Professional Documents
Culture Documents
PHRASE SENTENCES
FOUNDATIONS OF LANGUAGE
SUPPLEMENTARY SERIES
Editors
VOLUME 17
THE STRUCTURE
UNDERLYING MEASURE
PHRASE SENTENCES
by
W. G. KLOOSTER
PREFACE VII
I. Introductory Remarks
2. Measure Phrases 5
2.1. Are MP's Noun Phrases? 6
2.1.1. Prenominal Adjectives in MP's 8
2.1.1.1. Unit nouns: integer-dependent and integer-independent,
continuous and discontinuous, abstract and concrete 8
2.1.1.2. 'Six dusty miles' and 'Ten crisp new dollars';
abstract and concrete MP's 14
2.1.2. MP's, Relative Clauses, and Prenominal Adjectives;
'Non-Particular' Units; Syncategorematic Adjectives 15
2.1.3. 'Each/Either/Some of the' + MP; Differentiation and
Individualization 18
2.1.3.1. Definiteness and indefiniteness; incorporation of the de-
finite element 19
2.1.3.2. Numerical elements and numbers 23
2.1.3.3. 'Part of the' + MP 26
2.1.4. Demonstratives, Possessive Pronouns, 'Principal', 'Chief,
'Utter' 27
2.2. Recapitulation 28
3. Parameter Adjectives and Measure Adjectives 29
3.1. Parameter Adjectives 30
3.2. Objective Gradables and Subjective Gradables; Neutral
and Non-Neutral Sense 34
3.2.1. 'Warm'; Non-Neutral, Subjective or Objective 36
3.2.1.1. 'Warm' and other gradables in 'hoe' questions 37
3.3. Bierwisch's Observations on Polarity and Orientation of
Adjectives; 'Warm' an Un oriented Adjective 39
3.3.1. An Additional Criterion for the Polarity Sign; '( + Pol)'
Adjectives That Do Not Take MP's 41
4. Semicopulas 43
4.1. Middle Verbs and Semicopulas 43
X THE STRUCTURE UNDERLYING MP SENTENCES
4.1.1. 'Differ' + MP 44
4.1.2. Passive and Manner Adverbials 45
4.1.3. Measure Semicopulas 46
4.2. Other Verbs Taking MP's 46
4.3. Semicopulas Taking Adjectival Complements 48
4.4. 'Cost' + Indirect Object 49
5. Recapitulation 53
V. EPILOGUE 237
BIBLIOGRAPHY 243
INDEX 245
CHAPTER I
1. INTRODUCTORY REMARKS
1 This term is used by Lees (1960), p. 8, for verbs that do not take manner adverbs freely and
which are, characteristically, followed by NP's but cannot be passivized. Examples are
resemble,fit, marry, weigh, cost. See also Chomsky (1965), p. 103, and his note 28 on p:2l8,
which contain some comments on the matter of middle verbs "taking manner adverbials
freely". Also, the note contains a few interesting remarks on the source of manner adverbials.
These remarks suggest that Chomsky himself considers the base rules sketched in the second
chapter of his book as far from adequate, at least with respect to the generating of adverbials.
2 THE STRUCTURE UNDERLYING MP SENTENCES
(12) Ze is twaalf.
'She is twelve.'
The special problems posed by simple MP sentences of the type of (13) will
also be discussed in this study.
Much discussion in recent years has been centered around the role of
semantics in the organization of transformational grammar. According
to what has been called by Chomsky the 'standard theory' (as presented in
Chomsky (1965», the base component generates deep structures that are
mapped into surface structures by rules of the transformational component.
The semantic component assigns semantic representations to the deep struct-
ures.
Synonymy of superficially differing sentences is a necessary condition for
their having identical deep structures. But it is not a sufficient condition;
deep structures of synonymous sentences need not be identical. We cannot
say any more than that the deep structures of nonsynonymous sentences
must differ. Furthermore, the 'standard theory', as it is presented in Chom-
sky (1965), states that the lexicon forms part of the base component. Con-
sequently, the terminal strings of deep structure derivations already con-
sist of phonologically specified elements. In 1968, Chomsky abandoned the
hypothesis that nonsynonymous sentences necessarily have different deep
structures (Chomsky (1968». His present view is that there are also semantic
rules operating upon final derived phrase markers.
A number of linguists, notably James D. McCawley,2 have challenged the
thesis that there is a level of deep structure with the characteristics it is said
to have in Chomsky (1965) or in Chomsky's more recent version of trans-
formational theory. They reject the idea that the lexicon is part of the
base component, and deny that there should be a difference between 'deep
structure' and 'semantic representation'.
This, of course, still does not mean that synonymy of superficially dif-
fering sentences is a condition both necessary and sufficient for their having
identical deep structures. Gruber, one of the first whose views developed in
the direction of generative semantics, puts it in the following way:
The base component generates an underlying language which has immediate semantic as well
as syntactic significance. By no means do we imply here, however, that all surface forms that
mean the same thing have the same representation in the base tree. i.e., in the semantic
language. Certainly it would be unusual in the generation of any language for every entity
generated by the grammar of that language to have an interpretation different from every
other such entity. A phrase-structure type grammar of which we propose the base component,
2 See, for instance, McCawley (I 967a, b) and (1968). Gruber's M.LT. dissertation (1965)
was written at a time when no other work in generative semantics had been done yet.
McCawley often refers to it in his articles. It is therefore not unreasonable to suppose, as
De Rijk (1968) has remarked, that generative semanticians, particularly McCawley, have been
influenced by Gruber.
4 THE STRUCTURE UNDERLYING MP SENTENCES
will generate an indefinite number of trees that are equivalent in meaning but different in form.
( ... ) in addition to the underlying semantic language ( ... ) there will still have to be a set of
postulates and rules for a calculus by which it can be demonstrated that certain trees of the
base· language are equivalent in meaning or that they imply or negate each other. (Gruber
(\967a), p. 48.)
In this study an attempt will be made to show that, in order to account for
a number of relationships among MP sentences, it will be necessary to as-
sume that they can be analyzed in a way that precludes the possibility of
lexical attachment on the level of base phrase markers in the way Chomsky
has proposed. In so far as the attempt will succeed, Gruber's proposals
concerning the functions of the lexicon in formal descriptive grammars
will receive new support.
I do not intend to argue that all of the underlying structures to be proposed
should be seen as semantic representations. Although it seems plausible
to me to maintain that the base component should consist of rules generat-
ing a semantic language, I do not believe that all of the underlying struct-
ures that I am going to discuss in the following chapters can be considered
structures directly generated by the base rules. I do believe, however, that
it can be correctly claimed that, for instance, the sentences (17}-(19), below,
all derive from the same base structures, and that no calculus in Gruber's
sense will be needed to relate them to each other, despite their widely vary-
ing surface structures. In short, although we may not know the exact form
of the base trees of (17)-{ 19), it is highly probable that they are identical.
2. MEASURE PHRASES
In Ross (1964) it is argued that MP's are not simply noun phrases, although
they do bear some resemblance to them. Ross calls attention to the fact that
sentences like the ones below, with some of the, either of the, each of the
preceding the measure phrase, are ungrammatical (as opposed to, e.g.,
I:~~~r I
some/each of the 50 children, either of the 2 children):
We may translate each as ieder or elk. Either of the two may be translated as
meaning 'one or the other of the two'. But since either may also mean 'one
and the other (of two)', as in There was an armchair at either end of the table,
the Dutch equivalent of either of the two can be een (or welke ook maar) van de
twee, or elk/ieder van de twee. Thus the Dutch equivalents of the first two
l
sentences of (25a) (with each and either), may be given as (25b):
ieder 1
I
(25b) *Deze do os is een (or welke ook maar) I· van de twee voet lang.
I
elk
each
This box is one (no matter which) of the two feet long.'
each
The word some may be translated in Dutch as enige ('a number of), but in
certain cases it must be rendered as een stuk ('a piece'), een deel ('a part'), wat
('an unspecified quantity (of a mass)'). In the former case, some occurs in
sentences like Some of these people went berserk. In the latter, it occurs in
sentences such as Some of the paint was left over, He took some of it, Some of
it is true, etc. The former some, however, cannot be used, in any case, as a
limitation within a group of two. Thus it would be nonsense to say, "Some
of these two people went berserk." Therefore, since the number of feet in
Ross's example is two, some, taken in the sense of 'a number of, would be
an irrelevant cause of ungrammaticality, having nothing to do with the
nature of MP's in particular. In order to eliminate this irrelevant factor, we
may either take some to mean 'part of, or increase the number of feet in
MEASURE PHRASE SENTENCES; SOME OBSERVATIONS 7
Ross's example. The Dutch equivalent of the third sentence of (25a) may
be given as (25c):
(25c) *Deze doos is een { :t~!} van de twee voet lang.
'This box is part of the two feet long.'
If the number of feet is, say, ten, we have an English sentence that under
either interpretation is ungrammatical. The two Dutch equivalents are also
ungrammatical:
(25d) *This box is some of the ten feet long.
(25e) *Deze do os is een { ::!} van de tien voet lang.
(25t) *Deze do os is enige van de tien voet lang.
As Ross also observes, in sentences like (26), demonstratives, possessive
adjectives, and words like utter, chief, p6ncipal, etc., are excluded before
MP's (though they may occur in such noun phrases as this/my child, an
utter child, the chieJ/principal reason (dit/mijn kind, een volslagen kind, de
hooJdreden/belangrijkste reden»:
I de voornaamste J 20 minuten.
J die
I
(26) *Het concert duurde mijn
r those
*The concert lasted ·1 my 20 minutes.
the principal
I shall comment on these interesting examples later on.
"It seems," Ross then goes on to say, "that in general, relative clauses
should also be excluded (cf. (27) and (28». But I do not know how to handle
such sentences as (29) and (30), because prenominal adjectives usually derive
from relative clauses." (Quoted examples are numbered according to their
order in the present text.)
(27) *My room is the two degrees which your thermometer is too crude
to register hotter than your room.
(28) *He came the 3 times that I will never forget too often to go away
a willner.
dusty
f
. weary
(29) We marched SIX ong11 miles.
maal ('time')
decibel ('decibel')
volt ('volt')
watt ('Watt')
farad ('Farad')
ohm ('Ohm')
bunder (,hectare', 2.471 acres)
In addition, there are one or two unit words which only occur in the
plural: koppen ('hands', as in een bemanning van 20 koppen ('a crew of
20 hands'», and, possibly, stuks ('pieces'; I am not quite sure whether the
s is a plural ending. 1 stuks sounds odd to me. The second s in stuksgewijs
('by the piece') is not a plural ending). The words koppen and stuks cannot
be preceded by such words as enkele ('a small number of', 'a few'), ettelijke
('a fair number of'), verscheidene ('severa!'). Such words may precede all
other unit words, however, provided they have the plural ending. Thus,
integer-independent unit words must occur in the plural in such sentences
as (35)-(37):
(35) Oit boek kost slechts enkele guldens.
'This book costs only a { few II b f } guilders.'
sma num er 0
(36) Oeze steen weegt verscheidene kilo's.
'This stone weighs several kilos.'
(37) Het duurde ettelijke jaren.
'It lasted a fair number of years.'
Not only enkele, ettelijke, etc., require the plural, but also words like
honderden (,hundreds of'), duizenden ('thousands of'), miljoenen (,millions
of'). Words of the latter type may precede koppen and stuks. We may call
the property of preceding words which causes the integer-independent
unit words to take on the plural form 'differentiation'.3 Later on in this sec-
tion, I will introduce a category NUMBER underlying words like enkele
and honderden. It is distinct from the category NE (numerical element),
to which integers belong (see p. 24). Thus we do not have:
(38) *ettelijke ton
'a fair number of ton'
(39) *veel gulden
'many guilder'
(40) *verscheidene jaar
'several year'
3 See, for this term, also Klooster and Verkuyl (1972). Another form of differentiation,
which may be related to the kind in question is discussed in Chapter IV, Section 1.1.2.
MEASURE PHRASE SENTENCES; SOME OBSER VA nONS 11
also call them 'continuous unit nouns' and 'discontinuous unit nouns',
respectively (keeping in mind, however, that 'discontinuous unit nouns'
refer to things that either mayor may not be contiguous on the time axis.)
We may now ask whether we can generalize the notion 'continuous unit
noun' so as to make it applicable to other nouns besides 'Type 1 nouns'.
It seems that we can. Consider, for example, nouns referring to units of
one-dimensional space ('one-space units'). Meters, feet, miles, etc., all are
continuously ordered members of the respective classes to which they
belong. That is, one-space units are contiguously located on the measuring
scale. The same holds for units of weight, temperature, velocity, voltage,
electrical resistance, electrical capacitance, and so on. Similarly, the
measurements of area ('two-space') and volume ('three-space') are ex-
pressed in units which we can think of as being contiguous. Monetary
units (as distinct from means of payment) are likewise continuously ordered.
That is, if we were to indicate sums of money on a calibrated scale, as an
economist might do, the intervals between the value-representing points
Vi and Vi+ 1 , and between the points Vi + 1 and Vi+ 2 , respectively, would be
contiguous (where Vi and Vi + l' as well as Vi+ 1 and Vi + 2 , differ by one unit
of value.)
It is relevant to note, at this point, the distinction between monetary value
and means of payment. Consider, for instance, the following sentences:
(49a) Ik betaalde de melkboer een gUlden.
'I payed the milkman one guilder.'
(49b) Ik betaalde de melkboer met een gulden.
'I paid the milkman with one guilder.'
Sentence (49a) does not necessarily mean that 1 gave the milkman one coin.
1 may have given him ten 10 cent pieces, or four quarters, or any com-
bination of coins together representing the value of one guilder. 1 may even
have given him more than one guilder, receiving back the change. But
sentence (49b), under normal interpretation, means that 1 gave the milk-
man a I guilder piece. Compare also, in this connection, (SOa, b) and
(Sla, b):
(SOa) Ik betaalde hem via de bank 10 gulden.
'I paid him via the bank 10 guilder.'
(SOb) *Ik betaalde hem via de bank met 10 gulden.
'I paid him via the bank with 10 guilder.'
(Sla) Betaal hem maar die 10 gulden.
'Just pay him those 10 guilder.'
(Sib) Betaal hem maar met die 10 gulden.
'Just pay him with those 10 guilder.'
MEASURE PHRASE SENTENCES; SOME OBSERVA TlONS 13
This is not surprising, in view of the fact that always when we express
the dimensions of something in terms of abstract units, we refer to con-
tinuously ordered units.
2.1.1.2. 'Six dusty miles' and 'ten crisp new dollars'; abstract and concrete
MP's
Returning to Ross's examples (29) and (30), we are now able to give the
following comment.
If we translate in Dutch (29) We marched six dusty (etc.) miles and
(30) That'll only cost you 10 crisp new dollars, the words miles and dollars
must be translated by nouns with the plural ending. Although myI (mile)
and dollar (dollar) are integer-independent words, the b sentences below are
ungrammatical.
I
stoffige
J vermoeiende
W h bb
I
(56a) e e enzesll ange mijlen gemarcheerd.
(translation of (29»
(57a) Dat gaat je alleen maar 10 spiksplinternieuwe dollars kosten.
(translation of (30»
1
I
stoffige
vermoeiende ..
(56b) *We hebben zes I mlJl gemarcheerd.
ange
I~~~!.iendI
(long), spiksplinternieuwe (crisp new) derive? Consider the following ex-
amples:
f dusty 1
'We marched six miles, which were 1~:; I.'
There seems to be nothing wrong with at least the Dutch version of the
sentences (61). The examples (62) and (63), on the other hand, sound odd.
(62) ?Dat gaat je 10 dollars kosten die spiksplinternieuw zijn.
(63) ?That'll cost you 10 dollars which are crisp and new.
MEASURE PHRASE SENTENCES; SOME OBSERVATIONS 17
een ('one'), welke ook maar ('any', 'one', 'no matter which'), ieder ('each'),
elk ('each'), also somehow require the plural if they occur in expressions
such as the ones italicized in (65)-(68). (Cf. Dougherty (l968b), where such
words are said to carry the feature [ + individual]. See also p. 61 in Chapter
II). Therefore, it may be that we have eliminated one factor contributing to
the ungrammaticality of(25a, b, f) by replacing the singular voet by the plural
voeten. But the "individualizing" effect of the italicized expressions in
(65)-(68) still plays a role in the ungrammaticality of these sentences. Their
presence causes the abstract units of length to be represented as individual
entities differing from each other in a respect in which in· fact they are
identical.
When we measure the length of a given object, we do so in terms of
standard distance between points on a straight line. We can say that, tor all
units called "foot" which are continuously ordered on a measuring rod, any
foot is identical to any other foot as regards distance. They are non-identical
only with respect to their location on the rod. Since the location on a measur-
ing rod of any pair of points having a distance of one foot is totally irrelevant
if we want to express the length of an object in terms ofa specified number of
feet, it is not necessary to individualize the units. Just as it is irrelevant, for
instance, which of two guilders we pay for something as far as value is
concerned, it is irrelevant which of two feet on a measuring rod we point to
in order to indicate the length of a foot-long object. In other words, the
length of a foot-long object is not any "particular" foot.
In the three next subsections (2.1.3.1-3), a few further facts bearing upon
the examples (25a-f) will be discussed.
The difference between, on the one hand, een (one) (and all other integers,
for that matter) and enige (some, a number of), and, on the other hand, iedere
(each, every) and elke (each, or in the case of two, either), is that the former
may occur in indefinite noun phrases and the latter can only occur in de-
finite noun phrases. In eliminating van de + integer, we have eliminated a
definite article (de). But in (69) and (71) there is still a definite element
present. It is incorporated in iedere en elke. In the second sentence of (70),
welke ... ook maar, which is equivalent to elke willekeurige, also a definite
element is present. The distinction between the two groups of expressions
pointed out above also holds for, on the one hand, enkele ('a few', 'a small
number of), ettelijke ('a fair number of), which may occur in indefinite noun
phrases, and, on the other hand, aIle (a/l), beide (both), de meeste (most),
which occur in definite phrases.
Simple declarative sentences with indefinite noun phrases as subjects
normally start with the word er ('there'). But er is excluded in sentences
with definite subjects. 7 Thus we have Er loopt een jongen op het gras (,There
walks a boy on the grass'), but not *Er loopt de jongen op het gras ('There
walks the boy on the grass'). The following examples 8 show the definite/
indefinite distinction between the two groups of words under consideration.
(73) Er liep een jongen op het gras.
'There walked one boy on the grass.'
(74) Er gingen enkele maanden voorbij.
There passed a few months.'
(75) Er waren ettelijke maanden verstreken.
'There had passed a fair number of months.'
(76) Er gingen enige kinderen met ons mee.
'There went a number of children with us.'
(77) *Er liep {~ldke
Ie ere
} jongen op het gras.
'There walked each/every boy on the grass.'
(7S) *Er lie pen aIle jongens op het gras.
'There walked all the boys on the grass.'
(79) *Er liepen beide jongens op het gras.
'There walked both boys on the grass.'
(SO) *Er gingen de meeste kinderen met ons mee.
'There went most children. with us.'
That aile and beide are definite is also borne out by (SI b, c, d) and (S2b, c, d),
7 See Bech (1952) on the Dutch word er. Also Kraak (1966). Chapter VI. Section 26.
8 Examples like (73)-(82d) are given in Kraak and Klooster (1968). Chapter 5.
MEASURE PHRASE SENTENCES; SOME OBSERVATIONS 21
which are synonymous with (Sla) and (S2a), respectively. The b, c and d
sentences contain the definite article de.
(SIa) AIle kinderen waren ziek.
'All children were ill.'
(SIb) De kinderen waren alle(n) ziek.
(SIc) De kinderen waren allemaal ziek.
(SId) Al de kinderen waren ziek.
(S2a) Beide kinderen waren ziek.
'Both children were ill.'
(S2b) De kinderen waren beide(n) ziek.
(S2c) De kinderen waren allebei ziek.
(S2d) Allebei de kinderen waren ziek.
The ungrammaticality of (69) *Deze doos is iedere voet lang (,This box is
each/every foot long') and (71) *Deze doos is elke voet lang (This box is
either/each/every foot long') can now be explained by the presence of an
underlying definite article, incorporated in ieder ( e) and elk ( e). The definite
article individualizes. It is because of this, that it is excluded before abstract
measure phrases, which do not refer to quantities of "particular" units. This
is demonstrated by (66) and (70), and by (6S) and (72).
Notice, meanwhile, that whereas the article is incorporated in words like
aile, iedere, elke, beide, as well as in enige and (in certain cases, see p. 23)
enkele, ettelijke, it is not incorporated in words like honderden (hundreds of),
duizenden (thousands of), tientallen (tens; in Dutch, the word twintigtallen
('twenties') is not as commonly used as scores is in English. In many cases,
however, we may translate tientallen as scores or dozens). Although enkele,
ettelijke, enige, as well as honderden, duizenden, etc., have in common the
property that they require that the unit word be in the plural, there is a sign-
ificant difference between negative sentences containing, on the one hand,
enkele etc., and, on the other hand, honderden, etc. With respect to negative
sentences, furthermore, words like honderden belong to the same group as
integers. The difference in question can be explained by the difference with
respect to incorporation or non-incorporation of the indefinite article. Con-
sider, for example, the following sentences.
(S3a) Hij had tientallen kleinkinderen.
'He had dozens of grandchildren.'
(S4a) Hij had tien kleinkinderen.
'He had ten grandchildren.'
'He had a IJ
small number
fair number of grandchildren.'
number
(83b) Hij had geen tientallen kleinkinderen.
'He had no dozens of grandchildren.' (did not have dozens of)
(84b) Hij had geen lien kleinkinderen.
'He had no len grandchildren.' (did not have ten)
I enkele I
I
(85b) *Hij had geen 1etll!lijke J kleinkinderen.
I
emge
small number
'He had no a fair number of grandchildren.'
I
number
I
emge
small number I
'He had not a fair number f of grandchildren.' (did not have
number etc.)
(86a) Het duurde tientallen minuten.
'It lasted dozens of minutes.'
I
(87a) Het duurde lien minuten.
'It lasted ten minutes.'
I small number I
emge
I
'It lasted no ten minutes.' (did not last etc.)
enkele I
I
(88b) *Het duurde geen ett~lijke f minuten.
emge
Notice that where the Dutch expression contains the plural meters, the
corresponding English expression contains the preposition of, but where
the Dutch expression contains the singular meter, the preposition is absent
in the English equivalent. It is not implausible to surmise that in these cases
the English versions more truly reflect underlying structure than the Dutch
equivalents, all the more so because the underlying indefinite article, which
is not apparent in the Dutch examples, is manifested as a word in their
English counterparts in the case of (89a)--(9l b).
The structure underlying expressions of the type of (89)--(93) therefore
might be something like I.
II INDEF + NE + N + PLURAL
where 'NE' stands for 'numerical element', which may dominate an integer
or paar as it occurs in een paar meter (a few meters).
Een paar in certain cases means a pair. If it has the latter meaning, it is not
an NE, but a NUMBER. Accordingly, in that case the English equivalent
contains the preposition of The noun followingpaar in Dutch then must be
in the plural. The expressions (96a) and (97a), below, are identical in form but
differ in meaning. In addition, we have (98a), which has only one meaning.
MEASURE PHRASE SENTENCES; SOME OBSERVATIONS 25
(99a)
I~;e I paar gulden (NE)
,I ~~:se I
deze
few guilder'
these
f 1I
het
(99b)
I dat paar voeten (NUMBER)
,I~~;t I
dit
pair of feet'
I~:~ I
thIs
(99c) * (NUMBER)
,I :~;t I
paar gulden
dit
pair of guilder'
this
The gender of the definite article and the demonstrative pronoun is deter-
mined by the gender of paar if it is a NUMBER (the gender of paar is neuter).
In general, if a NUMBER word precedes the unit word, the gender of the
article or pronoun preceding the NUMBER word is determined by the gen-
der of the latter. If a numerical element precedes the unit word, the
article or pronoun preceding the numerical element is in the plural (for all
genders de/die/deze), or, if the NE is 1, its singular form is determined by
26 THE STRUCTURE UNDERLYING MP SENTENCES
the gender of the unit noun. Thus, in the case of uur (neuter gender), we have
for instance, dat ene uur ('that one hour'), die twee uur ('those two hour'), and
in the case of gulden (masculine gender), die ene gulden ('that one guilder'),
die twee gulden ('those two guilder').
Besides integers and een paar ('a few'), the words hoeveel ('how many')
and zoveel ('that amount of) should also be considered NE's. If hoeveel or
zoveel precedes an integer-independent unit noun, the latter does not take
on the plural form (cf. Hoeveel kilo weegt hy?('How many kilo weighs he?'),
Hy weegt zoveel kilo (,He weighs so many kilo'». On the other hand, veel
('many'), as we noted (p. 10, example (39», is a NUMBER word.
2.1.3.3. 'Parto!(the)'+MP
Of the examples (25a)-(25f), we have not yet dealt with (25c, e) *Deze doos
is een deel/stuk van de twee (or lien) voet lang (This box is part of the two (or
ten) feet long').
As was the case with the other 'box' examples, if we substitute voet by
voeten, the resulting sentence is still ungrammatical. We do not have *Deze
doos is een deel van de twee voeten lang. In fact. the latter sentence sounds
even worse than the corresponding one with the unit word in the singular.
If we eliminate the definite article, getting *Deze doos is een deel van twee
voet lang (This box is part of two feet long'), we still have a bad sentence.
The cause of the ungrammaticality of this sentence comes into focus if we
try to analyze what it might mean. A re-wording of it might go like this:
"There is a pair of points on a two-foot segment which have a distance equal
to the length of this box'. A statement such as this is strange, for consider
the two logical possibilities there are: (a) the box is exactly 2 feet long, or
(b) the box is less than 2 feet long. If (a) is the case, there is only one pair of
points on a 2-foot segment which have a distance equal to the box's length.
If the sentence in question is to be understood that way, it is just an odd and
superfluous way of saying "This box is two feet long". If (b) is the case, there
is an infinite number of pairs of points on the two-foot segment having a
distance equal to the box's length. Now in logicians' talk, there is an x
means 'there is at least one x, there are one or more x's', but in ordinary
language there is an x means that there is one x. What our sentence conveys,
then, is that there is one pair of points on a two-foot segment which cor-
responds to the box's length, which obviously is nonsensical in the case of
(b) for the same reason as is any statement expressing length in terms of a
unique pair of points on a measuring rod. I think it must be concluded that
the nonoccurrence of sentences like (25c, e) is due to certain characteristics
of what measure phrases denote rather than that it would have anything to
do with whether or not MP's are noun phrases.
MEASURE PHRASE SENTENCES; SOME OBSERVATIONS 27
mijn
jouw
Uw
haar
(100) *Er liep kind op het gras.
ZIJn
ons
jullie
hun
my
your
her
'There walked child on the grass.'
his
our
their
As for the third sentence, containing the word principal, notice that if this
word is translated as voornaamste (a superlative form), it only occurs preced-
ed by the definite article. It can also be translated by hoofd-, as may be done
in the case of, e.g., principal cause (hoofdoorzaak). In that case it may be
preceded by the indefinite article: Dat was een hoofdoorzaak (,That was a
principal cause'). The third sentence of example (26) is ungrammatical for
two reasons. Apart from the fact that the definite article causes it to be un-
grammatical, the word voornaamste causes the sentence to present a set of
abstract units of time as if they, in themselves, could be distinguished in any
way from any other such unit. As in (65)-(68) *Deze doos is iederjeen (or
welke ook maar)/elk/enige van de 2 voeten lang, the units are presented as
"particular", that is, unique, units, where they cannot be.
The fact that words like principal (voornaamste, hoofd-), chief (voornaam-
ste, opperste, eerste, hoofd-, opper-) and utter (volslagen, volstrekt, uiterst,
totaal) are excluded before MP's (as Ross observes), is in no wayan indica-
tion that MP's are not noun phrases. A phrase such as de baard (the beard),
28 THE STRUCTURE UNDERLYING MP SENTENCES
for instance, is without a doubt a noun phrase. Yet the following sentences
are ungrammatical:
head ofcattle is that in the case of small numbers they do not occur in measure
phrases which are preceded by words like bijna (almost), or which contain
fractional numbers (*2t man, *bijna 3 man; an exception is formed by the
facetious expression anderhalve man en een paardekop (' 1t man and a horse's
head'), meaning 'very few people').
Continuous unit nouns have the property of referring to units which, if
represented on a scale, are identical in all respects except location. It is due
to this that only under certain circumstances may units or sets of units be
individualized (iedere meter die hij aflegde, 'each meter he covered'), de
laatste mijl ('the last mile'). Individualization is caused by the definite article
and other words which are definite. It causes a unit or set of units to be
presented as unique, i.e., as a "particular" unit or set of units.
'Differentiation', that is, not giving the quantity as a whole, but in the form
of the constituent parts, occurs in expressions like verscheidene kilometers,
honderden meters (several kilometers, hundreds of meters). (Verscheidene
etymologically comes from a word which means 'differentiated'.) Another
form of differentiation occurs in Hij wachtte een uur en nog een uur (He
waited for an hour and another hour). This type of differentiation will be dis-
cussed later on, when the relation between duren (last), and gedurende (for
(the duration of) will be dealt with.
In the case of individualization of a set of units, integer-independent unit
words preceded by an integer greater than I do not have the plural ending.
In the case of differentiation caused by such words as verscheidene, the plural
ending is always required. MP's may contain a NUMBER noun (aantal,
honderdtal(len) , honderden, tweetal, paar, etc.), or a numerical element
(NE), such as integers or paar (few as in a few). NUMBER nouns have a
differentiating effect, numerical elements do not.
The formal description of MP's will be discussed elsewhere (p. 180ff.).
As we saw from examples such as (5) Jan is 2 meter lang ('John is 2 meters
tall'), (6) Ze is 12 jaar oud ('She is 12 years old') and (7) De lat is
2 centimeter dik ('The lath is 2 centimeters thick'), there are adjectives which
take measure phrases in the same way as verbs like wegen and kosten do.
The class of adjectives that may occur in combination with MP's in this
manner (Le., in simple MP sentences) I shall refer to as 'measure ad-
jectives'.
From the examples given in the preceding sections, it can be seen that
ordinarily MP's in Dutch contain words referring to units of length,
weight, etc. (unit words). There are some exceptions. One is formed by
30 THE S T Rue T U R E UN DE R L YIN G M P SEN TEN C E S
sentences in which age of humans is referred to, in which case the unit
word may be omitted (cf. (12) Ze is twaalf (,She is twelve'». Other cases
in which the unit word may be omitted are to be found among MP sentences
pertaining to monetary value, if that value is expressed in guilders and
cents (e.g., Dat hoek kost twaalf vi}ftig (,That book costs twelve fifty'»,
and also among MP sentences pertaining to length and weight (e.g., Hi)
is een vijfentachtig (lang) (,He is one eighty five (tall)'), Hij weegt 81 ('He
weighs 81 '».
We may call length, monetary value, weight, etc., the 'parameters'
relevant to the units in question. The concept of 'parameter' is also
relevant in connection with measure adjectives, since they convey what
parameters we are talking about. However, there are adjectives which,
though pertaining to parameters, do not occur freely in association with
MP's. Adjectives of the latter type, together with measure adjectives,
constitute a class which I shall refer to as 'parameter adjectives'.
height, and those pertaining to monetary value cannot modify the words
hoogte (height) and prijs (price), respectively, and so on. Expressions like
are ungrammatical.
It should be noted, in this connection, that we must distinguish between
two words hoog, say, hoog 1 and hoog 2 • Hoog 1 is a measure adjective and
carries the meaning 'tall' or 'high' in relation to physical objects. Hoog 2
expresses extent or degree. Thus we have een hoge temperatuur ('a high
temperature'), een hoge prijs ('a high price'). A similar distinction should
be made with respect to groot 1 and groot 2 ('large', 'big', 'great'). Groot 1
is a parameter adjective (een grote kamer ('a large room')). Groot 2 expresses
again extent or degree, but it occurs with different parameter nouns than
hoog 2 does. We do not have, e.g., *een grote temperatuur ('a great tempera-
ture') or *Het wordt verkocht voor een grote prijs ('It is being sold at a great
price'). Rather, it occurs in such expressions as een grote hoogte ('a great
height'), een grote afstand ('a great distance'), etc. The words gering and
klein may be used as the antonyms of groot 2 , whereas the antonym of groot 1
is only klein. The antonym of haog in both senses is [aag ('low'). Hoog 2 and
its antonym laag have a paraphrase relation with the verbs stijgen ('rise')
and dalen ('drop', 'fall'), respectively; groot 2 and its antonym gering with
toenemen ('increase') and afnemen ('decrease'). Thus we have de prijzen
stijgen/dalen ('the prices rise/drop'), de afstand neemt tae/neemt af ('the
distance increases/decreases'), but not *de prijzen nemen toe/nemen af ('the
prices increase/decrease') or *de afstand stijgt/daalt ('the distance rises/
drops'). (See also (161)-(164) on p. 42.)
While, as we saw from (106)-(110), such parameter adjectives as warm,
hoog1 , etc., cannot modify nouns referring to the corresponding parameters,
an expression like (Ill), below, is perfectly correct:
But the verb wegen clearly does not have the "literal" meaning here. Or,
perhaps more accurately, it is a different verb, not having the meaning
of wegen used in connection with physical objects, a word with different
paraphrases. The sentences (l18a) and (119a), below, are grammatical,
but the corresponding b sentences are not.
All parameter adjectives are grad ables, i.e., they may occur in the com-
parative form or the superlative form, be preceded by te (too), erg (very), etc.
Gradables can be subdivided into two classes, which I shall call 'sub-
jective gradables' and 'objective gradables'. All measure adjectives are
objective gradables. All objective grad abies tum out to be parameter ad-
jectives, and vice versa, by the criteria that I am following. Parameter
adjectives, in the sense in which I prefer to use this term, refer to
properties which can be measured. Thus, beautiful is not a parameter
adjective; it is a subjective grad able, because beauty cannot be measured
objectively, or rather, no theory of beauty has been developed thus far
which would enable us to express beauty in terms of standard units.
The difference between objective gradables and subjective gradables can
be characterized in terms of possible occurrence on the place of the
blank in constructions of the form Ik vind x -er dan y ('I find x --er
than y (more - than y)'). Although, with a few exceptions,10 grad abies
may occur in the complement of vinden ('find', 'think', as in I find/think
her beautiful), only subjective grad abies may occur in the construction
indicated (cf. (1 27a)-(l 28c».
(l27a) Jan is langer dan Piet.
'John is taller than Peter.'
(l28a) Ingrid is mooier dan Marie.
'Ingrid is more beautiful than Mary.'
(l27b) Ik vind J an lang.
'I find John tall.'
(128b) Ik vind Ingrid mooi.
'I find Ingrid beautiful.'
(127c) *Ik vind Jan langer dan Piet.
'I find John taller than Peter.'
(128c) Ik vind Ingrid mooier dan Marie.
'I find Ingrid more beautiful than Mary.'
10 Exceptions are gradables like misselijk ("sick', 'nauseated'). Such words refer to experiences
which cannot be shared with others. Subjective statements like *Ik vind hem zeeziek ('I find/
think him seasick') are odd because the experience of seasickness cannot be subject to judg-
ments of this sort on the part of others than the one who feels seasick. Adjectives of this kind
may, however, occur in contexts like Ik voel me - ('I feel -').
MEASURE PHRASE SENTENCES; SOME OBSERVATIONS 35
Many objective gradables (and, among these, all measure adjectives) can
be used in a "neutral" sense, whereas subjective grad abIes invariably
lack this capacity. The difference between the neutral oud (old) and
the non-neutral oud (id.) is demonstrated in the following examples:
In (129), oud is used in the neutral sense. If we leave out the MP, we
obtain (131). Omitting the MP has the effect, as it were, of turning oud
into the non-neutral adjective; (130), in which oud is non-neutral, is
possible, but (131), in which oud is non-neutral as well, is odd because
oud now contradicts jonge (young).
Consider the difference between the a sentences and the b sentences of
the examples below:
Vol (full) is an objective grad able in (136), for we do not have (138)
alongside (137):
(136) Het glas is vol.
'The glass is full.'
(137) Het glas is nu voller dan het was.
'The glass is now fuller than it was.'
(138) *Ik vind het glas voller dan het was.
'I find the glass fuller than it was.'
But there is no neutral sense of vol; (137) does not entail (139a) or vice
versa; (139b), on the other hand, means the same as (137):
(139a) Het glas was leger dan het nu is.
'The glass was emptier than it is now.'
(139b) Het glas was minder vol dan het nu is.
'The glass was less full than it is now.'
The situation appears to be that if an adjective is a subjective gradable
it cannot be used in the neutral sense. But the converse does not hold.
is not answerable in the sense in which such hoe questions with objective
gradables like vol and lang are answerable:
The only possible answers to (148) seem to be sentences like Mooier dan
Ingrid ("More beautiful than Ingrid'), Minstens zo mooi als Ingrid ('At
least as beautiful as Ingrid'), but such answers do not sound very adequate.
Heel mooi ('Very beautiful') sounds odd as an answer. Hoe questions
containing subjective grad abIes which occur as the complement of vinden
sound somewhat better than such hoe questions not occurring in the com-
plement of this verb but, though better than (148), they still are unnatural:
Apart from warm and vol (and perhaps a few others), there is another set
of parameter adjectives which never occur in the neutral sense. It is con-
stituted by words like klein (small), jong (young), licht (light), etc., in short,
by the words which are the antonyms of parameter adjectives that can be
used in the neutral sense. (For ease of reference, I shall call parameter
adjectives which can be used in the neutral sense 'neutralizable parameter
adjectives'.)
In Bierwisch (1967) some examples are cited demonstrating that ad-
jectives which are the antonyms of neutralizable parameter adjectives, be-
sides not taking MP's, do not occur either with such expressions as dubbe! zo,
twee keer zo, half zo, tienmaa! zo (twice as, two times as, half as, ten
times as). I quote his examples here, adding the Dutch and English
equivalents:
clearly distinguished. We have only the sentences [I add the Dutch and
English equivalents, WGK]:
(l59a) Der Zug is 10 Wagen lang.
De trein is 10 wagons lang.
The train is 10 carriages long.
(159b) Der Fluss ist 250 Meter breit.
De rivier is 250 meter breed.
The river is 250 meters wide.
(l59c) Die Maschine fliegt 2000 Meter hoch.
Het toestel vliegt 2000 meter hoog.
The plane flies 2000 meters high.
(159d) Achim ist 35 Jahre alt.
Achim is 35 jaar oud.
Achim is 35 years old.
The following sentences are indisputably odd:
(l60a) *Der Zug ist zwei Wagen kurz.
*De trein is twee wagons kort.
*The train is two carriages short.
(160b) *Der Bach ist einen Meter schmal.
*De beek is een meter sma I.
*The creek is a meter narrow.
(160c) *Die Maschine fliegt 100 Meter niedrig.
*Het toestel vliegt 100 meter laag.
*The plane flies 100 meters low.
(l60d) *Katharina ist zwei Jahre jung.
*Catharina is twee jaar jong.
*Catharine is two years young.
Only the ( + Pol) marked elements of an antonymous pair of adjectives can
take an MP, and in this case the whole Adjectival does not indicate one
of the poles involved, but only the scale which is established by the
pair and a certain point within the scale. The orientation can be taken
as "neutralized"." (Bierwisch, op. cit., p. 9.)
As regards Bierwisch's last remark quoted above, we may add that not
all elements which he would mark ( + Pol) take MP's. An example is duur
(cf. (103) *Het boek is 12 gulden duur).
3.3.1. An Additional Criterion for the Polarity Sign; '( + Pol)' Adjectives
That Do Not Take MP's
It is not clear from what Bierwisch says whether we should mark warm
42 THE STRUCTURE UNDERLYING MP SENTENCES
( + Pol) or ( - Pol), since he only gives a criterion that works with oriented
adjectives. As we saw (cf. (I58a, b), warm belongs to the unoriented set.
Perhaps we can find an additional criterion. As was remarked on p. 31,
the word hoog 2 and its antonym laag have co-occurrence relations with
stijgen ('rise') and dalen ('drop'), respectively, while groot 2 and gering
co-occur with toenemen ('increase') and afnemen ('decrease'), respectively.
Now consider the following synonymous triples:
(I61a) De prijzen van de boeken stijgen/dalen.
'The prices of the books rise/drop.'
(I61 b) De prijzen van de boeken worden hoger/lager.
The prices of the books become higher/lower.'
(I 6 lc) De boeken worden duurder/goedkoper.
'The books become more expensive/cheaper.'
(I 62a) De lengte van het touw nam toe/af.
'The length of the rope increased/decreased.'
(I 62b) De lengte van het touw werd groter/geringer.
The length of the rope became greater/smaller.'
(162c) Het touw werd langer/korter.
'The rope became longer/shorter.'
(I 63a) De temperatuur van het warme water steeg nog iets meer.
'The temperature of the warm water rose some more still.'
(1 63b) De temperatuur van het warme water werd nog iets hoger.
The temperature of the warm water became a bit higher still.'
(I 63c) Het warme water werd nog iets warmer.
'The warm water became a bit warmer still.'
(I 64a) De temperatuur van het ijswater daalde nog iets meer.
'The temperature of the ice-water dropped some more still.'
(l64b) De temperatuur van het ijswater werd nog iets lager.
'The temperature of the ice-water became a bit lower still.'
(164c) Het ijswater werd nog iets kouder.
The ice-water became a bit colder still.'
As we can see from the examples (l61a)-(l62c), the "extent word" hoger
and the verb stijgen have a co-occurrence relation with duurder, while
the "extent word" lager and the verb dalen have a co-occurrence relation
with goedkoper. Likewise with groter/kleiner, toenemen/afnemen and langer/
korter.
Now the words duur and lang, as we have seen, should be marked ( + Pol).
The examples (161a)-(l62c) suggest that if an adjective is (+Pol), it
co-occurs with stijgen or toenemen, as the case may be, whereas if it is
MEASURE PHRASE SENTENCES; SOME OBSERVATIONS 43
4. SEMICOPULAS
4.1.1. 'Differ' + MP
There is a number of middle verbs which take MP's, but which, at first glance,
do not seem to have obligatory complements. Cf. the sentences (177) and
(178) :
(177) Deze latten schelen 2 millimeter.
'These laths differ by 2 millimeter.'
(178) Deze latten schelen.
'These laths differ (in length, width or thickness).'
Other such verbs are afwijken, (differ, deviate) and zich verheffen (have an
elevation (of)). The verbs considered here are similar to copula + measure
11 See, for the term 'specifying complement' ("specificerend complement"), Klooster, Verkuyl
and Luif (1969).
MEASURE PHRASE SENTENCES; SOME OBSERVA nONS 45
adjective constructions in that in the case of the latter, too, we may omit
the MP without the sentence becoming ungrammatical (cf. (129) and (130)
in Section 3.2). But, as we noted, if we leave out the measure phrase in the
latter type of constructions, we get sentences in which the measure adjectives
are no longer neutral. Sentence (130) Die persoon is oud (That person is
old') means that according to a certain norm, the person indicated has
passed the age under which he was not yet considered 'old'. Looking at it
this way, we can say that MP's (and certain other phrases) are obligatory
complements of constructions of the form copula + neutral measure ad-
jective. Something similar appears to be the case with verbs like schelen
(differ). If two objects differ to a negligible extent in length, we do not say
that they differ. Only if a certain norm (which may vary from case to case) is
exceeded, do we say that there is a difference in length. In other words, the
norm, or criterion, by which one decides to consider two objects as differing
or not differing in length, and which varies with the circumstances, deter-
mines the point at which one starts using the verb schelen. This norm, which
also may vary from person to person (because it depends on personal
judgment), introduces an element of subjectivity in schelen as it occurs in
(178). This subjective element is not related to the parameter of length, but
rather to whether or not the difference in length is important enough.
All (or almost all) laths differ in length, although the difference may be
microscopically small. But we only say that they differ in length if the dif-
ference for our purposes becomes appreciable. However, there is no such
norm involved if we state how great the difference is, as in (177). Therefore,
it seems best to distinguish between two verbs schelen (as we also do with
respect to copula + measure adjective constructions), one which is a meas-
ure semicopula and one which is not a semicopula at all.
The question may be raised which, if any, of the italicized verbs in the
examples below are semicopulas:
MEASURE PHRASE SENTENCES; SOME OBSERVATIONS 47
The way in which we may account for this difference will be discussed
elsewhere (p. 199ff.). More characteristics of semicopulas will be discussed
in Chapter II.
There are a few semicopulas which can occur with indirect objects. Such
verbs are kosten and schelen. In this section I shall discuss the differences
with the normal cases this gives rise to. I shall demonstrate them here with
the verb kosten.
Consider the following examples:
(l99a) Dat horloge kostte 300 gulden.
'That watch cost 300 guilders.'
(l99b) Dat horloge kostte me 300 gulden.
That watch cost me 300 guilders.'
Sentence (199a) may tell us either that, in the past, the price of the watch was
300 guilders, or that at some time in the past 300 guilders were actually
paid for it. Sentence (199b) can only be interpreted in one way; it means that
the person indicated by the indirect object me paid 300 guilders for the watch.
The difference between the two interpretations of kosten is even more
pronounced in the examples (200a) and (200b), below:
50 THE STRUCTURE UNDERLYING MP SENTENCES
kosten2 :
(207a) De parkeerboetes kostten hem veel geld.
'The parking fines cost him a lot of money.'
(207b) *Hij kocht de parkeerboetes voor veel geld.
'He bought the parking fines for a lot of money.'
(208a) Die alimentatie kostte hem veel geld.
'That alimony cost him a lot of money.'
(208b) *Hij kocht die alimentatie voor veel geld.
'He bought that alimony for a lot of money.'
(209a) Het s/echte moreel kostte de vijand veel verliezen.
'The low morale cost the enemy many casualties.'
(209b) *De vijand kocht het s/echte moree/ voor veel verliezen.
'The enemy bought the low morale for many casualties.'
In addition, while kosten t co-occurs with moeten betalen met (have to pay
with) as well as moeten betalen (haw! to pay), kosten2 only co-occurs with
moeten beta/en met in the following sentences:
kostent:
(2 lOa) Het werk kostte een tientje.
'The job cost a tientje (ten guilder note).'
(2 lOb) Hij moest voor het werk een tientje betalen.
'He had to pay a tientje for the job.'
(21 Oc) Hij moest voor het werk met een tientje betalen.
'He had to pay for the job with a tientje.'
kosten2 :
(211a) Het werk kostte hem zijn gezondheid.
'The job cost him his health.'
(211 b) *Hij moest voor het werk zijn gezondheid betalen.
'He had to pay his health for the job.'
(21Ic) Hij moest voor het werk met zijn gezondheid betalen.
'He had to pay for the job with his health.'
have Daarvoor moest hij betalen met 1500 gulden as a paraphrase of Dat
kostte hem 1500 gulden. In both of these sentences the MP is abstract, i.e.,
it does not refer to means of payment, although met precedes it in the former
sentences.
Kosten2 can take definite specifying complements where kosten j cannot.
Thus we have such sentences as (200b) Die dingen kostten Truus haar baan
(Those things cost Trudy her job'), with haar baan as a definite specifying
complement. Since kosten without an indirect object normally is interpreted
as kosten j, the b sentences below are ungrammatical under normal inter-
pretation:
Note that sentences like (212a) must be translated in English with that'll cost,
whereas a kostenj sentence such as Het stuk speelgoed kost twee dollar must
be translated as The toy costs two dollars, without will (or a contracted form).
This difference has to do with the fact that kostenj, by itself, does not imply
any event (except maybe the event of pricing), whereas kosten 2 always does,
the event being the act of paying.
Another important difference, finally, between kosten j and kosten2 is
that only kosten j may undergo reduction. Thus we have (214b) along with
(214a), but not (2ISb) along with (2ISa):
5. RECAPITULATION
The set of all parameter adjectives is identical to the set of all objective
gradables. No subjective grad abies (i.e., adjectives which may occur in
contexts such as Ik vind x -er dan y ('I find x ~r than y (more - than y)')
are neutralizable, and there are objective grad abies which are not either.
Among those objective gradables which are not are the antonyms of neutral-
izable parameter adjectives (kort (,short'), jong ('young'), etc.). Bierwisch
makes a distinction between two sets of adjectives, which he marks '( + Pol)'
and '( - Pol)" respectively. In addition, he distinguishes oriented and un-
oriented adjectives. The latter, whether marked ( + Pol) or ( - Pol), can al-
ways be preceded by twiee as, half as, etc., but of the oriented set, only the
ones marked ( + Pol) can be preceded by such expressions. In Section 3.3.1
I proposed a criterion by which to distinguish '( + Pol)' and '( - Pol)' ad-
jectives, irrespective of whether they belong to the oriented or unoriented
set. (There is a paraphrase relation between '( + Pol)' adjectives and stijgen
('rise') or toenemen ('increase') on the one hand, and between '( - Pol)'
adjectives and dalen ('drop') or afnemen ('decrease') on the other). Within
the set of oriented '( + Pol)' neutralizable parameter adjectives there is a
further subset which takes MP's. These are called 'measure adjectives'.
Simple temperature MP sentences do not contain a semicopula or a
measure adjective, but simply the copula followed by an MP. Nonetheless
there exists a temperature parameter adjective (warm), which, however, like
vol ('full'), is non-neutralizable. Unlike vol, it may also occur as a subjective
gradable. Since hoe ('how') in combination with objective gradables (i.e.,
parameter adjectives) questions quantificatory phrases, which, however,
cannot occur in the complement of vinden ('find', 'think'), a sentence like
* Hoe warm vind je het water? ('How warm do you find the water?') is un-
grammatical if warm is taken as a parameter adjective. If we take it as a
subjective gradable, the sentence sounds slightly better. A question such as
Hoe warm is het water? ('How warm is the water?') on the other hand, is
grammatical and can be answered with, e.g., 40° Celsius (but not with 40°
onder nul (below zero), since warm is '( + Pol)' and non-neutralizable). Here,
warm is taken as an objective gradable.
A subset of the set of all middle verbs is formed by semicopulas. These
have obligatory complements specifying the price, weight, meaning, name,
etc., of the things referred to by their subjects. Some measure semicopulas,
like some copula + measure adjective combinations, may undergo reduc-
tion in simple MP sentences. Reduced simple MP sentences have the form
NP + Copula + MP.
A middle verb like selielen ('differ (in length, width, etc.)') can either be
a semicopula or a middle verb that does not take complements. Like, e.g.,
oud zijn ('be old'), sehelen may occur either in the non-neutral sense, in which
MEASURE PHRASE SENTENCES; SOME OBSER VATIONS 55
Type!
(la) Jan WEEGT 80 kilo.
'John weighs 80 kilograms.'
(2a) Dat boek KOST 12 gulden.
