LINGUIST List 5.1180

Tue 25 Oct 1994

Disc: French clitics

Editor for this issue: <>


  1. "Julie Auger", French clitics

Message 1: French clitics

Date: Tue, 25 Oct 1994 10:45:23 French clitics
From: "Julie Auger" <>
Subject: French clitics

 This posting is a joint response by Philip Miller and Julie
Auger to Pierre Larrivee's reply in Linguist 5.1132.

 First, three quick remarks are in order. We would like to
thank John Koontz for correctly pointing out that our arguments
about the status of so-called pronominal clitics in French are
mostly language-internal, not universalist (Linguist 5.1151). We
also want to stress that the use of the term "pronoun" in our
previous posting should not be taken to imply that we analyze
the 'pronominal clitics' of French as words: we do not. As we
tried to make clear in that posting, we claim that 'pronominal
clitics' in French are affixes, not words or clitics. Finally, we
are a bit puzzled by Larrivee's comment concerning the types of
criteria used in our preceding posting to defend our affixal
analysis: if morphological, positional, and syntactic
criteria are not relevant, then what is?

 One passage in Larrivee's posting indicates that part of his
disagreement with our proposal may stem from a misunderstand-
ing of what we mean by "lexically attached". Two remarks
concur to create the impression that he equates lexical attach-
ment with lexical listing or with derivational morphology to the
exclusion of inflectional morphology: "the lexicon isn't a
grammatica cloaca" and "lexical attachment amounts to say
that the lexicon is deriving syntactic units, which mixes
unwelcomely levels of analysis". We obviously do not mean that
"je le mange" ('I eat it') should be listed as a lexical item in the
lexicon. What we mean is that verb-clitic combinations are
generated by the same types of rules that handle derivational and
inflectional morphology. Only totally unpredictable forms must
be listed in the lexicon, such as "chus" ('I am'). Other forms are
generated by productive morphological rules. We reject both a
lexical-listing analysis and syntactic generation of what we claim
to be morphologically complex words. The latter, syntactic
approach violates strict lexicalism (though it is accepted in the
GB tradition). The misunderstanding here may stem again from
the fact that we have kept the traditional term 'pronominal clitic'
to designate items which we believe to be lexically attached.
Note that in any case, it is completely normal for inflectional
morphology to have important effects on the possible syntactic
contexts in which a word can appear. Finally, even though
Larrivee does not seem to like crosslinguistic comparison, we
feel compelled to point out that the morphological mechanisms
necessary to handle the morphologically complex verbs of
French in our analysis are all independently motivated by
the existence of numerous highly-inflected languages (as pointed
out by e.g. Koontz and Guy in their previous posts) and by the
existence of languages where subject affixes fill the role of true

 As we tried to make clear in our previous posting, one
main motivation for proposing a morphological treatment of
French "clitics" concerns the fact that no syntactic account has
yet been proposed which can derive the facts. We will not
repeat these arguments here (cf. our previous postings and the
references in it). Simply claiming that "the existing arguments
for an inflectional treatment of French clitics cannot be seriously
used to promote such a treatment" does not suffice to convince
researchers that the inflectional approach must be abandoned.
Empirical problems must also be identified. In this vein,
Larrivee discusses cases of what he calls "multiple dislocation"
as well as cases where non-clitic elements double dislocated
objects. Unfortunately for him, and as Koontz has already
pointed out in his own posting, these examples are no more
problematic for our inflectional approach than for a syntactic
approach. Very little work has been done on apposition and on
multiple dislocation, and we will not venture an analysis here
without more serious investigation.

 The question of whether a certain affix (in the case at hand a
'pronominal clitic') is an agreement marker or an argument
marker (the latter would be called an 'incorporated pronoun'
under certain analyses) depends on whether the affix bears
argument status or not, that is, whether it fills the subject or the
object role itself, or is just a grammatical morpheme which must
share some features with the subject or object. In the terms of
Jaeggli 1982, for instance, a 'clitic' that 'absorbs case' is not
an agreement marker but an argument marker: inflecting the
verbal stem for the 'clitic' removes the case marking potential of
the verb and forbids the presence of the corresponding NP with
argument status. A 'clitic' that is an agreement marker does not
absorb case in this sense. Thus a verb with such an agreement
marker requires a corresponding NP to appear in normal
argument position. Using GB terminology descriptively, without
taking position on the appropriate analysis, the required argu-
ment NP can be a full NP or it can be pro, i.e. a null pronoun.
Furthermore, if the 'clitic' is an agreement marker, then it must
appear doubling any full NP in an argument position. All theo-
ries have to allow for the distinction between agreement markers
and argument markers in this sense, given that both situations
are well attested cross-linguistically.

