‘French dislocation is by essence an interface phenomenon’ (216). In French dislocation: Interpretation, syntax, acquisition, Cécile de Cat sets out to substantiate this claim by offering a thorough and all-encompassing analysis of left and right dislocation phenomena in spoken French, the latter of which is illustrated in (1), with the dislocated phrases given in bold.

1. (1) Elle_i était vraiment bien, son_jexposition_i, àJulia_j. (1, ex. (1.1b)) she was really good her exhibition to Julia ‘Julia's exhibition was really good.’

De Cat's main claim is that French dislocation is best analyzed as a pragmatically motivated phenomenon requiring minimal syntactic machinery. As she argues, spoken French is a discourse-configurational language in the sense of É. Kiss (Reference É. Kiss1987), where topics (and foci) are obligatorily dislocated. This is demonstrated by showing that in a large corpus of spoken French, all sentences containing individual-level predicates, which force a topic interpretation of their subjects, occur with dislocated subjects. In addition, de Cat shows that dislocated noun phrases have the syntactic and semantic properties usually associated with topics. Thus, the use of indefinites in dislocation is restricted to generic indefinites (as in (2a); cf. (2b), which can have only a specific reading) and d-linked indefinites, mirroring the constraints on indefinites in topic positions. Also, dislocated expressions are interpreted as restrictors on the domain of application of the predicate (in the sense of Reinhart Reference Reinhart1981), just like topics are.

1. (2)

   1. (a) Unenfant_i, il_i arrive pi il te pose une question. (21, ex. (2.32)) a child he arrives then he to.you asks a question ‘A child arrives and he asks you a question.’

   2. (b) *Unenfant_i, il_i est arrivé pi il t' a posé une question. (22, ex. (2.33)) a child he is arrived then he to.you has asked a question

This obligatory dislocation, however, is not derived by movement or driven by a [+topic] feature. As de Cat convincingly shows, French left and right dislocation do not show movement effects according to any of the traditional diagnostics, including weak crossover, parasitic gap licensing, reconstruction, relativized minimality, islands, and narrow scope with respect to negation. Rather, topics are base-generated in their surface position, adjoined to a maximal projection with root properties. Consequently, the ‘resumptive’ pronouns (e.g. elle and son in (1)) that are associated with the dislocated elements are not real resumptives, as they are not remnants of movement. They are true pronouns, which simply co-refer with the dislocated elements.

This claim has several implications. First, as the author shows, as long as the referent is recoverable from the context, dislocated elements can be dropped without causing ungrammaticality or a change in meaning. Second, dislocated elements and their associated pronouns do not need to agree in number, gender or case, since they are not syntactically linked, cf. (3). Any pronoun that can co-refer with a noun phrase across sentences can also be used to ‘resume’ it in dislocated constructions.

1. (3) Le lait_i, j'adore ça_i. (109, ex. (4.14c)) the milk I.love that ‘I love milk.’

Third, de Cat argues against a morphological analysis of subject clitics as agreement markers in spoken French (e.g. Jaeggli Reference Jaeggli1982, Roberge Reference Roberge1986, Miller & Monachesi Reference Miller, Monachesi and Godard2003). According to the morphological account, sentences like (4) do not involve dislocation, but rather a noun phrase (les clitiques ‘the clitics’) in the canonical subject position and a subject clitic that acts as an agreement marker.

1. (4) Les clitiques_i ils_i comptent pour du beurre. (1, ex. (1.2)) the clitics they count for some butter ‘Clitics don't count.’

De Cat provides convincing arguments to show that the morphological analysis of subject clitics is empirically not tenable and has unwelcome theoretical consequences. First, she argues, such an analysis makes the agreement system redundant for the first and second person plural forms, where the verb carries overt person and number marking. Second, subject clitics can move in syntax, which is unexpected of agreement morphemes. Equally unexpected is a third phenomenon, namely that subject clitics can be separated from the verb to which they belong by other clitics, such as y ‘there’ and en ‘of it/them’, object clitics and negation (e.g. Je la luidonnerai, literally ‘I it to-him will-give’; 15, ex. (2.15a)). Fourth, spoken French does not allow subject doubling. If the subject noun phrase is focused, as in (5), where chaque ‘every’ can bind the subject noun phrase if and only if the latter is focused (Zubizarreta Reference Zubizarreta1998), no resumptive subject clitic is acceptable.

