LINGUIST List 5.1088

Thu 06 Oct 1994

Disc: French clitics

Editor for this issue: <>


  • Philip H. Miller, On the affixal status of French clitic pronouns
  • Lambrecht Knut, Re: 5.1074 French clitics

    Message 1: On the affixal status of French clitic pronouns

    Date: Wed, 5 Oct 1994 09:12:50 +On the affixal status of French clitic pronouns
    From: Philip H. Miller <>
    Subject: On the affixal status of French clitic pronouns

    This posting presents a joint response by Julie Auger and Philip Miller to the recent discussion on LINGUIST concerning the status of clitic pronouns in French, specifically, the postings by Steven Schaufele and Pierre Larrivee.

    In his summary (LINGUIST 5.1036) of the responses he got concerning topicalization and clitic doubling in French and Italian, Schaufele notes that various linguists have suggested that clitic pronouns are being reanalyzed in French as prefixed agreement morphemes.

    Larrivee (LINGUIST 5.1074) suggests that this analysis is problematic on the basis of various phenomena, concerning (i) the possible morphological complexity of clitic pronouns; (ii) the mobility of clitic pronouns; (iii) the fact that subject clitics are restricted to subject function.

    There has been an extensive debate on the status of pronominal clitics in French (and the other Romance languages) over the past 25 years. This debate has brought to bear on the question a huge amount of complex data which cannot be ignored in this context.

    Both authors of this posting have independently concluded from these data that pronominal clitics in French should be analyzed as lexically attached inflectional affixes (see references listed below).

    A first point that we would like to make, specifically addressing Schaufele's post, is that it is necessary to distinguish affixal status from agreement marker status. It is clearly possible for a language to have affixal pronominals ("incorporated pronouns") which are not agreement markers. This is the case, for instance, in many Bantu languages where the object marker, though it is clearly word internal, can be incompatible with a clause internal overt object NP (cf. e.g. Chichewa (Bresnan and Mchombo 1987), and also the discussion by Benji Wald in a recent response on LINGUIST 5.968 to a query by Dan Everett). Consequently, the question of whether the clitic pronouns in French are lexically attached or not is independent of the question of whether they are agreement markers. The complex data concerning clitic doubling are only relevant to the latter question, not the former (obligatoriness or optionality with various types of overt NPs, definite, indefinite, pronominal, etc. and depending on the syntactic function of the NP; prosodic properties of the doubling overt NP; locality constraints between the clitic bearing verb and the doubling overt NP; etc.). Julie Auger has argued in detail that subject clitics in spoken Quebec French are being reanalyzed as agreement markers, though not the other clitics. The doubling properties in Quebec French clearly indicate that reanalysis of object clitics as agreement markers has not (yet) taken place.

    The second question then is the issue of lexically attached inflectional status vs. independent word status. Note that under the latter hypothesis, French pronominal clitics should be analyzed as postlexical clitics, i.e. prosodically deficient words which are prosodically attached to a neighboring host word.

    The references below provide numerous types of data arguing in favor of lexically attached inflectional status. Here is a very brief summary:

    (i) Morphophonological idiosyncrasies, which cannot be accounted for on the basis of productive phonological rules, in the realization of the clitics and, more remarkably, in the realization of the stem to which they are attached (e.g. the pronunciation of 'je suis' (I am) as 'chuis' in France and 'chu' in Quebec, both of which are impossible for the phonetically identical 'je suis' form of the verb 'suivre' (to follow).

    (ii) Gaps in the paradigms, where there is no grammatical form with clitics. In some cases 'strong' pronouns can be used instead in the position of the usual overt NP argument, or one has to resort to a completely different construction. (E.g. '*Marie me te/lui decrit.' vs. 'Marie me decrit a toi/elle' (Marie describes me to you/her); '*Chante-je?' vs. 'Est-ce que je chante?', or 'Je chante?', or 'Je chante-tu?', (the last, derived from 'je chante-ti?', exists only in Quebec French) (Am I singing?)). Note that it is hard to see how these gaps could be explained in syntactic terms, especially for the cases where a strong pronoun alternative is possible. Similarly, the existence of well formed alternatives to the impossible clitic clusters shows that the problems are neither semantic nor pragmatic.

    (iii) Complex properties of the ordering and compatibility of the clitics among themselves, for which no satisfactory syntactic analysis has been proposed to date; the plausibility of the existence of a non ad hoc syntactic analysis is greatly cast into doubt by the variation in clitic ordering attested between otherwise extremely similar dialects (cf. e.g. Cummins and Roberge 1994 on this).

    (iv) The fact that pronominal clitics must appear attached to a Verb, to the exclusion of any other category (note that this cannot be explained as a consequence of VP initial attachment, since in those -cases where there is a preverbal element in the VP, the clitics attach to the verb in modern French ('Il faut ne rien lui dire' vs. '*Il faut ne lui rien dire'). Usually, items which are argued to be postlexical clitics do not impose such constraints on the category of the word to which they are prosodically attached.