 Auger argues that only subject clitics in Quebec Colloquial
French are agreement markers, because they are the only ones
which truly allow a lexical argument to cooccur with them and
because they are the only ones to show up in all contexts where
regular agreement markers usually occur. Sentences containing
both a subject clitic and a lexical subject such as (1)a below are
thus analyzed as basic SVO sentences and that is all, really, that
there is to say about them. We would like to mention the
possibility that, in this system, a truly left-dislocated version of
(1)a might look like (1)b, even though we do not want to com-
mit ourselves to this analysis at this point.
 (1) a. Pierre il veut pas
 'Pierre he wants not' = 'Pierre doesn't want to'
 b. Pierre, lui, il veut pas
 'Pierre, him, he wants not'
 = '(As for) Pierre, he doesn't want to'

 Sentences like (2), containing an object clitic and a lexical
object, however, are pragmatically marked and are thus analyzed
as dislocations.
 (2) a. Je l'ai vue, Marie
 'I her have seen, Mary' = 'I have seen her, Mary'
 b. Marie, je l'ai vue
 'Mary, I her have seen' = 'Mary, I have seen her'

 There is no room here to discuss the arguments behind this
distinction; let us just mention that such criteria as word order,
occurrence of pronominal clitics in extraction contexts (i.e.,
relative clauses and wh-questions), and intonation, clearly distin-
guish subjects and objects (cf. Auger 1994 and in press).

 This distinction between agreement marker and argument
marker has the following consequence concerning null argu-
ments: Quebec French is a null-subject language [i.e. in descrip-
tive GB terms, it allows pro subjects], but not a null-object
language [i.e. doesn't allow pro objects]. So, Larrivee is wrong
when he says that, in our analysis, "il lui a parle" does not bear
case; "lui" bears (or 'absorbs', if that terminology is preferred)
dative case and that's why it does not allow doubled indirect
objects, but only dislocated ones. It may be correct, however,
depending on your analysis, that "j'ai mange" '1sg.ate' does not
have any case-marked subject. It is, however, a grammatical
sentence, just as "ho mangiato" can constitute a grammatical
sentence in Italian. Here, the issue of whether case marking is
present or not in a sentence consisting of only a verb with
its inflectional markers depends on one's analysis of null argu-
ments. In a classic GB type analysis, the argument position
would be filled by pro, and would be case marked.

 Our conclusion is that syntax cannot give a principled
account of French pronominal clitics because they are not syn-
tactic but rather affixal elements. And we agree with Janda &
Kathman's 1992 conclusion that any attempt to have syntax
handle these facts would impose an unfair burden on syntax. A
diachronic approach to the study of Romance clitics such as
the one adopted by Fontana 1993 lends strong support to the
approach that we have been promoting. In this respect, we
would like to concur with Koontz's remark on the real possibility
that traditional French orthographic conventions play a role in
the gut feeling certain people have against the inflectional
analysis of the 'clitics'. The French tradition of writing the
pronominal clitics as separate orthographic words goes back to
Old French. At that period, both subject and object pronouns had
radically different properties from those that their present day
descendants have, and were clearly independent words. It should
be recalled in this context that French orthography is in general
extremely conservative, keeping traces of morphological distinc-
tion that have been lost in the spoken language for up to 700

 One final note. It is simply not true in normal 20th century
French (as we stated in our initial post) that 'words can be
inserted between them [the clitics] and their verbal head', as
claimed by Larrivee in his latest message. Larrivee's example,
"ne pas dire" ('not to say') is irrelevant since it involves 'ne'
which is not a pronominal clitic. Providing an analysis of 'ne' is
beyond the scope of this post. Let us simply note that in any
case 'ne' is disappearing from the spoken language (see Sankoff
& Vincent's 1980:299, who report an "omission" rate of 'ne' of
99.5% in Montreal French and Miller 1992,p.224, fn. 65 for
some comments).

 In conclusion, what proponents of a syntactic analysis of
French pronominal clitics should do is show that a principled
syntactic analysis is possible and that it does not require positing
ad hoc rules and principles only for the purpose of handling the
pronominal clitics. All such attempts have so far failed.

Philip Miller English and General Linguistics,
 Universite de Lille 3

Julie Auger Department of Linguistics,
 McGill University

 New references
 Fontana, Josep M. 1993. Phrase Structure and the Syntax of
 Clitics in the History of Spanish. Ph.D. dissertation,
 University of Pennsylvania. (IRCS Report 93-24)
 Jaeggli, Osvaldo. 1982. Topics in Romance Syntax.
 Dordrecht: Foris.
 Janda, Richard D. & David Kathman. 1992. "Shielding
 morphology from exploded INFL". In Costas Canakis,
 Grace Chan, & Jeanette Denton (eds.), CLS 28.2:
 Proceedings of the Parasession on the Cycle in Linguistic
 Theory. Chicago: Chicago Linguistic Society, pp. 141-157.
 Sankoff, Gillian & Diane Vincent. 1980. "The productive use
 of 'ne' in spoken Montreal French". In G. Sankoff. The
 Social Life of Language. Philadelphia: University of
 Pennsylvania Press, pp. 295-310.
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue