Perfect Adjective Positions in French: a Diachronic Perspective

Transcription

Perfect Adjective Positions in French: a Diachronic Perspective
Perfect Adjective Positions in French: a Diachronic Perspective
Paul Boucher, ALPL Université de Nantes
[email protected]
Introduction
In a series of recent talks and papers, Denis Bouchard (1998, 1999, 2002) has
challenged so-called ‘transformational’ analyses of adjective positions such as
Cinque (1993, 1995) or Bernstein (1993), which involve movement of N to the
head of some functional projection between D and N.
In Cinque’s analysis, the difference in the position of Ns relative to As in
Romance as opposed to Germanic is due to an N-raising mechanism, which
applies in the former but not in the latter. This analysis assumes a universal
basic order, derived from Kayne’s (1994) Linear Correspondance Axiom (LCA)
and places all As in pre-N positions. Adjectives are assumed to be ordered
across languages in a series of functional projections such as Size, Color,
Nationality. This hierarchy of A positions is determined by the selectional
properties of the various functional heads. Romance Ns are then assumed to
raise to an intermediate F° (since Romance NPs generally show A-N-A order)
to check some semantic feature of N against the features of F°. Germanic Ns
are argued to remain in situ, since they are claimed not to possess this feature,
thus deriving the surface order A (A) N.
Bouchard (2002) argues, contra Cinque, for an approach based on the
principle of ‘substantive motivation’, which posits that formal tools must find
their external motivation in cognitive and physiological properties that are
logically anterior to linguistic theory. A ‘pure’ or ‘perfect’ derivation in this
perspective is one in which semantic and syntactic combinations are
isomorphic. He proposes abandoning the Linear Correspondence Axiom
altogether in favor of a single operation, Associate, which assumes a fundamental asymmetry between Functor and Dependent elements and states that
‘the Functor precedes or follows the Dependent’. Arbitrariness, he argues, is
the ultimately economical system. The difference between French and English
word order, he claims, is due to the distribution of number morphology: Number
is on D in French; it is syntactic and thus scopes leftward; whereas it is a
bound affix in English and thus scopes rightward. This, as we’ll see below,
conditions the way As and Ns interact in the two languages.
In this paper, I will look at changes in adjective positions in French between
the 9th and the 16th centuries. At first glance, neither of the two approaches
sketched out above succeeds in explaining two fundamental facts about Old
French (OF):
a) attributive adjectives were mostly pre-N, but there was considerable
flexibility;
b) pre- versus post-N position did not correspond to a change in
meaning, contrary to Modern French.
Cinque’s approach, based on a N-movement operation for checking purposes,
has no explanation to offer for the fact that the vast majority of attributive
adjectives in OF were pre-N. What’s more, by focusing on abstract semantic
features linked to the various adjective categories to establish his functional
41
hierarchy, he sidesteps a crucial issue, that is, number agreement. The OF
nominal system included explicit marks of case, number and gender on N, D
and A, whereas Modern French marks number explicitly only on Ds and only
marginally on Ns, with As agreeing only in gender. The well-documented loss
of explicit number agreement on Ns during the Middle French period (cf.
Ménard (1973), Gougenheim (1974), Foulet (1967), among many others), in
fact coincides with a gradual shift of adjective positions from mostly pre-N to
mostly post-N, with a parallel semantic specialisation of each position. This
would seem to contradict the fundamental assumption that N-movement is
triggered by the need to check agreement morphology. It almost looks as if the
opposite were true.
Bouchard’s approach, on the other hand, places number agreement at the
center of the A-N word order problem. It thus offers an interesting perspective
on the link between loss of number marking on French Ns and changes in word
order. On the other hand, it does not explain why the Functor/Dependent
relationship should change over time, nor why the pre-N adjective position
should have become specialised for certain semantic relations.
In section one of the paper, I will look at each of these approaches in turn.
In section two, I will examine data from the earliest recorded texts in French up
until the Middle French period in order to reach a proper representation of
adjective positions in DP. Finally, in section three, I will attempt to offer a
satisfactory analysis of the phenomena referred to above, based on two
principles:
a) changes in Functor/Dependent order is linked to the position of
Tense in the clause;
b) loss of semantic content in a lexical item corresponds to change
in its distribution.
The first of these is predicted by Bartch & Vennemann’s (1972) Principle of
Natural Serialization. The second is the object of current work by Anna
Cardinaletti and Michael Starck (1999) on defective lexical items and the
relationship between semantic content and distribution.
1.
Two conflicting analyses of adjective positions
Cinque’s analysis has gained wide acceptance throughout the generative
syntax community, as demonstrated by the frequent references to it in college
linguistic textbooks1 as the “standard” analysis of adjective positions. The
success of this approach is due in part to the lack of viable alternative analyses
to date, as well as to its “naturalness” in the context of the verb-raising
hypothesis generally adopted since Pollock (1987). The only serious criticism
in recent years has come from Denis Bouchard, who criticizes Cinque’s
approach on a number of points (see Bouchard 1998 and 2002).
As mentioned above, Cinque assumes a universal, hierarchical order for
adjective positions, which is claimed to correspond to world knowledge notions
such as Size, Color, Nationality. This hierarchy is determined by selectional
properties of the functional categories. However, argues Bouchard, this
functional hierarchy is merely a descriptive list, not an explanation.
Adjectives are argued to appear in the Spec of each specific FP because
they agree with some semantic feature in the head and Nouns have some
42
particular feature which is also present in F° and which must be checked.
Movement of N somewhere in the vicinity of the functional category is assumed
to be the only way to check this feature. However, says Bouchard, the semantic
feature supposed to trigger N-raising is merely stipulated; no explanation is
given for why this and not some other feature triggers movement. There is a
second order feature [+strong] on this feature of N in French, but not in
English. This [+strong] specification has the property of forcing the feature to
be checked. The [+strong] nature of the feature that French, for instance, is
supposed to possess, as opposed to the [-strong] feature of English, for
instance, is again a mere stipulation, claims Bouchard.
