vor) comprises the transitivity profiles of both fürchten and
ängstlich sein: sich fürchten vor can be used synonymously with (i) fürchten,

Brought to you by | University of Queensland - UQ Library


Authenticated
Download Date | 7/11/15 10:56 PM
184 Regina Pustet
' i
(ii) ängstlich sein. There is a cline between the three lexemes with respect to
their compatibility with one specific type of argument, which in this case can
be characterized äs the target of an emotion. Expression of a target is obliga-
tory with fürchten, optional with sich fürchten (vor), and not admissible with
ängstlich sein. This triple transitivity differentiation goes unnoticed if the data
are analyzed in terms of the traditional concept of valency. In the notation used
in the Appendix, lexemes of the type fürchten are marked äs [+transitive], lex-
emes of the type sich fürchten (vor) äs [itransitive], and lexemes of the type
ängstlich sein äs [—transitive].
Occasionally, target phrases are coded by means of pluralization of the sub- ·
ject. This strategy is reserved for semantically symmetrical relations such äs
'to be alike', 'to be different', 'to be equal', 'to be similar', in which the ar-
guments or sets of arguments involved are seen äs somehow antagonistic but
fulfilling the very same type of semantic role. An example of this construction j
is these two cars are alike. .j
• · \
3.2. l .2. Energy release. Unlike transitivity, energy release has not yet been ']
generally acknowledged äs a factor which is relevant for lexical categoriza- (
tion. The necessity for positing this semantic parameter becomes evident when
concepts such äs 'to run' (itr.), 4to grieve', and 'to be tall' are compared. Intu-
itively, in the case of 'running', there seems to be "going on" a whole lot more
than in the case of 'grieving'; but, on the other hand, in the case of 'grieving'
there seems to be "going on" a whole lot more than in the case of 'being tall*.
Such differentiations can be translated into a scalar concept of energy release.
Within the sample of verb-adjective minimal pairs investigated, the fluctua-
tions with respect to energy release are more subtle, but regulär enough to jus-
tify the application of this parameter. For example, the Turkish verb öjkelenmek
expresses both the inner, not necessarily visible emotional state 'to be angry',
äs well äs its outer manifestation 'to be (visibly) enraged'. The correspond-
ing adjective öfkeli 'angry' cannot be interpreted äs indicating that the emotion
'anger' results in visible activities. In the same vein, Wierzbicka (1995) ar-
gues that Russian emotion verbs imply a higher degree of agency than their
English translations, which, in most cases, happen to be adjectives. The very
same agency distinction can be observed within some English verb-adjective
minimal pairs, such äs to grieve vs. säd, to rejoice vs. happy.
The terms energy release and agency are, in principle, equivalent. Since the
term agency, however, tends to be associated with the concept of animacy (cf.
Lyons 1977: II, 483-484), the more general term energy release is introduced
at this point. The feature value [+energy release] also covers concepts such
äs to burn and to be afire. The notion of energy release is best defined äs the
presence of an input of physical energy that has an impact on its environment
in that it results in something that can be perceived by human sensual organs.

Brought to you by | University of Queensland - UQ Library


Authenticated
Download Date | 7/11/15 10:56 PM
Ho w arbitrary is lexical categorization ? 185

Notionally, a variety of degrees of energy release can be distinguished in


real-world concepts. For the purpose of this study, however, a binary model of
representation has been chosen. The dividing line drawn between the values
[+energy release] and [—energy release] coincides with the semantic differ-
entiation exemplified by the English minimal pair to grieve vs. to be sad.
grieve is classified äs [dbenergy release], while to be sad is assigned the value
[—energy release]. 'Being sad' is just a state, while 'grieving' implies a higher
agency level, which may result in perceptible outer manifestations of an inner
emotional state.

3.2. l .3. Perfectivity. Of all the semantic parameters that have been claimed
to correlate with lexical class distinctions, time-stability is by far the most
prominent (e.g., Givon 1979, 1984, Pustet 1989, Stassen 1997). The analysis
of the verb-adjective minimal pairs compiled for the purpose of this study sug-
gests that time-stability should indeed be considered a necessary component
of any semantic approach to the verb-adjective distinction. However, time-
stability does not seem to be more relevant in this context than any of the three
other parameters transitivity, energy release, and processuality.
Stassen (1997: 162) advocates Splitting the concept of time-stability into
two subprinciples: the iNGRESSiVE PARAMETER and the PERMANENCY PA-
RAMETER. Former work on the subject deals with the Permanency Parameter
only, which states that concepts differ with respect to the degree to which they
intrinsically imply an endpoint. Likelihood of termination, further, covaries
with lexical class membership. Concepts expressed by nouns do not usually
imply termination, while concepts expressed by verbs usually do. Concepts
expressed by adjectives are hybrid in this respect: most of them admit both
permanent and transitory interpretations.
However, inclusion vs. non-inclusion of a starting point in the semantic
scope of a concept also has some bearing on lexical categorization. Stassen
(1997: 164—166) shows that in languages in which an Option between a nom-
inal and a verbal coding strategy exists for predicative adjectives, the former
always expresses stativity, while the latter conveys a dynamic meaning, i.e., the
notion of "becoming".
Within the verb-adjective minimal pairs investigated, the Permanency Pa-
rameter manifests itself in the frequent occurrence of a state vs. personality
feature distinction, äs in the case of to envy vs. envious. Envious can be in-
terpreted both äs a transient state and äs a personality feature, while to envy is
intrinsically compatible with the former reading only. The Ingressive Param-
eter leaves its traces in differentiations of the type state/property vs. inception
of the latter, äs in the Spanish minimal pair igualarse 'to be equal; to become
equal' vs. igual 'equal'..

Brought to you by | University of Queensland - UQ Library


Authenticated
Download Date | 7/11/15 10:56 PM
186 Regina Pustet
l

Presence vs. absence of starting point and/or endpoint can be regarded äs


facets of the same basic principle, that of boundedness in time. The concept of
boundedness employed here is derived from Langacker (1987). The Opposition
of boundedness vs. non-boundedness in time, in turn, is commonly referred
to äs aspect. The imperfective Signals lack of bounding in time. It presents
an event äs being in progress; its boundaries, i.e., beginning and end, are not
profiled. The perfective indicates that one of these boundaries, or both, are
part of the scope of a given predication. As Langacker (1987: 81) puts it:
"A perfective process is so called because it is bounded; i.e., its endpoints
are included within the scope of predication in the temporal domain ... [A]n
imperfective process ... profiles a stable Situation that may extend indefinitely
far beyond the scope of predication in either direction."
Henceforth, PERFECTIVITY will be used instead of time-stability äs cover
term for the semantic dimension of bounding in time. This is a somewhat un-
orthodox but hopefully not incomprehensible usage of the term perfectivity.
For the Standard definition of perfectivity, cf. Comrie (1976: 44-45). Nom-
inal concepts are, äs a rule, intrinsically imperfective or atelic, while verbal
concepts are intrinsically perfective or telic. A special instance of perfective
verbal concepts are those which explicitly include a starting point and/orend-
point in their scope. To redden' is ingressive, 'to find' is terminative, and
'to explode' is ingressive-terminative. Lexemes expressing adjectival concepts
have a strong inclination for being intrinsically unspecified äs to boundedness
in time. Such adjectives are marked äs [dbperfective] in the Appendix. An ex-
ample is jealous\ this lexeme can be given both a transitory and a permanent
reading, in the sense that denotes both a passing mental state äs well äs a per-
sonality feature. In contrast, furious designates a transitory state only, while
wooden implies permanence.

3.2. l .4. Processuality. The definition of processuality employed in the ana-


lysis of the verb-adjective minimal pairs contained in the sample is strongly
inspired by Langacker's concept of sequential scanning (1987: 72-74), which
refers to the mental Operation of "sampling" real world events, entities, etc.
at certain intervals. If the resulting temporal cross-sections are all identical to
each other, i.e., if no change occurs from one cross-section to the next, the re-
spective event, entity etc. is, by definition, non-processual; if the cross-sections
are not homogeneous, the event is processual. This simplified version of Lan-
gacker's concept of processuality is sufficient for present purposes.
States, such äs 'tired', are intrinsically non-processual, while actions, such
äs 'to eat', are intrinsically processual. The parameter processuality is relevant
for a contrastive semantic analysis of verbs and adjectives: in four of the seven
languages included in the sample, some of the verbal members of the verb-
adjective minimal pairs investigated express both a state or a property äs well

Brought to you by | University of Queensland - UQ Library


Authenticated
Download Date | 7/11/15 10:56 PM
Ho w arbitrary is lexical categorization ? 187

äs a process that results in the latter state or property. An example is Spanish


iguolarse 'to be equal; to become equal'.

3.2.2. Independence of the parameters. The four parameters transitivity,


energy release, perfectivity, and processuality are conceptually autonomous.
Each of them can be the sole factor distinguishing between verbal and adjecti-
val semantics within a given minimal pair. Both the positive and the negative
values of each parameter are cognitively compatible with the positive and the
negative values of any of the other parameters, äs the existence of lexical items
for each logically possible combination proves (cf. Table 2). The only ex-
ceptions in this respect are the pairings [-perfective]/[+energy release] and
[—perfective]/[+processual], for which no lexical realization has been found.
In constructing examples for these two combinations, complex predications
must be resorted to. The clause 'the universe continues to expand', for exam-
ple, covers both of the above feature combinations. The predicate 'continues to
expand' äs a whole can be interpreted äs [—perfective] (according to some ver-
sions of the Big Bang theory); 'continues to expand' must, further, be ascribed
the feature values [+energy release] and [+processual].
In essence, the three semantic dimensions transitivity, perfectivity, and pro-
cessuality correspond to Croft's (1991: 62-67) criteria valency, persistence,
and stativity, respectively.

