Natural language abounds with utterances that – at least on the surface – seem fragmentary or incomplete. Consider, for example, B's reply in the dialogue in (1).
- (1)
A: Who killed Mr. Blue?
B: Mr. Green.
By uttering the phrase Mr. Green, the second speaker in this dialogue wants to convey something like Mr. Green killed Mr. Blue. In other words, there is a discrepancy between the meaning of this sentence and its phonetic form: the former is clausal and the latter phrasal. One of the central research questions surrounding fragmentary utterances such as these is how to map sound/syntax onto meaning. It is this question that forms the main focus of The syntax of nonsententials.
Ever since the earliest generative work on fragments, two possible approaches towards answering this question have been explored. The first is to assume that in a dialogue such as (1) ellipsis has taken place. Speaker B's reply has the syntactic structure of a full clause, but part of that structure remains unpronounced (cf. (2)). The mapping from syntax to semantics is now entirely compositional, just as in a non-elliptical reply.
- (2)
A: Who killed Mr. Blue?
B: Mr. Green killed Mr. Blue.
The alternative line of analysis, sometimes referred to as the nonsentential approach, is to assume that what you see is what you get – i.e. at all levels of syntactic representation, B's reply in (1) simply consists of a nominal phrase. As a result, the propositional meaning of this utterance does not fall out from the syntax–semantics mapping, but instead is derived via pragmatic or discourse-related inference rules. As is clear from the title of the book, this is the approach adopted and defended in The syntax of nonsententials.
The book consists of twelve chapters, together with an ‘Introduction’ and epilogue (‘Epilogue: Wherefrom and whereto?’) by the editors. As pointed out in the introduction (4f.), the goal of this volume is twofold. On the one hand it wants to develop and defend a nonsentential analysis of fragmentary data within the theoretical framework of the Minimalist Program. On the other hand, it wants to extend the coverage of the analysis by looking at different registers, child language, second language acquisition, aphasia, pidgins and creoles. This double goal is reflected in the organization of the book: the first five chapters are devoted to outlining and defending the nonsentential analysis as well as evaluating ellipsis-based alternatives, and the next seven chapters apply the nonsentential analysis to the domains mentioned above. In the epilogue, the editors attempt to bring all the strands together and identify a number of core issues for future research. In what follows I summarize and discuss each part of the book separately, focusing first – and more extensively – on the nonsentential analysis.
In chapter 1, ‘Toward a nonsentential analysis in generative grammar’, Ellen Barton gives a historical perspective on the nonsentential vs. ellipsis debate. She introduces and evaluates some of the earliest proposals in both theories, thus setting the scene for the rest of the volume.
The main theoretical contribution of the book lies in chapter 2, entitled ‘The syntax of nonsententials: Small clauses and phrases at the root’, by Ljiljana Progovac. She focuses on data such as those exemplified in (3).
- (3)
(a) Nice lady!
(b) Problem solved.
(c) Him worry?!
Progovac's main claim about such examples is that they involve a base-generated phrase (such as the Noun Phrase (NP) in (3a)) or a small clause (e.g. the V(erb)Ps in (3b–c)). The reason why such phrases converge at the interfaces is because they do not contain any unchecked features. In particular, the verbs are tenseless and hence do not require the projection of a T(ense)P, and the small clause subjects have a default case feature, which obviates the need for merging a case assigner (i.e. finite T°). In English, default case is manifested as accusative on pronouns (cf. (3c)) and through determiner drop in lexical noun phrases (as in (3b)). Given that both tenselessness and caselessness of subjects require the absence of T(P), the two are predicted to correlate. As pointed out by Progovac (38), data such as those in (4) seem to bear out this correlation.
- (4)
(a) Problem solved.
(b) The problem is solved.
(c) ?*The problem solved.
(d) ?*Problem is solved.
These examples show that finite tense and nominative case are either both absent (as in (4a)) or both present (as in (4b)). Given that each of them independently requires the merger of T(P), not having the other leaves T° with unchecked (tense or case) features and causes the derivation to crash (as in (4c–d)).
Chapter 3, ‘“Small structures”: A sententialist perspective’, by Jason Merchant, is the only one that defends an elliptical approach to fragmentary utterances. The chapter focuses mainly on question–answer pairs such as the one in (5) and to a lesser extent on discourse-initial fragments like the one in (6).
- (5)
Abby: Who is Sarah bringing¿
Ben: Alex.
- (6)
[context: Abby and Ben are at a party. She sees an unfamiliar man with Sarah, a mutual friend of theirs, and turns to Ben with a puzzled look on her face. Ben says:]
Some guy she met at the park.