That book costs 12 guilders.'
(3a) Jan IS 2 meter LANG.
'John is 2 meters tall.'
(4a) Ze IS 12 jaar OUD.
'She is 12 years old.'
(lb) Jan IS 80 kilo.
'John is 80 kilograms.'
(2b) Dat boek IS 12 gulden.
'That book is 12 guilders.'
(3b) Jan IS 2 meter.
'John is 2 meters.'
(4b) Ze IS 12 jaar.
'She is 12 years.'
Type II
(Sa) De boot STEEKT 1 vadem.
'The boat draws 1 fathom.'
STRUCTURE UNDERLYING SIMPLE MP SENTENCES 57
A sentence like
(17) Jan zag Wim en Wim zag Jan.
'John saw Bill and Bill saw John.'
can be paraphrased as
(18) Jan en Wim zagen elkaar.
'John and Bill saw each other.'
More pairs like (17) and (18) can be thought up without any difficulty. No
doubt there is a regular relationship between elkaar (each other) sentences
and sentences like (17), although it is not entirely clear how this relationship
can be made explicit in the form of rules.
One can imagine a situation, for instance of people playing Monopoly,
in which someone observes:
(19) Een hotel kost vijf huizen en vijf huizen kosten een hotel.
'One hotel costs five houses and five houses cost one hotel.'
We do not have, however, (20):
(20) *Een hotel en vijf huizen kosten elkaar.
'One hotel and five houses cost each other.'
As I shall attempt to show in the next four subsections, the assumption of
an underlying element or string relatable to hebben in sentences containing
semicopulas is hardly escapable, if we want to determine the general
source of the ungrammaticality of elkaar sentences containing semicopulas.
In Subsection 2.1.1.1.1, below, I shall briefly sketch two proposals for
the derivation of each other sentences. (As I shall point out in Subsection
2.1.1.1.3, it does not matter with respect to the final point to be made which
hypothesis we should adopt.)
Semicopula sentences contain noun phrase complements (in particular,
MP's) which at first glance exhibit'a structural relation to the verb that is
indistinguishable from the relation the direct object has to the verb. Thus, if
we assume that a semicopula is simply the V in the base phrase marker, there
will be no difference between semicopula complements and direct objects.
This, in itself, constitutes an argument of sorts against describing semi-
copulas as directly dominated by V in the base phrase marker.
Each other phrases generally have the function of a direct or indirect
object, or they are the noun phrase in prepositional phrases. Although semi-
copula complements have none of these functions, the ungrammaticality
of each other semicopula sentences should not be explained in terms of the
function of semicopula complements. For, there are exceptions to the rule
60 THE STRUCTURE UNDERLYING MP SENTENCES
that each other phrases are (in)direct objects or the objects of prepositions.
These points will be discussed in Subsection 2.1.1.1.2.
In Subsection 2.1.1.1.3 it will be pointed out that the ungrammaticality of
each other semicopula sentences is caused by the fact that the condition of
co-reference is not met. (It is irrelevant which hypothesis concerning the
derivation of each other sentences we want to adhere to.) Hebben paraphrases
provide a crucial argument for this explanation.
The occurrence of elk- (each) in each other sentences containing a measure
semicopula is an extra source of ungrammaticality, since elk individualizes
(cf. p. 21). But we can isolate the factor consisting in the absence of co-
reference in each other sentences containing semicopulas such as betekenen
('mean'). Again, hebben paraphrases are crucial. This will be shown in Sub-
section 2.1.1.1.4.
I
s
~VP
NP
NP
~ Q
~
V NP
~ [:~~] I I
NP NP I zagen de ander
I I saw the other
Jan Wim elk
John Bill each
The noun phrase John and Bill each is generated by II, where (b) is a rule
schema:
II
(a) NP -+ NP' (Q(DADV))
(b) NP' -+ NP (NP)n
'Q' is a distributive quantifier (each, all, both, either, neither, respectively,
etc.) and 'DADV' is a distributional adverb (singly, alone, together, in con-
cert, mutually, simultaneously, at once, en masse, etc.). The feature complex
[+ indi~idualJ
- totalIty
becomes the surface f~rm each (in Dutch: elk, which presumably may 'fuse'
with de ander (the other) to form the word elkaar). Each is [+ individual]
for it co-occurs with alone. It is [- totality] since it does not co-occur with
simultaneously.
The transformational hypothesis requires rules which, in a structure like
John saw Bill and Bill saw John, conjoin the two subjects and convert the
'chiastically' co-referential noun phrases into something realized as each
other. Deriving each other from two noun phrases under 'chiastic' referential
identity with the two subjects is possible if the former two noun phrases are
(in)direct objects or if they are parts of prepositional phrases (where the
functions or prepositions associated with them must be the same)Y It is
also possible in certain cases in which it is not immediately clear what the
function of the relevant NP's should be said to be (see, for example, sentences
such as (78)-(80) on p. 72).
2.1.1.1.2. Assuming that semicopulas are simply V's leaves the ungrammatic-
ality of 'each other' semicopula sentences unexplained. As has been es-
tablished in Section 2.1 of Chapter I, there appears to be no reason to assume
that MP's are not simply noun phrases. On the contrary, the syntactic
behaviour and the appearance of MP's are such that it is quite reasonable
to claim that indeed they are noun phrases. No doubt they form a specific
class of phrases, which is why I shall keep referring to them with the symbol
'MP'. But apart from the fact that MP's are distinct from other NP's in a
number of important respects, they show the general characteristics typical
of NP's: they occur with prenominal adjectives, or as the antecedent of
relative clauses; they take the plural ending, which with MP's is of the same
13 See also Gleitman (1965), p. 280-283, Lakoff and Peters (1966), p. 78. Seuren (l969a),
p. 236-237. and Dik (l969a), p. 259-260, for some discussion of the derivation of each other
sentences.
62 THE STRUCTURE UNDERLYING MP SENTENCES
form as that in other NP's; they can be distinguished as to gender of the head
noun, and so on.
Should we assume that kosten is generated directly as the V in the sentence,
then it would be impossible to determine from the structure index that the
MP (given that it is an NP) is not a direct object. There is no structural
difference between, say, III and IV.
III s
~VP
NP
I~
V
een hotelNP
one hotel
I
kost
I
5 huizen
costs 5 houses
IV s
~
NP VP
I~
Jan V NP
John I I
zag Wlin
saw Bill
It might be objected that assigning to the phrase 5 huizen in III the structural
property of being dominated by the VP node may be unjustified (that is,
the question could be raised why the measure phrase could not be considered
a sister node of VP (immediately dominated by Predicate Phrase) and thus
be distinguished from direct objects). The question could conceivably be
relevant, since if it could be shown that MP's arc not in the VP, the un-
grammaticality of each other semicopula sentences could possibly be ex-
plained in terms of their grammatical function. 14
14 Thinking of a proposal in Chomsky (1965), we might consider NP's like 5 huizen in III as
constituents whose function can be further defined in terms of strict subcategorization of verbs.
Since middle verbs like kosten do not take manner adverbials and cannot undergo the passive
transformation, Chomsky suggests "that tl]e Manner Adverbial should have as one of its re-
alizations a "dummy element" signifying that the passive transformation must obligatorily
apply." (op cit., p. 103). Thus we might have a rule of the form ·Manner .... by.-.Passive'. If
b/--Passive occurs in a deep structure, the passive transformation thus must apply. We might
now try to explain in terms of a negative specification of semi-copulas for the strict subcate-
gorization feature [-NPManner) that semicopulas do not take each other phrases. But this
will not do, since there are verbs followed by NP which do not take manner adverbials but which
STRUCTURE UNDERLYING SIMPLE MP SENTENCES 63
But insofar as anything can be said about the structural relation to the
VP of direct objects, everything that may be relevant to the latter is also
true of the relation ofMP to VP. Lakoffand Ross (1966) have claimed that
do so replaces all of the constituents of the verb phrase and only these, in the
sense that elements which may occur after do so are "outside" the VP (are
not constituents of VP), and elements which cannot so occur are inside the
VP. Their claim embodies to some extent a departure from Chomsky (1965),
pp. 101-105, where a principle is proposed of strictly local subcategorization
to account for the fact that some adverbials are restricted in their occur-
rence with certain classes of verbs, whereas others are not.
Lakoff and Ross illustrate that the phrase do so "is a proform which may
substitute for a verb phrase" with the following examples:
(21) Harry forged a check, but Bill could never bring himself to forge
a check.
(22) Harry forged a check, but Bill could never bring himself to do so.
Do so cannot be used in this way, however, with verb phrases containing
'stative' verbs, like know. VP's containing adjectives may not reduce to
do so either. The Dutch equivalents of do so are doen + dat ('do' + 'that')
and doen + het ('do' + 'it'). Thus we have:
(23) Jan schreef, en hij schreef met een pen.
'John wrote, and he wrote with a pen.'
(24) Jan schreef, en hij deed dat met een pen.
'John wrote, and he did that with a pen.'
In the cases where the doet dat (does so) test does not work, we can use the
en wei test (I shall translate en wei as and ... so) :
(25) *Wim wist het antwoord en (hij) deed dat direct.
'Bill knew the answer and (he) did that immediately.'
(26) Wim wist het antwoord, en wei direct.
'Bill knew the answer, and immediately so.'
Sentences like (23) are called in Kraak and Klooster (1968) 'specifying
conjunctions' (op. cit., p. 258ff.). Obviously, the forming of doet dat sen-
tences is very similar to pronominalization. In specifying conjunctions, and
can occur in passive sentences and each other sentences (for example, Zij hoorden elkaar ('They
heard each other'; see also p. 70). We might therefore, embracing the PSR hypothesis, introduce
a strict subcategorization feature which in some way or other contains a representation of the
context typical of deep structures underlying each other sentences, for which semicopulas would
have to be negatively specified. But this solution would fail to explain why verbs followed by a
VP-dominated NP do not take each other phrases if they belong to the class of verbs not oc-
curring in nominalizations of the type Jans V-en van NP ('John's V-ing of NP'; see pp. 78-9),
i.e., to the class of transitive verbs.
64 THE STRUCTURE UNDERLYING MP SENTENCES
in a host of other cases, the conditions for pronominalization, the doet dat
substitution, and deletion are of the same sort; there must be some sort of
"identity", which for the time being let us call 'co-reference' or 'referential
identity' (although it may be maintained that the referents of, e.g., the two
phrases drank a glass of water, or of the pair 'Lolita' and it in John drank
a glass of water and Bill drank a glass of water and John bought 'Lolita' and
Bill bought it too, respectively, are not identical). Clearly, there can only be
referential identity between two strings of words if each of them is ex-
clusively dominated by one node, that is, if each of them is a constituent.
Thus in (27)-(30) the italicized parts are single constituents:
(27) Jan hielp zijn zoon en Piet zijn dochter. (second hielp deleted)
'John helped his son and Peter his daughter.'
(28) Jan helpt zijn zoon en Piet slaat hem. (zijn zoon = hem)
'John is helping his son and Peter is beating him.'
(29) Jan helpt zijn zoon en Piet ook. (second helpt zijn zoon deleted)
'John is helping his son and Peter too.'
(30) In de tuin hielp Piet zijn zoon en hielp Jan zijn dochter.
'In the garden Peter helped his son and John helped his daugh-
ter.'
Hence, for all we know, a structure of the fonn III would be correct if we
should generate kosten directly as a V.
Thus it would be difficult to explain in terms of syntactic function why
each other semicopula sentences are ungrammatical. On the other hand,
there is sufficient reason to assume that MP's are not direct objects in simple
MP sentences (see pp. 78-9). We need not, however, call upon this fact in
order to explain the ungrammaticality of (20). Indeed such an explanation
would remain unsatisfactory as long as it is not completely established that
elkaar co-occurs exclusively with (in)direct objects and NP's in preposi-
tional phrases (cf. my remark in the first paragraph on p. 61, and (78)-(80)
on p. 72.
As we shall see, there is another, more important condition which is
not met, namely, that of referential identity.
the claim that the structure underlying simple MP sentences with semi-
copulas contain an element (or string) relatable to hebben.16 (What has been
said here with respect to hebben paraphrases also holds for paraphrases
containing van constructions like De betekenis van dit woord is etc. (The
meaning of this word is' etc.). But as was remarked earlier, suchvanconstruc-
tions can be derived from structures which also underlie the hebben para-
phrases.)
strengthen our case, since, as we shall see, have in tum can be related to be.)
from objects the values of whose parameters are expressed in terms of ab-
stract units. This becomes clear in hebben paraphrases of (76), such as (76a):
(76a) Het is niet altijd zo dat een dichtregel een lengle heeft die een
vinger bedraagl en dat een vinger een [engte heeft die een dicht-
regel bedraagt.
'It is not always the case that one line of poetry has a length that
amounts to one finger and that one finger has a length that
amounts to one line of poetry.'
Sentence (76a), by virtue of analyzing lang zijn as lengte hebben, illustrates
by explicit formulation that the first occurring phrase een vinger (one finger)
and the last occurring phrase een dichtregel (one line of poetry) are both
specifications oflength, in the sense that they are used as unit words, and not
the names of physical objects (een lengte die een vinger bedraagt ('a length
that amounts to one finger'), etc.). The ungrammaticality of(8l), below, is
also related to that of (77), and forms an illustration analogous to the one
given earlier with kosten (cf. pp. 85-6):
(81) *Oe lengte die het ene object heeft is het andere object.
'The length that the one object has is the other object.'
I will now consider two possible types of underlying structure for simple MP
sentences, one in the section below, the other in the section following it.
After discussing the reasons for rejecting them, I will present arguments in
support of a third possibility.
!~NP
Jan V
JOhn! ~
heeft NP S
has_~:::::~~~=~~--:=~________~~ ~
een bepaald(e) {g,eWiCht} ~ ~VP
engte / /' '"
/ ,
a certain {~:::~:} {~~tg;;"lch~t'} y /~~,
die lengte : ~~____ ~~"'"
{that weight l bed~aagt {80 kilo }
that height J amounts to 2 meter
{ 80 kilOS}
2 meters
STRUCTURE UNDERLYING SIMPLE MP SENTENCES 81
i~
Jan
V NP
John i ~
heeft NP PP
has ///~"'" ~
~"--------':::> P MP
een {geWiCht} I /''',
lengte i
I /',
L_____ .::'"
a {we~ght} van { aD kilo }
heIght of 2 meter
{ aD kilOS}
2 meters
NP
~VP
Jan
l~
V Adjectival
John i ~
is Adj MP
is I
I
I
{zw~ar}
lang
{ heaVY}
tall
VIII 5
~VP
NP
l~
Jan V NP
John i ~
is NP 5
I
is
~VP
I
I
I
het NP
I
it
!
Jan Adjectival
John ~
Ad, J MP
/',
I
I
/ '
_ _ _ _ _ _ _ _ !tr.
/;' '"
{z~~~r}
~
{ 80 kilO}
2 meter
{h;:,?} { 80
2 meters
kilOS}
(Let US assume, for the moment, that [NP it SNP] structures are, in general,
justified, although it is not clear why underlying it could not be dispensed
with here.)
Ross's arguments for assuming underlying structures of the type NP 1
be [NP it [g NP 1 [v Adj v] s] NP] boil down to the following.
In several languages, it is possible with pronouns to refer to properties
expressed in the form of adjectives: 'pro-adjectives' are often phonologically
identical to pronouns which replace nouns. In German there is a rule moving
NP, prepositional phrases and adverbs to the front of a sentence: Bohnen/im
Garten/genusslich haben wir gegessen. Ross argues that these should be all
regarded as being dominated at some point in their derivational history by
NP, so the preposing rule should only mention that NP can be moved to the
front of the sentence. The fact that adjectives can also be preposed (Stolz auf
dich muss er sein!) can be automatically accounted for if the phrase stolz
auf dich is dominated by NP. Furthermore, in English and other languages,
an element which starts a noun phrase cannot be moved alone to the front
of the sentence by the question transformation (* Whose did you read
book?) the whole NP which the questioned element starts must be moved
with it ([NP Whose book NP] did you read?). That the same thing is true of
adjectives ([NP How old NP] is John? but not *How is John old?) again suggests
that adjectives must be dominated by NP in the underlying structure, even
though later rules remove all trace of this node, so that adjectives in the
STRUCTURE UNDERLYING SIMPLE MP SENTENCES 83
A difficulty would be that kosten and duur zijn (cost and be expensive) or
wegen and zwaar zijn (weigh and be heavy) differ too much phonologically
to be realizations of the same underlying morphological structures. The
disadvantages of V are similar. If we accept a structure like V, we must
require that hebben + gewicht (have + weight) may be converted into wegen
(weigh) (and in some cases into zijn + zwaar (be + heavy), cf. the examples
(90a)-(94b)). If we accept VII or VIII, we must require that zijn + zwaar
may be converted into wegen or hebben + gewicht. (I will not go into the
question of the treatment of the determiner in such cases.)
No matter which of the two alternatives indicated in the preceding two
subsections we choose, the difficulties in either case will be essentially of the
same nature:
(a) Both solutions necessitate rules that would be thrust upon us by one
of the two types of underlying structure we would opt for, while it is by
84 THE STRUCTURE UNDERLYING MP SENTENCES
no means clear which of the two is to be preferred over the other. The
choice, and the ensuing rules, would be arbitrary. In either case, moreover,
the rules would be just as complex, so that in that respect there is no criterion
for a choice either.
(b) The 'standard theory' requires that the terminal strings of deep
structures be strings of morphemes. Consequently, the transformational
rules operate upon structures with phonologically specified terminal strings.
As we noted, the strings that are to be related via transformational rules,
differ phonologically to such an extent that such rules, apart from being
cumbersome, would lead to trivial solutions. Instead of being related trans-
formationally, semicopulas, adjectives and hebben should be related in the
lexicon.
3.4. Underlying WITH String
The considerations (a) and (b), presented in the preceding subsection, tend
to cause us to look in another direction.
Suppose that we assign to simple MP sentences an underlying structure
which does not contain morphemes as terminal elements, a structure,
moreover, sufficiently abstract for us to be able to circumvent a choice
between an underlying have or an underlying adjective for all simple MP
sentences. Such a structure might roughly look like IX.
I do not want to imply that a structure like IX is a base structure. Quite
possibly, a number of structural changes will have to take place before
anything like IX will result. At present, however, this will not concern us.
IX s
~VP
NP
i~
(Jan)
V PP
I~
BE P NP
I
WITH
/~S
NP 1
I~
{WEIGHT} NP1 VP
HEIGHT I ~
{WEIGHT} V MP
Jan {weegt 80 kilo } HEIGHT I / /
/", " ,
L ______ -'>
is 2 meter lang
John [",:eighS 80 kilOS} BE <{ 80 kilO}
IS 2 meters tall 2 meter)
STRUCTURE UNDERLYING SIMPLE MP SENTENCES 85
i~
V
(Jan) PP
I~
BE P NP
I~
WITH DET N,
~I
ARTREL {WEIGHT}
I HEIGHT
S
~VP NP
~N,
DET V
~MP
I I , /
",,//"-.,
"'------_'\.
WEIGHT}
{ BE
HEIGHT
<{280meter)
kilO}
IXa(i)
NP1 BE [NP 6 ~ [vp WITH [NP [OET ART REL OET] N NP] yp] s] NP]
(something like 'John be [NP 6 ~ with [NP height 6 height be MP s] NP] s] NP]',
or 'John be [NP 6 ~ with [NP [OET ART REL OET] height NP] s] NP]', where, in
the latter structure, REL dominates 'DET height be MP'.)
However, we can ignore elaborations of this kind, and I shall neglect
them throughout this study. There will be no problems in adapting the
rules to be proposed later on so as to allow for Ross's claim.
The words written in capitals in IX and IXa stand for elements that do
not have the status of morphemes, but of semantic categories, and therefore
must be considered arbitrary symbols, not "spellings". They are supposed
to form part of the vocabulary of the base component, just as the categories
NP, VP, V, etc.
Possibly, it would be justifiable to use the symbol 'TENSE' instead of
'BE'. The function of the copula be/zijn does not seem to be much more than
that of being the tense carrier. The further specification 'WITH' of the
more general category 'P' must be regarded as the category representing a
specific relationship which is often expressed in Dutch by means of the
word met. The parameter categories WEIGHT and HEIGHT are tentative.
I am aware that there is some unclarity here as to the criteria by which
one should distinguish between sy'ntactic and semantic categories. But
then, problems of this sort are not entirely new or surprising, for, as
Chomsky notes, "it is clear that the intuitive notion of grammatical well-
formedness is by no means a simple one and that an adequate explication of
it will involve theoretical constructs of a highly abstract nature, just as it is
clear that various diverse factors determine how and whether a sentence
can be interpreted." (Chomsky (1965), p. 151.) In fact, "it should not be
taken for granted, necessarily, that syntactic and semantic considerations
can be sharply distinguished." (id., p. 77). However, theoretical considera-
tions with regard to the 'syntax' j'semantics' distinction need not concern us
here directly, as we are at present dealing with highly tentative descriptive
devices, which only after having more fully explored their possibilities we
may fruitfully discuss in this respect.
Lexical insertion, in structures like IX and IXa, does not necessarily take
place before application of transformational rules. 20 More on the subject
of lexical insertion will be said later. At present, it suffices to say that the
categories symbolized as BE and WITH form a string which may become
manifest as the word hebben (have). Strings roughly like WITH + WEIGHT
and WITH + HEIGHT may under certain circumstances turn into the
adjectives zwaar (heavy) and lang (tall), respectively. The latter two words in
that case must be taken in their neutral sense. Parameter adjectives occurring
in the non-neutral sense have a more complex underlying structure. The
WITH string underlying such a non-neutral adjective contains some element
of degree and/or norm.
Measure adjectives, then, and in general parameter adjectives, are analyz-
ed as prepositional phrases. The verb hebben is analyzed as BE + WITH,
and semicopulas as verbs which are partly copulas, i.e., as BE + WITH + NP.
Later on, I shall propose a number of transformations which, among
other things, can express the relation between structures of the type of IX
and X or, with some alterations, between IXa and X (whichever may be the
more correct alternative):
X s
NP
~VP
I
~PP
I
I
(j~n) V
I~
BE P NP
I~
WITH NP MP
A
,,
I
/
./
4 _ _ _ _ _\.
{WEIGHT} ({ 80 kilO})
HEIGHT 2 meter
The fact that the following a examples are synonymous with the correspond-
ing b examples is evidence for the relation of hebben to met (with):
(l04a) Een man die een gewicht van 80 kilo heeft
'A man who has a weight of 80 kilos'
88 THE STRUCTURE UNDERLYING MP SENTENCES
NP)
SD.: X - [NP - [ [+ Wh
{AUX
[
be}) -X) ) _ X OPT ~
NP S X be 5 NP
2 3 4 5
sc: 2 o 4 5
STRUCTURE UNDERLYING SIMPLE MP SENTENCES 89
way in which honger and dorst are analyzed). The differences in meaning
are often highly subtle, and it is by no means always clear how one should
go about formulating them in terms of underlying strings. We have a
somewhat clearer case in such pairs as Ik heb geluk-I am lucky. The senten-
ces Ik ben gelukkig and I am lucky mean totally different things and accord-
ingly the analysis of gelukkig (happy) here should include an abstract noun
category representing a meaning clearly different from the one represented
by the category underlying the word geluk in Ik heb geluk.
A close approximation of the Estonian sentence (116a), below, is given,
in Lehiste (1969), in the form of (166b); that of the synonymous sentence
(117a) in the form of (117b). In Estonian, there is no surface verb 'have'.
The word on is the third person singular of 'be' in the present tense.
Noormees ('the young man') is the singular of the nominative and uhke
hoiakuga ('proud bearing') is the singular of the comitative. Noormehel in
(117a) is in the adessive case and uhke hoiak is in the nominative case.
(116a) Noormees on uhke hoiakuga.
(116b) 'The young man is with a proud bearing.'
(117a) Noormehelon uhke hoiak.
(l17b) 'The young man has a proud bearing.'
The closest approximation, apparently, of the comitative case ending is
the preposition with. The verb have presumably expresses adequately the
relation of 'possession' (lIse Lehiste's characterization) given with the ades-
sive case ending. The comitative may correspond to the abstract relational
category symbolized as WITH.21
strings but reject rules relating strings like weegt and is zwaar. We would then
be forced to assign dissimilar underlying structures to simple MP sentences;
some with V's dominating semicopulas, others with Adj's dominating
parameter adjectives. In that case we could no longer state the generaliza-
tion that ( 1b) Jan is 80 kilo and (3b) Jan is 2 meter are both instances of
optional reduction. We would have to maintain that the structure under-
lying (l b) is distinct from the one underlying (la) with wegen, and either
that the sentences (1 b) and (3b) are not reductions at all, or that (lb) and (3b)
are instances of deletion of underlying adjectives, in the former case
obligatory, in the latter case optional. But it cannot be reasonably main-
tained that in the former sentence obligatory deletion has taken place of an
underlying adjective zwaar (the non-occurrence of *Jan is 80 kilo zwaar
notwithstanding), whereas in the latter sentence the adjective lang has been
optionally deleted. Similarly, although *The toy is 2 dollars expensive is
ungrammatical, it is implausible that The toy is 2 dollars is an obligatory
reduction, while John is six feet one is an optional reduction.
calculus of the kind envisaged in Gruber (I 967a) (see pp. 4 and 303). TO corresponds to what
in Fillmore's terminology would be Dative, under which the Estonian adessive would have to
be subsumed (see Fillmore (1968a)). As is remarked in Fillmore (l968b), "relations between
verbs that are like the converse relations of the theory of predicates involve position-switching
for subject and object ( ... )" (op. cit., p. 337). The same could then be said of such elements as
WITH and TO. Perhaps these categories should be seen as V's (see p. 220), which would be
more or less in keeping with proposals in Lakoff (1965).