3.2.3. Statistics. The sample investigated allows for the following general-
izations regarding verbal vs. adjectival semantics within minimal pairs. The
summarizing Statements (i) to (iv) pertain both to the wide spread of individual
parameters occurring across languages to differentiate adjectives and verbs, äs
well äs to the ways in which the parameter values are distributed over verbs
and adjectives.
(i) TRANSITIVITY: Target phrases may occur both with verbs and with adjec-

Table 2. Lexical representations of combinations of feature values


Transitivity Energy release Perfectivity Processuality
_
·+ l
+
Transitiv- + v

ity • . .<,
,.·'·'·. \' ·- . >: · '·
Energy + to pierce jump • · * ' , ·* • < -
l
l :

release — to lose tofall


" . ' .
i
JL , *

Perfectiv- + to pierce tojump to pierce dizzy


ity to resemble wooden wooden
Process- + to pierce tojump to pierce tofall tofall *

uality — to hold to stand to hold* big dizzy wooden ,

Brought to you by | University of Queensland - UQ Library


Authenticated
Download Date | 7/11/15 10:56 PM
188 Regina Pustet
9 f

tives. There are adjectives and verbs that require a target phrase, adjectives
and verbs which can optionally be combined with a target phrase, and adjec-
tives and verbs that do not admit a target phrase. The members of a given
verb-adjective minimal pair may either behave uniformly with respect to their
compatibility with a target phrase, or they may diverge. In the latter case, the
overall transitivity value of the verb always exceeds that of the adjective. Tran-
sitivity distinctions within verb-adjective minimal pairs can be observed in all
of the sampled languages.
(ii) ENERGY RELEASE: The feature [+energy release] is quite atypical for
adjectives, but a pervasive characteristic of verbal semantics. The members of
a given verb-adjective minimal pair may either behave uniformly with respectc
to the parameter energy release, or they may diverge. In the latter case, the
overall energy release value of the verb always exceeds that of the adjective.
Distinctions pertaining to energy release within verb-adjective minimal pairs
can be observed in all of the sampled languages.
(iii) PERFECTIVITY: The members of a given verb-adjective minimal pair may
either behave uniformly with respect to perfectivity, or they may diverge. In
the latter case, the overall perfectivity value of the verb always exceeds that of
the adjective. Perfectivity distinctions within verb-adjective minimal pairs can
be observed in all of the sampled languages. 0

(iv) PROCESSUALITY: The members of a given verb-adjective minimal pair


may either behave uniformly with respect to processuality, or they may diverge.
In the latter case, the overall processuality value of the verb always exceeds that
of the adjective. Processuality distinctions within verb-adjective minimal pairs
can be observed in all of the sampled languages except English, German, and
Carinthian Slovene.
In Table 3, which presents an overview of the logically possible pairings
of feature combinations within verb-adjective minimal pairs, those which oc-
cur in the sample are marked with "yes", while those which do not occur are
marked with "no".
Thus, none of the three parameters transitivity, energy release, and perfec-
tivity provides necessary or sufficient criteria for defining verbal vs. adjectival
semantics. The limited sample reproduced in the Appendix seems to suggest
that the feature [—processual] is a necessary, though not a sufficient, condi-
tion for adjectival Status since it does not contain any adjectives which can be
classed äs [+processual]. However, there are in fact adjectives which can be
ascribed the feature [+processual], such äs 'unstable' and 'variable'.
The general formula to be derived from the data is given in (6):

(6) Within a given verb-adjective minimal pair, the transitivity value and/
or energy release value and/or perfectivity value and/or processuality
value of the verb either equals or exceeds that of the adjective.

Brought to you by | University of Queensland - UQ Library


Authenticated
Download Date | 7/11/15 10:56 PM
How arbitrary is lexical categorization ? 189

Table 3. Logicallypossible pairings offeature combinations within verb-adjective min-


imal pairs

Adjeclive
Transitivity Energy release Pcrfectivity Proccssuality
± - + ± - ± - ± -

Transi-
tivity

yes
Energy
release
yes
Verb
yes
Perfec- no
tivity
no

+ no no
Process- ±
uality

Note: * indicates logical combinations which have not been dealt with because they
are systematically excluded from the scope of the investigation by the definition of
verb-adjective minimal pairs given in Section 3.1.2. Blackboxes indicate constellations
whose existence would disprove formula (6).

In Table 4 the number of examples of the four types of semantic divergences


within partially synonymous minimal pairs are contrasted with respect to their
relative frequency in the sample investigated. For instance, in English (cf. first
column) a total of 20 verb-adjective minimal pairs show a semantic differenti-
ation in terms of transitivity; there are four minimal pairs whose members di-
verge with respect to energy release, and three minimal pairs whose members
diverge with respect to perfectivity. There are no processuality distinctions
within English verb-adjective minimal pairs.
Given the fairly large size of the sample, the systematic correlations summa-
rized in (i) to (iv) should not be coincidental; formula (6) is, therefore, an aspir-
ing language universal. If this formula withstands further empirical scrutiny,
the question of whether the ostensibly chaotic crosslinguistic Variation in the

Brought to you by | University of Queensland - UQ Library


Authenticated
Download Date | 7/11/15 10:56 PM
190 Regina Pustet
' l

Table 4. Semantic divergences within partially synonymous minimal pairs


English French Spanish German Slovene Swahili Turkish
Transitivity 20 18 10 16 4 7 6
Energy release 4 4 2 3 1 1 2
Perfectivity 3 10 3 6 3 5 7
Processuality 0 1 4 0 0 1 3

surface structure of lexical class Systems obeys, at least to some extent, any
semantic principles must be answered in the affirmative. In its present form,
this Statementis valid only within the microcosm of lexical categorization, i.e.,
within verb-adjective minimal pairs. In addition, it must be kept in mind that
the data are derived solely from surveying the behavior of verbs and adjec-
tives in predicate position. Nonetheless, the regularities that have been seen in
this microcosm yield the basic ingredients of a more comprehensive semantic
theory of the macrocosm of lexical categorization.

4. Implications for lexical class theory: Towards a scalar model of the


lexicon 3

The claim that categorization within verb-adjective minimal pairs correlates


systematically with the presence or absence of certain semantic features has
been anticipated by Wierzbicka (1995). This study, which focuses on the
semantic domain of emotion in English and Russian, reveals regulär differ-
ences in agency, transitivity, or volition2 between verbal and adjectival quasi-
synonyms. Wierzbicka (1995: 242) arrives at the strong conclusion that lexical
categorization is completely iconic, i.e., that it is always indicative of semantic
differentiations:
... formal distinctions in grammar reflect differences in conceptualization, often
very subtle differences. Far from being semantically arbitrary they are extraor-
dinarily sensitive - so much so that when we come across two quasi-synonyms
belonging to two different parts of speech, we can regard the formal difference äs
a clue suggesting some difference in meaning.

Do the promising regularities for lexical minimal pairs have any impact on
lexical categorization in general? That is, does the macrocosm of lexical cat-
egorization abide by the same rules äs the microcosm? A superficial glance
at the evidence seems to discourage such speculations. Formula (6) is valid
for minimal pairs only, in particular, in predicate position; it cannot be directly
applied to the verb-adjective distinction in a language äs a whole, let alone at
a crosslinguistic level. Presumably, in the majority of languages which distin-
guish the categories verb and adjective, there are verbs and adjectives which

Brought to you by | University of Queensland - UQ Library


Authenticated
Download Date | 7/11/15 10:56 PM
How arbitrary is lexical categorization ? 191

can be grouped into pairs that reverse formula (6). Some English examples il-
lustrate this: to rise is less transitive than worth; to resemble is less "energetic"
than furious\ to resemble is less perfective than dizzyi to resemble might be
interpreted äs being less processual lhanfurious, if an extremely fine-grained
concept of processuality is employed.
These facts, however, cannot not be taken äs implying that the distribution of
feature values outside the restricted context of verb-adjective minimal pairs is
entirely random. A first hint at the possibility that the four semantic parameters
introduced above do indeed play an essential role in structuring the lexicon äs a
whole is that the phenomenon of double categorization in the verbal-adjectival
domain, in all the languages investigated, is conspicuously linked with specific
semantic domains, particularly those which represent emotions, bodily states,
and resemblance. Do concepts susceptible to double categorization share any
fundamental commonalities? With respect to the parameters transitivity, en-
ergy release, perfectivity, and processuality, prima facie evidence suggests that
the members of verb-adjective minimal pairs are distinguished by a marked
avoidance of clusters of either positive or negative feature values in their se-
mantic profiles; there is always some kind of balance between positive and
negative values. Positive values in one semantic dimension are compensated
by negative values in another. The opposite is true for prototypical verbs and
adjectives: the former tend to display positive values in all four dimensions,
the latter, negative values in all four dimensions. Thus, to hit is [+transitive],
[+energy release], [+perfective], [+processual], to run (itr.) is [—transitive],
[+energy release], [+perfective], [+processual]; big, on the other hand, is
[—transitive], [—energy release], [iperfective], [—processualj. Thus, it can be
hypothesized that lexical categorization is a multifactor scheme that emerges
when the above parameters are conflated into a single scale in which positive
and negative values interact in a cumulative fashion. In short, such a modified
application of formula (6) to the language data might enable us to explain why
certain concepts figure either äs verbs or äs adjectives only, while others are
lexicalized äs members of both categories.
A suitable method for uniting the four parameters is converting the binary
values given in Table 3 into numerical values, and adding these up. The fea-
ture value + is represented by the numerical value l, - is represented by 0,
and ± is represented by 0.5. For example, glad is characterized by the fea-
tures [±transitive], [—energy release], [+perfective], [—processual], and thus
receives the numerical value 1.5.
This method yields remarkably homogeneous values for the verb-adjective
minimal pairs in all the languages investigated. For the sample reproduced in
the Appendix, the values for füll synonymies ränge, in all languages with the
exception of Swahili, between 1.0 and 2.Q (cf. Table 5). The values of the se-
mantic overlaps between partially synonymous members of minimal pairs, i.e.,

Brought to you by | University of Queensland - UQ Library


Authenticated
Download Date | 7/11/15 10:56 PM
192 Regina Pustet

Table 5. Total ofexamples offull synonymies

1.0 1.5 2.0 2.5 Total


English 2 6 4 0 12
French 1 5 1 0 7
Spanish · 2 4 0 0 6
German 2 6 1 0 9
Slovene 5 2 1 0 8
Swahili 0 6 0 0 6
Turkish 3 3 2 0 8.

