Towards a unified account of adjectival agreement in French


Jane Shelton

University of Newcastle upon Tyne
(received March 1997, revised May 1997)
[Download this article (98K, Rich Text Format)]

Copyright Notice:
First published in Web Journal of Modern Language Linguistics in association with the publishers (to be announced). © 1997 Jane Shelton.
The moral rights of the author(s) to be identified as author(s) of this work are asserted in accordance with §§.77 and 78 of the Copyright, Designs and Patents Act 1988. This work may be reproduced without the consent of the author, in part or in whole in any manner and in any medium subject only to the two following conditions:
(a) no charge shall be made for the copy containing the work or the excerpt,
(b) a copy of this notice shall precede the work or the excerpt.

Grammatical gender and its associated agreement, whereby target[footnote 1] elements exhibit overt gender agreement with a controller element in various structural domains, is a widespread phenomenon in human language. In comparison to English which has three grammatical genders which can be strongly linked to natural gender distinction and has only a pronominal agreement system (see Corbett 1991), French has a dual gender system based mainly on formal phonological and morphological criteria (Mel'cuk 1974; Tucker et al. 1968) and a much wider range of target elements and structural domains in which agreement occurs. This article focuses on three target elements which exhibit gender agreement in French; determiners, past participles and adjectives, as exemplified in (1) and suggests an analysis of adjectival agreement which brings it into line with analyses already proposed in the literature for determiners and past participles, thus providing a more unified account of agreement in French.

(1a) La petite fille / le petit garçon
The(f) little(f) girl(f) / the(m) little(m) boy(m)

(1b) La fille intelligente / le garçon intelligent
The(f) girl(f) intelligent(f) / the(m) boy(m) intelligent(m)

(1c) La fille est intelligente / le garçon est intelligent
The(f) girl(m) is intelligent(f) / the(m) boy(m) is intelligent(m)

(1d) La boîte que j'ai ouverte est ici
The(f) box(f) which I have opened(f) is here

In spite of the pervasive nature of gender agreement in languages, theoretical interest in this phenomenon from the point of view of incorporating agreement facts into current linguistic theory does not date back very far. Prior to the late 1970s the prime attraction of grammatical gender was in its origin (Wheeler 1898, Fodor 1959, Ibrahim 1973, Greenberg 1978 etc.) or in the classification of nouns within a gender system (e.g. Parsons 1960, Givón 1970 Guthrie 1948, Baetens-Beardsmore 1971) although some descriptive works on agreement in various languages did surface (e.g. Greenberg 1966). Languages like French, however, with comparatively simple gender systems, did not receive a great deal of attention. Descriptive and prescriptive grammarians alike tended to (and, indeed, still do tend to) treat grammatical gender in French as a peripheral, accidental phenomenon meriting only passing attention. This passing attention has typically been in the form of lists of structures where agreement is found, with varying degrees of comment on the obligatory or optional nature of agreement in these structures. Thus are found statements such as "Adjectives agree with the noun" or "Past participles in the passive structure must agree in gender with the subject of the passive sentence" and so on. On the whole, grammarians have been content with what is essentially a stipulative account of agreement and little attempt has been made to either link structures together in a uniform account of agreement or to account for agreement per se.

There have, in the last fifteen years or so, been an increasing number of attempts to incorporate agreement into theoretical accounts of language and language acquisition (e.g. Cornish 1986, Chomsky 1989, Pollock 1989, Karmiloff-Smith 1979; Lapointe 1988, Zwicky 1986; Perlmutter 1983) and it is now well established that agreement in natural language has much wider implications for syntactic theory than was previously assumed.

This paper attempts to place gender agreement found on determiners, past participles and adjectives in French in the wider context of syntactic theory within the Universal Grammar/Principles and Parameters framework (Chomsky 1989) with the intention of providing not only an adequate descriptive and explanatory account of agreement phenomena found in French but also some of parametric differences between languages like French and English.

In French all nouns are assigned one of two grammatical genders, commonly termed 'masculine' and 'feminine'.[footnote 2] Several associated elements, including determiners, past participles and adjectives, act as gender targets and show overt phonological agreement with the controller noun. French may be said to have a comparatively opaque gender agreement system in that overt agreement in French is only partial. For example, variation is found in nominative pronouns, but only in the third person where il alternates with elle. There is no category in French such that all members of that category show overt gender variation in all circumstances.

For the purposes of this discussion, it will be assumed that the agreement relation is with a category in a given structural position, rather than with the individual members of the category. That is to say that, even when there is no overt morphological reflection of agreement for some members of a given category, agreement within that category will nevertheless be considered to exist provided that at least one member of the category does reflect overt agreement with the noun in that structural position. Thus, in the case of French adjectives, the fact that only some (e.g. vert(e) - 'green' / grand(e) - 'big') show phonologically realised gender variation is considered sufficient evidence that the class of adjectives enters into an agreement relation, in spite of the fact that other adjectives, such as rouge - 'red', efficace - 'efficient' etc. do not show overt agreement.

Such an approach is justifiable on several counts:

a) If one were to sub-divide a category, simply on the grounds of presence/absence of overt agreement, such a division may be extremely hard to support empirically. Although as noted above colour adjectives in French may or may not exhibit overt gender variation any grouping which placed 'vert' and 'rouge' into different sub-categories is highly counter-intuitive. Furthermore, Determiners in French show gender variation in the singular (le/la 'the', un/une 'a' ) but not in the plural, where only number agreement is overt (les 'the' / des 'some'). Yet, a division based on plurality alone is clearly not sufficient since, in colloquial French, most adjectives (phonetically) exhibit only gender variation and not number variation when the noun is plural.

b) Morphological variation for gender within a given category varies greatly across languages. For example, French numerals show variation only in the singular (un/une 'one'). However, as Corbett (1991) points out, in some languages, all numerals show gender variation. Any maximally unified theory of gender agreement would presumably want to show the similarity between the two languages, i.e. that the general category of numerals can enter into an agreement relation with a noun, rather than to mask that similarity by positing that only part of the category (i.e. individual lexical items) agree in French.

c) Orthographically, word-final schwa is present in some words (e.g. aimée 'loved(f)', although synchronically, it is not phonetically realised. Diachronic evidence shows that, prior to the fifteenth century, a word-final schwa which was preceded by a vowel was pronounced (Price 1971). Such evidence, together with a synchronic phonological analysis of French, implies that lack of overt agreement on particular lexical items is phonologically rather than syntactically driven and that it is phonological change which has produced an opaque system rather than any change in agreement processes.

d) Further evidence that phonological, rather than syntactic processes influence overt realisation of agreement comes from analysis of different dialects of French. In some dialects of French, notably the Midi dialects, speakers still pronounce a word-final schwa after a fricative or affricate thus making a distinction between, say noir
'black (m)'
and noire
'black (f)'.
Although, in principle, dialects may differ in which syntactic processes they employ, it would seem odd to suggest that phonological differences between dialects necessarily imply syntactic differences.