1. (5) Sa_i mère (*elle) accompagnera chaque_i enfant. (24, ex. (2.38)) his mother she will.accompany each child ‘His mother will accompany every child.’

In the traditional, movement-based accounts, left dislocation has been straightforward to derive, as syntactic movement is generally assumed to go leftwards. Right dislocation, which, as de Cat's corpus shows, is as frequent in spoken French as left dislocation, has always been more problematic for movement accounts to explain. Under de Cat's adjunction analysis, right and left dislocation receive a uniform account. The relatively small differences between them are derived from the differences in the nature of the two peripheries they target. Thus, the fact that left and right dislocation show different prosodies – left dislocation being more prominent with high pitch and a clear break from the sentence, and right dislocation being prosodically non-salient – falls out from the fact that the left and right peripheries of utterances are prosodically different. The left periphery is typically associated with higher pitch and more emphasis, whereas the right edge is usually characterized by lower pitch. Consequently, dislocation to the left periphery is preferred when the topic requires prosodic emphasis, whereas right dislocation is applied when no such prominence is necessary. According to the author, this functional/prosodic difference also accounts for the slightly different semantics involved in right and left dislocation. The fact that left-dislocated elements are often interpreted with a contrastive reading is suggested to be associated with the increased prosodic prominence of the left periphery. Thus, contrastive elements that require prosodic prominence will be dislocated towards the left periphery. Unfortunately, the difference in agreement patterns, whereby number, gender and case agreement between the resumptive and the dislocated element is preferred (although not required) in right dislocation, but dispreferred (although not impossible) in left dislocation, remains unexplained.

De Cat's adjunction analysis is backed up by acquisition data. Analyzing the early productions of four children from France, Canada and Belgium between the ages of 1 year and 9 months to 3 years and 1 month, the author shows that early dislocated structures exhibit the same prosodic and semantic–pragmatic properties as in the adult language; for example, they are used to express topics and are realized using dislocation prosody. This constitutes strong evidence in favor of the base-generation and adjunction account of dislocation, because children around their second birthday, who readily produce dislocated structures, do not show any evidence for the presence of a C(omplementizer)P(hrase) layer or other functional projections beyond that, where topics are traditionally assumed to move.

One of the particular strengths of the research reported in the volume is that it uses experimentally elicited judgements as well as corpus data as its empirical basis. Well-controlled acceptability judgements, here obtained from over a hundred speakers, and spontaneous production from diverse speakers representing different dialects are rarely used in the formal literature and thus constitute a uniquely reliable empirical background to the analysis being presented.

However, the author does not go all the way towards implementing experimental method, failing to conduct statistical analyses on the large quantity of data she reports. Using large corpora and responses from numerous speakers necessarily introduces variability into the data. In the absence of the appropriate statistical tests, it becomes impossible to decide whether different patterns in the data derive from experimental noise (i.e. are simply due to chance) or whether they represent true linguistic variation. Various remarks in the text indicate the author's difficulty in analyzing and interpret empirical variability:

The author's treatment of empirical data suffers from another flaw, as the definition of the scope of her investigation, i.e. spoken French, becomes circular. Central to her analysis is the claim that spoken French is a discourse-configurational language, where topics are obligatorily dislocated. Potential counter-examples, i.e. sentences where the topic is not dislocated (but, for example, appears in the canonical subject position, as in (2a) above), are dismissed as too formal in register to be part of spoken French. While the specific examples the author cites might be judged too formal for a variety of reasons (some of which might be independent of the absence of dislocation, e.g. choice of vocabulary), such an ad hoc and circular treatment of counter-examples leaves the reader with a feeling of unease and invites a definition of spoken French that is more precise than the brief discussion in chapter 2 of what is not spoken French.

These empirical issues aside, de Cat's monograph presents a thorough, carefully argued and coherent analysis of dislocation phenomena in French, displaced from syntax into the realm of the pragmatic/discourse interfaces.