Table 6. Total ofoverlaps in partial synonymies

1.0 1.5 2.0 2.5 Total


English 4 9 7 4 24
French 3 7 11 0 21
Spanish 4 8 2 0 14
German 9 5 4 0 18
Slovene 0 4 2 0 6
Swahili 4 6 1 0 11
Turkish 4 6 4 0 14 :S

Table 7. Total offull synonymies plus total ofoverlaps in partial synonymies

1.0 1.5 2.0 2.5 Total


English 6 15 11 4 36
French 4 12 12 0 28
Spanish 6 12 . 2 0 20
German 11 11 5 0 27
Slovene 5 6 3 0 14
Swahili 4 12 1 0 17
Turkish 7 9 6 0 22

of the readings which are common to the members of such minimal pairs, also
ränge between 1.0 and 2.0 in all languages, with the exception of Carinthian
Slovene and English (cf. Table 6). There is a marked overall predominance of
the value 1.5, äs the figures in Table 7 reveal.
In contrast, prototypical adjectives, such äs big, are characterized by the
value 0.5, while prototypical verbs rank either at 4.0, like to A/i, or at 3.0, like
to run (itr.), depending on their transitivity value. This, in turn, means that the

Brought to you by | University of Queensland - UQ Library


Authenticated
Download Date | 7/11/15 10:56 PM
How arbitrary is lexical categorization ? 193

members of verb-adjective minimal pairs occupy a coherent space between


prototypical verbs and adjectives in the overall numerical scale, which reflects
traditional intuitions about their conceptual Status: "... the fact that it is diffi-
cult sometimes to distinguish states from qualities, on the one hand, and from
actions and processes, on the other, would suggest that states have a certain on-
tological ambivalence" (Lyons 1977: II, 441). The overwhelming majority of
concepts contained in the sample must be classed äs states; prototypical adjec-
tives and verbs, on the other hand, designate properties and events, respectively
(cf. Section 2).
The numerical scope of the verbal-adjectival gray zone, i.e., the section
in the lexicon in which verbal-adjectival double categorization occurs, may
vary from language to language. The Indo-European languages included in
the sample all show values ranging between a minimum of 1.0 and a maxi-
mum of 2.0, with the exception of English (cf. Table 7); the values for En-
glish verb-adjective minimal pairs may reach a maximum of 2.5. In Swahili
and Turkish, this type of double categorization is even more pervasive than
in Indo-European languages. Both languages, especially Swahili, extend the
gray zone into the domain of "prototypical" adjectives. Examples are Swahili
imarika (verb) vs. imara (adjective) '(to be) firm', and Turkish kibirlenmek
(verb) vs. kibirli (adjective) '(to be) proud; (to be) arrogant'. All four lex-
emes are characterized by the feature combination [—transitive], [—energy
release], [iperfective], [—processual], and thus receive the numerical value
0.5.
The four semantic parameters can, of course, be applied to nouns äs well. A
prototypical noun such äs tree is [—transitive], [—energy release], [—perfect-
ive], and [—processual], and thus receives the numerical value 0.0. Between
nouns and adjectives another categorial gray zone exists, which has been exam-
ined in detail by Wierzbicka (1986). Categorial differentiation in this section of
the lexicon is, obviously, sensitive to an additional semantic parameter, which
can be characterized in terms of the contrast of characterization vs. typification
(Wierzbicka 1986). Inclusion of this parameter in a general scalar model of
the lexicon does not affect formula (6) and the theoretical conclusions derived
from it because both verbs and adjectives are, in any case, [+characterizing]
(or [—typifying]). Thus, the proportions of the numerical correlations between
verbs and adjectives obtained by means of the above methodology are not dis-
torted.
The general hypothesis that results from the present study is that double
categorization is the symptom of a semantic overlap between two lexical cat-
egories. Any type of double categorization is confined to a limited, coherent
space in the lexicon of individual languages. The backbone of this theoretical
approach is a semantic scale which allows for positioning categorial gray zones
between the semantic spaces covered by the prototypical instantiations of the

Brought to you by | University of Queensland - UQ Library


Authenticated
Download Date | 7/11/15 10:56 PM
194 Regina Piisiet
\
lexical categories involved. A strenger form of ihis hypothesis is the following
prediction:

(7) A concept whose numerical value is eilher above or below the values
that mark the upper and lower boundaries of a categorial gray zone
between the lexical classes X and can be realized äs a member of
category X only, if its value is below that of the lower boundary of the
gray zone, and äs a member of category only, if its value is above
that of the upper boundary of the gray zone.

Thus, in English, all concepts that rank below the value l .0 should be adjectives
(or nouns), and all concepts that rank above the value 2.5 should be verbs. None
of the languages contained in the sample seems to refute the strenger claim
formulated in formula (7). This, of course, does not amount to stating that any
concept whose numerical value falls within the scope of a gray zone must have
a semantic counte aIt in the opposite lexical class. On the contrary, all the
sampled languages are furaished with numerous "single" verbs and adjectives
whose numerical values coincide with those of verb-adjective minimal pairs.
The English examples to adore and dizzy illustrate this. The feature profile of
to adore is [-{-transitive], [—energy release], [±perfective], and [—processual],
which yields the value 1.5, that of dizzy is [—transitive], [—energy release],
[+perfective], and [—processual], which yields the value 1.0.
The approach outlined so far interprets the lexicon äs a semantic scale com-
posed of (at least) four parallel subscales. In this scale, the parts of speech
gradually blend into one another. The most obvious manifestation of this merg-
ing scenario is the phenomenon of double categorization, which is nothing but
a structural reflex of the hybrid Status of lexemes which are conceptually inter-
mediate between t\vo categorial prototypes.
The explanatory power of any of the four *semantic parameters is inferior to
that of the joint, multifactor approach. Neither transitivity, nor energy release,
nor perfectivity, nor processuality is capable of accounting for the categoriza-
tion patteras investigated in this study convincingly, when applied in Isolation.
The advantages of the scalar model of the lexicon developed above, äs com-
pared to the traditional prototype approach, are self-evidenL This model does
not deny the existence of prototypes, but rather redefines them by ascribing spe-
cific semantic features to them. Thus, it establishes a link between the hitherto
more or less monolithic prototypes, and thus portrays the lexicon äs a coherent
System. Moreover, the model offers a natural explanation for the occurrence
of double categorization. Verb-adjective minimal pairs are, in this modified
perspective, not just a case of structural idipsyncrasy, which comes in handy
äs proof of the tenet of the total arbitrariness of lexical categorization; rather,
their existence is the almost necessary consequence of a partial overlap of the

Brought to you by | University of Queensland - UQ Library


Authenticated
Download Date | 7/11/15 10:56 PM
Ho w arbitrary is lexical categorization ? 195

categorial prototypes VERB and ADJECTIVE in a universal semantic scale, in


which these prototypes occupy neighboring positions.
Although the hypothetical formula (7), if universally valid, increases the
predictability of language-internal categorization considerably, the approach
outlined above fails to completely eliminate arbitrariness. It does not allow
for any predictions concerning the category membership of a lexeme which is
not part of a lexical minimal pair, but whose numerical value falls within the
scope of a categorial gray zone. The behavior of the concept 'to glow; aglow',
which is located in the verbal-adjectival gray zone in all of the seven languages
investigated is a case in point (cf. respective entries in the Appendix): 'to glow;
aglow' is categorized äs a verb in German, Carinthian Slovene, and Swahili,
and äs an adjective in French, Spanish, and Turkish. Categorization within gray
zones, therefore, seems to be arbitrary.
However, the scalar model of the lexicon developed above can be conceived
of äs, theoretically, embracing an infinite number of subparameters; future re-
search is likely to reveal additional semantic parameters whose inclusion in
the overall scale narrows down the scope of the gray zones, and thus further
enhances the predictability of category membership. A potential candidate for
inclusion in the set of relevant parameters might be gradability, äs posited by
Croft (1991: 62-67). Gradability is traditionally considered a criterion that is
capable of separating adjectives from verbs äs well äs from nouns, although,
upon closer inspection, it turns out that many adjectives are not gradable. It
should also be kept in mind that even if a semantic criterion X is found to cor-
relate with lexical class distinctions to a certain degree, this does not automat-
ically imply that criterion X is "utilized" by the human brain in the process of.
segmenting concepts into categories such äs the time-honored parts of speech.
The minimal pair approach, on the other hand, offers an empirical tool that
discloses semantic criteria "at work": in cases in which there are systematic
meaning differences between verbs and adjectives such äs those described by
formula (6), one cannot help but conclude that lexical categorization is indeed
sensitive to specific semantic distinctions. Gradability, however, does not seem
to operate äs such a differentiating parameter within verb-adjective minimal
pairs in the seven languages investigated.

5. Synopsis and prospectus


Methodologically speaking, the theoretical approach presented in Section 4
employs both inductive and deductive reasoning, but is, äs a whole, strictly
self-contained in that the parameters derived inductively from the data base
constitute the sole basis for the deductive part of the argumentation. This the-
oretical model confines the occurrence of verb-adjective minimal pairs, which
are often cited in support of the claim that the surface realization of lexemes

Brought to you by | University of Queensland - UQ Library


Authenticated
Download Date | 7/11/15 10:56 PM
196 Regina Pustet
' t
is totally arbitrary, to a coherent, and restricted, section in the lexicon of indi-
vidual languages, and advocates the view that lexical categorization is, by and
large, non-random.
The semantic scale noun-adjective-verb which emerges äs the result of the
present study is paralleled by the well-known, presumably universal, surface
structure scale noun-adjective-verb. It has been pointed out repeatedly in the
literature that the morphosyntax of adjectives, in any one language that is fur-
nished with an independent category of adjective, is, for the most part, a mix-
ture of nominal and verbal grammar (e.g., Pustet 1989, Schachter 1985, Stassen
1997, among others). This places the adjective in an intermediate position be-J
tween the noun and the verb in structural terms. Thus, at the most fundamental
level of linguistic description, semantic structure indeed seems to reflect mor-
phosyntactic structure.
However, a final, and essential, question has not yet been touched upon:
does the semantic scale proposed above haye the potential of accounting for
those Systems of lexical categorization which are radically different from the
standardizing triple-class System, and which therefore pose enormous prob-
lems for a unified non-generative approach to the parts-of-speech issue? Many
languages display indistinctions between two of the three major lexical cate-
gories noun, verb, and adjective; some are even credited with an indistinction
between all three categories (for example, Nootka, a Wakashan language of
British Columbia; cf. Jacobsen 1976, Schachter 1985, Swadesh 1939). Cur-
rent dogma, however, tends to acknowledge at least the universality of the
noun-verb distinction, although the degree of morphosyntactic differentiation
between these two lexical categories may vary considerably from language to
language (Dixon 1977, Lyons 1977: II, Schachter 1985, among others). The
patterns of partial indistinction documented so far are not entirely random: ei-
ther the nominal or the verbal class, or both classes, may absorb the inventory
of adjectival lexemes in a given language; but nominal and verbal lexemes are
never grouped together in a single class that contrasts with the adjectival class
(Hengeveld 1992). These facts comply with the semantic scale posited above:
only categories which are adjacent to each other in this scale can be classed
together.
However, languages may not only differ with respect to the number of lex-
ical classes they distinguish, but with respect to the size of these classes äs
well, äs Dixon (1977) shows particularly for the category of adjective. Thus,
"languages which have no adjective class, or only a small closed class, tend
to distribute some of the normal adjective types amongst the other parts of
speech" (Dixon 1977: 29). It should be noted that the scalar model of the lexi-
con proposed in Section 4 is compatible with fluctuations in the size of lexical
classes. What is more, the analysis of Section 4 implicitly points to the exis-
tence of Systems of lexical categorization containing very small lexical classes.