    (v) The fact that pronominal object clitics in French cannot have scope over a coordination of hosts ('*Marie le voit et entend' vs. 'Marie le voit et l'entend' (Marie sees and hears her.)). In fact, we have argued that even for subject clitics, wide scope over coordination is abnormal in colloquial French ('%Je mange du pain et/pis bois du vin. 'I eat bread and drink wine').

    (vi) The types of phonological rules affecting clitic+host combinations. It appears that there are lexical phonological rules which apply between clitics and host (e.g. that liaison phenomena are of a word internal type, rather than of a postlexical type). These data are among the least reliable of those presented here.

    (For the relevance of these tests, see among others Zwicky and Pullum 1983).

    Let us reconsider the data proposed by Larrivee in this context.

    >Take third person >nominative clitik "il". It has its own morphology (number (vs Ils), gender >(vs Elle) and case (vs Le (accusative) and Lui (dative?)).

    The only case where a productive morphological relationship between these forms can be suggested is the case of the singular/plural alternation which shares the orthographic -s suffix with the regular sg/pl alternation for nominals. This could just as well be analyzed as a frozen remnant of previously productive morphology, all the more so that other forms show irregular alternations for sg/pl. This is the case for all 1st and second person forms (e.g. je,me/on, tu,te/vous, ...) and for the dative and reflexive forms of the 3rd person (lui/leur, se/se).

    > It can move around (Il le dit/Le dit-il?).

    To which we can add postverbal object clitics in imperatives ('Dis-le!' vs. '*Le dis!' (Say it!)).

    This in fact is one of the most convincing arguments against the affixal analysis we propose. However there a number of counterarguments.

    a) The form of certain clitics changes depending on whether they are pre or postverbal. ('Tu me regardes.' vs. 'Regarde-moi', similarly for te/toi, and, in many varieties of French, for 'y' and 'en' which are pronounced [zi] and [za~] postverbally 'Donne-moi-z-en'). Moreover, the pre- and postverbal forms are not related by productive phonological rules: if we assume 'me -> moi' under accent by such a rule, we would predict that 'le -> loi' in similar cases, which is not the case ('Prend-le!' vs. '*Prend-loi!' (Take it!)). b) The ordering of clitics before and after the verb is different ('Marie me le donnera.' 'Donne-le-moi!' (Marie will give it to me.; Give it to me!) c) As noted above, for most verbs, subject clitic inversion is impossible with 'je' ('*Chante-je?' (Am I singing?) vs. 'Puis-je?' (May I?)), which is hard to understand if 'je' is an independent word: syntactic fronting of the verb should apply independently of person. There is nothing phonologically illformed about the resulting sequence that would explain how it is filtered out by prosodic cliticization. If these forms are produced by the morphology, this is simply a case of gaps in a paradigm. Note that the restrictions on subject clitic inversion are much stronger in Quebec Colloquial French: only the 2nd person pronouns can be inverted. d) There are a number of languages with what have been called 'mobile position affixes' which have been documented. I.e. it does not appear to be crosslinguistically impossible for an affix to appear as a prefix in certain forms, but as a suffix in others. (cf. e.g. Fulmer 1991, Noyer in press).

    Finally, we fail to understand the relevance of the third argument ("it ['il'] is restricted to subject function")

    Julie Auger Dept. of Linguistics, McGill University

    Philip Miller English and General linguistics, Universit de Lille 3

    Auger, Julie. 1993. "More evidence for verbal agreement-marking in Colloquial French". In W. J. Ashby et al. (eds.), Linguis- tic Perspectives on the Romance Languages. Amsterdam: Benjamins, pp. 177-198. Auger, Julie. 1994. "Subject-clitic inversion in Romance: A morpho- logical analysis". Paper presented at LSRL 24, Los Angeles. Auger, Julie. 1994. Pronominal Clitics in Quebec Colloquial French: A Morphological Analysis. Ph.D. dissertation, University of Pennsylvania. Auger, Julie. in press. "Les clitiques pronominaux en francais parle informel: une approche morphologique". To appear in Revue quebecoise de linguistique. Cummins, Sarah & Yves Roberge. 1994. "A morphosyntactic analysis of Romance clitic constructions". In M. Mazzola (ed.), Issues and Theory in Romance Linguistics. Washing- ton, D.C.: Georgetown University Press, pp. 239-257. Fulmer, S. Lee. 1991. "Dual-position affixes in Afar: an argument for phonologically-drivent morphology". In A. I. Halpern (ed.), The Proceedings of the Ninth WCCFL Stanford: CSLI, pp. 189-203. Miller, Philip. 1992. Clitics and Constituents in Phrase Structure Grammar. New York, Garland (Doctoral Dissertation, University of Utrecht, 1991). Miller, Philip H. 1992. "Postlexical cliticization vs. affixation: Coordination criteria". Proceedings of the 28th meeting of the Chicago Linguistic Society, pp. 382-396. Miller, Philip H. and Ivan A. Sag. to appear. "French clitic movement without clitics or movement". To appear in Revue quebecoise de linguistique. Noyer, Rolf. in press. "Mobile affixes in Huave: optimality and morphological well-formedness". To appear in E. Duncan et al. (eds.), Proceedings of the Twelfth WCCFL. Stanford: CSLI.