For the purposes of this discussion then, it will be assumed that a single instance of overt agreement in any given category in any given syntactic position will be sufficient to assume that the category itself is in an agreement relation with another element. Of course, this assumption has the important consequence that any account of grammatical gender agreement in French must include an account of the syntactic relationship between target and controller.

In French, although the past participle does not agree with a direct object in complement position, past participle agreement (PPA) can occur when a DP[footnote 3] direct object has been extracted from the base complement position and moved to a position preceding the participle.

Thus, while (2b) is ungrammatical, PPA in (3-5) is possible.

(2a) Jean a ouvert les fenêtres
'John has opened the windows'

(2b) *Jean a ouvertes les fenêtres
'John has opened(f) the windows'

(3) Jean lesi a ouvertes [ti]
'John has opened(f) them'

(4) Les fenêtresi que Jean a ouvertes [ti]
'The windows which John has opened(f)'

(5) Je ne sais pas combien de fenetresi Jean a ouvertes [ti]
'I don't know how many windows John has opened(f)'

PPA cannot, however, occur when the subject of the clause is expletive 'il', as in (6) nor when the DP complement in question is not the complement of the past participle as it is in (7a), but is in fact the complement of another, lower verb, as in (7b):

(6) Quelle chaleur il a fait(*e)
'How hot it is'

(7a) Les musiciensi que j'ai entendus [ti] jouer
'The musicians who I heard playing'

(7b) La symphoniei que j'ai entendu jouer [ti]
'The symphony which I heard playing'

PPA also occurs in several other constructions, namely: the passive; 'unaccusatives' which take être as their auxiliary instead of avoir (see Burzio 1986); and all pronominal verb structures where the pronominal clitic cannot be analysed as a d-structure indirect object, divided here for the sake of argument into Reflexive/Reciprocal, Middle, Neutral and Inherent. (For fuller discussion of these different structure types, see e.g. Wehrli 1986, Lyons 1982)

(8a) Les fenêtres sont ouvertes par Jean. (PASSIVE)
'The windows are opened by John.'

(8b) Marie est morte hier soir. (UNACCUSATIVE/ÊTRE VERBS)
'Mary died yesterday.'

(8c) Marie s'est lavée. (REFLEXIVE)
'Mary washed herself.'

(8d) Les deux filles se sont comprises. (RECIPROCAL)
'The two girls understood each other.'

(8e) Une telle phrase ne se serait pas dite pour plaisanter. (MIDDLE)
'Such a phrase wouldn't have been said lightly.'

(8f) La porte s'est ouverte. (NEUTRAL)
'The door opened.'

(8g) Elle s'est mise à pleurer. (INHERENT)
'She started to cry.'

Furthermore, if we take three Romance languages, French, Spanish and Italian, we find that past participle constructions can be formed with one of two auxiliaries: E-aux (Etre, Essere, Ser) or A-aux (Avoir, Avere, Haber). Although auxiliary selection for a given construction differs between these languages and is clearly language specific, the differences appear to be systematic and PPA, when it can occur, appears to follow a uniform pattern, as shown in Fig. 1, taken from Shelton (1996).

Past Participle Agreement across Italian, French and Spanish
PassivesE: +PPAE: +PPAE: +PPA
Pronominal SEE: +PPA1E: +PPA2A: -PPA
UnaccusativesE: +PPAE: +PPA, A: -PPAA: -PPA
Wh-MovementA: +PPAA: +/-PPAA: -PPA
Weak CliticsA: +PPAA: +/-PPAA: -PPA
Relative ClausesA: +/-PPA3A: +/-PPAA: -PPA


E = E-aux
A = A-aux
+PPA = past participle agreement is obligatory
-PPA = past participle agreement is not possible
+/-PPA = past participle agreement is optional


(1) Excluding those interpretable as indirect objects
(2) Agreement is only optional in the absence of a clitic which is co-referential with the subject. If such as clitic is present, PPA is with the subject/this clitic as in:
a) Il vino che Maria ha comprato
The wine(m) that Maria has bought(m)

b) Il vino che Maria si e comprata
The wine(m) that Maria(f) herself has bought(f)

(3) According to Kayne (1989), for some speakers agreement within some pronominal constructions can be optional. However, an informal survey of native speakers revealed that they were significantly less ready to accept optionality for these constructions than they were for those involving A-aux.

The traditional analysis of structures involving PPA has been to say that the past participle agrees directly with the DP in question, without saying very much about the nature of the agreement process. Thus, PPA is said to be with the clitic in (3), the DP head of the relative clause in (4), the wh-phrase in (5), the 'subject' DP in (8a) and (8b) and with the subject and/or the pronominal clitic in (8c-g), depending on the analysis of the latter constructions.

Such an analysis is unsatisfactory in several ways:
1. As Fig. 1 shows, languages differ as to whether PPA is possible in a given structure. Under a traditional analysis an account for this would reduce to stipulation.
2. Under the traditional analysis, agreement is not local in the sense that Subject-Verb agreement is local. A maximally unified theory of agreement would ideally require all agreement relations to be similar in terms of locality restrictions.
3. No real account is given for the alternation of the auxiliaries être and avoir in structures involving PPA. Yet, as Fig. 1 shows, the link between auxiliary selection and PPA in French does not appear to be arbitrary and, in fact, appears to form part of a general pattern found in Romance languages. Any theory which tries to account for PPA should also provide a satisfactory account of auxiliary alternation.
4. As Kayne (1989) points out, any maximally unified theory of agreement would result in PPA receiving a similar analysis in all of the cases cited, whereas the traditional approach requires three if not four different accounts.
5. Again as Kayne notes, the traditional analysis fails to account for the fact that, if expletive 'il' is present in subject position as shown in (6), PPA is blocked.

Until fairly recently, within Government and Binding theory, tense and agreement inflection was considered to operate via the functional category of Infl which projected to IP (see Radford 1988 for a clear discussion of the mechanisms involved). Pollock (1989) argued that the two sets of features, Tense and Agreement, considered to operate within IP should more properly be analysed as separate syntactic heads with their own maximal projections, TP and AgrP. He proposes that a number of differences between French and English can be accounted for by positing differences in the properties of the two functional heads, T and Agr.