Brought to you by | University of Queensland - UQ Library


Authenticated
Download Date | 7/11/15 10:56 PM
How arbitrary is lexical categorization ? 197

The comparison of Swahili and Turkish with Indo-European languages shows


especially that the size and numerical scope of a categorial gray zone may vary
from language to language; this, in turn, should also affect the size and numeri-
cal scope of the "pure" lexical classes. On this basis, the existence of extremely
small lexical classes is predictable. The position of the cutoff points or gray
zones between lexical classes in the overall semantic scale is, obviously, äs ar-
bitrary äs the ascription of lexical class membership within a categorial gray
zone (cf. Section 4).
Thus, lexical categorization in typological perspective seems to obey the
following basic pattern (also compare Croft 1995: 504-506): the grammat-
ical domain äs a whole is structured by an underlying, semantically defined
scale. This semantic superstructure is universal and non-arbitrary, but the cuts
by which individual languages partition the scale, through surface structure
distinctions, are arbitrary.
In sum, the minimal pair approach seems to offer a viable method for un-
covering potential semantic factors underlying lexical categorization in gen-
eral, i.e., at a crosslinguistic level. As a next step, the noun-adjective interface
might be investigated for semantic regularities of the kind observed within the
verb-adjective minimal pairs dealt with in the present study. The methodology
outlined here, however, might be suitable only for languages with large vo-
cabularies. Attempts at tracking down verb-adjective minimal pairs in certain
African languages whose lexical inventories are more restricted than those of
the seven languages discussed in this paper did not yield any convincing data.
Only languages with large vocabularies seem to be able to afford the "luxury"
of double categorization. Further, it should be kept in mind thät in applying the
minimal pair approach in search of a universal semantic theory of lexical cat-
egorization, the choice of languages to be investigated is additionally limited
in a rather obvious way. Verb-adjective minimal pairs, for instance, will, of
course, occur only in languages which distinguish the classes verb and adjec-
tive in morphosyntactic terms, and not in languages which treat both lexeme
types uniformly.
The four semantic parameters which appear to be decisive for the phenom-
enon of lexical categorization correspond - perhaps coincidentally, perhaps not
- to the four fundamental quantities that the most basic descriptive model of
the world man has devised employs, physics: matter, space, time, and energy.
Matter and space relate to transitivity, perfectivity and processuality relate to
time, and the connection between energy and the semantic parameter energy
release is self-evident.
However, lexical categorization is presumably a function of the interaction
of a larger array of semantic parameters; supplementing the model outlined
above with additional parameters might^ increase the predictability of lexical
categorization, to the effect of limiting the ränge of phenomena which must

Brought to you by | University of Queensland - UQ Library


Authenticated
Download Date | 7/11/15 10:56 PM
198 Regina Pustet
' l
be labeled "arbitrary" in this domain of human language. After all, we have
to remember that in the history of science, "arbitrariness" has only too often
turned out to be just another word for "we don't understand".
Received: 17 January 2000 Ludwig-Maximilians- Universität München
Revised: 17July 2000 University of Colorado

Notes
Correspondence address: Institut für Allgemeine und Indogermanische Sprachwis-
senschaft, Ludwig-Maximilians-Universität München, Geschwister-Scholl-Platz J,
D-80539 München, Germany; e-mail: pustet@lrz.uni-muenchen.de <- \ &_
I am indebted to the following persons for language data: Hassan Adam, Olga
Chapado, Gwen Frishkoff, Mehmet Hacisalihoglu, Maria Jernej, Bill Raymond,
Chris Searles, Isabelle Süthold-Ferchaud, and §ebnem Uzunlar. Any errors are, of
course, my responsibility.
Abbreviations: 3 3rd person, COP copula, INF infinitive, OBJ object, PRS present,
SBJ subject, S G Singular.
1. This refers to grammatically unmarked forms. Passive transformations etc. are not
considered here.
2. For the analysis of the sample reproduced in the Appendix, the factor volition is of
minor relevance. One of the few minimal pairs within which different degrees of vo-
. · · · .
litionality can be observed is German hungern (verb) *to hunger (for food etc.)' vs.
hungrig (adjective) 'hungry'. Hungern is usually understood äs [—volitional], but
can be interpreted äs [+volitional], while hungrig can be interpreted äs [—volitional]
only.

References
Bloomfield, Leonard (1933). Language. New York: Holt.
Bolinger, Dwight (1967). Adjectives in English: Attribution and predication. Lingua 18: 1-34.
Comrie, Bernard (1976). Aspect. Cambridge: Cambridge University Press.
Croft, William (1991). Syntactic Categories and Grammatical Relations. Chicago: University of
Chicago Press.
— (1995). Autonomy and functionalist linguistics. Language 71: 490-532.
Crystal, David (1967). English. Lingua 17: 24-56.
Dixon, R. M. W. (1977). Where have all the adjectives gone? Studies in Language 1: 19-80.
Givon, Talmy (1979). On Understanding Grammar. New York: Academic Press.
— (1984). Syntax. A Functional-Typological Introduction. Volume L Amsterdam: Benjamins.
Hengeveld, Kees (1992). Non-verbal Predication. Berlin: Mouton de Gruyter.
Jacobsen, William H. Jn (1979). Noun and verb in Nootkan. In Barbara S. Efrat (ed.), The Victoria
Conference on Northwestern Languages, 83-155. Victoria, B.C.: British Columbia Provin-
cial Museum,
Kornfilt, Jaklin (1997). Turkish. London: Routledge.
Langacker, Ronald (1987). Nouns and verbs. Language 63: 53-94.
Lyons, John (1977). Semantics. 2 volumes. Cambridge: Cambridge University Press.
Pustet, Regina (1989). Die Morphosyntax des "Adjektivs" im Sprachvergleich. Frankfurt am
Main: Lang.
Rösch, Eleanor (l973a). Natural categories. Cognitive Psychology 4:, 328-350.

Brought to you by | University of Queensland - UQ Library


Authenticated
Download Date | 7/11/15 10:56 PM
How arbitrary is lexical categorization ? 199

— (1973b). On the internal structure of perceptual and semantic categories. In Timothy E.


Moore (ed.), Cognitive Development and the Requisition of Language, 111-144. New York:
Academic Press.
— (1975a). Cognitive representations of semantic categories. Journal of Experimental Psychol-
ogy: General 104: 192-233.
— (1975b). Universals and cultural specifics in human categorization. In Richard W. Brislin, S.
Bochner, & Walter J. Lonner (eds.), Cross-cultural Perspectives on Learning, 177-206. New
York: Wiley.
— (1978). Principles of categorization. In Eleanor Rösch & Barbara B. Lloyd (eds.) Cognition
and Categorization, 27-48. Hillsdale, N.J.: Erlbaum.
Sapir, Edward (1921). Language. New York: Harcourt, Brace, & Co.
Schachter, Paul (1985). Parts-of-speech Systems. In Timothy Shopen (ed.), Language Typology
andSyntactic Description, Volume l, 3-61. Cambridge: Cambridge University Press.
Stassen, Leon (1997). Intransitive Predication. Oxford: Clarendon.
Swadesh, Morris (1939). Nootka internal syntax. International Journal of American Linguistics
9: 77-102.
Wierzbicka, Anna (1986). What's in a noun? (Or: How do nouns differ in meaning from adjec-
tives?). Studies in Language 10: 353-389.
— (1995). Adjectives vs. verbs: The iconicity of part-of-speech membership. In M. E. Lands-
berg (ed.), Syntactic Iconicity and Linguistic Freezes, 223-245. Berlin: Mouton de Gruyter.

Appendix
·
For each language, the column on the left contains the verbal members of min-
imal pairs, while the adjectival members appear in the column on the right.
Unless otherwise indicated, the entries in the two English columns are also the
translations for the entries in the respective columns for the other languages.
In many cases, there are multiple synonymy relations: there mäy be more than
one adjectival equivalent for a given verb, or vice versa.

Abbreviations:
prf = perfective
prc = processual
trn = transitive
erl = energy release
> = overall numerical value ascribed

For the use or non-use of shading and underlining see Section 3. l .2, second to
last paragraph.