    Message 2: Re: 5.1074 French clitics

    Date: Tue, 4 Oct 1994 12:11:42 -Re: 5.1074 French clitics
    From: Lambrecht Knut <lambrechspot.Colorado.EDU>
    Subject: Re: 5.1074 French clitics

    The recent posting by Pierre Larrivee about French clitics reminds me of an exchange on the Linguist some time ago, about the same topic, triggered (as far as I remember) by a contribution by Julie Auger. I don't think it'll be very useful to start another debate on the century-old question of whether French "clitics" are pronouns or "morphemes" (to use Pierre Larrivee's term), and I have tried to justify my skepticism in a reply to the posting by Julie Auger.

    But I can't help reacting to the following remark by Pierre Larrivee:

    > There is one further cost to this analysis. Dislocations in > general can be put pretty much anywhere in a sentence ("(Les enfants) je les > ai vus (les enfants) qui foutaient le camp (les enfants)"); so French would > then not only be pro-drop but also non-configurational!!! I invite linguists > to ponder seriously on such a proposal, and i hope those few elements here > will help them make an enlightened choice.

    Independently of the issue of whether LES in the above example is a morpheme/inflection marker/clitic/bound pronoun or not, I take issue with the claim that detached (dislocated) NPs can go "pretty much anywhere in a sentence". I have discussed the issue of the syntactic status of left and right detached constituents in some detail in my 1981 monograph 'Topic, antitopic, and verb agreement in non-standard French, and a little more rigorously in my 1986 UC Berkeley dissertation, and I think I have shown that the position of such constituents is severely constrained, in a way that is exactly what we would expect from topic constituents.

    Essentially, LEFT-detached NPs are outside of (what used to be called) S', i.e. they precede the unit [COMP - S] (i.e. the TOP position of Chomsky 197(?), "On WH-movement", sorry I forgot the exact reference). They have severe restrictions on embedding, as shown e.g. in

    *Les films que Pierre il aime le plus ont tous ete interdits.

    with variation depending on the degree to which the propositional content of the embedded clause is asserted or presupposed, and with a certain subject-object asymmetry ("subject" dislocated NPs are easier to embed than "object" ones). (This is of course water on the mill of those who want to see detached NPs as "regular subjects".) So while in the relative clause island above the detachment is very bad, it's less bad, and sometimes perfectly ok, in complement clauses, such as

    Je crois que les enfants ils ont encore casse de la vaisselle.

    but less good when the NP corefers with an object pronoun, as in

    ?Je crois que la vaisselle les enfants ne l'ont pas encore faite.

    and positively awful in

    ?? Avant que la vaisselle je la fasse je veux finir de manger.

    More interesting perhaps is the case of RIGHT-detached constituents, such as the two occurrences in Pierre Larrivee's example sentence. These constituents are chomsky-adjoined to the minimal S, and as a result they have no restriction whatsoever on embedding:

    Les films dont je lui ai parle, a Pierre, (ils) ont tous ete interdits.

    and they show no subject-object asymmetry. The right-detached NP (I call it "antitopic NP") is always POSTFOCAL , i.e. it must follow the focus accent in the clause, which typically coincides with the last constituent of S, so we get:

    Quand est-ce que tu as vu les enfants? Je les ai vi HIER, les enfants (ok) # Je les ai vus les enfants HIER (bad)

    Qu'est-ce qu'ils ont fait les enfants? Ils ont casse de la VAISSELLE, les enfants. (ok) *Ils ont casse les enfants de la VAISSELLE. (real bad)

    This is so because detached constituents are topics, hence outside the focus domain, which in French is the minimal clause.

    There is one apparent exception: right-detached NPs can precede direct object complements as long as these are clausal, even if the clausal complement is focal:

    Je lui ai DIT, a Pierre, que je n'avais pas envie de le VOIR.

    but not:

    *Je lui ai dit a Pierre quelques MOTS.

    This is reminiscent of the well-known fact that locality constraints are different between V+NP and V+S', as observed also in English (see e.g. Peter Sells' "Lectures", somewhere on "adjacency"). Nevertheless it's an exception to my nice simple position rule, and it bothers me. (Anyone out there who can help me to be less bothered?)

    Pierre Larrivee's relative-clause example is NOT a counterexample. The example he gives (Je les ai vus les enfants qui foutaient le camp) does not involve a complex NP relative but a perception-verb relative construction, in which the relative clause is a sister (sic) to the "main" clause (see the section on French relative clauses in my dissertation), hence the right-detached NP "les enfants" is in its expected post-clausal and post-focal position.

    Enough already. There is much more to say on these constructions, but the above may give an impression. I hope to present a coherent argument on all this soon in a book I'm working on ("The syntax of pragmatics in spoken French", mmmmhhh!).

    The moral is: detached NPs are extra-clausal or extra-sentential, except in very specific cases, where the NP bound to a subject pronoun does behave more or less like a subject NP. (But not in all respects! see my earlier posting in reply to Julie Auger).

    Knud Lambrecht

    (UT Austin, presently CU Boulder)