Pollock also argues, as does Chomsky (1989), for the existence of a second AgrP, lower in the structure, above VP. Chomsky assumes that Agr in general is linked to case assignment and that, under normal conditions, the higher Agr (AgrS) is linked with nominative case assignment while the lower Agr (AgrO) is linked with accusative case assignment[footnote 4]:

(9) diagram

Kayne (1989, 1991) argues for a treatment of some PPA structures which crucially relies on the positing of this additional structure above the VP containing the past participle as well as analysing all the controller DPs in question as being base-generated as direct object complements of the participle verb. Within the current UG framework, Subject-Verb agreement is considered to be a local specifier-head agreement relation at some point in the derivation between a functional category, Agr and the DP subject in specifier position of the projection of Agr. Verb agreement with the subject occurs when the verb moves to the Agr node either to collect agreement features base-generated in Agr or, if the verb is base-generated with features, to have those features 'checked', either before or after Spellout (Chomsky 1993). If subject-verb agreement is linked with AgrS, Kayne argues, then a maximally unified account of agreement should involve the linking of object-verb agreement with AgrO.

Kayne argues that PPA in the constructions he examines is not directly associated with the extracted DP in its final position but rather with an intermediary trace of the DP created by movement through a position within AgrOP. The exact location of the trace depends on the structure itself and may be adjoined to AgrOP, in the Specifier of AgrOP or in AgrO itself. PPA results from one of two agreement relations, within AgrOP, Specifier-Head or Head-Head and occurs when the past participle moves to AgrO to collect its features or to have them 'checked', as in (10) and (11). Note the use of I(P) to represent both T(P) and AgrS(P) where the distinction between the two is not relevant to this discussion.

(10) diagram

(11) diagram

Kayne claims that the advantage of such an analysis over the traditional one is four-fold.
1. The analysis moves towards a maximally unified theory of PPA through the consistent involvement of an Agr projection.
2. Given the locality restrictions apparent with subject-verb agreement, it would be reasonable to suppose that such restrictions be extended to include PPA. An analysis such as the one proposed by Kayne provides exactly the locality conditions required, since all agreement occurs within the maximal projection of AgrO. The traditional analysis, on the other hand, violates such locality restraints since agreement would necessarily be across several maximal projections.
3. The difference between the Spanish paradigm, where PPA is absent in these structures and the French/Italian paradigm where PPA is present is accounted for by assuming that non-PPA languages lack the additional structure posited for PPA languages. Thus, in non-PPA languages like Spanish, there might be no lower AgrOP and the object complement would, presumably, move directly to its final position from where it cannot enter into any local agreement relation with the past participle.
4. Lack of PPA in constructions like (6) falls out naturally from the assumption that they involve adjunction to AgrOP which Kayne argues is an A-bar position. Replacement of the expletive in (6) would require improper movement at LF from an A-bar to an A-position.

Shelton (1996) expands on this work and shows that all PPA structures exemplified in (2)-(8) including unaccusative and pronominal verbs can be analysed in much the same way with a d_structure DP object forced to move for case-theoretic reasons and involving either Spec-Head or Head-Head agreement within AgrOP. Furthermore, building on work by Lois (1989), she shows that it is not necessary to assume (as Kayne does) language specific c-selection properties of auxiliaries to account for the auxiliary choice and occurence of PPA. Rather, that auxiliary selection and related PPA in the different languages falls out naturally from requirements forced by auxiliary theta-grids and the type of chain (A-, A'- or Head-chain) that is created during the different derivations. It is also suggested that PPA is linked with case-assignment properties of AgrO in that a possible parametric variation between languages may be the ability of AgrO which is inert for case (Chomsky 1989) to remain active for agreement in some languages, whereas in others, absence of case assignment may mean absence of Agr. This in turn may be linked to the relative 'strength' of Agr along the lines of Pollock (1989).

As far as Determiners are concerned, within the standard analysis (Chomsky 1981) of the maximal projection of N, that is, NP, Determiners are analysed as being in the Specifier position of the NP:

(12) diagram

Within this analysis, an account of agreement of Determiners with the head noun is relatively simple. Whether the agreement relation is considered to be a direct relation between a Specifier and a Head, or whether it is considered that the projection of an abstract functional category Agr (Pollock 1989) mediates agreement, is not necessarily important at this stage. The important point would be that analysis of this agreement relation would be one of Specifier-Head which would tie in with the analyses of Subject-Verb agreement and extracted object-main verb agreement proposed by Pollock (1989), Chomsky (1991) and Kayne (1989).

More recently, it has been suggested that Determiners should be analysed as being heads which take NPs (or, as will be shown, APs which ultimately contain an NP) as complements in preference to an analysis which considers them as NP specifiers (see Abney 1987, Radford 1993 among others).

(13) diagram

Within this analysis, the nature of the agreement relation between the Determiner and the Noun would be analysed as one of Head-Head (or, possibly, Head-Complement) agreement, one of the two possible agreement relations suggested by Chomsky (1991). The motivation for this type of agreement is, according to Abney, the need for the whole DP to carry the semantic interpretation of the noun for structural interpretation elsewhere in the phrase. Whether this agreement is one of selectional restriction such that c-selection by the head requires a complement which shares the gender/number features of the head or whether it is the result of a separate feature matching agreement function which percolates feature information up the tree is not relevant to this discussion.

Within the traditional analysis of NP then, Determiner-Noun Agreement was assumed to be a case of Spec-Head agreement. Revision of the status of the Determiner and the DP proposal described above changed the status of Determiner-Noun agreement to one of Head-Head (or Head-Complement) agreement. This, in itself, was not seen to be problematic, given that justification can be found elsewhere in linguistic theory for heads to put selectional restrictions, including agreement ones, on their complements, including the complement head. In short, either the NP or the DP analysis was seen to produce an adequate account of Determiner-Noun gender agreement in French and to tie in with proposals for past participle agreement given above.

Adjectives, on the other hand, are not so easily dealt with if the traditional analysis if followed. Both pre- and post-nominal adjectives would be treated as adjuncts to N-Bar (e.g. Radford 1988) as in (14) and, as such, obviously cannot enter into a Specifier-Head relationship with the head noun:

(14) diagram

Furthermore, adjectives in French also show overt agreement in the predicative position of copular être as shown in (15) and, again, are not in a Specifier-Head relationship with the head noun, neither are they, in this position, analysable as adjuncts to Nbar. Yet, agreement does occur in these structures and must be accounted for.

(15) diagram

If the traditional structural analysis of nouns and their projections is adhered to, determiner and adjectival agreement forces an account which posits three different agreement relations between
a) Determiners and Nouns,
b) Nouns and Adjectives in pre-/post-nominal position and
c) Nouns and Predicative Adjectives.

Myles (1990) suggests that agreement in all three cases may be accounted for by positing that it is government by the controller noun which permits elements to show grammatical agreement with the noun. Government by the controller noun certainly holds for determiner and pre- or post-nominal adjectival agreement if the definition of government is that proposed by Rizzi (1990):

Head Government: X head-governs Y if
(i) X [[propersubset]] {A,N,P,V,Agr,T}
(ii) X m-commands Y
(iii) no barrier intervenes
(iv) Relativized Minimality is respected

Antecedent Government: X antecendent-governs Y iff:
(i) X governs Y and
(ii) X is co-indexed with Y

In (14) above, the noun head-governs the DP and both AP adjuncts. However, in (15), the head noun does not govern the predicative AP for two reasons: the noun does not m-command the AP (the maximal projection NP does not contain the AP) and minimality is not respected since the verb would presumably qualify as a closer head governor. Looking to antecedent government does not save the situation even if we allow the fact that the NP is coindexed with the Adjective since the NP cannot be classified as the antecedent of the Adjective nor of the AP in (15) given that they are not of the same category and the latter are not traces.