One could also argue that a ‘transformational’ analysis is simply no longer
tenable in the Minimalist framework, which eliminates Move operations
altogether in favor of multiple Merger operations which insert copies at various
levels of the configuration. It also dispenses with Deep and Surface levels,
since only minimal requirements of legibility of the interfaces need be satisfied
(Chomsky 1999, 2000).
Bouchard (2002) argues that a more parsimonious application of Bare
Phrase Structure principles yields a more satisfactory analysis. Multiple Merger
operations, for instance, can be eliminated by abandoning the Linear
Correspondence Axiom altogether. In its place he proposes a single operation,
Associate, which subsumes three principles:
• There is a fundamental asymmetry in semantic combinations: in any
combination, one term is functor, the other dependent;
• There is a fundamental asymmetry in temporal juxtapositions: A either
precedes or follows B;
•
Linearization and semantic combination must be matched, so their
asymmetries must be matched.
This results in the Linearization Parameter which states that ‘the Functor
precedes or follows the Dependent’. Arbitrariness, again, is the ultimately
economical system for Bouchard.
At the root of this approach is the principle of ‘substantive motivation’, which
posits that formal tools must find their external motivation in cognitive and
physiological properties that are logically anterior to linguistic theory. A ‘pure’
or ‘perfect’ derivation, therefore, is one in which semantic and syntactic
combinations are isomorphic.
1.1. Perfect Adjective Positions
Bouchard’s most thoroughly worked out illustration of his theory concerns
adjective positions in Modern French. The difference between pre-N and postN adjective positions is explained in terms of the type of semantic relationship
holding between the ADJ and the N, where N is a functor and ADJ a
dependent.
• Common Ns are analysed as a network of interacting elements: the
characteristic function, f, the time interval, i, the possible world, w, and a
variable assignment function, g.
43
•
Pre-N adjectives are claimed to be ‘non-intersecting’2: ADJ modifies a
sub-element in the network of N. The association of a pre-N ADJ and a
noun forms a complex head: [AN]
(1)
b.
c.
d.
a.
le future president ‘the future president’: futur modifies the time
interval ;
de parfaits scélérats ‘perfect scoundrels’: parfait modifies the
characteristic function;
de faux diamants ‘false diamonds’: faux modifies the possible
world function;
ce présumé communiste ‘that alleged communist’: présumé
modifies the value assignment function.
• Post-N adjectives are ‘intersecting’ since they modify the whole network
of N. If N has already formed a complex head through association with a
pre-N adjective, then the post-N adjective intersects with this complex
head: [[AN]A].
(2)
•
•
a.
b.
c.
d.
le fils ainé ‘the eldest son’
un fait anachronique ‘an anachronistic fact’
un oiseau captif ‘a captive bird’
un homme marié ‘a married man’ and so on.
The Linearization Parameter applies to the unmarked case, that is, the
post-N adjective in ModFr, where the relationship between Functor and
Dependent is “whole-to-whole”. If it is a “whole-to-sub-element” relation,
then there is an ‘Elsewhere’ application of the LP: the order will be the
reverse of the setting for the language. So in ModFr, such Dependent
ADJs must precede the Functor Noun, rather than follow it. (Bouchard
2002)
Labelling is done by the full projection of the properties of the heads,
making these properties accessible at the phrasal level. Thus, for
instance, les lignes parallèles colorées would be labelled as follows.
DN
ru
DN
colorées
ru
DN
parallèles
ru
les
lignes
In this representation, D cliticizes to the first item to its right, in this case, N3.
This forms a complex functor category which is associated with two post-N
adjectives, each of which is a dependent. If N is first associated with a pre-N
adjective, then D cliticizes to this complex category, forming a new complex
category. Post-N adjectives are then associated with this.
44
DAN
ru
DAN
colorées
ru
DAN
parallèles
ru
les
AN
ru
petites
lignes
1.2. Number
The question is then, why isn’t this ‘perfect’ (because most economical) order
also universal? Why should it be different in English, for instance, or more to
the point here, in Old French? The first part of the answer concerns the role of
intonation.
In English, the difference between a ‘non-intersecting’ and an ‘intersecting’
adjective is made through stress placement. An intersecting adjective is
adjoined to the N head projection and gets independent stress (3a); while a
non-intersecting adjective forms a complex head with N and is de-stressed
(3b).
(3)
a.
b.
a SIMple SERvant = ‘un doméstique simple (d’esprit)’
a simple SERvant = ‘un simple doméstique’ (et rien d’autre)
The second part concerns the scope of Number in the two languages. The
difference between French and English word order is derived from the
difference in number coding in these languages. Number must be coded
somewhere on the nominal expression for reference, or ‘extensity’ (Wilmet
1986) to occur. In French, Number is coded on the Determiner. Since D in
French is a clitic, Number is a sort of ‘left-edge’ inflection, to quote Miller
(1991). As Number in French is expressed on a syntactic element, D,
government is rightward and Number has scope over the nominal expression to
its right, as we saw above.
In English, Number is coded on the Noun; it is a ‘right-edge’ inflection.
Since Number in English is a bound morpheme, Williams’ (1981) Right-Hand
Head Rule applies, and Number governs leftward. It thus takes scope over the
nominal expression to its left. This principle is termed the Number Scope
Condition by Bouchard (1998:161:35).
45
DAN
ru
those
AN
ru
small
AN
ru
parallel
AN
ru
colored
lines
According to Bouchard (2002), since one never has a “bare N” in English,
but always a [N+Num], a post-N Adj cannot modify the N alone, as it should. It
must appear in an Elsewhere position with respect to the Linearization
Parameter setting of English. The adjective must be pre-nominal in order to
“dip into” [N+Num] and modify the noun.