If our analysis of adjectives like tall as prepositional phrases is essentially correct, Fillmore's
remark no longer holds from a linguistic point of view that a predicate with one argument has the
simple predicate adjective or intransitive verb Pa (where' P' stands for 'predicate' and 'a' is
the argument; Fillmore (1968b), p. 375-6), for tall should then be seen as a two-place predicate
WITH NP • NP ' The McCawley-type base structure representation of NP I is tall would then be
Is Iv WITH vl NP I NP2 sl·
It is worth noting that the surface preposition with occurs also in numerous types of sentences
where it cannot be assumed to correspond to the category which I have labelled WITH. Nor,
of course, is 'have' always to be analyzed as BE WITH. (Cf. She was with me, She had me.)
See also, on 'have' and 'be', Bach (1967).
92 THE STRUCTURE UNDERLYING MP SENTENCES
zwaar and duur (Dat boek is iets zwaarder/duurder (,That book is a little
more heavy/expensive') vs. *Dat schip is iets stekender (That ship is a little
more drawing'), *Dat geschut is iets dragender (,That artillery is a little more
more carrying'». It would therefore be rather arbitrary to prefer 2 over 1;
under 1, no underlying adjectives were assumed where they never manifest
themselves whereas under 2 they were assumed in all cases.
Thus, should we abandon the idea of an underlying WITH string, or
some other underlying string serving the same explanatory purpose, a
choice would be forced upon us between two possible explanations, neither
of which is really satisfactory.
On the other hand, if we should substitute "WITH string" for "measure
adjective", the latter explanation would no longer be based on an arbitrary
claim, since, as we saw, there is evidence supporting the assumption of such
a string.
5. RECAPITULATION
In the first section of this chapter it was pointed out that there can be made
a distinction between two types of simple MP sentences. Type I sentences
may undergo reduction, whereas Type II sentences cannot. In order to be
able to formulate a rule or rules for reduction in simple MP sentences, and
also to account for other phenomena exhibited by them, a number of
pieces of evidence have been presented in favour of a certain uniform type
of underlying structure in simple MP sentences.
One similarity in the behaviour of simple MP sentences consists in the
fact that they all have the same paraphrase relation to hebben, which more-
over can be used as evidence in support of the claim that the structure under-
lying MP sentences contains a WITH string.
Hebben paraphrases reveal the source of the ungrammaticality of senten-
ces with each other as the complement of either a semicopula or a copula +
96 THE STRUCTURE UNDERLYING MP SENTENCES
I s
NP VP
v
~PP
I~
BEP NP
I
WITH NP NP
I
WEIGHT
P
~NP
I~
WITHNP NP
I
WEIGHT
I
AMOUNT
IlIa PP
~NP
P
I~
WITHNP NP
LENGTH
\ I
AMOUNT
+ lang, +
IIlb PP
~NP
P
I
WITH NP MP
I
LENGTH
IV S V S
~VP
NP
~VP
NP
V
~PP --;;.. ~PP V
I~
BE P NP
I~NP
BE PP
I ~Y
WITH NP
~~
P NP Y
I
WEIGHT
I
WITH
I
WEIGHT
2nd step
VI s
NP
~VP
--~ v --..,.
~PP
V NP
I~
BE P NP
~y
I
WITH
I
WEIGHT
VIII s
VI~
NP VP
~VP
NP
~NP
V
~NP
V
V
~~Y ~ --".
~~ V Y ~
I~
BE NP P
~I
NP P BE
WEIGHT
I I
WITH
I
WEIGHT
I /
WITH /
---------- ------\/
\,,'
wegen
say that, of the nodes immediately dominated by VP, only the leftmost node
is replaced by a new one. There is some inconsistency in this.
What could be the possible answers to these objections?
First, one could say that the string BE WITH WEIGHT (or WEIGHT
WITH BE) is a string underlying a verb and not a verb phrase. This
argument would be along the lines of Gruber (1967a), where he objects to
considering enter anything else but a V. The lexical entry of enter, according
to Gruber, should look something like IX:
IX
VP
~
v PP
GO
I P NP
INTO
I
'II' enter 'II'
v
~V
P NP
I
INTO
.....
GO
I
/'
'.......
,....,." ....
,./
'*' enter *'
.......
XI a b
VP VP
~PP ---;.
V
~PP ---;.
V
I~
BE P NP
I~
BE P NP
I~
WITH NP Y
I NP ~Y
WITH
WEIGHT
I I
WEIGHT
c d
VP VP
~PP
V ---.;.
~PP
V ---..;.
BE
I ~NP I
BE PP'
~NP
P NP
~Y ~NP
P
~Y
I I
WITH WEIGHT
I
WITH
I
WEIGHT
e (Cf.Y) f
VP VP
~PP
V ---.;. V
~PP ---.;.
BE
I~
PP' NP
I
BE PP'
~NP
~~
PNP Y P
~~
NP Y
WITH
I I
WEIGHT
I
WITH
I
WEIGHT
THE DERIVATION OF MP SENTENCES 105
9 h
VP VP
~PP
v ---~
~PP
V ---~
i (Ct. JlI)
VP
~PP
V
~ppl ~NP
VI
I P~~
BE NP y
I
WITH
I
WEIGHT
XII a b
VP VP
~PPV ---~
~PP
V ---~
I~
BE P NP
I
BE
~NP
I~
WITH NP Y P NP
~Y
I
WEIGHT
I
WITH WEIGHT
I
c d (Cf.]Z")
VP VP
~PP
V ---~
~PP
V ----+
BE
I~
pp' NP
I~
BEpp' NP
P NP
~Y ~NP
P
~Y
WITH
I I
WEIGHT
I
WITH WEIGHT
I
e f
VP VP
~PP ---~
~
VP. PP ---~
V pp'
~NP V PP'
~NP
IA
BE P NP
~I ~~
Y BE P NP Y
I
WITH WEIGHT
I I
WITH
I
WEIGHT
THE DERIVA TlON OF MP SENTENCES 107
vp
~PP VP'
v
~~ pp' NP
I~~
BE P NP Y
I
WITH
I
WEIGHT
22 According to the view presented in Gruber (1967a), however. no order reversal takes place
where it is not apparent in the morphemic order. But I think it will simplify a grammar if we
need not state in any way in every specific case whether or not order reversal is to take place.
As is pointed out on pp. 127~8. it may be preferable for productive affixes to have lexical en-
tries whose simultaneous environments include categorial variables or unspecified categories re-
presenting the classes of morphemes to which affixes are connected. rather than to postulate
108 THE STRUCTURE UNDERLYING MP SENTENCES
incomplete lexical items which after having been connected with other lexical items are at-
tached to the derived tree, as Gruber proposes. This means that we could introduce another
type of 'incomplete' lexical entries for affixes which could reverse the order of the affix node
or nodes and the as yet unlexicalized string of categories matching the categorial variables or
unspecified categories within the simultaneous environment (without affecting the order within
the latter string). Thus if A is the node underlying the affix and X the node underlying the
phonological string to which the affix is to be connected. where the prelexical order is A X. the
entry will change the order into X A, and lexicalize A. A further entry. applied afterwards. will
then lexicalize X. Thus order reversal as it is apparent in words consisting of two morphemes
will not be the result of attaching the polymorphemic word in stages. since order reversal will
already have taken place during lexicalization of A. If X is a string of two or more nodes (where
X corresponds to a string variable in the simultaneous environment). only the order of A X
will be reversed during the first stage. while a further entry (or further entries) will reverse the
order within X.
THE DERIVATION OF MP SENTENCES 109
I
bedragen ('amount to') can be reduced,
{ meet 21 m (m )
2 3 }
(2a)i. H
I
r is 21 m 2 (m 3 ) groot
et voorwerp weegt 30 k'l10
J
'
I
kOSI 100 gulden
21 m 2 (m 3 )}
j,
measures
{
'Th b' is 21 m (m 3 ) large
2
e 0 ~ect 'h s 30 k'llOS
welg '
costs 100 guilders
prijs
grootte
ii, De lengte bedraagt zo en zoveel eenheden,
helling
I
I
price
volume/area
'The length amounts to so-and-so many units,'
I
gradient
J 21 m (m
2 3
)
(2b)i, Het voorwerp is 130 kilo ,
100 gulden
prijs
grootte
ii. De lengte is zo en zoveel eenheden.
helling
price
volume/area
'The length is so-and-so many units.'
gradient
In certain cases reduction may also occur with human subjects if the para-
meter is price (Deze slaaf is 200 dollar (,This slave is 200 dollars'». Such
sentences are less common as humans are not usually bought and sold. If
the subject refers to an animal or animals, reduction in MP sentences
referring to price is possible as well (Dat hondje in de etalage is 75 gulden
(,That doggie in the window is 75 guilders'». Reduction in sentences with
kosten is not possible if kosten2 is meant (see p. 52).
Age is often treated as a main parameter of animals, particularly pets.
This is one of the many instances of the 'humanizing' of animals reflected
in language.
We may now replace A with the conditions under B:
B. Reduction may occur in the case of
(i) age: with human subjects and sometimes with subjects referring
to animals,
(ii) weight: with animate subjects and subjects referring to physical
objects,
(iii) height: with human subjects (Dutch: lengte),
(iv) price: with subjects referring to things that can be bought, ab-
stract or concrete,
(v) area: with subjects referring to physical objects,
(vi) volume: id.
(vii) quantity of units: with subjects containing parameter nouns.
There are a number of cases which do not meet B but are nonetheless not
entirely ungrammatical. Thus, for instance, (3) and (4) do not sound odd:
We might speculate, in this vein, that length and height in some way are
associated in our conceptual framework with the main axis of the human
body and that this is why the 'secondary' dimension called width or breadth
so often turns out to involve position relative to the observer. However, I
shall not here take this matter any further. A thorough inquiry into it
would involve many aspects beyond the scope of this study. I think, non-
etheless, that it is reasonable to claim that, in all cases where the maximal
axis is width, the position relative to the observer is relevant.
Returning to the question of the different degrees of acceptability of
THE DERIVATION OF MP SENTENCES 115
lang zijn
(30) een weg van 10 kilometer
'a road of 10 kilometers'
(31) een weg van 10 kilometer lengle
'a road of 10 kilometers length'
(32) *een weg van 10 kilometer lang
'a road of 10 kilometers long'
To some speakers, including myself, expressions like (26), (29) and (32)
sound quite acceptable. But the majority of my informants reject them. I
shall therefore regard them as ungrammatical.
The conditions for reduction in van + MP constructions are different
from those for reduction in simple MP sentences, as can be seen from
examples like the following:
(33a) een pauze van 20 minuten
'an intermission of 20 minutes'
(33b) *De pauze is 20 minuten.
'The intermission is 20 minutes.'
(34a) een bureau van 2 meter
'a desk of 2 meters'
(34b) *Het bureau is 2 meter.
'The desk is 2 meters.'
Are there cases in which 'depth' refers to the maximal axis? The maximal
measurement of a vertical mine-shaft is its depth. But 'hole-like' things,
such as mine-shafts, wells, pits, etc., are not objects proper in the sense
desks are, and cups, cupboards, houses, etc. (all of which have depth). We
do not use the word 'axis' in connection with 'hole-like' things, whereas we
do use it in connection with objects proper. This difference can be related
to the following observation. It is unthinkable that 'hole-like' things can
move unless their immediate environment moves, whereas on the other hand
it is not unthinkable that objects proper, regardless of their mass, move
without their immediate environment moving, too. Objects proper can be
imagined to have an axis around which they can be moved independently;
'hole-like' things cannot. There are no objects proper, at least that I have
been able to think of, whose maximal axis is referred to as 'depth'. It seems
clear that we must differentiate between axis measurements and other
spatial measurements. (Compare (36a) *een stuk karton van 0,5 millimeter
(axis measurement) and (35a) *een put van 10 meter (not an axis measure-
ment). (35a) is always unnatural, whereas (36a) is not when the topic of
discourse is thickness.)
If the van + MP construction refers to a non-spatial parameter or to
the main one-space parameter (i.e., to height of humans), the MP cannot be
followed by a parameter noun; if it refers to a spatial parameter, the MP is
either optionally or obligatorily followed by a parameter noun, depending
on whether it refers to either a two-space or three-space parameter, or to a
maximal axis parameter (provided, of course, the construction does not
refer to the main one-space parameter):
(37a) een vloer van 9 m 2 oppervlak
'a floor of 9 m2 area'
(37b) een vloer van 9 m 2
'a floor of9 m 2 '
(38a) een booms tam van 2 meter omvang
'a log of 2 meters circumference'
(38b) een boomstam van 2 meter
'a log of 2 meters'
(39a) een boot van 1 vadem diepgang
'a boat of 1 fathom draught'
(39b) *een boot van I vadem
'a boat of I fathom'
(40) een man van 2 meter (*Iengte)
'a man of 2 meters (height)'
liS THE STRUCTURE UNDERLYING MP SENTENCES
Of the b examples of (37)-(39), only the first one can be considered a correct
reduction of its corresponding a counterpart; (3Sb) is not synonymous
with (3Sa), and (39b) is ungrammatical.
We may conclude the following: The measure phrase is optionally fol-
lowed by a parameter noun if the expression refers to a maximal axis para-
meter (cf. (24), (25), (27), (2S), (30) and (31) or to a two-space or three-space
parameter. It must be followed by a parameter noun if the expression refers
to a non-main spatial parameter (cf. (36a) and (39a, b». It cannot be follow-
ed by a parameter noun if the expression refers to a non-spatial parameter
(cf. (41)--(45» or to the main one-space parameter (height of humans; cf.
(40».
There are MP sentences which have no van + MP counterpart. Thus,
alongside the a examples below, we do not have the corresponding b
constructions:
As for (46a, b), it can be remarked that schuld (,debt') does not refer to a
property; it refers to a relation. Therefore, it cannot refer to a measurable
property, that is, it cannot be considered a parameter noun.
The ungrammaticality of (47b) must be explained in another way. There
are only two cases in which constructions are possible of the form
NP + van + MP where the measure phrase refers to temporal units: (l) the
case in which the first noun phrase refers to humans (sometimes animates
in general), (2) the case in which the first noun phrase refers to things that
have duration (events). Inanimate nondurational nouns do not seem to be
associated with the notion of time as directly as are nouns referring to
humans and events. As we have postulated, age is a main parameter of
humans. A complete description of HUMAN nouns therefore should in-
clude (or at least imply), among other things, the fact that one of the main
parameters of humans is age. A complete description of the meaning of a
word can, in principle, be given in terms of the semantic categories present
in the entry of that word. Therefore the entry of say, man should somehow
indicate that a category AGE, and hence TIME, is involved. It should do
this either implicitly or explicitly. In the former case we shall need some
mechanism deriving the implication in question. In the latter, we may
simply include the category AGE in the entry of man. One of these two pos-
sibilities should be the case if we want to be able to explain in terms of
semantic categories why sentences like *Algernon's habit of scratching
himself behind the ear is twenty-one years of age are ungrammatical. Let us
assume, for the present, that the latter of the above possibilities is the case,
and that, consequently, words referring to humans must incorporate a
semantic category representing the notion of time. With regard to nouns
referring to events (such as operation), it seems clear that they, too, must
then incorporate such a category. Apparently, we must assume, further-
more, that no such category is incorporated in noun phrases referring to
inanimate non-durational entities (such as boek (book) and idee (idea», for
their meanings do not seem to be directly associated with the notion of time.
On the' other hand, we must explain why expressions like (47a) are
possible. For, if boek and idee do not incorporate a category TIME, one
would expect the same situation here as in the case of such words as
wedstrijd ('match'), pauze ('intermission'), etc., with respect to, for instance,
weight. We do not have *De pauze weegt 10 kilo (,The intermission weighs
10 kilos'). Presumably, this is because pauze does not incorporate categories
corresponding to the parameter of weight. A possible explanation for these
120 THE STRUCTURE UNDERLYING MP SENTENCES
After some time, a beard even stops being regarded as a thing with a
particular age. We do not have (SId) either:
122 THE STRUCTURE UNDERLYING MP SENTENCES
How should we account for the fact that simple temperature MP sentences
can only have the form NP+ Copula + MP?
The sentence
(S2) Het water is 30 graden.
'The water is 30 degrees.'
THE DERIVATION OF MP SENTENCES 123
P
~NP
I
WITH NP MP
I
TEMPERATURE
to which the application of this rule would be subject would, as far as I can
see, do no more than list a very small number of special cases, or perhaps
the transformation in question would be applicable only in the case of
temperature MP sentences. This would amount to essentially the same (but
in the wrong component) as what in fact the lexicon is supposed to do,
that is, provide us with "a list of basic irregularities" (to use Bloomfield's
(1957) formulation in his Language (p. 274)).
Let us call the set of MP sentences which cannot contain a semicopula
or a measure adjective the 'defective set of MP sentences' (or, for brevity,
'defective MP sentences'). The defective set of MP sentences contains as a
subset the defective set of simple MP sentences (or, 'defective simple MP
sentences'; that is, the set of simple MP sentences that only occur in the form
NP + Copula + MP). The complement of the latter set we may call 'de-
fective periphrastic MP sentences' (i.e., the set of periphrastic MP sentences
which have no counterparts in the form of simple MP sentences).
In general, there are no neutral adjectives or semicopulas for parameters
of recent invention, such as voltage, luminosity and electrical resistance. It
should be noted that only under special circumstances and in a highly re-
stricted number of these cases may the MP sentence take on the form of
NP + Copula + MP. In the remaining cases, which form the majority within
the defective set of MP sentences, only periphrastic forms containing
parameter nouns occur. We may speculate that, although temperature
sentences thus do not belong to the set of defective periphrastic MP
sentences, the fact that there is no temperature adjective or semicopula may
nonetheless have something to do with the historically relatively recent in-
vention of devices capable of giving an objective measurement of
temperature.
Most or all of the small group of defective simple MP sentences (except
temperature sentences) are acceptable only under special contextual and/or
situational conditions «*)Die peer is 75 watt (,That bulb is 75 Watts'),
(*)Die weerstand is 30 ohm (,That resistor is 30 Ohms')). I do not think the
absence of the parameter words in question in these cases should be account-
ed for in the lexicon, at least not in the way proposed above for simple
temperature MP sentences. Possibly some transformational mechanism
operating in cases where normally non-reducible simple MP sentences may
undergo reduction under the special conditions just mentioned should also
account for the defective cases under consideration.
It appears that defective simple MP sentences are highly exceptional.
As a rule a defective MP sentence is periphrastic. About the only really
acceptable type of defective simple MP sentences are simple temperature
MP sentences. The latter form an exception within the set of all simple MP
THE DERIVATION OF MP SENTENCES 125
sentences as well, namely, in that they are defective. It is because of this that
the lack of a lexicalizable WITH string should be considered an irregularity
in the case of temperature MP sentences, which must be accounted for in the
lexicon. If temperature MP sentences had not been able to have the form
NP+Copula+MP, they would have belonged to the set of defective
periphrastic sentences. For such sentences we need the sort of entries which
are necessary for all periphrastic MP sentences (including the hebben
paraphrases of simple MP sentences), such as entries for parameter nouns,
and an entry for hebben (cf. XIV).
XIV VP
~PP
V
I
BE P NP
I
WITH
# hebben #
There is one other fact about temperature MP sentences which can be ac-
counted for by assuming the existence of a phonologically null temperature
adjective. Reduction is prohibited in simple MP sentences if they occur as
restrictive relative clauses. Thus we do not have, e.g.,
(54) *een man die twee meter is
'a man who is two meters'
(55) *een steen die twaalf kilo is
'a stone which is twelve kilos'
But simple temperature MP sentences that occur as restrictive relative clauses
are grammatical even though they do not contain an overt adjective:
(56) water dat 30 graden is
'water that is 30 degrees'
It is highly improbable that the MP sentence in (56) has undergone reduction
at some point in the prelexical derivation, for we would then have to claim
that there is a transformational rule obligatorily deleting the underlying
WITH string in relative simple MP sentences only when the underlying
parameter noun is TEMPERATURE (and perhaps in one or two other
cases). A solution of this sort is repellent. This strengthens our argument
that the adjective in (56) must be considered present, but that it is a zero
morpheme.
126 THE STRUCTURE UNDERLYING MP SENTENCES
One difference between lang and op in lang zijn and opeten, as they occur
with their noun phrase complements, is demonstrated in the following
examples (in the "literal" translations the word order is preserved):
(57a) Jan is 2 meter lang.
'John is 2 meters tall.'
(57b) Twee meter lang kan Jan niet zijn.
'Two meters tall can John not be.' (Two meters tall John cannot
be.)
(58a) Jan eet de sandwich op.
'John eats the sandwich up.'
(58b) *De sandwich op kan Jan niet eten.
The sandwich up can John not eat.'
The status of lang differs from that of op in that with the former a noun
phrase is associated, viz., the complement twee meter, whereas op cannot
have a complement (in this case, de sandwich). In (58a), de sandwich is the
complement of opeten, or perhaps eten. This difference is a reflection of the
fact that op is an affix, while lang is not (neither is eten). Accordingly, eten
in the second schematized tree of XV is immediately preceded and followed
by a boundary symbol, while in XVI, below, it is lang which is immediately
preceded and followed by one in the second tree. In lang zijn, zijn in this view
should be looked upon as the affix. The combination lang zijn can be treated
in the same way as opeten in XV, as is illustrated in XVI.
A A
XVI
=IF zijn 'fI= + '1/= lang '1/= = =lF4I= lang 41= zijn =IF
~X
v
I
BE
'iF - - - - - z i j n 'iF
In XVII, the blank represents the string associated with the category that
may occur on the place of the variable X (where X is non-null). The strings
that may occur in the place of the blank will ultimately form a single 'word'
together with the affix zijn.
Note that, while a transformational cycle works on S's from bottom to
THE DERIVATION OF MP SENTENCES 129
top, the lexicalization cycle works, at least within each S, from top to bottom.
Working downwards during lexicalization is necessary in view of the fol-
lowing considerations. Suppose the V node were lexicalized before the
VP node. In that case, in a derived structure like [vp BE [pp WITH WEIGHT
MP pp] vp], BE would be lexicalized first. This would be wrong, since the
rest of the VP then could no longer be lexicalized, for there is no entry
containing the simultaneous environment WITH WEIGHT (cf. p. 100).
Instead, BE WITH WEIGHT should be lexicalized (as wegen). This dif-
ficulty can only be circumvented if we work from top to bottom.
The subtree containing the string of categories to which, ultimately,
lang zijn (first zijn, then lang) will be attached thus will be restructured com-
pletely after the PP cycle: first, during the VP cycle, the V node and the PP
node swap places, resulting in [vp [pp WITH LENGTH pp] BE vp] (with BE
lexicalized), then the P node and the NP dominating LENGTH swap places
during the PP cycle, thus giving [vp [pp LENGTH WITH pp] BE vp). That is,
in effect, the string BE WITH LENGTH will be reordered in the same
way as the string underlying wegen (cf. pp. 106-7), thus giving rise to
structures like the one represented under XVIII, below, which underlies the
sentence Het is twee meter lang ('It is two meters long').
XVIII s
~VP
NP
vp'
~MP
/"'-.. /\
/
pp' v
"'"
! I \
\
A
NP P
I
BE "
/ \ \
I
LENGTH WITH
I 1
I "
/ \
\
" I I I \
I " / I I \
II "I II I \
*
'.J L _____________ ~
# het # # # lang # zijn # # twee # 'IF meter
('it -long - be - two - meter')
(Of course, more trimmings will be necessary to get is, not zijn, in this
case. Notice that the NP dominating the parameter category in the lexical-
ized structure no longer forms a constituent together with MP. Compare, for
the way in which we arrive at this result, pp. 106-7.) As will be shown later
on, a post-lexical transformation can be set up which will turn the string
attached to the tree of XVIII into a string in which the elements, including
the boundaries, are put in the right order, that is, into a string of the type
130 THE STRUCTURE UNDERLYING MP SENTENCES
V
~NP
THE DERIVA TlON OF MP SENTENCES 131
This could be a possible entry, in the schema we are representing, and would require the
introduction into it of a V and an NP in order to complete it before attaching it to the derived
tree. But since both V and NP would have to be separate lexical items, they would both bring
with them # boundaries: Hence, there would be two between the V and the NP, forming two
words. Is this the way it should be? Conceivably it is. (Gruber (1967a), p. 123--4.)
It is only in the latter part of his study from which I have quoted above
that Gruber introduces the idea of entries with boxed-in simultaneous
environments. It is not clear which part or parts in the above entry should
be boxed in. We may replace Gruber's proposed entry by XIX:
XIX vp
v
~X
'/F----
Again, the blank here is associated with the X node. By XIX order-
reversal of the V and X will take place during the VP cycle, after which further
entries will introduce actual lexical items. We may formulate a convention
so as to prohibit application of XIX if the VP matches an entry causing a
VP to be lexicalized as a single word (as will be the case, for example, with
wegen and, possibly, lang zijn).
It may be of interest, in connection with the verb(")object order, and the
morpheme order in compound words in Dutch and German, to note that,
apparently, as Gruber observes, in Japanese there are some pieces of
evidence (stress, phonological rules and subsequent application of trans-
formations)
that post-positions, quantifiers and other things which manifest left-branching ( ... ) actually
form one word. (op. cit., p. 124.)