Brought to you by | University of Queensland - UQ Library


Authenticated
Download Date | 7/11/15 10:56 PM
200 Regina Pustet

<
. ·' r . „ Υ . X ' - N $. "* \ - / 'f ;"' \ - 'r , \ ' W ·'

v
:2S
N ; :
^ SP ^ t£ O
P ^ ^ti B ' ''-'"'' '^"' ·' ' " '
0 :
O C— Έ ^
°o £ cx cxiO
<3 cx CUCN ^^ P O Qt f\ C^l - '; ' ' ' ' - ' " ' ' v >V

1 ++ l Λ Q l l -H l Λ
^J«t l 1 1 1 1 A

co :'./.. ΐ''"·;·;·' - ^ ·' '. ν' ! ''χ'.ΐ" -V": ,

D
H I 1 4-4- l Λ : , ^) I 4 - 4 - 1 Λ

·§ 4*_?4l! .^Λ ΐ^ΐ^ν- · //Λ


g aS^SS'^^i^
a a cx cxoa
1
00 1 1 4-4- l Λ ^4-;T i -W-LA-r:V':";;" : : ; 'v-M^ ;

• *V S %
S^ ^O Γ^ " « *"* g^^ . y " > · ' ''.Χ'

: ;
S rt ' · ^ ' ' ' · ' - . '

V^ ^ ^^ ^^ II " · > t | A ' ·. x


^N^ ^^^/ ^Ta ^4*4 1 1 ' s '' ~. ' 1 * 1 1 f */\ *' '^>'
- ^^^ ' ·...-' · v" '
tu
ω 1 2**
O O rn
^M u -: . * ^\ '··*r ' * 'V<-'"
: ,;;·.^
'

O
V)
-S1
«X
o H cx cxcs
\ + + f * A c*j 1 4-4- l Λ
C ι **·* O

>^ > >%+ t !4i 1 Λ v„ ; >

'SbHi 8 Τ?^ΛΛ./· \":^;;/"· ,^'V'


GERMAN

ε P 'S t! Η
brennen

£3 D CX CXCS S <3 cx cxoa


1 4-4- l Λ 1 4-4- \ Λ '§)+,Τ^ΓΛ
·*· · \
!·§'"'' ^;,«/- r /''V-:rVv··'.···;:.^·:
ζ> ·* ^J ο ".
^5 E *^5 ^ S " . .
8 -H Τ-^^Λ ; ,;
SC
C/3 ^ δ § · -;ί ·'""·' -^' ^· ---
2
ί 1
<3
ρ _ι **-ι υ
a ω cx CXCM
\ 4-4- l Λ 1 a cx CX<N
1 4-4- l Λ
Si 5 cx cxr4 *?iS u
C ^
0> s fe *
CU Ci***^
^ I 4-4- l Λ §4- i 4i l Λ
•'S C ^-* Μί ^

,:
^ ' '"

1
s s
^^ . , '' * - "

aPTjt!
<υ cx B
cxts / . *ζ* O Cu CX *"·* * · ·'··,«'-*.·
β 1 4-4- l Λ §4i ι 4^ ι. Λ · ; V
. > - > ' ' · '
W fc Λ " ' ' '
FRENCH

/ ' · · ,

ε £3P T3α> *t3CX αCX<N P TU<u T2Du CU<N


C ^ — <4~t O '
-
«S, 1 4-4- l Λ 114·+ l Λ 84- Τ-^^Λ .

·*** ·· ^^ ^ fj *^j * j ^^^""u^ o


11
<3 + + f* Λ «l 1 S1 +S +cx o*c*
ip^VP
l Λ
^P^^^
cS S S o, CN £ S ο* α,*-< . %> B i> cx cx*-^
cl 1 ++ l Λ 8 -H L -H t Λ ;. &-« l -H l Λ
/* « ' ' · · . ' · ' . . ' . · ·· - ' · ' · / - ·
ENGLISH

^ O · '/' ·..,;'·'·*.; . . . . . · ·
g« j <+* p O es »—» H** P ^O * .
11 + + l Λ s * ++ ι Λ
ι*4 _^^ ^^4 C^l · '
^ bf\i te
ζ* Oc %) CX*™1·*
4
j
S <U Cu CXCN S CU &CN ' '
; 34- 1 41 l Λ .

Brought to you by | University of Queensland - UQ Library


Authenticated
Download Date | 7/11/15 10:56 PM
Ho w arbitrary is lexical categorization ? 201

S
O
.. .. z
u " Ck. W D. D,«— &« Ο CU Ο·*·*
• + I - Η Ι Λ .3 + l -H l Λ o^-H 1 -H l Λ <2>-Η l -H t Λ

o: ^hg i3p ft.


J -On D.
i gD.-^
O
S^-f l -H l Λ
2
ω
g" T-.<e
IM « ·· ν>·
4> D. D.·-· Ο CW W D. O.—·
+ l -H I Λ Ι -Η Ι Λ Ι -Η Ι Λ
CQ
Ο CQ
ί/5
g-cts
o D, HO
1

S d.—
-f l -H l Λ
α. υ D, D.—«
Ι -Η Ι Λ
2
ω
S ^ z°
^j .^ - ^
υ
^> Qi CL*··* "S S ft. D. D,^
l -H l Λ l -Ή Ι Λ 5 H- l -H l Λ
ω
2
U]

(U D. D<*~«
+ l -fi J Λ
υ 2 s
s* ω
δ o
2 l
Q) CU CU*·^ i3 CU Ο Οι fX**^
4- T - W ^ A l -H l Λ -H 1 -H l Λ + l ΉΙ Λ S + l·-« l Λ
U
Ϊ
Οϊ
··*-
ω S CL, O CU CU^
ο + l l Λ •f l -H l Λ

^> ·->
•^J OQ

O C
ι -Η·Ι-·Λ
O^ O VA4 C**^1^*

l ~H l Λ l -H l Λ -fi
l

χ A
O O
s: CQ
«
2 G
**» fc T3
*-« *Β h? ^
fc? CX^-i
S
c/5
<X) Z3 0> CX
δ + l -H l Λ
*3 CU O D« D.-^
+ : 1 -H l Λ

CQ CQ
ΪΛ
P "^ t3
ob B D, D,«
-H i -H l Λ Τ^^Λ- ^fl ^ Y-fff Λ

m
CQ
U J ° co
Ζ
ία IM UR2
i T ΐ-Η Ι Λ ^+
C
>5 ^3
t »^ϊ *r^ Tl
fit *-<
(^ *^
c? Ir*·
^\i ^
jj^*··*
· Ι '-Ή Ι Λ

s0
Ο Λ CU^-<
g Ή «g a
£3 (U Cu Cu»~-»
1 -H ) Λ -tt l -H l Λ

ac
C/i
CQ
J
Ο
2
ω ι -H ι Λ . r tr Λ

Brought to you by | University of Queensland - UQ Library


Authenticated
Download Date | 7/11/15 10:56 PM
202 Regina Pustet

. »i ι ^ \f\
T3 t« ι-* ·
<U CX. CXO l UcxS
l -H l Λ + T-H l Λ

Ο Α 3%
3 ^* £3
o So
-H + \ Λ ~·—\· '>VT J*i'~;*V "*

0
S S 1
e-e'egQ«—*
2 .t» « Cu Q
Q *t2 B Ό
Cw Cw-<
O
55- g
-ct: H"!·
o CX,Cu—
-H 1 - H . l Λ :-« l Λ·· ·.δ .l -H Ι . Λ

».'8
l
·** · ΛΑ v^

e-cti O
ο Λ CU'-^ .
C/5 . + ι .+ ι Λ } -Ή t Λ

^ O

+ C-χ O^ O< ΟΗ*·^


5 -H Ι -Η Ι Λ l -W l Λ
U]
Z U
ω <
o ·
CM
C/5

^,2 <
O
;. Z
*-^
t , t^ o
| ^ J^
o> ex ex»—·
H

χ <D CX Cu·-^
δ-H l -H t Λ + ". l -H l Λ

Z R: υu
<
s
cc 5 r; ^_ Cj-j ^> δ S o^
ω g S O CX CX<S +.2 fc;fc.
o 8+ l +.1 Λ l Λ
o
δ ·«
G
*$+** ^T+i't'Tv
di CX.O
1 l l l Λ

ac ^ β ' ^s
2 ° ^ -v "*-* CQ
z 'S
> H <3 ex. CXCN
g ^H ^ y

co 5+ l + l Λ A

ι H·* ^
^3 o Q4 Cj*'™^
^^ ^"^ ΪΜ? ^Sf
»S £ o> ou di*—«
-H Ι -Η Ι Λ -S.-H l -H Ι Λ l .-H i Λ

K m
U 0
Z
o:
&'vv
i3 »cω *tiex ex—ϊ
H \ Λ 1 -f l Λ + l -H l Λ

V β
ο., <υ Q* cu*—ι
-H l . -H l Λ » -H Ι - Λ + 1 l Λ
m CQ ^
O m Cw ^^ sg
* *·*( *^

Hr ι -f 1 -H J A

Brought to you by | University of Queensland - UQ Library


Authenticated
Download Date | 7/11/15 10:56 PM
Ilow arbitrary is lesical calegorizjation ? 203

;_. ι ι ι ιΛ 4^a ι ι ι ι Λ 4 l ++ l Λ
QCk
kKISH

4M + l Λ

E u - ^ Ι Ι Ι Λ 6 -»< -»M Λ

CQ A

Ι
-r
•^ S ο Ε. Β.Γ4
-M 1 -f· l Λ I -h l Λ c 1 -f -f l Λ

-H l ι Λ
u) u
/;
wU
^
0
• l + + l Λ

Ι Ι + Ι Λ
GERMAN

P*
Lj
^
2? S

5; :§ -H l -f l Λ
*L·^.