Before turning to the data on Old French in section two, we’ll briefly look at
two independent lines of research, which lend some indirect support to
Bouchard’s analysis. The first was developed by philologists working in the
1970’s. The second, more recent approach has been put forward by Anna
Cardinaletti and Michal Starcke in their work on the relationship between
substantive content and word order.
1.3. A Diachronic Perspective
Although fairly new to generative grammar, the ‘substantive motivation’
principle has long been familiar to logicians interested in natural language, like
Montague (1970), or philologists interested in word order change. Such
analyses can be traced back, via the Prague School (Daneš 1970, or
Trubetzkoy 1939) to Otto Behaghel’s Deutsche Syntax (1923) and even as far
back as Adelung’s Deutsche Sphrachlehre (1782).4 Vennemann (1974) argues
that “A theory of word order requires…a characterization of constituents as to
which specifies the other, or, in the language of logic, which is function and
which is argument, at the level of surface syntax”. (1974:342).
Lehmann (1972) reformulated Greenberg’s Universals with the aim of
correlating changes in the basic verb positions in the sentence with changes in
the positions of ‘specifier’ items relative to their ‘specified’ terms. He found that
in true verb-first languages like Arabic, where the Verb precedes its object
(VO), specifiers follow their arguments, so adverbs follow verbs, main verbs
follow auxiliaries or modals, adjectives or relatives follow nouns, standards of
comparison follow comparatives and NPs follow ad-positions. In true ‘verb-final’
languages (i.e. OV), like Japanese, the opposite is true. In this framework,
Bartsch and Vennemann (1972) worked out the Principle of Natural
Serialization as a theory of basic word order. This says that operators either all
precede or all follow their operands:
(4)
{ operator ({operand})} ⇒ { [operator [operand]] in XV languages
{ [[operand] operator] in VX languages
46
According to Vennemann, in order to test the validity of these principles “we
have to study word order change in the time dimension, diachronically; …word
order change provides perhaps the most powerful evidence for the universals
and the principle” (1974:349).
1.4. Cardinaletti & Starke
These authors develop a general theory of ‘structural deficiency’ and its effect
on word order, among other things, based on a careful analysis of strong and
clitic pronouns in Romance and Germanic. They first show that the traditional
bi-partite opposition needs to be enlarged to include a third type of pronoun:
weak pronouns. They show that the distinction between deficient (weak or
clitic) pronouns and non-deficient (strong) pronouns correlates with a regular
range of semantic (i.e. referential) oppositions, as well as with different
distributional, morphological, and prosodic patterns. For instance, deficient
pronouns tend to be “morphologically a proper subset of the corresponding
(5)
STRONG
WEAK
je-ho
he-mu
ho
mu
‘him’ Slovak ‘to-him’ SL
a loro
loro
‘to-them’ Italian
They also tend to have different distributions:
(6)
a.
b.
c.
Non gli dirò mai *gli tutto *gli
Non *loro dirò mai loro tutto *loro
Non *a lui dirò mai *a lui tutto a lui
not to-him I-will-say never everything
Clitic
Weak
Strong
Italian
As for semantics, C&S argue that strong pronouns have a referential index, a
range specification and a [± human] interpretation, whereas deficient pronouns
lack all three.
In section three of their article, C&S envisage extending their analysis to
other grammatical categories, arguing that ideally the same tripartition,
correlated with the same set of asymmetrical properties, should apply to
adverbs, verbs, adjectives, nouns, prepositions and so on. They give a few
examples of deficient adverbs to illustrate their claim, one of which we’ll
reproduce here.
The French adverb bien has two meanings and two corresponding
distributions (respectively illustrated by (7) and (8) below), which correlate with
the (in)ability to be coordinated, c-modified or focused, that is, with the classic
tests for clitic status developed by R. Kayne (1975).
(7)
(8)
Il a bien essuyé la vaisselle.
‘he has indeed dried the dishes’
Il a très bien essuyé la vaisselle.
‘he dried the dishes very well’
47
The ‘indeed’ version of bien cannot be moved to VP-final position (9a), nor be
coordinated (9b), modified (9c), nor focused (9d), contrary to the ‘very well’
version.
(9)
a. *
b. *
c. *
d. *
Il a essuyé la vaisselle bien.
Il a bien et vite essuyé la vaisselle.
Il a très bien essuyé la vaisselle.
Il a BIEN essuyé la vaisselle.
C&S argue that the strong/deficient dichotomy applies here and explains
straightforwardly the differences in stress, distribution and so on. The key
point, as far as we are concerned here, will be seen to be the difference in
semantic content. The ‘strong’ adverb, that is the ‘very well’ version of bien,
contains semantic features of manner, while the ‘deficient’ adverb, that is, the
‘indeed’ version, does not. This parallels the presence or absence of a
[+human] feature, which distinguishes strong and clitic pronouns. I will argue
below that just such a distinction explains the differences in distribution of
Adjectives in Modern French and explains, along with the loss of overt phi
features on Ns and As over time.
2.
Old French adjective order
2.1. The ‘old stage’ of the language
The oldest written document in French, Les Serments de Strasbourg (c. 842),
gives us a glimpse of what Yvon (1946-47) calls ‘l’état ancien de la langue’5
Louis-le-Germain and Charles-le-Chauve, grandsons of Charlemagne, swear
an oath of mutual fidelity, each addressing the other’s soldiers in their
vernacular tongue. The Romance half of the text is written in the lingua rustica
vel romana, or Vernacular Romance, though the somewhat stilted and artificial
phraseology is certainly not representative of the spoken language of the
period.6
The text7:
(10) Pro Deo amur et pro Christian poblo et nostro commun salvament,
‘for the love of God and for the salvation of the Christian people and for our own’
(11) d’ist di en avant, in quant Deux savir et podir me dunat,
‘from this day onward, to the extent that God gives me the knowledge and the power
(to do so),
(12) si salvarai eo cist meon fradre Karlo, et in aiudha et in cadhuna cosa,
‘(so) I’ll stand by ‘(this) my brother Charles, and with help and in all matters,
(13) si cum om per dreit son fradra salvar dift, in o quid il mi altresi fazet,
‘just as by law one should stand by one’s brother, on condition that he do as much for
(14) et ab Ludher nul plaid num quam prindrai qui meon vol cist meon fradre
Karle in damno sit.
‘and I will, of my own free will, make no agreements with Lothar which might be
detrimental to (this) my brother Charles’
(15) Si Lodhuvigs sagrament que son fradre Karlo jurat conservat, et Karlus
48
‘If Louis keeps the oath which he has sworn to his brother Charles, and if Charles’
(16) meos sendra de suo part non lostanit, si jo returnar non l’int pois, ne jo
ne neuls
‘my lord, on his part does not keep his, if for some reason I cannot prevent him from
doing so, neither myself, nor any of those’
(17) cui eo returnar int pois, in nulla aiudha contra Lodhuuvig nun li iver.
‘whom I can prevent from doing so, will give him any help in opposing Louis.’
Syntactic analysis
•
•
•
•
•
Verbs are all in final position: dunat (11); fazet (13); prindrai (14); sit
(14); conservat (15); jurat (15); lostanit (16); with one exception: a V2
construction: si salvarai eo (12);
Verbs precede auxiliaries or modals: salvar dift (13); returnar pois (16,
17);
Objects precede verbs: son fradra salvar (13); altresi fazet (13); nul plaid
prindrai (14); etc.
Adverbs precede verbs: num quam prindrai (14);
in nulla
aiudha…nun…iver (17) ;
Adjectives and genitive attributes precede nouns: Deo amur (10);
Christian poblo (10); commun salvament (10).
So, Bartsch and Vennemann’s (1972) Principle of Natural Serialization is
respected, as far as one can tell from such a short text. This seems to
correspond to the findings of other workers. Marchello-Nizia (1995) discusses
word order in Old French at some length (pp. 35-114). She quotes (p. 43)
studies by E. Richter (1903) and R. De Dardel (1989)8 showing that in ProtoRomance, OVS seemed to be the preferred word order, and gives an example
from Italo-Romance:
(18) Se pareba boues, alba pratalia araba
‘He prepared the oxen, he plowed the white fields’
(19) albo uersorio teneta, Negro semen seminaba
9
‘he held the white plow, he sowed the black seed’
She quotes G. Serbat (1980) on Latin: “The most striking characteristic of word
order (in Classic Latin) is undoubtedly the fact that the verb is usually found in
the clause-final position.”10 She also points out (pp. 44-45) that the OV order
was possible in all of the Old Romance languages up until the 16th century.
2.2. A Transitional Stage
Although written less than forty years later, La Cantilène de Sainte Eulalie (c.
878) clearly represents a more advanced stage of the language and can be
seen as a transition between Gallo-Romance and Classic Old French.
Vennemann (1974) argues that in the transition from verb-final to SVO, French
went through a long intermediate TVX period, by which he means that the verb
occupied the second position (V2) of the sentence, following an initial topic (T),
which could be the subject, and followed by some other element (X), which
49
could be the object. This text shows a majority of V2 constructions, with a few
traces of the old stage.
The text11 : La Cantilène de Sainte Eulalie (ca 878)
(20) Buona plucella fut Eulalia
‘Eulalie was a fair maiden’
(21) Bel avret corps, bellezour anima.
‘She had a beautiful body and an even more beautiful soul’
(22) Vouldrent la veintre li Deo inimi,
‘The enemies of God wanted to vanquish her’
(23) Voldrent la faire diaule servir.
‘They wanted to make her serve the Devil’
(24) Elle no’nt eskoltet les mals conselliers
‘She didn’t listen to these evil counselors’
(25) Qu’elle Deo raneiet, chi maent sus en ciel
‘who asked her to deny God who dwells in heaven above’
(26) Ne por or ned argent ne paramenz
‘Not for gold, nor silver, nor for jewelry,’
(27) Por manatce regiel ne preiement.
‘not for all the threats and pleadings of the King’
(28) Niule cose nonla pouret omque pleier
‘Nothing could ever make her yield’
(29) La polle semptre non amast lo Deo menestier.
‘nor bring this damsel not to love the service of God forever’
(30) E por o fut presente de Maximiien,
‘For this (reason) she was presented to Maximian’
(31) Chi rex eret a cels dis soure pagiens.
‘who was (the) king of the pagans at that (the) time’
(32) Il li enortet, dont lei nonque chielt, Qued elle fuiet lo nom chrestiien.
‘He ordered her (to give in) but it pleased her little to give up the title of Christian.’
(33) Ell-ent adunet lo suon element
‘ She gathered her forces’
(34) Melz sostendreiet les empedementz Qu’elle perdesse sa virginitét;
‘She preferred submitting to (the) torture, rather than lose her virginity.’
(35) Por os furet morte a grand honestét.
‘That’s why she died with great honor.’
(36) Enz enl fou lo getterent com arde tost.
‘They threw her into the fire so that she would burn up quickly.’
(37) Elle colpes non avret, por o noscoist.
‘But since she had committed no sins, she burned not at all.’
(38) A czo nos voldret concreidre li rex pagiens.
‘The pagan king couldn’t accept that.’
(39) Ad une spede li roveret tolir lo chieef.
‘He ordered them to cut off her head with a sword.’
(40) La domnizelle celle kose non contredist :
‘The girl didn’t complain about this (thing).’
(41) Volt lo seule lazsier, si ruovet Krist.
‘She wanted to leave the world; so she prayed to Christ.’
(42) In figure de colomb volat a ciel.
‘In the form of a dove, she flew up to the sky (heaven).’
(43) Tuit oram que por nos degnet preier
‘Let us all pray that she deign intercede for us’
(44) Qued auuisset de nos Christus mercit Post la mort
50
‘so that Christ should take pity on us after death’
(45) et a lui nos laist venirPar souue clementia.
‘and let us come to Him by his mercy.’
Verb position:
Attribute
Buona pucella
Bel
Bellezour
Rex
Verb
fut
avret
( )
eret
Subject
Eulalie (20)
corps (21)
anima (21)
pro (31)
Adverbial
Pro o
Enl fou
Por os
Por os
Ad une spede
A czo
In figure de colomb
Melz
Verb
fut presente
le getterent (33)
furet
noscoist (34)
li roveret tolir
nos voldret concreidre
volat
sostendreiet
X
de Maximiien (30)
Subject
Elle
La polle
Ellelle
elle
pro
chi
Tuit
Verb
no’nt eskoltet
non amast
ent adunet
fuiet
perdesse
ruovet
maent
oram
Object
le mal conselliers (37)
lo Deo menestier (29)
lo suon element (33)
lo nom chrestiien (32)
sa virginitét (34)
Krist (41)
sus en ciel (25)
que…(43)
Subject
Elle
La domnizelle
Niule cose
Il
(dont)
( )
( )
Object
colpes
celle kose
non la
li
lei nonque
por nos
a lui nos
Verb
non avret (39)
non contredist (40)
pouret omque pleier (28)
enortet (32)
chielt (32)
degnet preier (43)
laist venir (45)
Verb
Vouldrent
Voldrent
Volt
(Qued) auuisset
Object
Subject
la veintre
li Deo inimi (22)
la faire diaule servir ( ) (23)
lo seule lazsier (41)
de nos
Christus mercit Post la mort (44)
morte (35)
lo chieef (39)
li rex pagiens (38)
a ciel (42)
les empedementz (34)
Out of a total of 31 clauses containing tensed verbs, 20 are V2 constructions, 4
are topicalisation constructions with the VO unit moving to the clause initial
51
position and 7 place the verb in final position. In other words, alongside older
verb-final constructions, which may have taken on a stylistic value at this point,
since they persist throughout the classic Old French period in poetic texts, the
overwhelming majority of clauses are of the V2 type, which would correspond
to Vennemann’s description of Old French. As for the other, ‘specifierargument’ constructions, we find:
All auxiliaries and modals precede verbs:
Vouldrent…veintrei (22); voldrent…faire…servir (23); pouret…pleier (28) ; fut
presente (30); voldret concreidre (38); volt…lazier (39);
Most but not all verbs precede objects (except for clitic pronoun objects):
eskoltet le mal conselliers (24)
amast lo Deo menestier (29)
fuiet lo nom chrestiien (32)
adunet lo suon element (33)
perdesse sa virginitét (34)
sostendreiet les empedementz (34)
tolir lo chieef (39)
ruovet Krist (41)
However, attributes precede verbs, as seen above. Otherwise, two nonpronominal objects precede their verbs; this may be due to stylistic constraints,
as suggested above.
colpes non avret (39)
celle kose non contredist (40)
Likewise, adjectives, possessives and genitive attributes may precede or
follow their nouns:
mals conselliers (24); buona pucella (20); Deo inimi (22); Deo menestier (29);
grand honestét (35); arde tost (36); suon element (33) ; sa virginitét (34) ;
souue clementia (45) ;
But:
nom chrestiien (32); rex pagiens (39); manatce regiel (27); figure de colombe
(42);
Determiners – articles, demonstratives, quantifiers – precede nouns :
lo suon element (33); lo chieef (39); celle kose (40); une spede (39); cels dis
(31); etc.
We see a clear change from OV to VO order, though the old order can still be
found, possibly as a stylistic variant. As a general rule, the ‘operator-operand’
order also changes, though it is far from systematic and again, it’s hard to
judge the nature of this stage of the language from such a small sample.
Adjectives and genitive attributes may now be found in either pre- or post-N
position without any obvious change in meaning.
2.3. Classic Old French
52
The fact that the best-known texts of the ‘classic’ Old French period (late 11th
century to early 13th) are written in verse makes it difficult to determine the
natural position of adjectives in the spoken prose of the period. Word order
often changes for stylistic or rhyming purposes. On the other hand, a sample
taken from any text of the period yields approximately the same results, so
there was certainly a limit to the types of word order variation allowed. The
main patterns seem to have been:
A. (Determiner) - Adjective - Noun
des joenes genz; male creature; novele robe; bones oevres; povres chiens
(Roman de la Rose) ; blance image ; grant ennui et grant contrere ; fox
garçons ; bele suer ; caude eve ; bon cuer ; loiale espose ; virges poules ;
hardie beste (Roman de Renart) ; bone eloquence ; granz biens ; grevose
oevre ; nobles reis ; un vert chemin ; de blanc ivoire (Lais) ; etc.
B. (Determiner) – Noun - Adjective
maintes colours dyverses ; tens amoreus ; robe neve ; icele saison novelle ;
une perche graillette (Roman de la Rose); fer caut : un cardon asnin ; beste
fole ; fosse desfensable ; (Renart); tens anciënur ; une fille bele ;l’erbe drue ;
sa face vermeille (Lais) ; etc.
C. (Determiner) – Adjective – Noun – (et) Adjective
les douz chanz piteux ; noble chose et precieuse ; un seul sac estroit (Rose) ;
un mervellos mangier françois (Renart) ; bels sire chiers (Lais) ; etc.
D. (Determiner) – Adjective (et) – Adjective – Noun
baux tres doz sire (Renart) ;
E. (Determiner) – (Adjective) – Noun – Adjective (et) Adjective
l’eve grant et roide ; ymage maigre et chaitive ; la color pale et morte ; les
autres estudes droituriers et honestes (Rose) ; cest bon miel fres et novel
(Renart) ; uns vielz prestre blans e floriz (Lais) ; etc.