Even if government were the relationship necessary for agreement to occur, such a statement only describes the relation that must hold for agreement to take place. It does not account for why agreement takes place nor for the fact that, in French, overt agreement is both obligatory and optional, depending on the structure in question. Under an agreement-through-government analysis, this would be merely stipulatory. More so since other elements governed by the noun do not enter into agreement with it.

Both Abney (1987) and Radford (1993), among others, question the analysis of both pre- and post-nominal adjectives as adjuncts to N-bar. Radford analyses English adjectives and notes several differences between pre-nominal adjectives on the one hand and post-nominal and predicative adjectives on the other[footnote 5]. His observations would appear to be extendable to French adjectives (indeed he does so, in part, himself) and will be explored below.

Radford notes that grammatical agreement with the noun in French shows a degree of variation depending on the category and structural position of the agreeing element. For example, Determiners, Quantifiers and pre-nominal Adjectives invariably show strict syntactic agreement with the grammatical gender of the noun, even when the 'natural' gender of the referent is at variance with the grammatical gender. Thus "sentinelle" in (16) can denote a male or female referent, but has feminine grammatical gender. Determiners, Quantifiers and pre-nominal adjectives which co-occur with such a noun must always be +F even though the noun may, and indeed, usually does, denote a male referent:

(16) La(f) grosse(f) sentinelle(f) - 'the fat guard'
*Le(m) gros(m) sentinelle(f)

However, in a predicative position with copular être, as in (17), the adjective may exhibit grammatical agreement with the natural gender of the referent rather than with the grammatical gender. In other words, for adjectives in this position, strict grammatical agreement does not appear to be obligatory.

(17) La(f) sentinelle(f) est beau(m).
'The guard is handsome.'

An informal survey amongst native speakers revealed that, even when they were not prepared to accept sentences like (17) where the adjective agrees with the natural gender of the referent rather than with the grammatical agreement, neither were they prepared to accept the equivalent of (18) if the natural gender of the referent was known and was at odds with the grammatical gender

(18) La(f) sentinelle(f) est belle(f).
'The guard is handsome.'

Thus, predicative adjectives seem subject, at the very least, to some kind of semantic over-ride of grammatical gender where the natural gender of the referent is at odds with the grammatical gender to the point where, for some speakers, neither strict syntactic nor semantic agreement is available in these structures. Determiners and pre-nominal adjectives are never subject to the same type of semantic over-ride of agreement and strict syntactic agreement is obligatory in all instances for all speakers.

Radford also notes that some French adjectives which appear post-nominally exhibit similar agreement patterns as predicative adjectives. In (19), for example, gens which is +F for grammatical gender, may refer to a group of solely male referents, in which case, while the pre-nominal adjective must be +F, the post-nominal adjective is, for many speakers, obligatorily +M:

(19) Les vieilles gens heureux.
'The happy(f) old people(m).'

It should be pointed out, however, that this is the only French noun for which these comments hold true and even here, the situation is more complex than has been supposed[footnote 6].

Adjectival position can affect the semantic interpretation of the adjective. Radford notes that there are a number of adjectives in English which can occur both pre- and post-nominally with a difference in meaning according to which structural position they occupy . In addition, if the same adjective occurs in predicative position, the semantic interpretation is generally that of the post-nominal and not the pre-nominal adjective:

(20) Present students (antonym = past)
Students present (antonym = absent)
Most of the students are present (antonym = absent/*past)
The situation is similar in French:

(21) ma propre maison
'my own house'

ma maison propre
'my clean house'

ma maison est propre
'my house is clean/*my own'

Such adjectives might be considered to be homonyms and thus different lexical items altogether, producing a situation where certain pre-nominal adjectives cannot appear in the predicative position, a situation found in English with adjectives like mere and former. It is also the case in English, that some adjectives like alive can only appear in the post-nominal or predicative position but not in the pre-nominal position. Note that, in French, although it is possible, for purposes of emphasis, to place some normally post-nominal adjectives in a pre-nominal position, this is generally restricted to adjectives which "express a value judgement or [....] a subjective or emotional reaction" (Byrne & Churchill 1986). Other post-nominal adjectives cannot appear in pre-nominal position.

As Radford notes, an analysis such as the standard one which treats both pre- and post-nominal adjectives as adjuncts to N-bar and thus as having the same structural status, fails to reflect the semantic differences found in adjective distribution.

Pre-nominal adjectives can freely stack, within the bounds of relative positioning of semantic class and within the limits of production/stylistic constraints. (See Arnold 1989 for discussion of stylistic and syntactic constraints on adjectives). Post-nominal adjectives, on the other hand, cannot freely stack (native speakers prefer to resort to other constructions rather than stacking) and predicative adjectives cannot stack at all.

(22) le beau petit chat
'the handsome little cat'

??le chat méchant intelligent/le chat intelligent méchant
'the intelligent naughty cat/the naughty intelligent cat'

*le chat est intelligent méchant/le chat est méchant intelligent
'the cat is intelligent naughty/the cat is naughty intelligent'

A further difference, not mentioned by Radford, between pre-nominal adjectives on the one hand and post-nominal and predicative adjectives on the other is that only the latter two appear to be able to freely modify conjoined phrases. Native speakers will not, generally, accept a structure in which two conjoined nouns are modified by one pre-nominal adjective, even if the adjective is in the plural and the gender of the two nouns is not in conflict:

(23) *les grands bâtiments et livres
'the big buildings and books'

(Note that this type of structure is possible for some speakers on the condition that the gender of the two nouns is not in conflict and, crucially, that the two nouns together form some kind of semantic unit, such as is found in (24) where table et chaises might be construed as a single inseparable unit. One native speaker consulted would only accept this structure if she perceived the table as being singular - a judgement which seems to support the inseparable semantic unit interpretation).