. ,-
U kW UM··« Vj *^ *·* V W* WU—·
'ς <2,-Η I -f l Λ 1 -f -f Ι Α

'S'S.g.n
1 ++ Ι Λ
SPAMSH

1 +·« Λ

* v- .
3 3ξ "
.. 'S S 0 0. 0.<N
8 1 1 4M Λ b 8 1 1 ·« l Λ S,^ 8 l 1 -H l Λ § 1 ++ l Λ

Ί•Sa -P-na> D.
F ENC

.
b -f ι -t- ι Λ

sE -a* ' ex
S G.—
^ ^<
•H l -H l Λ

CQ
ENGLISH

aP-et:
w a, 8G.
+ ι+ ιΛ 1 | Λ

Brought to you by | University of Queensland - UQ Library


Authenticated
Download Date | 7/11/15 10:56 PM
204 Regina Pustet

Brought to you by | University of Queensland - UQ Library


Authenticated
Download Date | 7/11/15 10:56 PM
Ho w arbitrary is lexical categorization ? 205
• «· v - - .·- · - · .·,«.· · - - „

υ υ
u υ
<
·* c-ct
t: <
? P-r'e B
g ^1
*ί* ϋ o cx cx*·^ •^ ag -et:
o cx Βcx—
Ό
u
3C o χcx sex·— ^5 ts ^«o s?cxs?cx— C o o &· cx*·^
.v
-H 1 -H l Λ §-« l -H l Λ ^ 1 1 + l Λ . J3 Ι Ι - Ι - Ι Λ "g 11 11 11 Ι Α

u
I?
IRKISH

.j
H -Ή T -f fl\

1 1 V·? f Λ 'S 1 V + f Λ
P^
m CQ
SWAHILI

O O
O g"-n*t: ΒΌ
u ex cx—
1+ l -H l Λ
£ S <υ cx cx—
•a -H ι -f ι Λ

«M —^ ^^eA f ^
•Z *^^ ^^^ J^
^.tS 1> CX CX·^
H~4 1 1 -f Ι Λ
2
ω := | |
SLOVENE

| Ο
!^| «3 o CX CX***"* ίί > "S * S S CX CX<N
+ T -fr t* Λ iTil ι ι + ι Λ 5 ^ 5 Si-H-fH- l Λ

•S5
c^
.. ...
**· c — *ti 5*
• S o cx ex·"·
£ p-ct: B
:3 S 0 CX CX<N
A .1 l H- l Λ. s-Ή-ΗΗ- l Λ
U
2 Si ^C C

e — V" O »n ^O e — <*-> t> ΙΛ


OS, C *·* s t · Si R IM i·* «·«
U* •s:
·—« «3 U Qj M«^^ Q S f\ CX""*
O + l -Ή l Λ S H- l -H l Λ § -H V + f Λ • *
^

ς> 0
g .g «g g «ο
*2 O Oi CN*·"* ^
£ £ p -G <C
^* a §g-gUct- ^
O Ot C^***"4
Ό
'
.s>w -Sε o-c^ B
cx ω cx cx<s
Vi + Ι-ΜΙΛ -S 'S H- l -H » Λ "5 1 1 4- l Λ ^-H -H + l Λ
»-» *^ · ' . N, ' .·'··.·.;· " -
SPANISH

O. 'V' /.:. - ' - : · ' - . ' ...


** »x VJ · ·

1+ T -H
PΉ ^ o
l Λ
* **·' H **"* Λ. r?
S -h T -H l Λ
.h P-ct!
•>-H 1 +: l Λ .
^W -ft ^ ^S-| C^V·^
"!
. . . · : . .

^ ' · d ' '


3 » o <
^ C · —» ^ t j P ^ J *^P'** t5fcf
"C £ cx cx^ *C a o cx c
<s> m+. ι Λ < ι -H+ ι
3
o· ·
FRENCH

^ _ ». _
Λ Ι ^^ »^^ ^H^ ^^ ^^1
u. H *^ ^ bi
1fi V -8* f Λ > *3 (D Qj C^*"H jj* ?uj ^\ CL< Cu o*^
*5 -H 1 + l Λ ^L-H ++ l Λ
^ '§· * ' : - ' :; ·
1—ε —^ <*-«ex ocx2«o " ··'','·
-
-
'
' &) r^ ^-H ^*H O
^ C
•^ w ^Oi CXi i CU'"H
?
· '
*
·· · . ·-
.·"-··'
o H ^3 *t2 H
"C a υ cx CX<N ^
•8 -f 1 -W l Λ Ή 1 1 H- l Λ
B
m '-· ' "'· ./··'· · - . · · . | ':·^·\ ·,.··.·: . · . ·
ENGLISH

o . ·,'·'. •o . ; f ;. · ; . · _ . .. :; .· ·., -
<J 60 .. . ' .' Λ
**^* a o QI p i*··* · * v S Ρτ:*β HO
s + "" ΐ-Η ?"Λ"" S -H 1 -f 1 Λ : ;
** S 0 CX CXCM
a-H-h+ l A

Brought to you by | University of Queensland - UQ Library


Authenticated
Download Date | 7/11/15 10:56 PM
206 Regina Pustet

^e-sts c.> *£· ε *^-* ^ b ^*r "v^ .' " * · ^"--" ; '^' ^- - ' - » '
i τ ? + f"X ^x pl ^ f^j pt<^p - '··
^ l : Γ - Η . Ι -Λ·-.';-.;·. ..;^,-V-h:'i,{::
- '- .:'·'-,··.-,.

^.s -'ι".;~*[· '. .-" : ; :'·'.:' .· ..·-'. '.· ··"·'' ; "·· : ·' ^
3C
^ ^
<2 -v Q
1 g -et:
. ^ .Jrf* .'-·'· '·;', . " ' \.' .' :-·;<'·'.·.; . .'..\V "

Οβ
*** Q> OH 0-CS ^ ^3 o CXiOu·"* ' ' · ' '· '· '*> · · ''"·
Ε- *+ 1 + l Λ 2ί(·Ή·.| + l Λ;,, /.ν. ' .;/;·': -.L~:'::.'r -

: :
*Q · ·: - \ / : ' ·.' · ·'''···' · ·. V·"'-.'-' - .
, ^B &, o.
ε 'S t » -H.I Λ - i
α « .'; -Q'^S ' - V , ' .· ·';''. "^. ·".:· · ,;/ .(··
1
Ξ) j*
·*
Q
fc V.X*cx, ' -\."i · · ''·'=·' V ' . " ' " ;.'·'' '"'-••.-

11,7 Ϊ +
*^
s
»—*
vj - ' - A\ »3 ~- -. ^ . " * ι "·
O ^ ) ^H O ""*· ^ t -g C^«4 O
ϊϊ ζ5 C ^ S. o^ *^* fi ^ o. fe.
C/5 Ι Λ . 2» T 1 + l Λ ^ 1 1 + l Λ •§C*?*S ΐ 4 ί + - Η Λ «a.TT + f'A

. ω . > . ' · ' - "·"· ''-"-.


r? - ' ''. Ό····· ·.'- ..-?··· ' -"'· ' · '
*···* -^ ^ o* 0,2 "§ g 2*5*0/11-«
HS> ~H 1 "h 1 Λ ' 5ί-Η Ι -Η Ι Λ
**^ · , - -.' "
ω ·« u

I"1*W- · · ' · ' "


SLOVEN

.< ^
« e «r·« *
§ S « a , O.C4 . *o |^ j** Q^ f\^ CXC^ ' ·
§ + •• '1- . +
·*- *
*' Λ - s
- · <v<
·' > · v
"·· · -· * ··
2-H · .-H+ l Λ ' ·
fc. '. . v. · >'·

^*t ^^ ^^1* Τ·*Φ Vj^ 1^^^ ^ ^"^ ^M^ ^^"4 ^^ ϊ^^^

^fl T $* f" Λ ^4ί Τ + ?" Λ


Z ?U gt
< ^> <c ^ :^ .
2 4
O£l ε- « cL<s
(U ^5 S <*v § «?T2 v3 w«*
o Qt CXlT^
o £ + 1 + ?*Λ •ζ-Η -fi-f l Λ

ΐ·§1
^^ ^^ ^
Q ΛΗΛ
PΛ Ή δί
3 5 S,*> α. cl— ^ι i3 ^ CX f^t^j^
ft, M 1 -f l Λ "β Ι Ι -Η Ι Λ
^->
X £·
*** ftk O

CT
^*
*Si^
^3 0
§ t v ^"^
1 ~ — .TU^ -.
εχ
^| ^
»!b *" , o
i^ £Z P "C w u*«
δί S Ή f\ Ou^v| SJ *^ ^ pi^ Q^^s|
C/5 -ΗΙΛ ^-f-14-ΙΛ

Ig-O-BB
!/fi T ·& f" Λ
N
^n
v. 00
u S> o
Z
2
LU
αέ Ιε-βΐ
C»o p ^> CU ίετι^Η gj ή^ OM C^O^I
U δί + 1 + ?*Λ
^* " *
^^
•NA ^^ 3 ·' . · .
«§."§
^

^ P ^~* *C o m
έ**
b ' e ^ — ViP«o
Ί i
^O ϋ Ο f\ α S o> 0,0.-^ &·£ Q ^ o cw ο,—ί
δί-Η 1 + l Λ % 4ί ° T + f Λ ^c-H 1 + \ Λ ^-H 1 4- l Λ
*-*
3C w PO ^><·
£O
C/}
**· β -~
2 oo · ·
O δ! ε "C *2 ti
° . .Λ
5ϊ* ρ §_·ΰ2 *tj H
a
'
Z
U4 a-h V + ?"Λ afl 4i^?"A

Brought to you by | University of Queensland - UQ Library


Authenticated
Download Date | 7/11/15 10:56 PM
Ho w arbitrary is lexical categorization ? 207

J
*tv>
*>
§" τί
«3
+ l

ΙΛ
2
S M
D 0> CXi Ou,^
H l -H l Λ

CQ

Ou, DN— «
4- l -W l Λ

CQ ^
c: m 'S O «
ac
g l»
4-3 O* H) Gk CX**"**
+ \ -U l Λ
4) O, O.—*
l -H l Λ

Q
P 72 *tl H ^
V"8*f"A
ω
2
UJ
>
O
oo

* ..Z
c^
,_ .. ^ ^ j
* i··^ i "t ^^ 1Λ

+ ^ T -S4 f Λ

OS
UJ
g o.j-c^g^
0) O< fti'-H
O + l -H l Λ

K s?— *-»


00 ^ Q PQ
1-H
Z
S
00
CX CX^H
H l Λ

m ι Λ •l
<i> 04 ex«"·4 k. O CX I D.CX-
Q
l -Ή Ι Λ &J l -H l Λ -f l -fl l Λ

u
:: §
CQ t
00 .&1 C3
Q
C
2 ^> ; o u p
(U <U P
OS 5X>
[i, l -H l Λ
OQ

<L) CX CX—' <u ex cx<~*


-f l -H l Λ -H l -H J Λ

sO CD^
-
2
UJ aii? 4> CX CX*->
l -H l Λ

Brought to you by | University of Queensland - UQ Library


Authenticated
Download Date | 7/11/15 10:56 PM
208 Regina Pustet

- ,·. .· - · · ,..·,.>,...· · , · — ,—χ -·:.·»- τ, r.-·-.*?' ··<·*·-:-r. ?-f;. ·*,·:-· ;-.·.·-..·^.·-)>,-'Γ,Λί-.···;···;·· Γ;>· •^ rv^· '. -·«· ί";"- τ ,-"i ··,>·' ···—

Η '· · . ·. Η ' . · ·. Η \ · " · ' · ; ; { .'." /':·:-. --,·' ΐΊ . ''·' : '.·- /' ;', .'' ''"-. , '· /'. ''"'·.· ':. ' ··-'. " ' " " ? ' "' ' - > .' ...
: ;
j. Q '·· · ' · Ρ "· .-· ' ;'- .« ·' ··. ''.''·· ϊ·*f ^:·\·' . '"'.·· -V '"ι'.?.* ""ν*·: · ~'·-' '· « '<';V^ - \ . , V -
: Q ;

5ό ρ τ3 Έ 0
}i 5 Ο Ρτ Q*(S
^ Ρ τ: t} Η Ό 1· 3 ·Μ τ3 ts
c* +ϊ Ο Ο< Ο^^* ' · ·Ϊ3 Χ3 (L> Ο* Ο4^*< ' * · .
'8 <oV·?'^ '· ' ·'.
14
·- Χ; ν·Ύ - :.
· . ' "- , . . . ; . » . . .
;..' · .' : ·^ is'-W'
-8 § Έ U
..v
&-χ

·§-Ή-Η+ Ι Λ .^-Ή-Η-Η Ι Λ ;.-3^-H-HJ · Λ^:. Χ; :


- -../J · ;:..:;;·Τ..'· ·< %ν'" it 1 ,l/,4- J
v M :A
: : :
·'.···'. _. _. ·, ,·.'..·*'· ' , '-'-· ν* '',/.-. ,J :-".>':'; ;;V -:;'·'. ·,·;'·; ·'"'·, *,·.<; ·Λ ιΧ.·; '' ' ·'v ·"* '
Χ
co "8 -χ-χΐ ^ ' ' ·Λ s'S-.i-S · .^|:ν''/λχ^':^ ^'.>V.V'·'-.·"
2
OS ^ ο1^!"«^^ S^ | ^ ^ 0 3 | Ι'δ'&α,οί/, :| ^^l'S^&
? S 4> m
d* CX^-4
Η a] 1 1 + l Λ

3 _
SWAHILI

2 ο ^ ^ΓΝ
^ s 0 ο. ΡΗ<Ν £> 2 2? ^ ex. ο,οί 4
^5+-Η+ΙΛ
" · - < - ; · - - · · 7 ^ ' . °* /-^ ·-.·"'·---
^^§ΙΗΗ++Λ
— . -;./·' : * · ' > ; ^ . .**.-.,·*. '*»$-' *'^\*\ * Λ/.ν-*^··*^. ";··\^-
3 7 T + T Λ
' ' · · -· * .. \ , ' "· , ' \ ' ·· * ' . ' > :~

:
^ S <u Cu cud .. " ;,... * ' . - · , . - -; ;v V /·' >/" ; " / ·ν -· - ; - ·/ ; ^

:
ω
ζ v"| · " ' · . : ·· :^ : '^- : -;vv '•^-· ; '.!; ; νί: : ':? ί ί=?Κ· :
ο ? § "^ 5 Έ« P.CM ' ' ^ ·' " - ' - ' . · ' :·..·:· ·.·' , ' ' · *J\\'. *"·.·'''' :' ^S ώ ^) P-I CX<^^
CO ^ I-« + l Λ ;· ·. -' .,..·..· . ;..· ·; : . - / ,;;. ' ' '- .·:';;,';·,ν.1.. ' ' '--' : : S· 1 1 + l Λ

« ' · ' · '. ' ; · · '""''·· -·.· - '' V' v -r Λ··/'.. ':*'.: '··.·,.'V-:
;
£< S ex. D.CS
h OJ -U _L j A
'- -^
·. .·
' - · . .·.' .' · . " "\ :.\':-.'.TV; ,;'·.. /.·"- ·.' · '
· _ ; ' · . , : · , - ; ; · . · . . . . : \ .,·.',»·..·····-...;·
:W 1J 11 1 1 / \ . , ·' v \ - \ - " ' ; · ' ' }
4

u
GERMAN

c :=s ' " '· " · :·.· "··. . ·. - '· ··" ." ·''·';·' V'·"· §<
^ c — *t* y ^
^ ^ ω p, cl'-i
'§ :|45 41 + .f Λ , _ . :-; . : ' · ' " . • ' * : : ? V ! . · . ."Λ, ·\'.ν;··? '>-· " ' ! ,, ^ f i 1 + l Λ

χ—s
*> V. si
δ 2 ^^
SPANISH

ω k.
•S ^ c^ ^ ·»*»
^ P <UTS CX
5 5
Έ CX^-^
^> c^ ee ~H ~H "f* ~H Λ "§ 1 1 + l Λ

£··* 2 ^>
FRENCH

^"^ ^fk ^J

^-^ O D * - » £ v^ q3 o^o-ioi jtJt -2P τ; τ2 ι-ι


OJ Ou CX*· ^ 1

·§ 1 1 + l Λ

Ct ' R ·" · · O^ O^ ^ ' ·j ' . ' · " . ' ' ,% "· j


ϊκ »^ ' . ' *5^ β ' '',. T ' ' < " ' '"''/"' l .' - - 1 \f · ' ·

§-H 45 + f A Ifl-S + f/X §-H "*-S^f A; ! V - / · : "§ 1 T ? f Λ


>"* ,-*· - '" ' »* ^ . ' '- ' - v . , "" - , -.· . - ' . * . . · : ' > .
ac 1§ ··. ;.·' · ' · '. '/Λ, 'r, ."-V :"·-·· · ' " · ; · .·.·>'/.->:" γ "·;.;
:
*. ' ·'" :-
β * ". t '.'^

M, O
0 ^** A·} 1Q^ CL· CL· ^^l " ^ ·* · . ' ^ ' *v · » · ' · · * . ",'.··'
ω *Φ^ *"^™ 11 —
l* 1 /\ ' ' . s . * ' ' * · ·' ' · ' * · > , · > ' ·. ' a-« 'T·??* Λ

Brought to you by | University of Queensland - UQ Library


Authenticated
Download Date | 7/11/15 10:56 PM
How arbitrary is lexical categorization ? 209

to H ' H '·'
O - · '
1o 4g ° ,S ·' ' · ' ' *ί
Q :
S
^J
.Q ··. ; '

!
5
rjcgcclv-t^pir) - · "- ' ^Vi P
H v« u M
-H ° T ·? ?* Λ 5 -H S O S &> CX CX^-ί
1 *+ \ Λ
. ,
iI-H 3 O Cu fX*··* *'' *··
1 -H l Λ £
b
,-H
o cu cu—
1 -H 1 A
00 W) ' . '* . '. - H' · ·' '
•Ae o
C/5
c -i< O C \ · '* *i
Q
1
C 'S
2 S « Sl^tHP^
a « l - to.^
jaW
1S
^

οέ 5 H υ
.. ^^^ ri ·'
eo e *-· ~ ii *
S
5
^e** ^^1 f^

D C « Cu cu-< S?
H 5 -Ή M 4- l Λ -N T -fr F Λ

W) &Q
0
*"5
Ξ5 .c o ° .G OQ
o
s •· ^ 3 ti_-r> ^ C ^ "^ · · ^> *i
g
a

g^ SJ C3 ^·2 ^^ ^™^ γ^ £^ ^ * j
-Ui S 3 S cx cx^
\^ ^^ ^^ ^H c« ΐ·2 ^^ ^™^ ιί c^ ^*·
^ ^^SoE &i o<^ *ξ ^j ^™* ^ti

CX Du*-*
LL·

K·^
(Λ s: -H 1 + l Λ g^rS'3-H 1 + l Λ a-Hl -H l A
H % '
o ' .· . · , · · · · . Q ,··'. ··.' Q
1? H -r? t!
·§ iisliLcC, -1
1
;^
^ ' '· ' ' " '·" - •· - *^
R rs * B CX CX*-«
*» fl T-f l Λ -H Ι -Η Ι Λ S -tt 1 -ti l Λ
ω x
, ' ' · , ,;: ' . , ' . ' ' " · ' ' " '
1s
2 o O •^ 'C . ·
smrdet
SLOVE

Q
1-H <D Oj CX^^
1 + l Λ fi ?+.f*A '
v
: . ; , -H T ΗΗ* f A A

. " · ' -' \^ " "

A
'' 1 '·'·'.' ·' '·' ···:", X '
: - ' t '*.
fcC
ir ig-ct: 2 '>" ' "
•S S
, U.,, Q ^" % ,,· \ \ , ' .

dj CX QM^·* - »* ·. ' ·,>"'>-- c S <U Cu CU·—·


Ϊ5 1 1 -f Ι / Λ · · ;·- -·;' . ' : -: ν :
< -H l -H Ι Λ
-5!
O
•s:
u . · · . - ·· ' . · ' · ·'. . . ; ,", - ', -;-·"
; - · - ;> ' . , - ' , . .', ,. · *"2 '
·'.*':·.·;''
-
Q M
s: s:
N
^ζ v . '

Q- V··.·
sc
s; ·' · · * , - ' ' ' ' ' ·';"'»''"

1
M δ
tt!
5 D ex ex—ϊ 1-Ήes T*ι *e-f αι ^A ···· ··;···'< ' . ; —/:;-. -;;'·
• *

C Ο^ Γ^ι
5i Qi
»

ii i**<*
m v ^ ' ^ ' '

, . ' ", - · -
'

\, ,
'

- ·* '
"-
- :'-
"
· ·· · : - " ,
" * ' " ·

UJ
"C -fl 1 + l Λ s +3
:
*

' \ ,' ' - · '


.;' '- . ··- ;- "·.:
·; ' '\ '
:
s< -Η.Γ-Η,Ι.Λ.··'
S O CX Cu*"*<

-Q ,·'.· ; · ._ ' «
w
o
" * · ' · ' ' v ' ^, V ' ' j, .