He then goes on to say,
But while the order between the verb and its object is also left-branching (object followed by
verb) there is no similar evidence that we have formed one word here. It would be reasonable
to claim that Japanese is a language which likes to preform its constituents in the lexicon before
attaching to the derived tree. Suppose left-branching is normal for morphemes being attached
to the derived tree as one lexical item ( ... ). We could then explain the word order between the
object and the verb in Japanese if we claimed they were attached at once by a lexical entry ( ... ).
But these are treated as separate words by transformations and phonological rules, and hence
we see that the rule holds that two # boundaries make two words. It appears that the entry
must be doubly incomplete here, because both the verb and the object are open word classes.
Also they never appear except with nouns, whereas NP may appear elsewhere than as the ob-
ject of the verb; thus while articles and post positions can be treated as affixes, the object of a
verb cannot be productively treated so. (id., p. 124).
Clearly, a solution like the one proposed by Gruber, even in its amended
form, will still not be adequate to account for the order of constituents
132 THE STRUCTURE UNDERLYING MP SENTENCES
{sn
I
PAR
XX (i) S-+ NP VP} UNIT
(viii) ABSTR -+ EVENT
(ii) IDET
NP-+ Npn N VAL
NP S
(iii) VP-+ V PP
I
(iv) V-+ BE
(v) PP-+ P NP (ix) CONCR-+ {A~IM}
(vi) !WITH
P-+ AT
23 Since objects are either abstract or concrete, we might therefore, in a less sketchy proposal,
do away with either the category ABSTR or the category CONCR.
A probably apparent counterexample, also cited by Gruber, with respect to the claim that
there are no selectional restrictions requiring the class of animate non-human nouns is the
German word fressen (= Dutch vreten), which means the same as essen (eat) except for
fressen occurring, as a rule, with subjects referring to animals. But, as McCawley (1971) cor-
rectly observes, "it is not clear that [essen and fressen] refer to the same kind of eating (which
verb do you use in reporting a well-mannered chimpanzee's eating something with a knife
and fork?)" (op. cit., p. 290).
24 An argument which seems to support the idea that 'non-particular' units do not occur in
MP's with underlying indefinite articles is the fact that sentences containing covert indefinite
articles like Hi} leest (geen) boeken (,He reads (no) books') correspond to er constructions
of the type Er zi}n (geen) boeken die hi} leest ("There are (no) books which he reads'), while
a sentence like Hi} weegt (geen) 80 kilo (,He weighs (no) 80 kilo') does not correspond to
*Er is/zi}n (geen) 80 kilo ( 's) die hi} weegt ("There is/are (no) 80 kilo(s) which he weighs').
THE DERIVATION OF MP SENTENCES 135
tion may be that geen now is allowed to result from NEG+ DET as well as
from NEG+ART. There are, of course, interesting problems involved in
finding really adequate DET rules. But I shall not go into that matter
here.)
Rule (xiv) rewrites PAR as a string containing optionally the category
MAIN, which will be needed whenever the MP sentence refers to main
parameters (see p. 108).
XXI (x) ANIM -+ (HUMAN)
(xi) DET-+ (ART)
(xii) ART-+ (DEF)
I
SPACE
TIME
(xiii)
1
UNIT -+ TEMPE~ATURE
I TIME
SPACE
(xiv) PAR-+ (MAIN) l TEMPE~ATURE
1
(xv) I
JONES PACE
SPACE-+ TWOSPACE
THREESPACE
As has been observed earlier (p. 114), width appears to be related to the
notion of 'observer' if it is the maximal axis of a physical object. Depth,
whether or not it happens to be the greatest measurement, always is observer-
related (p. 112f.). It appears, furthermore, that for simple MP sentences
containing spatial parameters which are not main parameters, it is to a
certain extent acceptable to omit explicit mention of the parameter, if it
is associated with the maximal axis, provided the parameter does not
involve the observer (cf. p. 100f.).
Thus, in simple MP sentences it is, to a certain extent, possible to delete
lang and hoog, when used in connection with the maximal axis of a physical
object. But it is unnatural to do so in the case of breed and diep. In van + MP
constructions, it is possible to eliminate not only the parameter constituents
corresponding to lang and hoog, but also the one corresponding to breed; if,
however, depth is referred to, there can be no reduction. 'Depth' never refers
to a maximal axis. In all cases that we have a possibly or necessarily observer-
related axis parameter, we have a sentence referring to an object with at
least two dimensions. Therefore we may conclude that the notion of 'ob-
server' comes into play only if we are dealing with objects to which two-space
or three-space parameters are relevant.
136 THE STRUCTURE UNDERLYING MP SENTENCES
Length and height do not necessarily imply more than one dimension.
Width implies at least two dimensions, depth always three (recall our ob-
servation concerning depth and shape, or inner structure). There are other
types of measurements which can be classified according to the number of
dimensions they involve, such as radius, diameter, circumference, etc.
It therefore seems justifiable to introduce categories indicating the
minimum number of dimensions which are involved. Since with spatial
parameters it cannot be the case that no dimensions are involved, it will
suffice to have a category signifying that no less than two dimensions are
implied, and a category signifying that no less than three dimensions are
implied, '201M' and '301M', respectively. Thus we may automatically
assume that one or more dimensions are involved if no 'DIM' category
appears at all under the category SPACE.
We may assume that as far as measurements are concerned, 'length' is
the one unspecified term, in the sense that 'width', 'height', 'depth', etc., are
all to be considered as 'length' plus some further specification. It may be
useful, therefore, to introduce a tentative category 'SPEC', which dominates
all semantic categories incorporated in one-space parameters not involved
in'length'.
It is assumed here, that if a one-space measurement ('MSR') parameter
is unspecified in the sense that it does not incorporate SPEC (i.e., if it is
'length'), it always includes a category MAX. It will be assumed, further-
more, that 'length' in all cases is an axis word, even applied to abstract one-
space objects. Thus 'length' applied to a line segment is also assumed to
include MAX.
If we have a specified (SPEC) measurement, it may be non-maximal.
Hence our rules must allow for this possibility.
In view of these considerations, the rules for spatial categories may look
something like XXII (below). Note that (xvi), (xviii), (xxii) and (xxiii) of XXII
are optional. In certain cases it is unnecessary to indicate anything more than
that one-space, square or cubic units are involved. The category 'DIST'
(DISTANCE) in (xviii) is introduced in order to differentiate between
'length' and 'distance', and between the 'locational' and 'axis' meanings
of hoog(te} and diep(te}.
According to (xx), there are two types of TRANSV (TRANSVERSE,
'width') axes. One is observer-related, the other is not. In the case of 'depth'
(associated with the tentative category 'PERSP' (PERSPECTIVE)), the
category 'OBS' is always present. The category 'TRANSV' may also be
useful for the description of such words as across, which is related to width.
For example, the sentence Fred went across the road means that Fred went
'along the width' of the road. The category 'OBS' may also be needed in
THE DERIVATION OF MP SENTENCES 137
entries for such words as left, right, back,forward, sideways, etc. No doubt a
category like 'THICKNESS' will have to be replaced, eventually, by some-
thing more adequate and illuminating, but at present it will suffice.
I~~~IUS
(xx)
(xxi)
J OBS PERSP
l
3DIM ..... THICKNESS
RADIUS
I I
I
ASSTR
XX'Za,b
ONESPACE
I
PAR
~
SPEC (MAX)
I
SPACE
I
VERT MSR
I
I
ONESPACE AXIS
I
I
MAX hoogte ('height' )
I
MSR
I
AXIS
XX'Zc
ONESPACE
I
VERT OIST
I
XXJ'Ia
ONE SPACE hoogte ('altitude')
~
SPEC MSR
I
201M
I
AXIS
XX¥ib
ONESPACE
I
TRANSV
~
SPEC
MAX
breedte ('width')
I
201M MSR
I
OSS
~TRANSV AXIS
I
XXJlIIa,b
ONESPACE breedte ('width')
~
SPEC (MAX)
I
301M MSR
I IX!ZIl C, d, e, f
ONESPACE
OSS
~PERSP
~
SP1EC CMr)
diepte ('depth') 301M MSR
~ I
OSS PERSP {OIST}
AXIS
XXJl1[[
PAR
~
diepte ('depth')
MAIN TIME
AGE
I
leeftijd ('age (of humans)')
140 THE STRUCTURE UNDERLYING MP SENTENCES
x~
CONCR PP
P NP
I
WITH NP NP NP
A
OET N OET
A N
A
OET N
ABSTR
I I
ABSTR ABSTR
I
I I
PAR PAR
prI
~
MAIN
SPACE
~
MAIN
WEIGHT
I
THREESPACE
SPEC
~MSR SPEC
~MSR
~
SPEC MSR
201M
I 201M
I I
301M
CIRCUM
I OIA
I I
THICKNESS
schijf ('disc')
porate it. Hence we see that different ways of looking at things may be
reflected in differences between the lexicons of various languages.
It will also be possible, of course, to set up simultaneous environments
for semicopulas with the base rules given above.
It should be stressed that the rules (i)-(xxv) are tentative. Many problems
remain to be solved, and, even quite apart from that, it will certainly be
possible to set up alternative rules which may be simpler and more general.
However, for our present purposes they suffice, and I shall not here try to
improve upon them or consider alternatives.
Note, finally, that our base rules also generate structures that we do not
want. We can, however, rely on the filtering effect of the lexicon, which
permits only certain structures to qualify for lexical attachment. Thus we
have two components acting as filters, the transformational component as
THE DERIVATION OF MP SENTENCES 141
well as the lexical component. The objection to having rules with a filtering
effect can be answered with Chomsky's remark that the apparent absurd-
ity in this is "simply a corollary to the deeper absurdity of regarding the
system of generative rules as a point-by-point model for the actual con-
struction of a sentence by a speaker. ( ... ) But a generative grammar as it
stands is no more a model of the speaker than it is a model of the hearer.
Rather, ( ... ) it can be regarded only as a characterization of the intrinsic
tacit knowledge or competence that underlies all actual performance."
(Chomsky (1965), p. 139-140).
I am not convinced that Seuren's criticism of this answer applies here.
Seuren observes that a matter of principle is involved:
We can either allow structures to shoot up freely within ihe most general limits set for structures
and use grammatical rules as calibres, or filters, rejecting the structures that do not fit. Or one
can predetermine the growth of structures in such a way that only such structures will be
XXX N
~PP
CONCR
P
~NP
I
WITH NP NP NP
A
DET N
A
DET N
A
DET N
A8STR
I
ABSTR
I ABSTR
I
I
I
PAR
I
PAR
I
PAR
~
MAINSPACE
/\
MAIN WEIGHT MAIN
~MONVAL
I
THREESPACE
~
SPECMAX SPEC
A MSR SPEC
A MSR
I MSRI
201M
I
VERT
I
AXIS 301M
I I
AXIS
/\ I
OBS TRANSV AXIS
A
08S PERSP
bureau (desk')
142 THE STRUCTURE UNDERLYING MP SENTENCES
XXXI~
CONCR PP
P
I
WITH
---------------
NP
NP
~=:::::::::::--
NP
A
DET N
A
DET N
ABSTR
I ABSTR
I
I
PAR
I
PAR
MAIN
~~
TWOSPACE MAIN MONVAL
~
ONESPACE
ONESPACE
SPEC
~MSR
~
SPECMSR
I
VERT
I
AXIS
I
2DIM
I
AXIS
I
TRANSV
foto ('photograph')
agree that nonetheless the matter seems worth investigating. Quite possibly,
a base component generating a semantic language will become in some
sense simpler to the extent that we are able to have it avoid generating
meaningless or contradictory representations. But even so, it may not be
possible to dispense completely with components acting as filters. We
should not exclude beforehand the possibility that the filtering effect of
components may be a reflection of an aspect of linguistic competence that
can be related to performance just as adequately as (or even more ade-
quately than) a base component with the tendency to avoid the generation of
ill-formed structures.
XXXII S
~VP
NP
~~
DET N V PP
I~
BE P NP
I~
WITH NP MP
~N
DET
I ABSTR
ART
I
PAR
I
A
(MAIN) ...
contain a WITH string in which the category PAR appears. Instead of PAR,
it must contain the category VAL.
But the structures underlying other simple MP sentences, at some stage
preceding XXXII, probably will also have to include the category VAL,
since they can be paraphrased with sentences containing the verb bedragen.
Compare, for example, the following a sentences and their paraphrases
(the b sentences):
(59a) De stok is I meter lang.
'The stick is I meter long.'
(59b) De stok heeft een lengte die I meter bedraagt.
'The stick has a length that amounts to I meter.'
(60a) Het boek kost 12 gulden.
'The book costs 12 guilders.'
(60b) Het boek heeft een prijs die 12 gulden bedraagt.
'The book has a price that amounts to 12 guilders.'
It seems plausible to assume that, at some stage preceding XXXII, we must
have a structure containing VAL somewhere under the second highest NP
node in XXXI/. Therefore, we should now ask how one may arrive at a
structure of the form XXXIII, which is the lower part of the structure un-
derlying simple MP sentences not containing bedragen (cf. XXXII, above;
I have omitted irrelevant details in XXXIII):
THE DERIVA nON OF MP SENTENCES 145
XXXIII
~ NP
NP
~MP
~
DET PAR
I
ART
XXXIV
~NP
~REL
NP1
A
DET PAR S
I
I~
ART NP VP 1
A PAR v~PP
DET
ART
I BE
I~
P NP
I~REL
WITH NP2
A
DET VAL
I
S
I
ART
~VP
NP2
DET
AA VAL V PP
I
ART
IA
BE P MP
I
AT
(See, for the category REL ('RELATIVE CLAUSE'), pp. l7Iff., below.)
If we interpret 'value' as something definable as a point on a calibrated
scale, the preposition AT in XXXIV can be taken to represent the relation
146 THE STRUCTURE UNDERLYING MP SENTENCES
between the value and a certain point corresponding to the number of units
expressed in the MP.
By RELATIVE CLAUSE FORMATION, which we may state as XXXV,
a structure like XXXIV will be changed into XXXIVa:
A
DET PAR
I
S
I~
ARTNP, VP
DET
A~
PAR V PP
ART
I BE
I~
P NP
I~REL
WITH NP 2
A
DET VAL
I
S
I
ART
~VP
NP2
I~
PRO
V PP
I~
BE P MP
I
AT
tion as ART N (recall that we have assumed that if DET is not rewritten by
the base rules as ART, the noun phrase is considered to be 'non-particular').
RELATIVE CLAUSE REDUCTION (which I shall state below) applies
after RELATIVE CLAUSE FORMATION in the cycle. Hence, the lowest-
down REL may undergo reduction now that the former rule has applied.
When the cycle has been run through once, RELATIVE CLAUSE FOR-
MATION will once again apply, followed again by RELATIVE CLAUSE
REDUCTION. Thus, XXXIVa may now be changed by repeated applica-
tion of both rules into XXXVII.
XXXVII
~NP
~REL
NP 1
DET
A PAR
ART
I
PP
~NP
P
I~
WITH NP REL 2
DET
A VAL
ART
I
PP
P
~MP
I
AT
148 THE STRUCTURE UNDERLYING MP SENTENCES
(In the following it will be assumed that there are transformations which
operate cyclically. However, running through a cycle an nth time does not
automatically mean that a transformation in that case will operate on the
nth lowest S of the (restructured) tree; it will be assumed that any time a
transformation is applied, it must operate on the lowest S of the derived
tree (except of course ifthe S.D. contains an embedded S). For our purposes
this matter does not seem to constitute an issue of major importance. Al-
ternatives to this way of interpreting the cyclical mechanism will merely
require reformulations of certain rules without affecting the essence of the
proposals below.)
The encircled nodes in XXXVII must be pruned. I am following here
Gruber's criterion for tree-pruning, according to which a node is pruned as
soon as it has become irrelevant. A node is irrelevant if it no longer dominates
its left-branching 'head'. (Cf. p. 101. See also Ross (I966b), where a some-
what different criterion is proposed.)
I shall assume, furthermore, that REL is pruned if it comes to dominate
only PP. In general, there do not seem to be transformational or lexical
rules operating upon PP, the application of which would be contingent
upon the presence of a REL node immediately dominating it. Hence REL
may be assumed to be always irrelevant as soon as it comes to dominate PP
directly. XXXVII then, after tree-pruning will result in XXX VIla.
XXX VIla
~NP
~PP
NP
~~
DET PAR P NP
I
ART
I~
WITH NP PP
~~
DET VAL P MP
I
ART
I
AT
~T ;:
2 3
S.c. : o 3
The substring WITH DET (PAR MAIN Y PAR] in XXXIX will of course have
to be extended in the event that we reject PAR-MP CONTRACTION. In
that case we should have, instead, WITH DET [PAR MAIN Y PAR] WITH
DET VAL AT.
To recapitulate, we have now stated four prelexical transformations,
which may apply in the order in which they are listed below:
~N1
DET DET
~N1
~REL
ART ART
~REL
I I
S S
X
~
NP Y 2
I
NP 2
2 3 4 ;: 1 3
25 Gruber (ibid.) notes that "unrecoverable deletion may be avoidable if we simply allow the
generation of a noun phrase under the REL as well as a full sentence, and then allow this to have
a complete ambiguous interpretation." However, as we shall see, an unrecoverable deletion of
the type that Gruber's rather unsatisfactory proposal may require can also be avoided in other
THE DERIVA nON OF MP SENTENCES 151
~N,
DET
~REL
ART
I
NP2
XLII
# of #
The attachment of of, then, is like the attachment of case markings (... )."
Gruber's deletion rule will have to be restated in a more precise way, so
as to prevent the derivation of such sentences as *Dit is een boot van 1
vadem (This is a boat of I fathom'), *Het leger gebruikte geschut van 10
kilometer (The army used artillery of 10 kilometers'), which would be the
result of deletion of the subject NP and the categories underlying steken
('draw') and dragen ('carry').
For our purpose we shall need some other 'REL stripping' rule, deleting
PRO BE WITH in restrictive relative clauses, stated in such a way as to
restrict its applicability to embedded sentences in which the subject is co-
referent with the antecedent in the matrix sentence. For instance, it should in
some way relate a structure such as XLIII (below) to XLIV:
ways. although these do not account for the putative fact that the expressions John's book,
a book of John's, the book of John, etc., do not only have the reading "the/a book which John
has." They could. according to Gruber, have readings like "the/a book which John read",
"the/a book which John saw," or "wrote," or "burned," or "threw a pebble at." It seems to me
that if we want to avoid a mind-boggling set of recoverable rules relating these expressions to
John's book, etc .. we could interpret Gruber's list of possible readings as part of the class of
things that maybe referred to by the expression the/a book which John is (in some way)
associated I\·ith. which may be taken to correspond to the structure underlying John's book,
etc., thus granting that the latter is "ambiguous", albeit not homonymous.
152 THE STRUCTURE UNDERLYING MP SENTENCES
XLIII
~NP
~RELr
NP,
A
DET PAR
I
S
A
NP, VP
IA
PRO V PP
I~NP
BE P
,~
WITH NP2 RELr
/'\ ,
DET VAL S
~VP
NP2
I~
PRO
V PP
I~
BE P MP
I
AT
XLIV
~NP
~RELr
A ,
NP,
DET PAR MP
(e.g., 'a length lREL, MP RELJ', which may be lexicalized as 'a length ofMP'.)
In order to do this by means of a rule deleting only PRO BE WITH in XLIII,
we shall have to get rid of the string DET VAL PRO BE AT. Part of the latter
string can be deleted by RELATIVE CLAUSE REDUCTION, which will
turn the subtree XLV of XLIII into XLVI (in the latter, REL, S, and VP are
pruned).
We cannot delete DET VAL AT in XL VI by PAR-MPCONTRACTION,
THE DERIVATION OF MP SENTENCES 153
for the highest NP in XLVI, as it occurs in structures of the kind under dis-
cussion, either is not preceded by WITH (contrary to what the S.D. of this
rule requires), or, if it is, WITH will be deleted by our REL stripping rule,
which should delete PRO BE WITH. The reason that the REL stripping
transformation must delete WITH together with PRO BE is, as we shall see,
that by the same rule we must derive constructions such as van twee meter
[engte ('of two meters length'). In the latter type, too, WITH preceding PAR
has been deleted. Obviously, this cannot be the result of PAR-MP CON-
TRACTION.
XLVI
XLV ~
NP
~NP
~RELr
NP2
~PPNP2
~VAL SI
DET ;
~~
DET VAL P MP
~VP
NP2
I
AT
I~
PRO V PP
BE
I~
P MP
I
AT
S.c. : o 3
154 THE STRUCTURE UNDERLYING MP SENTENCES
r
NP,
~PAR
DET
NP
~RELr
NP2
~VAL
DET
I
S
~VP
NP2
I
PRO
~PP
v
I~
BE
P MP
I
AT
(The encircled node is pruned.)
RELATIVE CLAUSE REDUCTION can change XLVIII into XLIX:
XLIX
~NP
~RELr
NP,
~PAR
DET
I
NP
NP2
~VAL
DET
PP
P
~MP
I
AT
THE DERIVATION OF MP SENTENCES 155
S.C.: o 3
Application of L to XLIX will result in LI:
LI ~
NP
~RELr NP
~PAR
DET
MP
In LI, PP and NP will be pruned, in that order (we may assume that,
unless there are other requirements which decide which node is pruned
first, the pruning takes place, in principle, from bottom to top.)
We could now make use of a Gruber-type entry for the insertion of van
(of) in LI. It could be objected that the word van in such expressions as
Het boek van Jan (,The book of John'), for which Gruber's rule was intended,
may have an entry different from the one for van in van + MP constructions.
But, although the various types of van constructions probably will have
different sources, there is no need to assume that the word van in the latter
case should have an entry different from that for the word van in Het boek
van Jan, until evidence to the contrary should become available.
The entry for van, then, may be given as follows:
LII NP
N0~l
'*' '*' van
NP
As will be pointed out later, it may be necessary to have more entries like
the one above for special cases in which in Dutch the genitive morpheme
is zero.
We are now able, in principle, to derive expressions such as een [engte
van twee meter ('a length of two meters') from underlying structures having
156 THE STRUCTURE UNDERLYING MP SENTENCES
the form of XLIII ([NP DET PAR NP] [RELr PRO BE WITH DET VAL
lRELr PRO BE AT MP REd REd).26
I shall now discuss the way in which to derive van + MP constructions in
which the underlying antecedent does not contain a parameter noun (as it
does in een lengte van twee meter), and in which no overt parameter noun is
present, for instance, expressions like een steen van 2 kilo Ca stone of2 kilos').
LIII "'"
NP
~RELr
NP
I
s
~VP
NP
I~PP
PRO V
I~NP
BE P
I~RELr
WITH NP
A
DET PAR
I
S
1\ A
(MAIN) X NP VP
I~PP
PRO V
I~NP
BE P
I~RELr
WITH NP
DET
A VAL
I
S
NP
~VP
I~
PROV PP
I~
BEP MP
26 See p. 157.
I
AT
THE DERIVATION OF MP SENTENCES 157
LIV
~NP
~RELr
NP
I
S
~VP
NP
I~PP
PRO V
I~NP
BE P
I~RELr
WITH NP
A
DET PAR
I
NP
I \AA
(MAIN) Rlr
DET VAL S
~VP
NP
I~
PRO V PP
I~
,
BE P MP
AT
26 Cf. Wunderlich (1970). where constructions of this type having the form van+MP (or
rather. von + M p. since he uses German examples) are also transformationally related to
bedragen (German: betragen). Underlying bedragen is a string of the form BE WITH DET
VAL AT. which isa string that will be yielded by RELATIVE CLAUSE REDUCTION applied
twice to XLIII.
158 THE STRUCTURE UNDERLYING MP SENTENCES
LV
~ NP
~RELr
NP
I
5
~VP
NP
I~PP
PRO V
I~NP
BE P
I~
WITH NP RELr
A
DET PAR
I
NP
1\
(MAIN) X NP RELr
DET
A VAL S
PP
~MP
P
AT
I
We shall now restate EQUI-SUBJECT REL STRIPPING so that it may
also apply to structures such as L V, above.
2 OPT
3
S.c. : o 3
THE DERIVATION OF MP SENTENCES 159
Note that LVIa specifies that the PAR category present in the antecedent
noun phrase must be identical to the one occurring in the third term of
the structural description. This is in accordance with condition (vii) of C,
on p. 122 (no van + MP construction can be derived if the antecedent of the
underlying relative clause does not incorporate a parameter category identi-
cal to the one present in the WITH string; compare, for example, *een
bureau van 10 jaar (ouderdom/leeftijd) ('a desk of 10 years (age)'). See also
p. 119, in connection with nouns not incorporating the category TIME).
The variable U stands for DET, or in the case of sentences with 'double
WITH strings' (see Chapter IV), such as duration sentences (cf. p. 137,
second paragraph), for DET PAR WITH DET.