(24) de belles table(?s) et chaises
'some lovely table(s) and chairs'

Post-nominal adjectives on the other hand appear in constructions where the adjective clearly has scope over conjoined phrases:

(25) [les hommes et les femmes]intelligents
'the intelligent men and women'

While some native speakers do not seem particularly comfortable with such a structure, no native speaker was found who did not unreservedly accept scope of a predicative adjective over a conjoined phrase such as

(26) [les hommes et les femmes] sont intelligents
the men and the women are intelligent

Note that in both cases, the post-nominal adjective and the predicative adjective will allow the conjunction of two or more noun phrases even when the grammatical gender of the noun phrases is not the same, as found in (25) and (26). The grammatical agreement found on the noun will be +F if both or all the noun phrases have +F grammatical gender or +M in all other instances. Note that the fact that the pre-nominal adjective obeys this 'rule' does not save the structure in (23). Note also, that whereas post-nominal adjectives can modify conjoined phrases, these phrases must include a determiner:

(27) *les hommes et femmes intelligents

Within an analysis where adjectives are adjuncts to N-bar, it is not readily clear why these differences should occur. One might argue that French does not allow conjunction of N-bar thus structures like (23) and (27), seen below in tree form, would be impossible.

Post-nominal adjectives, however, cannot only be analysed as adjuncts to N-bar, since this would preclude structures such as (25) since N-bar does not contain a DP. Under the standard analysis, then, post-nominal adjectives would have to be analysed as adjuncts to either N-bar or NP to allow for structures like (25). The question then arises as to why right-adjunction but not left-adjunction of AP to NP is possible in French.

In the pre-nominal position, French adjectives cannot take complements or be post-modified whereas post-nominal and predicative adjectives can:

(28) a) le problème est difficile à resoudre
'the problem is difficult to resolve'

b) un problème difficile à resoudre
'a problem difficult to resolve'

c) *un difficile à resoudre problème
'a difficult to resolve problem'

Again, under the standard analysis, this would reduce to stipulation since there would be no reason for an AP, in pre-nominal position not to contain additional material in the same way that an AP in post-nominal or predicative position can.

As both Abney and Radford suggest, any analysis which treats pre- and post-nominal adjectives in the same way, will be inadequate in that the differences highlighted above would have to be accounted for by positing a significant number of stipulative filter and lexical rules, a situation to be avoided in a minimalist approach to grammatical analysis. Within such a framework, the standard analysis, which would treat both pre- and post-nominal adjectives as adjuncts to N-bar would thus fail.

A number of researchers have proposed different treatments of adjectives in an attempt to incorporate some of the facts mentioned above. The following provides an outline of three of these proposals: Giorgio and Longobardi (1991), Bernstein (1991) and Radford (1993) / Abney (1987). A revised proposal for adjectives which builds on work by Radford (1993) is then put forward.

Giorgio and Longobardi (1991) suggest that, in French, all adjectives are base-generated in a post-nominal position and that pre-nominal adjectives are the result of movement to Spec of NP, thus:

(29) diagram

Romance languages, Giorgio & Longobardi claim, project adjectives to the right of the head noun while Germanic languages project them to the left. This, it is claimed, accounts for the fact that, in Romance, the majority of adjectives remain in the post-nominal position and, in Germanic, the majority are in pre-nominal position. The difference in direction of projection is, it is argued, a parametric difference between the two languages.

However, Giorgio and Longobardi's account does not solve some of the differences discussed above.

a) There is no apparent reason why only some adjectives can move into Spec of NP and others cannot.
b) Since the Specifier position is filled by a maximal projection, AP, there is no motivated reason why the moved adjective cannot be post modified.
c) The semantic differences between pre- and post-nominal adjectives in relation to the predicative position remains unexplained.

Bernstein (1991) looks at agreement within the NP and suggests that it is not the adjective that moves to a pre-nominal position but that the noun itself moves into a higher Num node, across a base-generated pre-nominal adjective. Thus, for Bernstein, it is the post-nominal position which is derived:

(30) diagram

NumP would be a complement of the Determiner. Since it is the Noun which is involved in head movement and only the noun raises across the AP with the Determiner higher in the structure, it is unclear how the proposed structure can account for the fact that post-nominal adjectives can modify whole DPs or AP. Moreover, some adjectives, like petit ,which must normally remain in pre-nominal position in French, would have to 'block' noun movement while others require obligatory noun movement and there is again no reason why no material may intervene between an adjective and a noun which remains in situ but. Finally, the Adjective remains in an adjunct position within the NP and the problem of the definition of the agreement relation remains.

In an interesting departure from the traditional analysis, both Abney and Radford propose that pre-nominal adjectives, unlike their post-nominal counterparts should be analysed as heads which take either another AP or an NP complement, with the condition that the head of the lowest maximal projection within the AP complement should be a noun[footnote 7].

(31) diagram

Since the entire DP must carry the semantic interpretation of the ultimate head the features of that head must be passed up to the highest DP node. Both researchers assume that the features of the ultimate head are therefore percolated up through the intervening heads until they reach the Determiner. Radford accounts for the agreement found in Determiners and Adjectives in these structures in terms of a Head-Head agreement relation between the 'ultimate' head (the noun) and the 'immediate' heads (Adjectives, Quantifiers and Determiners). Adjectives such as petit which must appear pre-nominally would be presumed, as Abney suggests, to obligatorily c-select an NP or AP whereas other adjectives which can stylistically be placed before the noun would optionally do so.

One attraction of this type of analysis is that agreement within the noun phrase becomes essentially a one-to-one relation as is the relation between specifiers and heads. Agreement is motivated as the overt reflection of a necessary percolation of features for semantic interpretation. Furthermore, the fact that pre-nominal adjectives cannot take complements or post-modification falls out naturally if pre-nominal adjectives c-select an NP or another AP as Abney suggests. Under these circumstances, there would be no structural position available for any adjectival complement.

Radford goes on to argue that, given that the agreement discussed is one of head with complement and, as such is based on structural relations between elements, Determiner/Quantifier and pre-nominal Adjectival agreement with a noun is essentially a syntactic one and, as such, involves strict syntactic agreement. Thus, he claims, the invariability of agreement between these items is accounted for. On the other hand, post-nominal adjectives, he argues, share practically all the attributes of predicative adjectives and behave much more like them than pre-nominal adjectives do. In order to account for this, he suggests that post-nominal adjectives are, in fact, predicative adjectives of sorts and that the agreement relation between them and the noun mirrors that found between the predicative adjective and the subject noun phrase. As for the nature of such an agreement relation, he suggests that, while pre-nominal agreement is strictly syntactic, post-nominal agreement is "a relation of semantic compatibility" which is effected, presumably, by some kind of predication rule.

Such an analysis provides a neat account of the agreement differences between agreement of pre- and post-nominal adjectives with the noun gens. If agreement of pre-nominal adjectives is one of strict syntactic agreement and post-nominal adjectives are "semantically compatible" with the noun, then the agreement facts found in (19) are to be expected.