F. Other combinations
Clere estait l’eve et aussi froide ; clere et saine et bele estoit la matinee et
atrempee ; (Rose)
Goyens (1994) has drawn up statistics for 2493 NPs in a late 13th century
translation of a Latin prose text. Out of 323 NPs containing a modifying
adjective or participle, I found the following:
• 219 modifiers were in pre-N position (pattern A above);
•
45 were in post-N position (pattern B);
•
33 were in pre-N and post-N position (pattern C);
•
25 were in coordinated post-N positions (pattern E);
•
1 was a coordinated pre-N position (pattern D).
53
The same text translated into Modern (19th Century) French yielded the
following figures for a total of 339 NPs containing a modifying adjective or
participle:
•
39 modifiers were in pre-N position (pattern A);
• 254 were in post-N position (pattern B);
•
3 were in both pre and post-N positions (pattern C);
•
43 were in coordinated post-N positions (pattern E).
It is clear that whatever criteria are used to oppose the two positions, whether
semantic – intersecting and non-intersecting (Bouchard) or restrictive and
syllabic – mono versus polysyllabic (Marchello-Nizia)
categorial – adjective versus participle (Moignet, Foulet), there is an
overwhelming move from pre-nominal to post-nominal position for modifiers of
nouns as we go from Old to Modern French. In the next, and final section of the
paper, I will attempt to explain this change in word order.
4.
Analysis
In the first part of my analysis, I will look at the evolution of number agreement
in French and its influence on changes in word order. In the second part, I will
discuss Cardinaletti & Starcke’s ‘structural deficiency’ hypothesis as a way to
explain the semantic differences between pre-N and post-N adjectives.
4.1. ‘Semi-strong’ agreement morphology
While Old French retained two of the Latin cases, the cas sujet, or ‘subject
case’, and the cas régime, or ‘object case’, the actual agreement affix, -s, which
expressed both case and number, was confusing and intermittent. Only
masculine nouns and determiners were marked for case as well as number,
while feminine nouns were only marked for number. Moreover, the affix –s was
used in both cases:
(46)
(47)
li cuers (‘the heart’): Masculine, Nominative, Singular
* le cuer (‘the heart’): Masc; Accusative, Sg
li cuer (‘the hearts’): Masc; Nom, Plural
* les cuers (‘the hearts’): Masc; Acc, Pl
la dame (‘the lady’): Fem, Nom, Sg
la dame (‘the lady’): Fem, Acc, Sg
les dames (‘the ladies’) Fem, Nom, Pl
les dames (‘the ladies’) Fem, Acc, Pl
As this short list shows, it was only possible to determine the number of the NP
in OF if both the determiner and the noun were present. This situation
corresponds to what I’ll call ‘semi-strong’ agreement morphology. During the
course of the OF period, case morphology weakened and was lost. Over the
Middle French period, the number affix ceased to be pronounced, though it is
still written and can be reactivated in liaison contexts.
54
4.2. Number Scope in OF
We have seen that Old French tended to place adjectives in pre-N position,
while Modern French places them most often in post-N position. This makes
OF look like English, which is understandable if there is an overt number affix
on nouns in OF, as there is in English. However, we have also seen that there
was considerable flexibility for the adjective position in Old French, which has
never been the case for English. This is perhaps an inheritance from Latin,
which allowed considerable freedom in word order. But it may also be due to
the fact that OF, contrary to both Modern French and English had overt number
morphology on both determiner and noun, though of a somewhat sporadic and
confusing nature. I think that in fact, the key to the solution lies in the
combination of these two factors:
•
•
the ‘semi-strong’ nature of the agreement morphology precluded a rigid
Functor-Dependent order, while at the same time:
the double coding of number in OF meant that any given combination of
adjectives and nouns would always fall under the scope of a number
marker, either a syntactic item to its left, or a morphological item to its
right.
In such a system, it seems that ‘expressivity’ took precedence over the type of
strictly semantic constraints described by Bouchard. That is, positioning an
adjective to the right or left of the noun in OF resulted, not in a change in
meaning, but in a more or less stylistically expressive value. The intersectivenon-intersective opposition discussed by Bouchard for Modern French seems
to have developed only once the agreement morphology on the noun had
disappeared. Another factor in the change, discussed by Marchello-Nizia
(1995), may have been the loss of word stress in the transition from Old to
Middle French. In the Modern French oxytonic system, there is simply no other
way to express such semantic differences.
As for why English, which also evolved from an OV to a VO language, did
not also change its Adjective-Number order, Bouchard suggests that, again, it
is Number coding which provides the solution. English has retained a strong
number affix, as mentioned above. Thus, in Bouchard’s terms, there are never
‘bare Ns’ in English, but always complex [N+NUM] heads. In other words, the
number affix always forms a complex functor category with the noun in English.
Adjectives must be placed in pre-N position in order to intersect, partially or
completely, with this complex head, leaving intonation to indicate which type of
semantic relation is intended.
Finally, as to why Adjective-Noun order changed in Modern French, and
why the pre-nominal position came to be associated with ‘non-intersecting’
adjectives, I would suggest two directions of investigation.
4.3. Loss of inflection
First of all, it seems clear that the loss of case morphology on the determiner
and the eventual loss of a strong number affix on nouns in French contributed
55
to the change. One clear indication of the effects of this loss on word order can
be seen in the disappearance of possessive pronouns between D and N at the
end of the Middle French period. As the following examples from the two texts
given above suggest, possessives could (and quite often did) occupy a post-D
position.
(48) a
b
cist meon fradre Serments : (15)
‘this my brother’
lo suon element Cantilene : (36)
‘the her forces’
This type of construction is found today across Romance languages except for
French.
(49) a
b
la mia amica (Italian)
‘the my friend’
la meva casa (Catalan)
‘the my home’
(Examples from Longobardi 1996:29:72)
Adopting the analysis given in Haegeman & Guéron (1999:433), I propose that
the OF possessive occupied the Specifier position of an AgrP between D and
NP.