Application of EQUI-SUBJECT REL STRIPPING, as stated now, to
LV, will result in L VIla:
LVIIa
----------
~
NP
NP RELr
I
NP
~RELr
NP
~PAR
DET
I
NP
~~
(MAIN) X NP PP
~~
DETVAL P MP
I
AT
LVlIb
~NP
~RELr
NP
I
NP
~RELr
NP
~
DET PAR
I
MP
~X
(MAIN)
S.c. : o 3
LIX
~NP
~RELr
NP
I
MP
S.D. : X NP (
REL
r
I {
LX SPACE DELETION (prelexical)
- DET ( (
Y
TWOSPACE }
MAIN THREESPACE
SPACE MAX
AXIS
1 1
MAX SPACE
]
- MP
OPT
2 3 ;
S.c. : o 3
2 3 4 5~
S.C: 4 3 o 5
162 THE STRUCTURE UNDERLYING MP SENTENCES
Thus, if we have a structure like LXII (below), it will either undergo SPACE
DELETION, or (if we do not apply that rule) it must undergo MP
SHIff. Therefore, we will get either LXlIIa or LXIlIb:
LXII ~
NP
---------------
NP RELr
NP
I
~ DET
---------------I
NP
PAR
RELr
MP
I
ART MAIN
~TWOSPACE
LXlIIa LXIlIb ~
~NP NP
NP
---------------
RELr NP
I~
MPDET PAR
(SPACE DELETION)
I
MP
~
MAIN TWOSPACE
(MP SHIFT)
LXII/a, of the form [NP [REL MP REd NP] (where, as will be recalled, 'MP'
r
is an NP), it is the higher NP node which will be pruned, not the REL node.
The reasoning is analogous to what has been said with respect to the
configuration ~EL PP REd on p. 147, but here it is the REL node which is
irrelevant with regard to transformations and the lexicon.
Furthermore, if a given category A immediately dominates only some
other category B, and if A and B are categories of the same kind (for in-
stance, both are REL), then A will be pruned. Hence, the higher REL node
in LXII/a, as soon as it comes to dominate immediately the lower REL
node, will be pruned.
Any node which is generated as a preterminal node only, will be pruned
if it no longer dominates anything. The same goes for preterminal nodes
that are newly introduced by the transformational component. Therefore,
the encircled REL node in LXI/lb will also be pruned. The encircled NP
node in the latter structure will be pruned as well, since it now immediately
dominates only a category of the same kind.
The rule of MP SHIFT should also account for sentences like:
(61) De ball on zweefde op 300 meter hoogte.
'The balloon floated at 300 meters altitude.'
(62) De duikboot bleef op 300 meter diepte.
'The submarine remained at 300 meters depth.'
(63) Op twee meter afstand kon je hem niet meer horen.
'At two meters distance you couldn't hear him any more.'
One way to account for sentences such as these is to assume that their
underlying structure before MP SHIFT contains a PP roughly like LXIV:
LXIV
~PP
P
~NP
I~
ATNP REL
~N
DET
I
NP
r
I
ART
I
POINT
~RELNP
~PAR MPI
DET
r
ART
I I
ONESPACE
164 THE STRUCTURE UNDERLYING MP SENTENCES
I~
ATNP RELr
~N
DET
I
NP
I
ART
I
POINT DET
~PAR
MP
I I
ONESPACE
o
AT
I
REL
NP
r
~PAR
DET
I
MP
I
ONESPACE
italicized labels; see, for the elimination ofDET preceding POINT, p. 206f.).
Since we will then have a structure of the general form [pp PP' [REL NP REd pp],
the entry for van (which is in a later cycle because RELr is lower than PP)
no longer fits, which is as it should be. It would seem that the latter solution
is less costly than setting up a transformation deleting DET POINT, so
incorporation instead of deletion in this case may be preferable.
Notice that MP SHIFT, as stated above, need not be revised in order to
account for sentences such as (61)-(63).
To sum up, we now have the following order of rules:
RELATIVE CLAUSE FORMATION (obI.; p. 146)
RELATIVE CLAUSE REDUCTION (opt.; p. 147)
PAR-MP CONTRACTION (obI.; p. 149)
SIMPLE MP SENTENCE REDUCTION (opt.; p. 150)
EQUI-SUBJECT REL STRIPPING (opt.; p. 158)
SPACE DELETION (opt.; p. 161)
MP SHIFT (obI.; p. 161)
MP-TAIL SHORTENING (obI.; p. 160)
All of these rules apply within the same cycle. Some of them may be
assigned another place relative to the rest of the rules listed above. Thus,
it may be that the four transformations appearing at the bottom should
be placed between RELATIVE CLAUSE FORMATION and PAR-MP
CONTRACTION. However, the order of the first four rules is probably
correct. It is necessary, furthermore, that MP SHIFT precede MP-TAIL
SHORTENING, as it will otherwise be impossible to prevent deletion of the
parameter noun phrase in all cases where we want to derive van + MP +
parameter noun constructions. SPACE DELETION should precede MP
SHIFT, lest the parameter noun be ultimately deleted by the obligatory
rule of MP-TAIL SHORTENING in cases where optional SPACE
DELETION has not applied. The structural descriptions of the last three
rules, finally, are such that they cannot apply unless EQUI-SUBJECT
REL STRIPPING has been applied, and it, in turn, cannot apply if
RELATIVE CLAUSE REDUCTION has not been applied. Alternatives
to idiosyncratic rules are mentioned in Chapter V.
LXVII NP
~
NP RELr
~RELr NP
I
MP
~
DET QUANTITY
I
NP
I
ART
~N
DET
I
ART
I
MASS
I shall not try to answer the question as to the source of structures like
[NP DET QUANTITY [REL DET MASS REd Npl, since the possibilities
r
LXIX NP
~REL
REL
I I r r
NP MP
~N
DET
I
ART
I
MASS
LXX NP
M~
I NP
~
DET N
I
ART
I
MASS
That is, in the case of applied amount terms, MP probably should not replace
ART preceding MASS, if the English equivalents of Dutch applied amount
terms are any indication of the kind of underlying structure we should
assign to the latter. In an expression like two liters of water, water in all
probability should be seen as a complete noun phrase, not just a noun. A
structure such as LXX accounts for the presence of of in applied amount
terms in English, if we accept a lexical entry of the type of LII (p. 155) for
of Such an entry would require that water in two liters of water be a noun
phrase, not a noun. The fact that applied amount terms in Dutch do not con-
tain van can be accounted for by introducing an extra lexical entry applicable
in cases such as these. This entry is basically of the form of LII, but, for the
class of expressions under consideration the sister NP of the REL which
constitutes the simultaneous environments should be specified further as,
possibly, DET QUANTITY, or some such string, at any rate as something
which can be considered to represent the class of noun phrases which are
not followed by van when preceding, for instance, mass terms. The REL-
dominated NP will probably also have to have some further specification,
e.g., DET MASS. The REL node should then be associated with a zero
morpheme. This type of entry would thus account for the absence of a
genitive marker in, for instance, twee liter water. In some cases we find
the case marker -s representing the REL node in Dutch (2 bladzijden druks
168 THE STRUCTURE UNDERLYING MP SENTENCES
(two pages of print», but generally it is absent. For an expression like the
one just given a separate entry with a complete specification of the NP
dominating druk ('print') will be required, where, furthermore, the REL
node should be associated with the morpheme -s, specifying that it comes
after the NP druk, not before it.
The situation is exactly the same, as far as lexical entries are concerned,
with expressions such as the ones considered in Section 2.1.3.2 of Chapter I,
i.e., expressions like enkele meters (a number of meters), ettelijke meters
(afair number ofmeters), etc. (cf. (89a)-(93b) on pp. 23-4). That is, in the case
of these expressions we shall also need a lexical entry for a zero case
marker. This entry may be actually the same as the one suggested for the
zero case marker in Dutch applied amount terms.
Presumably, applied amount terms are essentially similar in structure
to expressions such as these. Therefore, we could replace I on p. 24 by a
structure like LXXI. (I was proposed for such expressions in Chapter I.)
NP
---------------I
LXXI
NP RELr
~N
DET NP
I
ART
~~
X
NUMBER DET N
I
ART MASS
~UNIT
I
PLURAL
the difference between, e.g., on the one hand a height, a length, etc., and on
the other hand height, length, etc. in MP + parameter noun constructions.
The most plausible solution would seem to be to regard spatial parameter
words as non-count nouns. This means that we should look upon an ex-
pression like een hoogte van 10 meter ('a height of 10 meters') as containing
a non-count noun hoogte ('height'), despite the fact that it is preceded by an
overt indefinite article.
As is well known, there are many constructions in which nouns appear
preceded by an overt indefinite article although normally they are not
preceded by one, for instance, the non-count noun enthousiasme in Hij
toonde een enthousiasme dat mij verbaasde ('He showed an enthusiasm that
surprised me'). Always when we are dealing with sentences such as these,
the presence of the indefinite article can be explained by the fact that it is
not really directly associated with the following non-count noun, but with
some other underlying constituent, such as, in this case, 'degree', or, in
other cases, 'great amount' or 'remarkable kind' (as in Wat een whisky!
(,What a whisky!', meaning What a great amount of whisky! or What a
remarkable kind of whisky!).
In the case of a height, a length, etc., we are probably dealing with
something similar. It may be that expressions like the latter two have an
underlying structure in which some element representable as 'amount of'
or 'measure of', or the like, is present and explains the overt indefinite article
in such expressions. Although at present I see no way of substantiating such
a claim, I shall assume that it will be possible to account for the occurrence
of the indefinite article in van + MP constructions in which van is preceded
by a parameter noun, by some rule stipulating that ART is manifested as
een before non-count nouns only if they are accompanied by some under-
lying attributive phrase referring to the quantity or kind of the non-count
noun (cf. Wat een (hoeveelheid/bijzondere) whisky! (,What a (quantity
of/remarkable) whisky!'), een enthousiasme dat mij verbaasde ('an en-
thusiasm that surprised me'), een hoogte van 2 meter ('a height of 2 meters'),
etc.). Suppose this is correct. We could then assume that it is also correct
that MP SHIFT should replace the category ART by MP in order to derive
MP + parameter noun constructions, thus relating, e.g., een hoogte van
10 meter ('a height of 10 meters') to 10 meter hoogte ('ten meters height').
It seems rather difficult to decide whether or not this course should be fol-
lowed. If MP + parameter noun constructions in Dutch are to be handled
in this way, there must be a difference in the transformational treatment of
such phrases between, for instance, French, and Dutch. For, a phrase like
twee meter breedte ('two meters width') translates in French as deux metres
de large ('two meters of width'), in which large should not be considered an
170 THE STRUCTURE UNDERLYING MP SENTENCES
adjective but a noun phrase (compare expressions like etre au large, etre
au large de Brest, du large, se promener de long en large, prendre Ie large, Ie
vent du large, etc., in which it is not an adjective either). The word de ('of')
in deux metres de large suggests that there must be a REL node dominating
an NP large, rather than that large represents just an underlying N.
Another consequence of treating Dutch MP + parameter noun construc-
tions in the way considered above (i.e., by replacing ART by MP) will be
that we will then be able to account for such phrases in Dutch which
differ from that of applied amount terms in that the disappearance of
the indefinite article is seen as the result of replacing ART by MP in the
former case, and as the result of deleting DET QUANTITY in the latter.
I am not sure as to whether this is justifiable or not. In the formulation of
MP SHIFT below, however, it is assumed that it is.
LXXIIb MP SHIFT (pre lexical)
{ {
1 MAIN THREESPACE }}l [11
l t J[
[ [ ART
r
TWOSPACE
S.D. : X REl r DET _ _ DET ONESPACE [ _ MP _ 1 REl r Z
[ ¢ [ DET MASS 1 RELr REl r 1
NP RELr REl r NP
2 3 4 5 oel ;
S.c. : 4 3 o 5
(where the large square brackets indicate that if the upper half of possible
categories inside a pair of them is chosen, it should be chosen for all pairs
of them. Likewise for the lower half. The second term of the S.D. is empty
if we choose the lower half of possible strings. This is indicated by '0'.)
LXXIIb will convert LXIX into LXX, LXII into LXIIIb, and LXIV into
LXV (see above).
We can now add QUANTITY DELETION to our list of transformations.
It should apply before MP SHIFT, but after EQUI-SUBJECT REL
STRIPPING. It does not matter whether it comes before or after SPACE
DELETION.
RELATIVE CLAUSE FORMATION (obI.; p. 146)
RELATIVE CLAUSE REDUCTION (opt.; p. 147)
PAR-MP CONTRACTION (obI.; p. 149)
SIMPLE MP SENTENCE REDUCTION (opt.; p. 150)
EQUI-SUBJECT REL STRIPPING (opt.; p. 158)
SPACE DELETION (opt.; p. 161)
QUANTITY DELETION (opt.; p. 166)
MP SHIFT (obI.; p. 170; see also p. 239)
MP-TAIL SHORTENING (obI.; p. 160)
THE DERIVATION OF MP SENTENCES 171
:]
3+[5 6 7 1
s.c. : REL REL 8
{: 6+[2 3 4 1
REL REL
(where the NP's with the subscript 'i' are identical; '+' means Chomsky
adjunction).
A rule of this kind would make the third rewriting possibility of the base
rule (ii) on p. 133, NP-NP S, unnecessary. It would relate, for instance,
both LXXV and LXXVI (on pp. 172 and 173, respectively) to LXXIV.
LXXIV S
S s ----s
~VP
NP,
~VP
NP2
~VP
NP,
~PP
V
~~
DET N V PP
~~
DET N V PP
I~
BE P .NP2
II~
PAR BE P NP,
II~
VAL BE P MP
I~
WITH
DET N
I
WITH DET
/''"--.N I
AT
I
PAR
I
VAL
LXXV s
A
NP1 VP
v
A PP
I~NP
BE P
I~REL
WITH NP2
DET
/ \N I
S
IA
PAR NP2 VP
IA
PRO V PP
I~NP
BE P
I~REL
WITH NP 3
DET
/ \N I
S
I~VP
VAL NP3
I~PP
PRO V
I~
BE
P MP
I
AT
indicated by the two analyses representable as [NP roET ART REL DET] N NP]
and [NP NP REL NP], respectively. This could mean that we should have two
different EMBEDDING transformations. However, since I do not know
whether or not such a solution is correct, I shall continue to indicate the
difference between the two kinds of relative clauses by means of the
subscripts Or' ('restrictive') and On' (,nonrestrictive'), and assume that both
types occur in the structure [NP NP REL NP]' No doubt the semantic
difference between restrictive and nonrestrictive relative clauses should
somehow be made explicit in the S.D. of EMBEDDING rules, if indeed
it is correct to assume that both types derive from sentence conjunction.
Clearly, an advantage of having some kind of EMBEDDING transforma-
tion is that we can now account for the synonymy between, e.g., De steen
THE DERIVATION OF MP SENTENCES 173
---------------
LXXVI S
NP VP
NP2
~REL V
~PP
/ \N
DET S
I I ~NP
BE P
IA
PAR NP1 VP
I~REL
WITH NP 3
v
~PP A
DET N
I
S
I~
BE
P NP 2 VAL
I~
NP VP 3
I
WITH
I
PRO
I
PRO V
~PP
I~
BE
P MP
I
AT
weegt 10 kilo (The stone weighs 10 kilos') and Het gewicht van de steen
is 10 kilo (,The weight of the stone is 10 kilos') in a very simple way;
EMBEDDING implies that both derive from the same underlying structure.
(I have been told that, of the two types, only the latter is possible in
Japanese.) Another advantage may be that we will never get embedded
clauses with wrong antecedents. Seuren (l969b) calls attention to the
fact that a base grammar like the one in Chomsky (1965) may generate a
string like the dog (the dog Past be angry) Past frighten the man, but also,
for example, the dog (the man Past be angry) Past frighten the man, "which
cannot be further developed into a surface structure since any trans-
formation incorporating the bracketed clause in [the latter string] requires
that the remaining and the deleted element [i.e., in this case, the antecedent
noun phrase and the subject of the embedded clause, WGK] should be identi-
cal - a condition arising from the general requirement of recoverability of
deleted elements." (op. cit., p. 51). There is no way to prevent Chomsky's
base from generating strings like the latter. However, with an EM-
BEDDING-type rule we will never get such a string.
We cannot derive van constructions like het gewicht van de steen by
means of the EQUI-SUBJECT REL STRIPPING rule given earlier. There
must be a rule that eliminates BE WITH, as well as the object NP coref-
erent with the antecedent. Thus constructions like het gewicht van de steen
and, e.g. het boek van Jan ('the book of John'), de poten van de tafel ('the
174 THE STRUCTURE UNDERLYING MP SENTENCES
legs of the table'), could be derived via this rule from structures we may
loosely represent as the book - John be with the book and the legs - the table
be with the legs, respectively.
In order to derive such expressions we shall need a rule that could be
formulated as follows:
LXXVII EQUI-OBJECT REL STRIPPING (prelexical)
~VP
NP
~pp
v
I~
BE P NP,
I
WITH
I
PRO
LXXIX
~NP
~REL
I
N~
r
®
I
NP
S.c.: 3 5 2 4 6
(where 'PRT' stands for 'PARTICLE'. The category PP corresponds to the
PP' in XVIII on p. 129.)
Notice that if the input starts with X NP # - #PRT - # V #, it will
always continue with a non-null fourth term, i.e., a boundary symbol. If
the first three terms are NP # # - # PP # - BE #, the fourth term will
always be empty. (This is simply an automatic consequence of the two
different ways in which affixation can take place.) Therefore it will suffice
to put the fourth term in parentheses. As indicated above, the complement
NP (the fifth term) mayor may not be empty.
BOUNDARY UNTANGLING will convert a string like (63) into (64),
and a string like (65) into (66):
(63) #Jan# - #op - #eet# - # - # de # # sandwich #
'John up eats the sandwich'
(64) #Jan# - #eet# - #de# # sandwich # - #op - #
'John eats the sandwich up
(65) #het# # - #lang# - is# - #twee# # meter #
'it long is two meter'
(66) #het# # - is# - #twee# # meter # - #lang#
'it IS two meter long'
BOUNDARY UNTANGLING must not be allowed to operate if the
sentence is embedded. For, in that case, opeet and lang is should remain
single 'words'. Therefore we must state the following condition:
LXXXIa Condition on BOUNDARY UNTANGLING
Transformation blocks if the substring starting with the NP of
the first term and ending in the fifth term is dominated by an
embedded S.
The rule of BOUNDARY UNTANGLING of course will have to be stated
THE DERIVATION OF MP SENTENCES 177
differently if it should tum out that zijn must not be considered an affix.
We have discussed a total of thirteen transformational rules in this
subsection and the preceding ones. They are listed under LXXXII:
LXXXII List of transformational rules
Prelexical
1. EMBEDDING (opt.; p. 171)
2. RELATIVE CLAUSE FORMATION (obI.; p. 146)
3. RELATIVE CLAUSE REDUCTION (opt.; p. 147)
4. PAR-MP CONTRACTION (obI.; p. 149)
5. SIMPLE MP SENTENCE REDUCTION (opt.; p. 150)
6. EQUI-SUBJECT REL STRIPPING (opt.; p. 158)
7. EQUI-OBJECT REL STRIPPING (opt.; p. 174)
8. SPACE DELETION (opt.; p. 161)
9. QUANTITY DELETION (opt.; p. 166)
10. MP SHIFT (obI.; p. 170)
11. MP-TAIL SHORTENING (obI.; p. 166)
Postlexical
12. BOUNDARY UNTANGLING (obI.; p. 176)
13. REL REORDERING (obI.; p. 175)
The EMBEDDING rule cannot be part of the same cycle as the rules 2-11,
for if it were, we could derive dubious sentences such as *De stok heeft een
lengte die twee meter is ('The stick has a length which is two meters'), in
which SIMPLE MP SENTENCE REDUCTION has taken place in a
sentence which ultimately will become a relative clause (cf. p. 125, examples
(54) and (55)). Such a sentence could be derived in the following way:
LXXXIII(=LXXIV) s
~---------------
~
s
~
s
~
s
~PP
v
~~
DET N V PP
~~
DET N V PP
I~
BE P NP,
II~
PAR BE P NP,
II~
VAL BE P MP
I~
WITH
DET N
I~
WITH
DET N
I
AT
I
PAR
I
VAL
LXXX/lla s
~s
~VP ~VP
NP1
~PP
v DET
An NP2
N V PP
I~
BE P NP 2
II~NP
PAR BE P
I A
WITH DET N
I~REL
WITH NP 3
PAR
I DET
A N
I
S
I~VP
VAL NP3
I~
PRO
V PP
BE
I~
P MP
I
AT
RELATIVE CLAUSE REDUCTION=>
LXXX/lIb s
s S
~VP /,,-,----
NP1 NP2 VP
~PP
V
~~
DET
N V PP
I~
BE
P NP
II~
PAR BE
P NP
2
I~N WITH
I~ NP PP
WITH DET 3
PAR
I ~~
DET N P MP
I I
VAL AT
LXXXIIIc s
s s
~VP ~VP
NP2
NP,
~PP
v
~N V~MP
DET
I~
BE
P NP
I I
PAR BE
2
I~N
WITH DET
I
PAR
LXXXII~
NP, VP
~PP
v
BE
I~
P NP
I~REL
WITH NP2
~N
DET
I
S
I~
PAR NP VP 2
I~
PRO
V MP
I
BE
LXXX/IId would lead to lexicalized structures like the one cited above
(* De stok heeft een lengte die twee meter is). It cannot undergo any further
transformations (aside from postIexicaI ones). Therefore, EMBEDDING
must be outside the cycle of the rules 2-11. That is, if it has been applied
(once or several times), it may not apply again after the cycle of 2-11 has
been run through once or several times.
180 THE STRUCTURE UNDERLYING MP SENTENCES
According to the base rules (ii), (vii) and (xiii) in Section 5, the structure
underlying abstract measure phrases should contain at least the following
type of configuration:
LXXXIV NP
~N
DET
I
ABSTR
I
UNIT
I
SPACE
TIME
TEMPERATURE
INDEF + NE + N + PLURAL
(where 'NE' stands for 'Numerical Element'. NE was assumed to dominate
integers or words like paar as in een paar meter.) In Section 5 of the present
chapter, it was assumed that if DET in a measure phrase is not rewritten,
it refers to a 'non-particular' set of units, that is, to a set of units which are
undetermined as to their location on a measuring scale. Insofar as this
assumption is valid, we could replace LXXXV by 'DET+NE+N+
PLURAL'. But it is also possible that NE is dominated by DET. In that
case LXXXV should be restated as something like [DET X NE Y DET] + N +
PLURAL. However, as we shall see in this subsection, there are more
possibilities, and we shall be able to restate LXXXV in a different way.
Jackendoff (1968) proposes the following structure underlying a phrase
like three men:
THE DERIVATION OF MP SENTENCES 181
LXXXV/a NP
~
Det N PP
I
Art
I~
three of NP
~N
Det
I
Art
I
men
I
Indef
LXXXVlb s
NP
---------------
~
NP
5
VP
I
(were) many
I~
NP
the men VP
the men
I I
left
significant with respect to the structure underlying the phrase ten dollars:
(71) a number of dollars amounting to ten
We could then consider the underlying structure LXXXVII for ten dollars,
since the part amounting to ten represents an underlying relative clause.
Note that if we interpret VAL as something definable as a point on a
calibrated scale, the preposition AT, as it occurs in structures such as
LXXXVII, can be taken to represent the relation between the value and a
LXXXVII NP
~RELr
NP,
~
NP
RELr
I
S
~N
DET
I
NP
~VP
NP,
I I
ART QUANTITY DET
~N I V~PP
PRO
I
NUMBER
I~
ART MASS UNIT
I~
BE P NP
I
PLURAL MONVAL
I I~
WITH NP RELr
I
DOLLAR DET
~N I
S
ART
II~ VAL NP VP
I~PP
PRO V
BE
I~
P NP
I~
AT DET N
I
ART
I
NE
I
DEF
1
INTEGER
I
10
THE DERIVATION OF MP SENTENCES 183
LXXXVIIa NP
NP, RELr
~RELr
NP
I
NP
A
DET N
I
NP NP
~RELr
I I
ART QUANTITY DET
~N /\
DET
I N S
I
NUMBER
I~
ART MASS UNIT
IIA
ART VAL NP VP
I I
PLURAL MONVAL PRO V
I~PP
I
DOLLAR
I~NP
BE P
I~
AT DET N
I
ART
I
NE
I
DEF
I
INTEGER
I
10
184 THE STRUCTURE UNDERLYING MP SENTENCES
LXXXVIIb NP
NP1
~RELr
NP
DET
A N
I
NP NP
I I
ART QUANTITY DET
~N /\
DET N
I
NUMBER
I~
ART MASS
UNIT
I I
ART VAL
I
PLURAL MONVAL
I PP
I
DOLLAR P
~NP
I~N
AT DET
I
ART NE
I
I
DEF
I
INTEGER
I
10
THE DERIVATION OF MP SENTENCES 185
'MP-TAIL SHORTENING'=>
LXXXVIIc
NP,
~RELr
NP
~N
DET
I
NP
I
ART QUANTITY
i ~N DET NP
II~
NUMBER ART MASS UNIT DET
~N
I
PLURAL
I
MONVAL
I
ART NE
I
I
I I I
DOLLAR DEF INTEGER
I
10
QUANTITY DELETION=>
LXXXVIId NP
RELr
I
NP
RELr
I
NP
~N
DET
~N I
ART
I
NE
DET
I~UNIT
ART MASS
I
DEF
I
INTEGER
I
PLURAL
I
MONVAL
I
10
I
DOLLAR
186 THE STRUCTURE UNDERLYING MP SENTENCES
'MP SHIFT'=>
LXXXVIIe NP
NP RELr
~N
DET
I NP
I
ART
I
NE
~N
DET
I
DEF
I
INTEGER
I
ART MASS
~UNIT
I
10 PLURAL
I I
MONVAL
I
DOLLAR
LXXXV/If NP
RELr
I
NP
DET NE N
I
ART
I
INTEGER MASS
~UNIT
I
10
I
PLURAL
I
MONVAL
I
DOLLAR
as well, for the same reason. The remaining encircled NP node will then be
pruned too, since after pruning of the previously mentioned NP, it will
dominate only REL r •
Alternatively, we could have the rule of NE INCORPORATING change
LXXX VIle into LXXXVIlg:
LXXXVIlg RELr
I
NP
~N
DET
I~
ART NE N
I~
INTEGER MASS UNIT
I
10
I
PLURAL
I
MONVAL
I
DOLLAR
LXXXVIII NP
NP RELr
~N
DET
I
NP
I
ART
I
QUANTITY
~N
DET
I
NUMBER
I
ART
~UNIT
MASS
I
PLURAL
(67)--(70), and 3, the similarity between the Lakoff-Carden proposal and the
one sketched here.