However, an informal survey of native French speakers revealed that Radford's assumption about pre- and post-nominal adjectives does not seem to hold with any noun other than gens. For example, if Radford is correct then the prediction would be that speakers would accept all three sentences in (32) as grammatical if the natural gender of the referent docteur was female:

(32) a) le gros docteur
'The(m) fat(m) doctor'

b) Le docteur intelligente
'The intelligent(f) doctor'

c) Le docteur est intelligente
The doctor is intelligent(f)'

While speakers were, reluctantly, prepared to accept (32c) as a possible French sentence (they preferred to re-word the sentence into something along the lines of "Le docteur, c'est une femme intelligente") none of the native speakers surveyed would accept (32a) or (32b) as sentences of French if the referent was female. Interestingly, as with (32c), speakers preferred to find alternative structures for the other two sentences. In several instances, speakers were prepared to change the grammatical gender to match the natural gender, producing, for example, (33b) from (33a):

(33) a) Le docteur intelligente
b) La docteur intelligente

If Radford's "semantic compatibility with the noun" for post-nominal and predicative adjectives were the correct account then native speakers' intuitions about (32) and (33) are unexpected, since it would be anticipated that (32b) and (32c) would be equally acceptable. By extension, if post-nominal and predicative adjectival agreement is solely based on "semantic compatibility", we might expect to see evidence of this with all nouns. Let us assume that masculine grammatical gender is the "unmarked" form for French. Such a statement is supported by the fact that, if the grammatical gender of a referent is unknown, masculine agreement takes place. If the masculine form is in some sense a "neutral" form as well, then one would expect that, if semantic compatibility drives the agreement relation between post-nominal and predicative adjectives and a noun, then structures like (34) should be possible

(34) a) La table est rond - 'The(f) table(f) is round(m).'
b) la table rond - 'the round(m) table(f).

However, native speakers do not accept (34) as grammatical structures in French.

More intuitively appealing would be the notion that all agreement is syntactically based but that, under certain conditions, semantic factors can over-ride syntactic agreement.

Copular être and other verbs which take adjectives in the predicative position (e.g. sembler) are arguably analysed as being raising verbs. That is, verbs which have no external theta role themselves and which take a clausal complement from which the subject is raised into the empty d_structure-generated subject position to fulfil case assignment requirements (Couquaux 1981; Burzio 1986). Let us assume this to be the case and, further, that verbs like être take a small clause as their complement. On the assumption that the maximal projection of the small clause is the same as that of its head, in this case an AP, we would suppose that the ultimate subject is base-generated in the Specifier position of the AP and assigned a subject theta role by the adjective.

(35) diagram

We could then argue that agreement between the DP and the adjective is a local Specifier-Head relationship which exists within the small clause complement, in this case between the head adjective and the trace of the DP generated in Specifier position of the maximal projection, AP, rather than a predicative relationship between the s_structure subject and the adjective across être. Thus, the D_ and S_ structures of (36a) would be (36b) and (36c) respectively:

(36) a) La table est grande.
(36) b) [ [e] être [[la table] grand]

c) [ [la tablei] est [ [ti] grande]

Gender agreement, then, under this analysis would be the result of a general specifier-head agreement relation as supposed by, e.g. Georgopoulos (1991).

However, since the functional category Agr is now commonly held to play a crucial role in both Subject-Verb and Object-Verb agreement (see Chomsky 1989; Pollock 1989; and Kayne 1989) we might assume, as does Chomsky (1991), that Subject-AP agreement in predicative constructions follows the same lines, i.e. that the AP small clause is dominated by an AgrP as in (37a). The subject in the small clause which originates in the Specifier of the AP passes through the Specifier position of the AgrP which immediately dominates the AP on its way to receive case. The Adjective itself converges with Agr at some stage in the derivation to picks up the agreement features which are generated as an affix in Agr or to have already present features 'checked' , in much the same way as Verb raising is considered to take place in Subject-Verb and Object-Verb agreement described above. Since the DP subject cannot receive case within this lower AgrP, it must move up to the higher AgrSP to satisfy its case requirements.

(37) diagram

Such an an analysis has several advantages:
a) It would be predictable that only one adjective can appear in predicative structures: only one Agr node is available for adjectival movement; only one Adjective can raise to that Agr node.
b) Predicative adjectival agreement mirrors agreement facts found elsewhere, i.e. subject-verb and object-verb agreement.
c) In line with Corbett's Agreement Hierarchy (Corbett 1991), predicative adjectives in French may agree with the semantic, or natural gender of the DP rather than obey strict syntactic agreement; pre-nominal adjectives must always obey strict syntactic agreement, even if grammatical gender is at odds with natural gender, as outlined above. If we suppose that semantic over-riding of grammatical agreement, if it can occur at all, can occur only within Spec-Head agreement relations, and not in Head-Head agreement then the above analysis is further supported. Note that if the DP hypothesis is adopted, semantic over-riding of Spec-head, but not Head-Head agreement is also found in English and French collective nouns in Subject position:

(38) a) This/*These class is/are good at French.
b) La moitié des sénateurs a/ont voté pour le projet de loi.

Half the sentators has/have voted for the bill.

Given the strong similarities between post-nominal and predicative adjectives, it seems reasonable to suppose that, as Radford suggests, they do have the same structure. Suppose that post-nominal adjectives are also the head of a small clause with the nominal element they modify generated in Spec(AP) and assigned a theta role by the adjective. Suppose also that the post-nominal AP small clause is, just like its predicative counterpart, the complement of Agr as in (37a) and that the DP subject of the small clause again moves into in the [Spec,AgrP] as in (37) above. However, suppose also that, in the case of post-nominal adjectives, the DP stays there as in (39).

(39) diagram

One difficulty with this analysis is that DPs require case to be visible and we do not assume that the Agr node is, in itself, capable of assigning case to the DP in its specifier position. Were it so capable, then the DP in predicative constructions would not be able to move into the higher subject position and receive nominal case from the Tense node. However, in constructions containing the post-nominal adjective, the AgrP itself, in these structures, is always in a case-marked position and we might assume that AgrP is in some way transparent for case and is able to pass on any case it receives to the DP in its specifier. In predicative adjectival structures, the AgrP is not in a case-marked position and thus the DP could not get case via a transparent Agr node and would be forced out of the AgrP to move into a case-marked position.

Within such an analysist the difference between post-nominals and predicatives would reduce to the fact that in predicative constructions, the DP moves out of the AgrP node to a higher position in order to get case whereas in post-nominal constructions the DP remains in situ. The similarities between them would then stem from the fact that they share the same structural makeup and that agreement in both constructions results from specifier-head agreement within AgrP.

Crucially, a major problem for analyses which hold APs to be adjuncts to N-bar is resolved. As has been noted, post-nominal adjectives can modify more than an NP. In the following sentences, the post-nominal adjective modifies a DP, an AP and an NP respectively.

(40) a) [[les hommes et les femmes] intelligents]
the men and women intelligent

b) [le [beau [[petit garçon] intelligent]
the handsome little boy intelligent

c) [le [petit [garçon intelligent]]
the little boy intelligent

Recall that the AP=Adjunct analysis would have to stipulate that an AP can right-adjoin (but not left-adjoin) to a DP in French. Under the revised analysis, the facts in (40) fall out quite simply from the assumption that adjectives can assign a subject theta role to any of these three categories and thus that any of the three can be generated in Spec(AP).

Furthermore, the fact that, unlike pre-nominal adjectives, additional material can follow both post-nominal and predicative adjectives also falls out naturally from this analysis since both the latter are free to c-select further material in their complement positions.

If the above account is correct, it might be supposed that the difference between post-nominal adjectives and predicative adjectives is not one of structure per se but depends on the fact that, in post-nominal adjectival structures, the DP subject is physically present in Spec(Agr) whereas, in predicative adjectival structures, Spec(AgrP) only contains a trace of the moved DP subject thereby perhaps allowing a 'weakening' of syntactic agreement in favour of semantic agreement.

An alternative supposition may be that syntactic agreement is required with post-nominal adjectives because agreement is bound up with case-assignment to the element in [Spec,AgrP] for post-nominal adjectives, but not for predicatives where the DP receives case through movement to the higher subject position and agreement is not linked to case-assignment.

To conclude, it has been suggested that two types of agreement relation operate with regard to adjectives, determiners and past participles in French; Head-Head and Specifier-Head agreement. The notion of different types of agreement relation is not new. Georgopoulus (1991) and Iatriadou (1991) both suggest that nominal agreement may be different to past participle and subject-verb agreement and Chomsky (1989) also suggests that there may be two fundamental agreement relations; Specifier-Head and Head-Head agreement. Ideally, a maximally unified theory of agreement would propose agreement relations which were reflected elsewhere in the grammar. It was for this reason that the analysis of past participle agreement as a Specifier-Head relation was so attractive -- it mirrors the agreement relation proposed for Subject-Verb agreement (Chomsky 1989).

Georgopoulus (1991) points out that the Papuan language has an agreement relation between the verb and its complement in situ, unlike French where object-verb agreement only occurs when the object has been extracted from its base-generated position and moved to the left of the verb, never when the object remains in base position. Furthermore, as Haegeman (1991) notes, in West Flemish, the following examples illustrate the apparent existence of a Head-Complement (or Head-Head) agreement pattern:

(41a) (H71a) ..[ [ da [ den inspekteur da book gelezen eet]]]
that the inspector that book read has

(41b) (H71b) ..[ [dan [ [d' inspekteurs da book gelezen een]]]
that the inspectors that book read have

Haegeman notes that "the perfective auxiliary eet/een agrees in number and person with its subject 'den inspekteur/d'inspekteurs', illustrating Specifier-Head agreement. Furthermore, the complementizer da agrees in number and person with the subject and with the inflection: da is third person singular, dan is third person plural. The head of the CP C agrees with the head and the Specifier of its complement IP" (Haegeman 1991).

Let us assume, therefore, that there are a minimum of two agreement relations in natural language; Specifier-Head agreement and Head-Head (or possibly, Head-Complement) agreement. Let us also assume that languages may have one, both or neither of the two types of agreement relation (the latter found in languages such as Malay where no evidence of agreement of one element with another exists) and that the presence or absence or either or both agreement relations may be a parametric variation in natural language.

A further difference between languages may be the degree to which agreement is overtly reflected in the language and whether or not agreement is intrinsically linked with case assignment or not. Note that, in the analysis suggested above, agreement in French takes place whether or not Agr is a case-assigner (some PPA, predicative adjectives where Agr does not assign case to the DP in Spec position and post-nominal adjectives where the Agr node only indirectly assigns case to its Spec). What agreement exists in English and which involves Agr seems to be possible only if the Agr node is also a case assigner, e.g. in Subject-Verb agreement, or possibly in possessive constructions (cf Abney 1987). Head-Head agreement of Determiner and the head of its complement is possible in English.

Pollock (1989) suggests that differences found in French and English regarding movement possibilities may be linked to the fact that Agr is 'strong' in French, but 'weak' in English. It is possible that the 'strong' Agr in French is strong precisely because a) agreement is overt in both head-head and specifier-head structures and is found in a number of structures headed by different categories and b) French Agr is operative whether or not it is a case assigner. That is, the Agr node can survive without being +case. In English, a) agreement is not particularly overt and not found in many structures and b) English Agr and Case do not operate separately. That is, English Agr cannot survive without being +case, making it 'weaker' than its French counterpart. Chomsky (1989) suggests that Agr nodes that do not assign case are either absent or inert for case. It is feasible to suggest, perhaps that a parametric difference between French and English is whether or not Agr is absent if there is no case to assign (English) or merely inert for case (French). In turn, the survival of the Agr node in French may be linked to the relative strength of agreement features in French and the fact that its presence is required for feature checking prior to Spellout (Chomsky 1996).


Abney, S., 1987. 'The English Noun Phrase in its Sentential Aspect'. Unpublished doctoral dissertation, MIT.

Arnold, D., 1989. 'Theoretical and Descriptive Issues in Machine Translation'. Unpublished doctoral thesis, University of Essex.

Baetens-Beardsmore, H., 1971. A gender problem in a language contact situation. In: Lingua 27.141-59.

Barlow, M., & Ferguson, C., 1989. Agreement in Natural Language: Approaches, theories, descriptions. Stanford, CA: Centre for the Study of Language and Information.

Bernstein, J., 1991. DPs in French and Walloon: Evidence for parametric variation in nominal head movement. In: Probus 3.2 101-126.

Burzio, L., 1986. Italian syntax. Dordrecht: Reidel.

Byrne, L. S. R., & Churchill, E. L., 1986. A Comprehensive French Grammar. Oxford: Blackwell. (3rd Edition, revd by G. Price.)

Chomsky, N., 1981. Lectures on Government and Binding Theory. Dordrecht: Foris.

-- 1989. Some notes on economy of derivation and representation. MIT Working Papers in Linguistics 10: Functional Heads and Clause Structure. 43-74.

-- 1996. The Minimalist Program. Cambridge, Massachusetts: MIT Press.

-- and H. Lasnik, 1993. Principles and Parameters theory. In: J. Jacobs, A von Stechow, W. Sternefeld and T. Vennemann (eds). Syntax: An International Handbook of Contemporary Research. Berlin: de Gruyter. 506-69

Corbett, G., 1991. Gender. Cambridge: Cambridge University Press.

Cornish, F., 1986. Anaphoric Relations in English and French: A Discourse Perspective. London: Croom Helm.

Couquaux, D., 1981. French Predication and Linguistic Theory. In: R. May & J. Koster (eds), Levels of Syntactic Representation, 33-64. Dordrecht: Foris.

Fodor, I., 1959. The Origin of Grammatical Gender. In: Lingua, 8.1-41,186-214.

Georgopoulos, C., 1991. On A- and A'-agreement. Lingua 85.135-170.