(50)
DP
ru
D
AgrP
ru
lo/cist
Spec
Agr'
ru
meon/suon Agr
NP
ru
N
fradre/element
Adjectives, which agreed in number and gender with their nouns, could then
occupy a second, lower [Spec, Agr] position.
(51)
DP
ru
AgrP1
ru
D
cist
Spec
nostre
Agr'
ru
AgrP2
ru
Agr
Spec
56
Agr'
ru
commun Agr
NP
salvement
This is essentially the situation prevailing today in Spanish, Italian or
Portuguese. In French, the gradual loss of agreement morphology on the
adjective can be thought to have lead to the weakening and subsequent
disappearance of the lower agreement projection. Adjectives subsequently
occupied the post-N position more and more frequently in order to fall under
the scope of the only remaining marker of number, the determiner.
By correlating Bartsch and Venneman’s Principle of Natural Serialization
and Bouchard’s Number Scope Condition we could perhaps further conjecture
that, as the position and/or strength of functional heads such as Tense°,
Number° or Case° vary in time, the overall economy of the clause is affected.
This seems not only to constrain operator–operand order across the board, as
we saw was the case in the transition from the state of the language
represented by the Serments to the one corresponding to the Cantilene, but
may also trigger the emergence of overt operators such as determiners or overt
pronominal subjects where none existed before. It is certainly not a simple
coincidence that as French lost its agreement morphology on the noun, it also
lost the ability to have referential Bare NPs, as well as ceasing to be a pro-drop
language.
In other words, the nature and position of the tense operator in the clause
may constrain the orientation of the other ‘operators’ of mood, reference,
quantity, or definiteness. As the position of T changes over time, so eventually
do those of C, D and also various adjectives, adverbs and other ‘modifiers’ of
lexical heads. This type of movement seems to be involved in the changes
observed for French, though much further research is needed before any
definitive statement can be made.
4.4. Loss of semantic content
The processes of change described in the preceding section do not, in
themselves, suffice to account for the A-N-A order characteristic of Modern
French, however. If it were just a question of a change in ‘operator-operand’
order, then we would expect French to be the mirror image of English and
maintain a strict N-A order. This is not the case, as we have seen. As we have
also seen, pre-N adjectives are semantically specialized for a ‘non-intersective’
meaning. I would like to argue that in fact a process similar to semantic
bleaching eventually lead to the creation of a set of ‘structurally deficient’
adjectives, in the sense of Cardinaletti and Starcke 1999. Indeed, the pre-N
adjectives described by Bouchard as being ‘non-intersective’ show all the
characteristics of the ‘weak’ pronouns and adverbs these authors discuss.
•
Semantic specialisation: pre-N adjectives modify a sub-element in the
network of N, as opposed to post-N adjectives, which modify the whole
network of N;
57
•
•
Distribution: ‘defective’ adjectives are only found to the left of N, with
which they form a complex head; this pattern of ‘raising’ and
‘incorporation’ is typical of defective items like clitics;
Absence of coordination, modification or focalisation:
(52) a. * un pauvre et petit homme
un homme pauvre et petit
(53) a. * un très pauvre homme
b
un homme très pauvre
(54) a * un PAUVRE homme
b
un homme PAUVRE
Again, the (a) patterns are typical of clitics and other defective items, while the
(b) patterns are typical of strong, or non-defective items. I would like to claim
then that the A-N-A pattern of French is not due to an N-raising mechanism as
claimed by Cinque, but rather to a series of converging phenomena which took
place over several centuries: the loss of agreement morphology, the acrossthe-board change of operator-operand order and finally the gradual loss of
semantic structure in certain adjectives, which lead them to become structurally
defective and so to occupy special, raised positions where they incorporate into
the N head.
5.
Conclusion
The analysis of word order change in French determiner phrases over time has
lead us to call into question the ‘standard’ analysis of Adjective-Noun positions
in Romance languages. We have discussed the advantages but also the limits
of an alternative analysis proposed by D. Bouchard, based on the principle of
‘substantive motivation’. We have seen the importance of considering the loss
of inflection morphology, not only throughout DP, but across the major
functional head positions. Finally, we have considered the role of ‘structural
deficiency’ in determining changes in word order. To my knowledge, the
hypotheses of Cardinaletti & Starcke have never been applied to diachronic
data before, and so the present paper offers a novel approach to a number of
difficult problems currently being investigated in the generative syntax
framework.
References
Abney, Stephen. 1987. The English NP in its Sentential Aspect. Phd Thesis:
MIT.
Bartsch, Renate and Theo Vennemann. 1972. Semantic structures: a study in
the relation between semantics and syntax. Athenäum-Skripten
linguistik, 9, Frankfurt am Main: Athenäum.
Bernstein, Judy. 1993. Topics in the syntax of nominal structure across
Romance. PhD Dissertation, City University of New York, New York.
58
Bouchard, Denis. 1998. The distribution and interpretation of adjectives in
French: A consequence of Bare Phrase Structure. Probus 10, 139-183.
Bouchard, Denis. 1999. Cliticization and Universal Grammar. Talk at
l’Incontrato Grammaria generativa, Sienna, Italy, February, 1999.
Bouchard, Denis. 2002. Perfect tools: from functional motivation to substantive
motivation. Talk at GLOW-2002, Amsterdam and Utrecht, April, 2002
Boucher, Paul. 2002. Determiner Phrases in Old and Modern French, in From
NP to DP, Volume I: The syntax and semantics of noun phrases, M.
Coene and Y. D’hulst (eds.), 47-70, Amsterdam: Benjamins.
Boucher, Paul. (To appear). Definite Reference in Old and Modern French: The
rise and fall of DP. In Grammaticalization and Parametric Change, F.
Roca (ed.), Oxford University Press.
Cardinaletti, A. and Michal Starcke. 1999. The typology of structural deficiency.
In Henk van Reimsdijk (ed.) Clitics in the Languages of Europe. Mouton
de Gruyter, 145-233.
Chomsky, Noam. 1995. The Minimalist Program. MIT Press
Cinque, Guglielmo. 1993. On the evidence for partial N movement in the
Romance DP. ms, Università degli studi di Venezia.
Cinque, Guglielmo. 1995. On the evidence for partial N-movement in the
Romance DP. Italian Syntax and Universal Grammar. Cambridge
University Press.
de Murçin. 1815. Serments prêtés à Strasbourg en 842 par Charles-le-Chauve,
Louis-le-Germanique et leurs armées respectives, Paris : Didot l’Ainé.
Delfitto, Denis and Schroten, Jan. 1991. Bare Plurals and the Number Affix in
DP. Probus 3, 2, 155-185.
Foulet, Lucien. 1977. Petite syntaxe de l’ancien français. Paris : Champion.
Goyens, Michèle. 1994. Emergence et evolution du syntagme nominal en
français. Bern.
Haegeman, Liliane & Jacqueline Guéron. 1999. English Grammar: a generative
perspective. Blackwell.
Harris, Martin. 1977. A Typological Approach to Word-Order Change in
French. In Romance Syntax: synchronic and diachronic perspectives,
ed. M. Harris, 33-53, Salford.
Jensen, Frank. 1990. Old French and Comparative Gallo-Romance Syntax.
Tübingen: Niemeyer.
Kayne, Richard, 1975. French Syntax: the Transformational Cycle. Cambridge,
Mass. MIT Press.
Kayne, Richard. 1994. The Antisymmetry of Syntax. MIT Press.
Kleiber, Georges. 1990. Sur le démonstratif de notoriété en ancien français.
Revue québécoise de linguistque. 19. 11-32.
Langford, Chad. 2001. Perspectives diachroniques sur la position de l’adjectif
épithète en français. Talk at the International Colloquium
français et à travers les langues. Caen, Juin 2001.
Lehmann, Winfred. 1973. A structural principle of language and its
implications. Language, 49, 47-66.
Marchello-Nizia, Christiane. 1992. Histoire de la langue française aux XIV° et
. Paris. Dunod.
Marchello-Nizia, Christiane. 1995. L’évolution du français: ordre des mots,
démonstratifs, accent tonique. Paris: Colin.
59
Ménard, Philippe. 1994. Syntaxe de l’ancien français. Bordeaux. Bière.
Miller, Philip. 1991. Clitics and Constituents in Phrase Structure Grammar. Phd
Thesis, Rijksuniversiteit te Utrecht.
Moignet, Gérard. 1973. Grammaire de l’ancien français. Paris : Klincksieck.
Pollock, Jean-Yves. 1997. Langage et cognition. Presses Universitaires de
France.
Radford, Andrew. 1997. Syntactic Theory and the Structure of English.
Cambridge University Press.
Schroten, Jan. 1991. Interaction between morphological structure and logical
form interpretation: bare plural nouns in Spanish and Dutch. Recherches
de Linguistique française et romaine de l’Université d’Utrecht 10, xxx, 2939.
Schroten, Jan. 2001. L’absence de déterminant en espagnol. In Typologie des
groupes nominaux. Eds.Georges Kleiber, Brenda Laca et Liliane
Tasmowski, 189-204, Presses universitaires de Rennes 2.
Vennemann, Theo. 1974. Topics, subjects and word order : from SXV to SVX
via TVX. 339-375, in Historical Linguistics, Vol. 1; J. Anderson and
Charles Jones (eds), North Holland
Williams, Edwin. 1981. On the notions ‘lexically related’ and ‘head of a word’.
Linguistic Inquiry, 12, 245-274.
Wilmet, Marc. 1986.La détermination nominale. Paris : Presses universitaires
de France.
Yvon, Henri. 1946-47. L’assiette du nom des origins au XIV° siècle. Romania
49, 289-316.
Notes
1
See for instance the discussions in Haegeman & Guéron, 1999, pp.422-430, 453-464;
Radford, 1997, pp.157-158; or Pollock, 1997, pp.212-213.
2
Or more exactly: “they ‘intersect’ with a sub-element of N” (D. Bouchard p.c.). Bouchard
admits that this use of the term ‘intersect’ is a bit far from the usual (set-theoretic) notion of
intersection.
3
D is analysed as a clitic in French by Bouchard, following Miller (1991).
4
Daneš, František. 1970. “One instance of Prague school methodology: functional analysis of
Method and Theory in Linguistics. Paul Garvin (ed), 132-146. The
Hague: Mouton. Trubetzkoy, Prince N. 1939. “Le rapport entre le déterminé et le déterminant
in Mélanges de linguistique offerts à Charles Bailly, 75-82. Genève : Georg. See Vennemann
1974 for discussion.
5
Yvon says: « Cet état qui rappelle exactement celui du Latin peut être considéré comme l’état
ancien. » ( 1946-47 :290)
6
The text was written down before being spoken aloud by Louis-le-Germain and depends
heavily on standard Latin oath-taking ceremonial formulas of the time. See de Murçin 1815 for
examples of equivalent Latin texts of the same period and discussion.
7
English translation based on the Modern French translation in de Murçin 1815.
8
Dardel, Roland de. 1989. L’hypothèse d’une base OVS en protoroman. In Probus 1/1? 121143. Richeter, Elise. 1903. Zur Entwicklung der romanischen Wortstellung aux der
Lateinischen. Halle.
9
From the Enigma de Verone, 8th-9th centuries, in Monteverdi, Testi volgari italiani dei primi
tempo
60
10
“Le trait le plus frappant dans l’ordre des mots (en latin classique) c’est sans doute que le
verbe est le plus souvent à la fin de la proposition. Serbat, Guy. 1980. Les structures du
Latin. Paris : Picard, p. 141.
11
The text as well as the Modern French translation I based my translation on can be found at:
http://www.restena.lu/cul/BABEL/T_CANTILENE.html.
61