1. As was remarked before (pp. 167-8), it may be necessary to have more
entries like LII for lexicalizing the REL node. Thus, for the Dutch lexicon
and for certain cases in English, we shall probably need a REL-entry
associated with a zero phonological string. Thus we could account for the
difference between e.g., many people and a large number of people in the
lexicon, rather than state exceptions to Jackendoff's of-dropping rule.
(We can also account in this way for the superficial difference between, e.g.,
French beaucoup de monde and English many people.)
2. Will it be possible to account for Jackendoff's examples (67)--(70)
(Guess what we don't have any of. .. , We don't have any insect repellent,
etc.)? It will be, if the pre posing of what does not prevent lexicalizing the
REL node. According to our proposal, what we don't have any of would
have an underlying structure roughly like that of LXXXIX (p. 190).
Rules such as NEGATIVE PLACEMENT, DO-SUPPORT, and
SOME=>ANY will yield something like 'we do not have any [REL, WH ART
MASS RELJ'. Setting up an appropriate REL-entry and one for what will
enable us to get 'we do not have any of what'. A postlexical rule might now
prepose what, yielding what we do not have any of. The English lexicon should
furthermore indicate that the genitive marker is zero if ART QUANTITY
is lexicalized as some or any and is not followed by WHo
This admittedly sketchy description of how to derive (67) shows again
the necessity of having at least two REL-entries stipulating exactly when
the genitive marker should be zero and when it should not. It appears
possible, at any rate, to account for Jackendoff's examples without his
proposed of-dropping rule, which would have to specify with what word
classes it should apply (a great number of men, but not *many of men). It
would seem that a solution in terms of REL entries will be less costly than
adding an extra rule to the transformational component for eliminating of.
190 THE STRUCTURE UNDERLYING MP SENTENCES
LXXXIX
~NP
I
S
~S
Q
~S
NEG
~VP
NP
I
(we)
~
(have) NP
~RELrNP
~N
DET
~NP
ART
I I~
QUANTITY
WH NP
~
(some)
~NP
DET
I
ART N
I
I
MASS
The simplicity criterion of the lexicon (cf. Gruber (1967a), 2.2.2) which is a
matter of counting types of configurations occurring in it, makes it clear that
it will always be more economical to account for rules with a relatively small
domain (and with exceptions) in the lexicon rather than in the transform-
ational component.
But we also have Guess what we don't have three of Presumably the
structure underlying the what clause of this sentence should be something
like LXXXIXa (p. 191).
NE INCORPORATING here cannot apply, since WH occurs between
the NE and ART PLURAL. The rest of the derivation is similar to that of
what we don't have any of
3. A similarity between the Lakoff-Carden proposal and the one present-
ed here is that in both cases the noun preceded in the surface structure by
the quantifying word is related in the underlying structure to the latter by
a verb phrase containing the quantifying word, at least in the case of
THE DERIVATION OF MP SENTENCES 191
LXXXIXa
~NP
S
I
Q
~S
~S
NEG
~VP
NP
I
(we)
~
(have) NP
~RELr
NP
~N
DET
~NP
I
ART
I~
NE WH NP
I I
D\)GER I +5 ~
(three) PLURAL
numerals. That is, the Lakoff-Carden proposal implies that ten dollars is
related to a structure like 'dollars be ten'. In LXXXVII we have something
similar: here we could express it as '(a number of) dollars which be (with a
value which is at) ten'.
xc (NP)
~(N)
(DET)
~N
(NE)
~
(MASS) ABSTR
I
(PLURAL) UNIT
I
MONVAL
I
I
{ GUILDER}
DIME
~N
DET
NE N
~
(MASS)
ABSTR
I
(PLURAL) UNIT
I
MONVAL
I
I
GUILDER
"*' gulden,,*,
XClb N
(MASS) ABSTR
I UNIT
I
(PLURAL)
I
MONVAL
DIME
I
'* dubbeltje '*
THE DERIVATION OF MP SENTENCES 193
(MASS) ABSTR
I
(PLURAL)
I
UNIT
I
MONVAL
I
GUILDER
'*' gulden,*,
The cyclical ordering of lexical rules causes entries like XCla to take
precedence over XClb and XCII, since entries for integer-independent
unit words are in a cycle applying to higher nodes than the ones for integer-
dependent unit words (see pp. 128-9).
NE N
~
(MASS)
ABSTR
I
(PLURAL)
I
UNIT
I
WEIGHT
I
KILO
"'kilo '"
THE DERIVATION OF MP SENTENCES 195
XCIIlb S
~VP
NP
I~
HUMAN
V NP
I~
BE DET N
NE N
~
(MASS)
ABSTR
I
(PLURAL) UNIT
I
TIME
I
I
AGE
I
YEAR
XCI~
NP VP
~PP
V
I~
BE P NP
I~
WITH NP NP
DET
~~
N DET N
I NE
ABSTR N
I
PAR
~
(MASS)
ABSTR
~
MAIN WEIGHT
I
(PLURAL)
I
UNIT
I
WEIGHT
I
KILO
{ ,*,kiIO *}
'*' ¢ '*'
196 THE STRUCTURE UNDERLYING MP SENTENCES
Other zero entries, such as those needed to account for (76)-(79), can be
given in a similar way (but here it will not be necessary to indicate in the
peripheral environment whether or not we have a case of SIMPLE MP
SENTENCE REDUCTION). There are other cases of MP reduction,
which occur under different conditions, but those can be dealt with in the
lexicon in rather obvious ways. Therefore, I shall not dwell on the subject
any longer.
7.4. Recapitulation
It will be noted that the three aspects of the description of measure phrases
(namely, the derivation of unreduced forms, the distinction between
integer-dependent and integer-independent unit nouns, and the description
of reduced forms) all rely rather heavily on the lexicon. This is not sur-
prising, for the presence or absence of the genitive marker as well as both the
'integer-dependent/independent' distinction and the reduction phenomena
appear to be highly idiosyncratic and hence language-specific.
As has been shown, there are certain points in my proposal for the
description of MP's (and related structures) which are similar to some of
those in the proposals put forward by lackendoff and by Lakoff and
Carden. One similarity is that there is some underlying element directly
relatable to of in lackendoff's proposal as well as in mine. Another simil-
arity is that the quantifying word in MP's (dozens, three, etc.) appears in
the underlying structure of a predicate starting with 'be' (where the unit
noun occurs in the subject) in both the Lakoff-Carden proposal and mine.
lackendoff's of-dropping rule can be dispensed with if the claim is correct
that the lexicon can, and should, deal with the presence or absence of a
phonologically non-null genitive marker. The Lakoff-Carden proposal does
not account for the differences between integers and differentiating words
like ettelijke ('a fair number of) and verscheidene ('several'), that is, between
NE's and NUMBER constituents. Furthermore, the derivation of MP's
can be handled by (slightly altered) transformations already available for
the derivation of all kinds of MP sentences, save one extra transforma-
tion ('NE INCORPORATING'). The approach suggested in this section
implies that numerals such as they occur in measure phrases and, in general,
in prenominal position, are all derived front underlying proper nouns, viz.,
number names. Hence, on the deepest level no distinction need be made
between number names and prenominal numerals, which opens the way
to new generalizations.
Apart from the fact that a systematic distinction can be made between
NUMBER constituents and NE's, expressible in structural terms in the
lexicon which at the same time explain the fact that they are akin to one
THE DERIVA TlON OF MP SENTENCES 197
DURATION SENTENCES
1. 'DUREN'
In the following subsections I shall discuss a number of facts about sentences
referring to duration. As we saw (Chapter I, Section 3.1), the semicopulas
duren ('last'), steken ('draw'), reiken ('reach', 'have a range or) and dragen
('carry') are similar in that they take adjectival complements. But there is a
difference between the adjective occurring with duren and those occurring
with the other three (cf. (106}-(126b) of Chapter I). After having dealt with
duration sentences, I shall add a few remarks on the other three semicopulas
(Section 2).
1.1. Double WITH Strings
1.1.1. A Comparison with 'Simple' Semicopulas and Related Structures
As we observed in Chapter I, Section 3.1, the adjective lang, used in connec-
tion with duration, differs from other parameter adjectives in that it may
occur as the complement of the corresponding semicopula duren (Het duurt
lang vs. *Het weegt zwaar, *Het kost duur). 'Simple' semicopulas (i.e., semi-
copulas not taking adjectival complements, as opposed to 'complex' semi-
copulas like duren) may be followed by veel ('much') instead of an adjective
(Het kost veel vs. *Het duurt veel).
Suppose we have the following pair of entries for duren:
I 5
NP VP
~PP
v
I~
BE P NP
I
WITH NP PP
DET
A PAR P
~NP
I I~
TIME WITH NP NP
DET
~~PAR DET N
'*' durenl '*' I ~
LENGTH AMOUNT (X)
DURA TION SENTENCES 199
NP VP
v
~PP
I~
BE P NP
I~PP
WITH NP
A~
DET PAR P NP
I
TIME WITH
I NP MP
DET
~PAR
I
LENGTH
N~
~
DET PAR
Rr
NP
r
I
LENGTH
~PAR
DET
I
TIME
200 THE STRUCTURE UNDERLYING MP SENTENCES
~RELr
N~
~PAR
DET
I
S
I
LENGTH
~VP
NP,
DET
A~
PAR V PP
II~
TIME BE P NP
I
WITH PRO
I
Let us assume that the structure underlying duration is something like V:
NP
V ~RELr
NP,
~PAR
DET
I
S
TIME
I~
NP, VP
PRO
I~V PP
I~
BE P NP
I~PAR
WITH DET
I
LENGTH
Both REL constituents of IV and V derive from the same underlying relative
clause (via RELATIVE CLAUSE FORMATION), namely, [REL r DET
TIME BE WITH DET LENGTH RELJ. Suppose that the structure immedi-
ately underlying duration is one derived by RELATIVE CLA USE RED U C-
TION from V(V/).
Assuming that III and IV are essentially correct, we could account for the
fact that duration and length of time are semantically closely akin, or even
equivalent, but that they cannot be substituted for one another in all sorts
DURATION SENTENCES 201
VI NP
~PP
NP
~
DET
PAR P
~NP
II~
TIME WITH DET PAR
I
LENGTH
of contexts (e.g., The length of time for this pedal-chord is indeterminate but
not *The length of time of this pedal-chord is indeterminate, although The
duration of this pedal-chord is indeterminate is correct and means the same as
the first sentence; but *The duration for etc. is bad).
It seems clear that the noun duur and the verb duren must be related in
terms of underlying structure. Therefore it is plausible to assume that the
prelexical terminal strings underlying duration sentences must contain a
substring representable as (WITH) DET TIME WITH DET LENGTH.
2. As I mentioned earlier, een lange duur ('a long duration') is correct,
although we do not have expressions such as *een zwaar gewicht ('a heavy
weight') or *een dure prijs ('an expensive price')'. If we take the following
diagram to represent roughly the entry for the noun duur as it occurs in een
lange duur, we are in a position to explain this.
VII NP
~PP
NP
DET N
~NP
P
I
PAR
I~
WITH NP NP
I
TIME
~N
DET
~N
DET
I
PAR
~(X)
AMOUNT
Or, to put it differently, that the underlying structure of (4b) does not
contain a string WITH DET LENGTH.
It would seem, in view of this, that we may have the following entry for
duur as it occurs in (4b) :
VIlI NP
~PP
NP
DET PAR P
~NP
I
TIME
I
WITH NP w
~PAR
DET
I
LENGTH
duur taking MP's (or, put differently, not incorporating DET AMOUNT).
However, we do have simultaneous environments in which WITH DET
WEIGHT and WITH DET PRICE are not followed by DET AMOUNT,
namely, those associated with wegen and kosten, respectively.
The following sentence is also ungrammatical:
(7) *De pauze duurde 5 minuten lang. (cf. (3b»
'The intermission lasted 5 minutes long.'
The fact that (7) is ungrammatical can be explained in terms of the economy
principle in the following way. The underlying structure of (7) would have
to be something like NP BE WITH DET TIME WITH DET LENGTH MP,
where BE WITH DET TIME would have to underlie duren. The economy
principle accounts for the non-existence of a semicopula duren having such
a simultaneous environment and matching the structure of a simple dura-
tion MP sentence if we assume that the entry II for duren2 (p. 199) is correct.
F or, the entries of this hypothetical duren and of duren 2 would both consist
of subtrees ending in the string NP BE WITH DET TIME WITH DET
LENGTH MP. The economy principle dictates that of two such entries, only
the latter can exist (i.e., only the one having BE WITH DET TIME WITH
DET LENGTH in its simultaneous environment).
In short, we can account for the ungrammaticality of sentences like (5),
(6) and (7) in a simple and uniform way (viz., in terms of the economy
principle) if the double WITH string hypothesis is correct.
Having discussed a number of facts that seem to support the double WITH
string hypothesis, I shall now add a few remarks that have some bearing on
the points considered above, after which, in the following subsection, an-
other type of duration sentences will be examined.
In connection with (7), above, the following should be added.
In certain cases it seems that lang is not wholly excluded in simple dura-
tion MP sentences, especially with MP's referring to relatively long periods.
For example, a sentence like William York's 'Four-Pedal-Appointed-Pro-
fessor-Give-Away'duurt twee uur lang is not entirely unacceptable. How-
ever, it seems that lang here is not a neutral adjective. It somehow emphas-
izes the fact that two hours is a relatively long time. It is not clear to me how
this phenomenon should be handled in the lexicon.
Finally, in connection with the entries given above, we may add the
following.
In Chapter III no attention was given to the DET node in entries for
words showing no trace of it. In the entries for wegen, zwaar, and the like,
I left out the DET node in order to keep matters simple. However in more
precise representations of such entries DET should be present, as it is in the
206 THE STRUCTURE UNDERLYING MP SENTENCES
diagrams above. If it were not, they could not match any derived structure.
Structures directly underlying parameter adjectives and semicopulas
always contain DET as a result of the base rule NP-+DET N, and of the
fact that DET in such structures is not deleted by any transformation.
Hence we should now ask how one might deal with the determiner during
lexicalization of words like the measure adjective lang.
One way to explain the fact that the determiner has no phonological
correlate in lang is to assume that it is simply incorporated. Alternatively,
we could suppose that DET in the reversed string LENGTH DET WITH
somehow corresponds to a zero phonological element occurring in lang.
A third possibility is to assume that the element DET is omitted during
lexicalization of lang. This course is followed by Gruber.
Gruber (1967a) proposes a lexical restructuring rule of 'morpheme
omission'. He offers as an argument against ordinary transformational
treatment of word extension (such as affixation) the fact that for trans-
formations to apply it would often be necessary to specify that a good deal
of structure is not present. Thus we do not want to derive from He snatches
bodies the sentence *He bodies-snatches (rather, He body-snatches), or, to
give an example of a different kind of word extension (extending the mean-
ing of a word like can) from John put the fish into the big cans, the sentence
*John big canned the fish. Having cited these examples he then points out
that "if we were to treat these things by transformations, we would have to
specify either a series of things that must be absent in order for the trans-
formations to apply (not always possible, as in the case of obligatory articles)
or to specify ad hoc a large number of obligatory deletions)" (op. cit.,
p. 136). Therefore, some deletions should take place during lexical attach-
ment if we accept Gruber's rather plausible argument. It would seem that
his solution presents no special problems with regard to DET. It can be
eliminated as follows. The first step in restructuring the tree to be lexicalized
in the case of lang would take place as indicated below:
IXa PP IXb PP
P
~NP ---..,. ~NP
pp'
I~
WITH
NP MP
~~
PN NP MP
~N
DET
I
WITH
I
LENGTH DET
I
I
LENGTH
DURATION SENTENCES 207
PP' NP
P
~N MP
WITH
I I
LENGTH
The encircled NP node will also be pruned. Thus, after order-reversal and
attachment of the phonological string, we get the following configuration:
[Xd PP
~MP
PP'
N
~P
LENGTH
I I
WITH
................. ,..... /////
# ,'a.,,~ #
Phrases like gedurende een week ('for a week'), with an indefinite de-
terminer, will be referred to as 'duration measuring adverbials' (DMA's),
as opposed to phrases like gedurende die week (,for (the duration of) that
week'), gedurende het weekend (,for (the duration of) the weekend'), with
definite determiners, which will be called 'duration dating adverbials'
(DDA's). In the latter, the determiner has a role in the identification of an
interval along the time axis.
Verb phrases of simple MP sentences (such as duurde een week) will be
called'MVP's'.
The following pieces of evidence can be presented in support of the
proposal that sentences like (8) and (9) are transformationally related to
each other.
1. If the gedurende phrase cannot contain a definite determiner, the
corresponding MVP cannot either:
(10) *Gedurende de week lag Lex plat.
(11) *For the week Lex lay flat on his back.
(12) *Het platliggen van Lex duurde de week.
(13) *Lex's lying flat on his back lasted the week.
2. DMA's and MVP's behave in the same way with constituents in post-
position:
(14) gedurende drie jaar achtereen
'for three years consecutively' (for three consecutive years)
(15) ... duurde drie jaar achtereen
' ... lasted three years consecutively' (three consecutive years)
(16) *gedurende drie jaar gevangenisstraJ
'for three years of imprisonment'
(17) *... duurde drie jaar gevangenisstraJ
' ... lasted three years of imprisonment'
3. In DMA's as well as in MVP's the conjunction of MP's is excluded:
(18) *Gedurende 2 uur en 3 uur regende het.
'For 2 hours and 3 hours it rained.'
(19) *De regen duurde 2 uur en 3 uur.
'The rain lasted 2 hours and 3 hours.'
But the italicized phrase is possible in sentences like 2 uur en 3 uur is
samen 5 uur ('2 hours and 3 hours makes 5 hours together'). To some
speakers (18) and (19) may sound acceptable if the italicized phrase is
interpreted as an expression synonymous with
DURATION SENTENCES 209
x s
~VP
NP
s
I V
~PP
//A", I ~
// ',BE P NP
/
L __________ -' ", I ~
(Lex liggen plat)
(Lex lie flat on his back) A
WITH NP
~
PP
DET TIME P NP
,~
WITH NP MP
/'.,
~
DET
LENGTH
//
L -'
"
_______"-.::..
(een week)
('a week')
S.c.: 2+5 3 0 0 6
XII S
~VP
NP
i~
(Lex)
VP PP
/ //"--... , ~
,~------.....:::.
(iiggen plat) P NP
('lie flat') I ~
WITH NP PP
DET
A~
TIME P NP
I~MP
WITH NP
, /.... ,
~ /
/ , ....
~:.<-----.::.:::::..
DET LENGTH (een week)
('a week')
We can infer from this that the expressions gedurende twee weken, twee
weken lang and twee weken are all synonymous in (24)-{26). In our article,
Verkuyl and I tried to account for this by having prelexical transforma-
tions drop either WITH DET TIME or the entire double WITH string
underlying gedurende. Deletion of WITH DET TIME thus would give
lang twee weken, which could be turned, post-lexically, into twee weken
lang, while deletion of the entire string underlying gedurende would
result in simply twee weken.
A disadvantage of this solution is that one would have to set up a rule or
rules in the transformational component for the sole purpose of handling
DMA's and DDA's, which seems rather ad hoc. A less costly alternative
would be handling the above phenomena in the lexicon.
In order to be able to suggest a solution along these lines, let us first con-
sider the following facts.
Note, first, that, although we have (27), (28) is ungrammatical:
(27) een langdurig verblijf
'a long-lasting stay' (a prolonged stay)
(28) *een twee weken langdurig verblijf
'a two weeks long-lasting stay'
Instead of (28), we must have
(29) een twee weken lang verblijf
'a two weeks long stay'
Secondly, langdurig, like most other compound adjectives starting with
212 THE STRUCTURE UNDERLYING MP SENTENCES
lang- and ending in the suffix -ig, cannot be preceded by erg, whereas lang
itself can. Thus, although we have (30), we do not have (31)-(37):
(30) erg lang
'very long'
(31) *erg langdurig
'very long-lasting'
(32) *erg langharig
'very long-haired'
(33) *erg langbenig/langpotig
'very long-legged'
(34) *erg langschedelig
'very long-skulled' (dolichocephalous)
(35) *erg langvezelig
'very long-fibrilled'
(36) *erg langstelig
'very long-stemmed'
(37) *erg langorig
'very long-eared'
etc.
An exception is langdradig ('long-threaded', meaning prolix), but this word,
as opposed to the italicized ones above, is clearly a metaphor.
Thirdly, lang- in langdurig, etc., must be taken to be the affix, whereas
-durig, etc. cannot be, for, although we have, e.g.,
(38) lang- en kortharig
'long- and shorthaired'
we do not have
(39) *langharig en -benig
'longhaired and -legged'
Typically, it is the affix in compound words which can occur separately in
conjunctions of the above type.
By 'affix' I mean a phonological string occurring (if it is a productive
affix) in an incomplete entry (i.e., in an entry with one or more blanks),
or (if it is a non-productive affix) a proper substring y of a phonological
string xyz (where either x or z, but not both, may be empty) such that:
(I) y can be identified as corresponding to a proper substring Y of the
terminal string of categories XYZ of some simultaneous environment (where
again X or Z, but not both, may be empty), (2) X and Z, insofar as they are
not empty, each occur as the terminal strings of simultaneous environ-
DURATION SENTENCES 213
ments in complete entries, and, (3) Y does not occur as the terminal string
of any simultaneous environment.
In the fourth place, lang- can only occur as an affix preceding N + -ig
if the latter string cannot occur alone (cf. *schedelig, *stelig, *orig, and
*werpig (from langwerpig, 'oblong')). Words like benig, potig, harig, vezelig
are only apparent counterexamples. The first two words have meanings
(viz., 'bony' and 'strong-limbed', respectively) that are at best remotely
related to the word-parts in (33) which are identical in form. The word
harig also has a meaning that differs from that of the part -harig in
langharig: een langharige linguist ('a long-haired linguist') refers to a
linguist with long hairs on the head, whereas een harige linguist ('a hairy
linguist') refers to a linguist with hairs all over his body in abundant
quantity. Likewise, the meaning of vezelig ('fibrous') differs from that of the
corresponding word-part in langvezelig (een vezelige substantie ('a fibrous
substance') but not *een langvezelige substantie ('a long-fibrilled substance')).
Moreover, although we have, e.g., both baardig ('bearded') and met een
lange baard ('with a long beard'), we do not have *Iangbaardig ('Iong-
bearded'), contrary to what we would expect if it were the case that lang-
could occur as an affix with strings of the form N + -ig which could also
occur by themselves.
Note, finally, that langdurig is not synonymous with lang, for (40) and (41)
are not paraphrases of each other:
(40) Hij verbleef er lang.
'He stayed there long.'
(41) Hij verbleef er langdurig.
'He stayed there very long.'
It can be illustrated that (40) and (41) are not synonymous by the fact that
lang may be preceded by erg whereas langdurig, as we have seen, cannot.
Sentence (40) is related to the English expression his long stay there, while
(41) is related, more or less, to his prolonged (protracted) stay there. Lang-
durig is equivalent to erg lang. Hence *erg langdurig is excluded for two
reasons: it is ungrammatical for the same reason as (32)-(37), and erg is
superfluous. Lang- in langdurig may be assumed to incorporate an element
underlying erg. If this is correct, lang- is not a productive affix in langdurig,
although it is in the other cases. On the other hand, it is clear that, if this
analysis is correct, the difference between lang- in langdurig and lang- in
the other compounds is very small; they differ only in that the former in-
corporates the element underlying erg whereas the latter does not.
The main point of the observations above is that there must be a relation
between words like lang (the temporal measure adverb and the non-neutral
214 THE STRUCTURE UNDERLYING MP SENTENCES
XIII PP
P
~NP
I~PP
WITH NP
~~
DET X P NP
I~
WITH NP NP
~PAR DET~N
DET
I~
LENGTH
AMOUNT MODIFIER
I
GREAT
+ lang.-- - - - - - - - - . +
XIIfa
PP
~NP
P
I~PP
WITH NP
A~
DET PAR P NP
II~NP
TIME WITH NP
~PAR DET
DET
~N
I~
LENGTH AMOUNT MODIFIER
A
GREAT VERY
XIIfb
PP
~NP
P
I~
WITH
NP PP
~~
DET
PAR P NP
II~
TIME
WITH NP NP
~PAR DET
DET
~N
I ~
LENGTH AMOUNT MODIFIER
GR~-X
'*' lang '*'
(where X may be empty or a category like VERY.)
216 THE STRUCTURE UNDERLYING MP SENTENCES
x/v
PP
~NP
P
I~
WITH
NP ~
~~
DET
PAR P NP
II~
TIME
WITH NP MP
~
DET PAR
I
LENGTH
X/Va PP
P
~NP
I~
WITHNP PP
~~
PAR
DET
I
TIME
LENGTH
But, as is pointed out in Klooster and Verkuyl (1972) with regard to this
claim, there are cases which seem to be counterexamples. For instance, we
have a sentence like
(42) Gedurende vier middagen wandelden ze naar het strand.
'For four afternoons they walked to the beach.'
which does not correspond with
(43) Hun wandeling naar het strand duurde vier middagen.
'Their walk to the beach lasted four afternoons.'
Sentence (42) can only mean that the persons referred to took repeated
walks to the beach, that they did so on four afternoons, and either that
each walk to the beach took them one such afternoon or that they spent
each of the four afternoons taking several walks to the beach. Sentence (43),
on the other hand, means that the persons referred to took one single walk
that was interrupted three times, and reached the beach at the end of the
last of four afternoons, having spent those afternoons walking in the
direction of the beach.
We have a similar case in the sentences (44) and (45):
(44) Jarenlang werd Yvonne de U.S.A. uitgezet.
'Yvonne was deported from the U.S.A. for years.'
(45) *Het uitzetten van Yvonne uit de U.S.A. duurde jaren.
'Yvonne's deportation from the U.S.A. lasted years:
These incongruities seem to be caused by the fact that underlying the
italicized constituents of (43) and (44) is a structure containing categories
which cause them to have a momentaneous and a terminative aspect,
respectively (cf. Verkuyl (l972a».
The non-durative aspect of the italicized part of (42) wandelden ze naar
het strand is described in Verkuyl (1972b) in terms of categories occurring
in the structure underlying sentences that can be said to have such a non-
durative aspect. Together the categories in question, occurring in a certain
configuration, can be seen as the complex of categories constituting the
non-durative aspect.
According to Verkuyl, a sentence of the form NP1 - [yp V - NP 2 (or
QC) - (NP 3) vp] (where 'QC' is either a 'Quantifying Complement' or a
directional Prep Phrase (such as naar het strand), and where the paren-
thesized NP is the indirect object), should conform to the structural de-
scription, or schema, of the form XVa, below, if it can be said to have a
non-durative aspect (I have abbreviated 'SPECIFIED QUANTITY OF
X' as 'SQ-X' in XVa):
DURATION SENTENCES 219
XVa l 5Q-X ]}
[[ 5Q-X )-[ [VERB] - { NP2 NP2 -([5a-x» ]]
5 NP, NP, VPV V QC NP3 NP3 VP 5
The fact that we do not have (45) *Het uitzetten van Yvonne uit de U.S.A.
duurde jaren is accounted for in Klooster and Verkuyl (1972) by means
of a condition on nominalization of the subject S in structures of the type
of X which can be formulated as follows:
XVI NOMINALIZA TION of the subject S blocks if, in structures
where the matrix S dominates a VP containing the categories BE
WITH DET TIME WITH DET LENGTH, the VP of the subject
S is non-durative.
We can now replace XVI by XVIa:
X VIa Condition on NOMINALIZATION
Nominalization of the subject S blocks if, in structures where
the matrix S dominates a VP containing BE WITH DET TIME
WITH DET LENGTH, the subject S matches the non-durative
schema XVa.
One thing that is not accounted for by ADVERBIALIZA TION is the fact
that the tense-less subject S in X acquires the same tense that is carried by
duren in (9) Het platliggen van Lex duurde een week. In order to account
for this we might replace X by a structure roughly like XVII:
XVII
NP VP
I I
V
S2
~VP
NP
I
TENSE
I
S3
~NP
WITH
I
PAST
/ /\ , ~
/ '
/ '\ NP PP
l~ ______ ~~~ ~
(Lex liggen plat) DET
~NP
TIME WITH
~MP
NP
~
DET
LENGTH
27 See. for a more detailed discussion of tense and time reference. McCawley (in press). and.
XVIII S
~VP
'"
NP
---platliggen-
(Hef
----~----
---- VP/ '
van-Lex)
~
PP
V
I~
WITH NP
I~
TENSE NP PP
I ~~
PAST DET TIME WITH NP
~MP
NP
~
DETLENGTH
~X
VP
I v
I
TENSE
#----zijn#
xx S
~VP
NP
i~
(L~X) V VP
l~
TENSE VP PP
I .c::~~~~~~~. ~NP
PAST (Ijggen plat) P
I~PP
WITH NP
A~
DET
TIME P NP
l~MP
WITH NP
DET
~LENGTH
s.c.: [5 + [2 + 4]]3 6
VP VP VP VP
A solution along these lines can probably be worked out in a satisfactory
way.
The fact that there must be a transformational relationship between
duren MP sentences and gedurende MP sentences provides additional evi-
dence in favour of the double WITH string hypothesis. For, it is plausible
to assume that duurde in (9) Het platliggen van Lex duurde een week in-
corporates a string which is related to lang in the paraphrase Lex lag een
week lang plat. As the entries XIII, XIIla, b, and XIV (which account for a
relatively large number of facts) make clear, it is the string WITH DET
LENGTH which directly corresponds to the phonological string lang in
various ways. This string must also be present in the underlying structure
of (9). Moreover, as we have pointed out already, both gedurende and duren
in MP sentences exclude lang. The double WITH string hypothesis neatly
covers these facts if it is assumed that the explanation is to be found in the
incorporation of WITH DET LENGTH in gedurende as well as duren z (cf.
II and XIVa, respectively).
DURATION SENTENCES 223
XXII
s
~VP
"
NP1
//~'"
.£~------~
"-
A
(de pauze)
(the intermission) I A
V PP
BE P NP
IA
WITH NP2 RELr
I I
TIME S
A NP2 VP
IA PRO V PP
IA BE P NP
I
LENGTH
I
S
A
NP3 VP
IA
PRO V PP
IA BE P NP
I I VAL S
A NP4 VP
IA PRO V PP
IA BE P MP
,,
I
/~,
,
/
/
/
"- _____ ..a.
AT (5 min.)
DURATION SENTENCES 225
XXIII "'"
~P
~LrI
NP2
I
TIME
S
~
VP
NP2 ~
I
PRO V ~
PP
I
BE I
WITH
~p
NP3
NP
I
LENGTHP ~
NP
I
WITH NP4 ~
PP
I
VAL PI
MP
/"',
/ ',.
AT (S"""inin.)
XXIV
~p
~ELr
NP2 I
TIME
I ~
S
VP
NP2 ~
PRO
I VI ~
PP
NP
BE P ~
WITH
I NP3
MP
/'',..
'/'-''-,-
I (~minuten)
LENGTH (5 minutes)
226 THE STRUCTURE UNDERLYING MP SENTENCES
xxv s
~VP
NP,
L~~~~~" ~PP
(de pauze) V
(the intermission) I ~
BE P NP
I~PP
WITH NP2
I~NP
TIME P
I~MP
WITH NP
I
3
/ ,./',
/ '
~~-----~~
LENGTH (5 minuten)
(5 minutes)
XXVI ~
NP
~PP
NP,
I~NP
TIME P
I~RELr
WITH NP 2
I
LENGTH
I S
~VP
NP2
I~
PROV PP
I~
BE P NP
I~REL r
WITH NP3
I
VAL
I
S
~VP
NP3
I~
PRO
V PP
I~
BE
P MP
"", " ,
I
AT
/' / '
L-.::.. ____ ~~
(5 minuten)
(5 minutes)
The structure underlying een pauze van 5 minuten presumably should look
like XXX (p. 229).
EQUI-SUBJECT REL STRIPPING can simplify XXX so that we get
XXXI (p. 230).
After RELATIVE CLAUSE REDUCTION, we get XXXII (p. 231).
MP-TAIL SHORTENING now obligatorily reduces XXXII to XXXIII
(p. 232); which, after RELATIVE CLAUSE REDUCTION, will be reduced
to XXXIV (p. 233).
The rule that should (again) apply is EQUI-SUBJECT REL STRIPPING.
It should delete the italicized categories in XXXIV. If we assume that
a noun like pauze, referring to something that has duration, has a simul-
228 THE STR UCTURE UNDERLYING MP SENTENCES
XXVII
~NP
~PP
NP1
I~NP
TIME P
I~RELr
WITH NP2
I
LENGTH
I
NP
~RELr
NP3
I
VAL
I
S
~VP
NP3
I~
PROV PP
I~
BE
P MP
I L~~~'::::-'"
AT ( 5 minuten)
(5 minutes)
XXVIII
~NP
~PP
NP1
I~NP
TIME P
I~RELr
WITH NP2
I
LENGTH
I NP
~PP
NP 3
I~MP
VAL P
I ""
/A-..,
-.. -..
L _ _ _ _ _ _-.,.
AT (5 minutM)
(5 minutes)
DURA nON SENTENCES 229
XXIX
xxx
NP
~
,~;
RELr
/' ", I
L ________ .:-...
(een pauzej S
~VP
NP1
I~PP
PRO V
I~NP
BE P
I~RELr
WITH NP2
I I
TIME S
NP2
~VP
I~PP
PRO V
I~NP
BE P
IA
WITH NP3 RELr
I
LENGTH
I
S
~VP
NP3
I~PP
PRO V
I~NP
BE P
I~RELr
WITH NP4
I
VAL
I S
A
NP4 VP
I~PP
PRO V
I~
BEP MP
I " ,A,'-,
/
AT (5min~;
230 THE STRUCTURE UNDERLYING MP SENTENCES
XXXI
NP
~RELr
NP,
"
/,"'--,
' ...
I
~::----:~
(een pauze) S
~VP
NP,
I~PP
PRO V
I~
BE P NP
I~RELr
WITH NP2
I
TIME
I
S
~VP
NP2
I~PP
PRO V
I~
BE P NP
I~RELr
WITH NP 3
I
LENGTH
I
NP
~RELr
NP4
I
VAL
I
S
~VP
NP4
I~PP
PRO V
I~
BE P MP
I /"'-',
L-:. _____-
/ ~~
AT (5 minuten)
DURATION SENTENCES 231
XXXII
NP
~RELr
NP1
-' /'..
-'
L.:_____
, "-
.:::~
I
(een pauze) S
~VP
NP1
I~PP
PRO V
I~
BE P NP
I~RELr
WITH NP 2
I
TIME
I S
~VP
NP2
I~
PROV PP
I~
BE P NP
I~RELr
WITH NP3
I
LENGTH
I
NP
~PP
NP4
I~
VALP MP
/",
I ",,"" -' "
"
AT ~------~
(5 minuten)
232 THE STRUCTURE UNDERLYING MP SENTENCES
XXXIII NP
~RELr
NP1
'" '
/ /
(een pauze)
"-
"
L /_ _ _ _ _ _ -!:Io.
I
5
~VPNP1
I~PP
PRO V
I~
BE
P NP
I~RELr
WITH NP2
I
TIME
I
5
~VP
NP2
I~PP
PRO V
I~
BE
P NP
I~
WITH NIP RiLr
LENGTH MP
XXXIV NP
A
NP RELr
/ A'" I
/ ---.::.:~ s
reenpauzeA
/\p
NP VP
I'
PRO VIA
NP
BE P ~
I
WITH NP2 A PP
I NP
TIME PI ~
WITH NP3 I
RELr
I
LENGTH
MP
/",
/ ,
fs~mTnutenJ
XXXV
234 THE STRUCTURE UNDERLYING MP SENTENCES
XXXVII
As was already mentioned, duren and the verbs dragen, steken and reiken
behave in a similar way. Therefore it seems plausible to assume that the
double WITH string hypothesis holds in all these cases. This means that we
can have three pairs of entries analogous to the pair for duren.
The only difference between the adjectives that may occur as the comple-
ment of dragen, steken and reiken on the one hand and lang on the other,
is that the adjectives corresponding with the first three verbs may not modify
the related parameter nouns, whereas lang may modify duur (not *een ver
bereik, *een diepe diepgang, although een lange duur is correct).
We may account for this by introducing entries for bereik and diepgang
which incorporate ver and diep, respectively. Thus the entries for bereik
and diepgang might look something like XXXVIII and XXXIX.
Note that XXXVIII and XXXIX are of the same type as VIII on p. 204.
If we exclude from the lexicon entries for bereik and diepgang of the type
of VII, the fact that ver and diep cannot modify bereik and diepgang, re-
DURA TION SENTENCES 235
XXXVIIla, b NP
~PP
NP
DET PAR P
~NP
I
ONESPACE
I
WITH NP w
I ~
DETPAR
{ RANGE}
REACH
I
ONESPACE
I
DIST
# bereik #
XXXIX NP
~~PP
NP
DET PAR P
~NP
I
ONESPACE
I
WITH NP w
I
DRAUGHT
~
DETPAR
I
DEPTH
41= diepgang #
of expressions referring to range, reach and draught. Thus the upper half of
possible strings of categories between large square brackets in LXXII on
p. 170 (the earlier given statement of MP SHIFT) should be restated in
such a way that the third term of the S.D. gets the form XL:
XL
{ TWOSPACE } }
{ MAIN THREESPACE
_ DEf ONESPACE (WITH DET PAR) _
DET MASS
Note that the optional rule of SPACE DELETION (p. 161) need not be
restated. Its structural description is such that it cannot apply in the cases
considered in this section.
CHAPTER V
EPILOGUE
presentations. But in Gruber (1967a, b) they do. Cf.also the diagram given in Gruber (1967a),
p.6.
238 THE STRUCTURE UNDERLYING MP SENTENCES
necessary, and so on, until no further operations could take place in either
component. At this stage the 'derived phonologically terminal trees' entered
the phonological component which turned them into pronounceable strings.
As seems clear, it will not be necessary, at least in the case ofMP sentences,
to organize a grammar in this way, i.e., to view lexical insertion rules as a
kind of 'anywhere rules', like tree-pruning. Instead we may make a clear
distinction between prelexical and postlexical transformations, the former
operating upon structures not even partly lexicalized, and the latter operat-
ing upon completely lexicalized structures. As McCawley (1968) remarks
(p. 78-79), having all lexical insertions take place after the cycle but before
the post-cyclic rules would be a possibility worth investigating in view of the
fact that transformations which can be argued to be prelexical are also rules
that appear to be in the cycle. As we have seen, if a rule of EMBEDDING
is assumed to be present in the transformational component, it cannot be
part of the same cycle as the other prelexical rules that we considered, al-
though it will have to precede the latter. Therefore it might be that there
is more than one cycle of prelexical transformations, or that there is one
prelexical cycle preceded by rules that are not ordered cyclically.30
The transformational rules that have been proposed above will not
stand, at least not in their present formulations. However there appears
to be some independent evidence for most of them if put in one form or
another. Thus, the rule of EQUI-SUBJECT REL STRIPPING can prob-
ably be formulated in such a way as to relate not only various measure phrase
constructions but also constructions containing human propensity nouns
(a man who has great charm-a man of great charm) and other construc-
tions of the same general type (eyes that had an astonishing intensity-eyes
of an astonishing intensity; a theoretical construct which has a highly abstract
nature-a theoretical construct of a highly abstract nature; an object that
has an irregular shape-an object of irregular shape). Similarly, there is
independent justification for EQUI-OBJECT REL STRIPPING (see p. 174).
As was demonstrated, rules like MP-TAIL SHORTENING, QUANTITY
DELETION, and MP SHIFT, with some revisions, may serve during the
derivation not only of various constructions in which MP's mayor may not
appear, but also of measure phrases themselves. The rule of RELATIVE
CLAUSE REDUCTION, as it has been stated in the preceding chapter,
can be regarded as supported by much more evidence than has already been
30 Lakoff (I 970b) argues that there must be global derivational constraints in order to solve
the problem of avoiding unwanted sequences of phrase markers during application of rules
in the transformational component. If it proves possible to formulate well-formedness condi-
tions on configurations in non-adjacent trees in the derivational sequence of MP sentences
there will thus be a third and rather more attractive possibility.
EPILOGUE 239
Nip
~RELr
NP
~
DETPAR
I
MP
I
LENGTH
Such entries would solve the problem of the elimination of DET in a simple
way. Many rules, then, may be merely lexical regularities.
The base rules discussed in Chapter III are, as has been stressed,
tentative, and can certainly be improved upon. Some of them generate
structures matching proposed lexical entries without any intervening trans-
formations. This may prove incorrect in at least a number of cases. Thus
it is possible with these rules to generate strings directly underlying semi-
copulas and copula + adjective constructions (albeit in trees not matching
240 THE STRUCTURE UNDERLYING MP SENTENCES
the stipulated peripheral environments) while it may turn out that the
element labelled 'BE' should not be generated as part of a VP containing
a prepositional phrase but as a higher verb ('TENSE') never followed by
PP in base structures. (The possibility of regarding 'be' as a tense carrier
has been considered in Chapter IV.)
Possible shortcomings of this kind, however, do not, from a general point
of view, weaken the case made out for base rules generating semantic
representations of measure phrase sentences whose relationships to phono-
logical representations are expressed in terms of transformational and
(directly or indirectly) of polycategorial lexical attachment rules. It seems
probable that a full-fledged grammar should contain a base component
generating structures which are even more abstract (i.e., farther removed
from surface structure) than the ones considered in the present work. At
least part of the unclarity as to the criteria by which to distinguish between
syntactic and semantic categories (see p. 86) may be attributed to this.
McCawley (1971), in his discussion with Katz (1970), calls attention
to the fact that more recent papers by Lakoff and himself have made ex-
plicit the claim that three non-terminal node labels ('S', 'NP', and 'V')
will suffice and that "these categories match in slightly broadened form
categories used by logicians: S corresponds to the use of 'sentence'
(= 'propositional function'), V to 'predicate' (taken as including 'operator'),
and NP to 'argument'." (op. cit., p. 292; what McCawley is thinking of
when he uses the phrase "in slightly broadened form" is what is referred to
in Lakoff(1970a) as 'natural logic'.) In the sense that conditions of semantic
well-formedness can be stated in terms of S, V and NP, the distinction be-
tween 'syntactic' and 'semantic', as these terms are used by McCawley and
others (but not, for example, by Seuren), is invalid. It is obvious that a
matter of terminology is involved, for clearly it is justifiable to apply the
term 'syntactic' to rules determining the correct surface ordering of various
constituents, as opposed to 'semantic' in the case of rules generating se-
mantic representations. 31 Part of the reasons for assuming the validity
31 As another instance of different uses of the terms let me cite the following examples (italics
are mine):
(a) "The meaning of a word must be formalized in terms of the same sort of structures as we
have in syntactic construction, i.e., in terms of a tree of elemental semantic categories."
(Gruber (I 967a), p. 50.)
(b) '" believe that these considerations indicate that syntactic and semantic representations
are objects of the same formal nature, namely, ordered trees, whose non-terminal nodes are
labeled by syntactic category symbols, and that in each language there is a single system of
transformations which convert semantic representations of sentences into their superficial
form: these transformations include 'lexical transformations', i.e., transformations which
replace a portion of a tree by a lexical item." (McCawley (I967a), p. 55.)
EPILOGUE 241
Note, incidentally, the difference between McCawley's view and Gruber's with respect to tree-
restructuring preceding lexical attachment.
As De Rijk (1968) notes, McCawley's using the phrase 'syntactic category symbols' while
Gruber refers to the same objects as 'semantic categories' must be "merely a matter of
terminology, not necessarily reflecting any substantive difference between their views."
(p. 8). The fact that the terms 'syntactic' and 'semantic' could be interchanged in the above
quotes illustrates what McCawley means when claiming that the syntax/semantics dichotomy
is invalid.
242 THE STRUCTURE UNDERLYING MP SENTENCES
Bach, Emmon, 1967, 'Have and be in English syntax', Lang. 43, 462-85.
Bech, Gunnar, 1952, Ober das niederliindische Adverbialpronomen 'er', Travaux du Cercle
Linguistique de Copenhague, Vol. VIII, Amsterdam.
Bendix, Edward Herman, 1966, Componential Analysis of General Vocabulary: The Semantic
Structure of a Set of Verbs in English, Hindi and Japanese, The Hague.
Bierwisch, Manfred, 1967, 'Some Semantic Universals of German Adjectivals', FoL. 3,1-36.
Bloomfield, Leonard, 1957, Language, London (revised edition).
Carden, Guy, 1968, 'English Quantifiers', Report No. NSF-20, MLAT, Cambridge, Mass.
Chomsky, Noam, 1965, Aspects of the Theory of Syntax, Cambridge, Mass.
Chomsky, Noam, 1968, 'Deep Structure, Surface Structure and Semantic Interpretation', un-
published paper, in L. Jakobovits and D. Steinberg (eds.) t.a.
Dik, Simon c., 1968, Coordination; Its Implications for the Theory of General Linguistics,
University of Amsterdam Doctoral Dissertation, Amsterdam.
Dik, Simon c., 1969a, 'Seuren over Coordination', De Gids 132, 243-62.
Dik, Simon C., 1969b, Relatieve Termen, Amsterdam.
Dougherty, Ray c., 1968a, A Transformational Grammar of Coordinate Conjoined Structures,
unpublished M.LT. Doctoral Dissertation.
Dougherty, Ray C., 1968b, 'Coordinate Conjunction', unpublished M.LT. paper.
Dougherty, Ray c., 1969, Review of Simon C. Dik, Coordination, Lang. 45, 624-36.
Fillmore, Charles J., 1968a, 'The Case for Case', in Emmon Bach and T. Harms (eds.),
Universals in Linguistic Theory, New York.
Fillmore, Charles J., 1968b, 'Lexical Entries for Verbs', FoL. 4, 373-93.
Gleitman, Lila R., 1965, 'Coordinating Conjunctions in English', Lang. 41, 260-93.
Groot, A. W. de, 1959, 'Een nieuwe Nederlandse grammatica', De Nieuwe Taalgids 52,141-44.
Gruber, Jeffrey S., 1965, Studies in Lexical Relations, unpublished M.LT. Doctoral Dis-
sertation.
Gruber, Jeffrey S., 1967a, Functions of the Lexicon in Formal Descriptive Grammars, Technical
Memorandum 3770/000/00, Santa Monica, Cal. (System Development Corporation).
Gruber, Jeffrey S., 1967b, 'Disjunctive Ordering among Lexical Insertion Rules', unpublished
M.I.T. paper.
Hertog, C. H. den, 1903-1904, Nederlandsche spraakkunst; Handleiding ten dienste van aan-
staande (taal)onderwijzers, Amsterdam (2nd. edition).
lackendoff, Ray S., 1968, 'Quantifiers in English', FoL. 4, 422-42.
Jakobovits, L. and Steinberg, D. (eds.), t.a., Semantics: An Interdisciplinary Reader in Philos-
ophy, Linguistics, Anthropology, and Psychology, London.
Katz, Jerrold J., 1967, 'Recent Issues in Semantic Theory', FoL. 3, 124-94.
Katz, Jerrold J., 1970, 'Interpretative Semantics vs. Generative Semantics', FoL. 6, 220-59.
Klooster, W. G. and Verkuyl, H. J., 1972, 'Measuring Duration in Dutch', FoL. 8, 62-96.
Klooster, W. G., Verkuyl, H. J., and Luif, J. H.l., 1969, Inleiding tot de syntaxis, Culemborg/
Keulen.
Kraak, A., 1966, Negatieve zinnen; Een methodologische en grammatische analyse, University
of Amsterdam Doctoral Dissertation, Hilversum.
244 THE STRUCTURE UNDERLYING MP SENTENCES
I. John W. M. Verhaar (ed.), The Verb 'Be' and its Synonyms. Philosophical and Grammat-
ical Studies. Part I: Classical Chinese. Athapaskan. Mundari. 1967, VIII + 100 pp.
Dfl.23,-
2. Nicholas Rescher, Temporal Modalities in Arabic Logic. 1967, IX+50 pp. Dfl. 16,-
3. Tullio de Mauro, Ludwig Wittgenstein. His Place in the Development of Semantics. 1967,
VIII+62pp. Dfl.19,-
4. Karl-Otto Apel, Analytic Philosophy of Language and the Geisteswissenschajien. 1967,
X + 63 pp. Dfl. 16,-
7. Hugo Brandt Corstius (ed.), Grammarsfor Number Names. 1968, VII + 123 pp. Dfl. 32,-
8. John W. M. Verhaar (ed.), The Verb 'Be' and its Synonyms. Philosophical and Grammat-
ical Studies. Part III: Japanese. Kashmiri. Armenian. Hungarian. Sumerian. Shona. 1968,
VIII + 125 pp. Dfl. 28,-
9. John W. M. Verhaar (ed.), The Verb 'Be' and its Synonyms. Philosophical and Grammat-
ical Studies. Part IV: Twi. Modern Chinese. Arabic. 1969, VIII + 125 pp. Dfl.28,-
10. F. Kiefer (ed.), Studies in Syntax and Semantics, 1969, IX + 242 pp. Dfl.50,-
II. A. C. Senape McDermott, An Eleventh-Century Buddhist Logic of 'Exists '. 1969, X + 88 pp.
Dfl.25,-
12. Karl Aschenbrenner, The Concepts of Value, Foundations of Value Theory, 1971, XVII +
462 pp. Dfl. 100,-
15. H. J. Verkuyl, On the Compositional Nature of the Aspects, 1972, XIII + 185 pp. Dfl. 53,-
In Preparatioll:
13. F. Kiefer and N. Ruwet (eds.), Generative Grammar in Europe.
14. John W. M. Verhaar (ed.), The Verb 'Be' and its Synonyms. Philosophical and Grammatical
Studies. Part V: Urdu/Turkish/ Bengali/ Amharic/Indonesian/Telugu/ Estonian.
16. Charles H. Kahn, The Verb 'Be'in Ancient Greek.
SOLE DISTRIBUTORS IN THE U.S.A. AND CANADA:
Volumes 1-12: Humanities Press / New York