Giorgio, A., & Longobardi, G., 1991. The Syntax of Noun Phrases: Configuration, Parameters and Empty Categories. Cambridge: Cambridge University Press.

Givón, T., 1970. The resolution of gender conflicts in Bantu conjunction: when syntax and semantics clash.In: Papers from the Sixth Regional Meeting, Chicago Linguistic Society. 250-61.

Greenberg, J., 1966. Universals of Language. Cambridge MA: MIT Press. (Second edition. First published 1963.)

-- 1978. How Does a Language Acquire Gender Markers? In: J. Greenberg (ed.), Universals of Human Language 3. Stanford: Stanford University Press. 47-82.

Guthrie, M., 1948. Gender, number and person in Bantu languages. Bulletin of the School of Oriental and African Studies. 12.847-58.

Haegeman, L., 1991. Introduction to Government and Binding Theory. Cambridge, MA: Blackwell.

Iatriadou, S., 1990. About Agr(P). Linguistic Inquiry 21.551-557.

Ibrahim, M., 1973. Grammatical gender: its origin and development. The Hague: Mouton.

Karmiloff-Smith, A., 1979. A Functional Approach to Child Language. Cambridge: Cambridge University Press.

Kayne, R., 1989. Facets of past participle agreement in Romance. In: P. Benincà (ed.), Dialect variation and the theory of grammar. Dordrecht: Foris. 85-103.

-- 1991. Romance Clitics, Verb Movement, and PRO. Linguistic Inquiry 22/4.647-686.

Key, M. R., 1975. Male/Female language, with a comprehensive bibliography. Metuchen, NJ: Scarecrow.

Lapointe, S. G., 1988. Towards a unified theory of agreement. In: M. Barlow & C. A. Ferguson (eds.). Agreement in Natural Language: Approaches, Theory, Descriptions. Stanford: Center for the Study of Language and Information. 67-87.

Lois, X., 1990. Auxiliary Selection and Past Participle Agreement in Romance. Probus, 2.2.233-255.

Lyons, C., 1982. Pronominal voice in French. In: N. Vincent & M. Harris (eds). Studies in the Romance Verb: Essays offered to Joe Cremona on the occasion of his 60th birthday. London: Croom Helm.

Mel'cuk, I. A., 1974. Statistics and the relationship between the gender of French nouns and their endings [1958]. In: V. J. Rozencvejg (ed.). Machine Translation and Applied Linguistics. Vol. I. Frankfurt: Athenaion. 11-42.

Myles, F., 1990. 'Error and order in the acquisition of French as a second language'. Unpublished doctoral thesis, University of Sheffield.

Parsons, F. W., 1960. An introduction to gender in Hausa. African Language Studies 1.117-36.

Perlmutter, D. (ed.), 1983. Studies in Relational Grammar. Chicago: The University of Chicago Press.

Pollock, J.-Y., 1989. Verb movement, Universal Grammar and the Structure of IP. Linguistic Inquiry 20.365-424.

Price, G., 1971. The French Language: Present and Past. London: Edward Arnold.

Radford, A., 1988. Transformational Grammar. Cambridge: Cambridge University Press.

-- 1993. Head-Hunting: On the trail of the Nominal Janus. In: G. Corbett, N. M. Fraser, & S. McGlashan (eds.). Heads in Grammatical Theory Cambridge: Cambridge University Press.

Rizzi, L., 1990. Relativised Minimality. Cambridge, Massachusetts: MIT Press.

Sadler, L., & Arnold, D., 1994. Prenominal adjectives and the phrasal/lexical distinction. Linguistics 30.187-226.

Shelton, J. F. (1996) 'Second Language Acquisition of Grammatical Gender in French'. Unpublished doctoral thesis, University of Essex.

Spilka, I., 1976. Contrastive Analysis and French Gender, remarks on a study by Elaine Tarone, Uri Frauenfelder and Larry Selinker. Working Papers in Bilingualism 8.112-27.

Tucker, G. R., Lambert, W. E., & Rigault, A. A., 1977. The French Speaker's Skill with Grammatical Gender: An Example of Rule-Governed Behavior. The Hague: Mouton.

Wehrli, E., 1986. On some properties of French clitic SE. In: H. Borer (ed.). The Syntax of Pronominal Clitics New York: Academic Press. (Syntax and Semantics. 19.)

Wheeler, B. I., 1898. The Origin of Grammatical Gender. In: Journal of English and Germanic Philology 2.528-45.

Zwicky, A., 1986a. Agreement Features: Layers or Tags? In: Working Papers in Linguistics, Columbus, OH 32.146-148.

-- 1986b. German Adjective Agreement in GPSG. In: Linguistics: An Interdisciplinary Journal or the Language Sciences, 24/5.957-990.


[1] It should be noted that, in much of the literature on grammatical gender, the terms 'controller' and 'target' gender are used. Briefly, this is taken to mean that one lexical item is in some sense more dominant in terms of the gender system and that it 'controls' the gender marking on other elements, which are deemed to be 'targets' for gender agreement. Usually, the controller is taken to be the lexical item which is invariant for gender: the noun or a projection of it, including both N-bar and NP. As Barlow and Ferguson (1989) observe, if we assume that agreement involves controller and target genders, then the notion of directionality of agreement, from controller to target, is tacitly assumed. Some theoretical frameworks, such as Lexical Functional Grammar, do not assume directionality, but rather a feature-checking device which ensures that all elements involved in an agreement relation carry the same features. For these theories, the notion of controller v. target gender is of lesser importance, often retained for expository purposes rather than from any theory-internal reason. As will be seen, the theoretical analysis assumed in the following pages suggests that the noun or noun phrase does not play a directional role in gender agreement but rather, agreement is the result of a structural relationship between two elements. For convenience' sake, and where the directionality is of no theoretical importance, the terms controller and target will be retained to indicate noun(phrases) and other elements respectively.[Return to text]

[2] These terms are not really fully indicative of the basis of the French gender assignment. For discussion of this area, which is outside the scope of this paper, see e.g. Corbett (1991), Key (1975), Shelton (1996) or Spilka (1976).[Return to text]

[3] For ease of reference, the term DP includes clitics found in constructions like (2) above, and "moved DP" covers object-movement, clitic movement and wh-movement, unless otherwise stated. However, it is by no means clear that clitics can only be considered to be maximal projections even though they do replace them. Clitic adjunction to functional heads would suggest that clitics behave in ways which would give them the status of a head (see e.g. Chomsky 1995). This will be dealt with at a later stage.[Return to text]

[4] Tree diagrams are pruned where structure is not needed for illustrative purposes.[Return to text]p> [5] A particularly clear exposé of the various properties of English adjectives can also be found in Sadler and Arnold (1994).[Return to text]

[6] See Byrne & Churchill (1986). [Return to text]

[7] The 'ultimate' head in Radford's terminology.[Return to text]

Return to The Issue 2 Contents Page |