> · * ' · " " . . ' * .' ζ$ ; < .·'/' ^ *H


§ ,·"' ' · " / - " . ' ^
t5 ft, - ; f'- " ' " J , g ft.
^ ^^ g^^ £J ^^NJ ^ ^^^^ e*^ ^ ' ^. , .'
| V* r. ^ ^v , ****
",-<<fc! P
*ft- § 4>

^ ϋ <D Qi Qt i»~< *^2 *«3 ^ Oi f^4 ν«-4 - S -<u o^ 0**·^ ^ CX.
t G±*-*
tf
ι -f ι Λ ίδί, ι ι -f t A ; v - H l -H l Λ g -H l H4 l Λ
,v
z
C/3
Q ' *'· ' · '· . >'· . - · ' . :'·; " · · - ' · . ' ; ' ' ; · .'.·'.' o. .''; .··}·.. ·. .
^
Q • * " . · l*
2 "ρ τ: t: a Ό 8 : : : ;
f^. IlsUcLS""· -"· '· · ^ --'" :""V>?.· -·
;
. -'''·' S iilsUcL-« ·
2 "o^ -H
CX»
C/5
Z^ <t) Q t (X^H

l H- l Λ •-H ι + ι Λ -'' :"· ^ ..,;·; - V '-{'/. '"··'·. : '·· 5 - H l -WM Λ


C * v ' --
- \, . ,
'','''%"/'''.'-·'
^, , -- \ . ϊ ·. '
. ^ v
"
'!-'·'^
,· '
Λ ;- ,
ai·-· 'κ.'' "'··: '
O • S, ' ' :.'
:
]§ g T2 *t3 H '^"· ' ' ·. -^ · " ; "'Λ '''·' - f · '' ^·' :: ' ; v^ %- -- l gTs'c a ..
s -Hu OΙ -ΗCM {ΐ**^
δ ι ι -f ι Λ,.· ; : : -':-".:. .;;v;^·, ,···'· ;; •'••'•: !
- "s" · ' -N· -
8l
ζ» <D O^ Ο ι —·* . '
ΙΛ
fc.3 -
οΞ
' "3 v·' ' ' . , · . ·. : · . . . · · · · « ·\'·\ ' '
u
2
ί τ ΐ et e^t: a^
l^ft 0" ··".
P -53O *ttCX aCX'*^
Ό · **·
· . ; ; % 0. '··-.,:···· .
g ·α *tt H "Υ ' . ' : .'-- -' '.'··
· · * . · : · 'i.-. '·:
''
*> 6 o *G *t3 a
*5
Ο ^ Λ- "··;
UJ
oi £3 4> Cu Cu-^ J^ ^ w &) CX CX11^4 '·
u. «0 -Ή Ι + l Λ -H 1 + 1 A . §,-H I -h.f A'; g •H ^ T - f t f A
'. - ·" ' / V ·» '' ' "· - " - '· '· W
o '·· ·'·· ' c
o< *ΓΓ ' ' »Si *T?
6 *c *G a tt
»" >. '' " . ' · , . - ' , ' --v -

P^jj "C *t3 B •S 'SU&'-M ' · ·=''"·* '·"" ·'' '· ·.-·· - · · ·'"''· a
Λ E T! *t3 u
i jik CL· CX^"^
1
Ό
^ 1 1 + l Λ 7 T + Γ Λ . '·- ;/·.;'::;' , '.·' ''_· δ -H T -Ή ^Λ g -H T % f Λ
I : . ""'·' ' · ·. _ · ' .·; ·. ·' * s

00 o* •^ O^ ' '· ' , · · ' ; ·' ·' tl H^


^v
0 § β'β'Β ^^4
2
UJ a -H
U <U CX Cu^
i + l Λ · S^ T + fK ' . : . a^ T-ttfA..;

Brought to you by | University of Queensland - UQ Library


Authenticated
Download Date | 7/11/15 10:56 PM
210 Regina Pustet

5<^> E -B<υ t:
Ou· H
Q.O O O, O.—*
δ 1 l l l Λ

χ
C/3 α ,*-
•^ o S'S
.S Q ·|¥Ί:·.·'·-·
Ρ ~* *βou a
c
D
H

2S
α
ω Cu Ο««—ι l 'eg
a· o, α.'—«
C/D S 1 + l Λ ΙΛ

»
S
P T3 *t2 £
S QJ
<U Q* CU«—*
O*
Ι + Ι Λ -Ή Ι -Η l Λ
α <x> Q
2

Ο ^5 cxi cLi^-H
(Λ Ί-Η l + l Λ

g Γ^ETJ*« g
Ο r^"i ^iX|
§ -Η V+* f Λ S-H I + l Λ

Ου
UJ
-dt:
<u o
^ 5 C
s t3 —
*-· <t!
1
fi fc ^*
Ο Ι:Λ §fi T + l Λ

1^1'S U P-^
5-H T-H t Λ
SPANISH

-8
S
g SE -aδ ts ^
α, α,~«
ι Λ 5 -H l + l Λ

*** Γιι^ *4„4

5* C *·« ^
5-Η Τ + Ι Λ
v.
^
C
^ ο*3 ι« ν<
Ο «—<
*S Ρ Τ3
S ϋ . CL 4) D-, •Βΐ.
Ο CU
1 l Λ Ι + Ι Λ 1 + l Λ
^
Ο
,
B afe,^
e -~«Vi P Ό'
'S -H 1 + l Λ -H 1 4- l Λ

«s τίοΙ tj+α,Hcucs
*^k * ·

Ι Λ -H l -h

Brought to you by | University of Queensland - UQ Library


Authenticated
Download Date | 7/11/15 10:56 PM
Ho w arbitrary is lexical categorization ? 211

S e Έ <ti y »*"} ·£ί p ,£ <£ ^ ίο .U α Ε t2 H ""i


-^ S (U D. tX «-Ϊ S 1> CX CX S S O O, O.
S l l + Ι Ι Λ 2 1 1 +1 l Λ 3 l l -H l Λ

s
l/>
^
2
06
D
Η -.3* 1 T + f Λ
2 2
UJ 3 UJ
0 £ 0
P^ ^
^ τ- ^*ΐϊs αt
c ^b" C fc S tc "-*
*-« *t! H
si ^ α -*
.s£ C IM<ti»ίt>Μ^ I
fc Ό <U Oi &**···* iy w <D CX CU^^ ·ί* (U Οι CX^D
rS-HI-HIA κ^-ΗΙ-ΗΙΛ SlI-fMA
SWAHILI

: ;
3 ( V^ B '"'" :·;··· ;;;7;:·- 7™::/
:
Ifi T-8*?*A .:.·. .
υ · ' '· . .',:,· '·, ' · '._·
SLOVENE

o <S ω eil ο«·—ι Λ' ' t I•^ ue^^


O Cu ^
Cu*·^
S4H 1 + l Λ . : - . ^1 1 + l Λ

•^ P * *t! H Ό ^***
^ +3g τιo^ *t;^| H
*"^ v^ 0^ Οι CXO^
Ό
C^CI^ ' ' ' ' 'S S O CX CU*—*
g 1 1 -H l Λ S 1 I -tt l Λ 5 1 1 4" l Λ
R ' R ^ ' . , ·. ' '··''.'.....'.
GERMAN

« ' ·'·' · '


ΛΙ US
^^ *"^ 5? (IX^^—4
5?
·< (U O< Ou^** ^* *^ί <U CU fXCS '' ' '' ' ^H^ Q^ P|
£-H 1 4- Ι Λ ·£ £ + l -M Λ '.."·- «S r i..+ J A . .
o
•2 1
R e-cts ^
^ h <u o, 0.0
• ^ 1 l -H l Λ

X
'
«
2
co

•1 I
|l·· .1 : 1
··. · - . · : " . .·
'H}! Tr"^C -1^- «*^
V ^^
P C
^^ fi l^ ^WM
H -*^^ ^^
P S^
£^ Γ"ί ^
t^V^^i P
*^ ^
* ' ' ' '
^s i3 O l^u Ou1**^ -^j t^ ρ*Χ| Ou^^™"* Q ft ^ £ij C^CI^
δ -H 1 -H l Λ "55 -H 1. -H 1 A i: l ! -Ή Ι Λ .
- ^n . * ' . · - ' · ' · '" ·
^ 03 .
FRENCH

^> <o O ' . ' " , ·

^* S O CX Ou*·^ ·ίϊ O QJ Cu CuO$ " "


S -H 1 + Ι Λ 2 + 1 + l Λ .. . . ; . . · . - . . · · -
1
Q tj
| 1
β
i <3
c g

*^* O G JD^ST t^ ^ p^4 {Χ*Ή


"δ-ΉΤ-^^Λ ^ - Η ^ Τ + ^Λ S-nT-it?"A &-Η?-^^Λ W _O ^ · ^ 1 1 | ιA

3C
co
• ·
UJ

Brought to you by | University of Queensland - UQ Library


Authenticated
Download Date | 7/11/15 10:56 PM
212 Regina Pustet

g
δ P τ:
<u t:
o, Hcxo
"Ί -cs: S
P -nω t2α- α,
H "Ίο "S
S ex cx
o l l - Η Ι Λ S, l l -H l Λ S l l -H l Λ

Z Z
ω ω
o o
"S S <3 ex cx>—' s .5 ex ex,«—· gfc £3
e 4) CX CX O
-H 1 -H Ι Λ 5c-H l -H l Λ S l l -H l Λ

£l P -s<U <eCX CX.O


O
"ί "δ g «c<U «tsD, CX«—
"C
a ' is6 o ' o, cx^
1 1 -H l Λ >-H l -H l Λ -H l Ή J Λ

E -a4> 'sPL« aCX—<


S
-H i -H l Λ

W *tt
— V·* "">·
*·* P
<i> Qi CuCD
-H l -H Ι - Λ l -H l Λ

» g^ o
sf S cu o-c^i
oo-h l H- l Λ
s;

i3 g -s«υ t:ex, aD,»-H <u Q, cx-^ £3 ω CX CXO


«-H 1 -H l Λ S-Ή l -Ή Ι Λ l l -H l Λ

•8 l -β
§P
s-- <uC^Cex ex—·
» * \ *

<υ ex ex*-« 3 hE-ets


w ex cxo
a-H 1 -H l Λ l -H l Λ l -H l Λ

—. <*- p »o
α α> ex cx-^ ex S.
t3-H l -H l Λ 1 -H l Λ

Brought to you by | University of Queensland - UQ Library


Authenticated
Download Date | 7/11/15 10:56 PM

You might also like