Merchant argues that the fragments in (5) and (6) are derived from fully sentential sources through a combination of movement and ellipsis. In particular, the D(eterminer)P Alex in (5) first undergoes A′-movement to the left periphery, with subsequent ellipsis of the I(nflection)P leading to the non-pronunciation of the rest of clause, cf. (7).
(7) Alexi [IPSarah is bringing ti]
In discourse-initial contexts like (6), where there is no linguistic antecedent, Merchant assumes that a maximally simple IP consisting of nothing but a pronoun and a copula is elided. This yields the representation in (8).
(8) Some guy she met at the parki [IPhe is ti]
The evidence in favor of this analysis is twofold. On the one hand, fragments often display connectivity effects linking them to a clausal source. For example, in languages with morphological case, a fragment generally shows up in the case that it would bear in a non-elliptical reply. On the other hand, the idea that the fragment has undergone movement is supported by the fact that it is sensitive to well-known restrictions on movement such as islands.
The next two chapters offer a direct critique of Merchant's paper. In chapter 4, ‘Neither fragments nor ellipsis’, Robert J. Stainton criticizes Merchant's analysis of discourse-initial fragments (cf. (6)–(8)). He points out that Merchant's account overgeneralizes in that it seems to allow for the ellipsis of maximally simple IPs of the type in (8) in contexts where this is not allowed (e.g. embedded clauses). Moreover, determining the possible content of such IPs turns out to be much more problematic than indicated: while Merchant only allows for it is and do it, Stainton argues that this yields only a subset of the possible discourse-initial fragments. Finally, he also notes some empirical disagreement about the island data.
In chapter 5, ‘Big questions, small answers’, Eugenia Casielles tackles Merchant's account of question–answer pairs and proposes a nonsentential account instead. She shows that there are fragments in which case connectivity is not respected, or that do not seem to have a straightforward clausal source, cf. (9).
- (9)
A: Who ate the cake?
- B:
(a) Me.
(b) *Me ate the cake.
Moreover, she points out that fragmentary answers to wh-questions are attested in child language at a stage when A′-movement – a necessary ingredient of Merchant's analysis of fragments – is still lacking. This concludes my summary of the first half of the volume.
As pointed out above, the main theoretical contribution of The syntax of nonsententials lies in Progovac's paper. Her small-clause analysis is explicitly adopted by all the other papers in the book (except Merchant's, of course), including those in the second part (to be discussed below). The analysis revolves around the claim that the machinery of the Minimalist Program allows one to construct utterances that do not involve the merger of TP. As long as the derivation does not contain any unchecked or unvalued features, it converges at the interfaces and the resulting utterance is well-formed. This is an interesting approach, which capitalizes on the derivational bottom-up nature of Minimalist theorizing. The consequences Progovac derives from this analysis, however, are much less worked out or well-established. She argues that from the absence of TP it follows straightforwardly that fragments are tenseless and caseless. While this certainly holds for the core case (finite tense and nominative), it raises questions in other areas. A first problem concerns Progovac's analysis of determiner drop as default case (cf. the discussion of (3b) above). This is not only an unorthodox approach that requires more substantiation than is given in the book, it also makes a number of incorrect predictions. Firstly, the contexts in which default case is attested are not identical to those where we find determiner drop. For example, while the former is found in coordinations, as noted by Progovac (50), the latter is not. Secondly, equating determiner drop with default case predicts there to be a distinction between subjects and objects. While the former are dependent on (finite) T for structural case, the latter are not. This means that only the former should be able to drop their determiner in fragments, contrary to fact (as acknowledged in footnote 10 on page 152).
A second area where the correlation between tenselessness and caselessness breaks down concerns embedded contexts. Specifically, while small clauses can be straightforwardly embedded, their subjects do not display default case in such environments. Progovac claims that an embedded small clause ‘becomes subject to sentential rules, including structural Case checking’ (41), but it is unclear that this is more than a reformulation of the facts. For example, the question arises as to what would cause a derivation containing an embedded small clause with a default case-marked subject to crash.
The two cases I have just discussed are indicative of a more general problem. In a number of respects, the nonsentential analysis remains too sketchy to be fully evaluated. The cause of this, I feel, is the editors' decision to include a paper (i.e. Merchant's) in which the ellipsis-based alternative is put forward. This leads to a 72-page excursion consisting of (i) an abbreviated version of an already published and widely accessible paper (Merchant Reference Merchant2004) and (ii) two commentaries on that paper that are mainly negative in tone, that both contain summaries of Merchant's paper, and that at times misrepresent the facts (e.g. Stainton's assumption (98) that the E-feature found in sluicing is the same as that found in VP-ellipsis) or even ignore them (e.g. Casielles not addressing Merchant's data from morphological case languages). Those 72 pages would have been much better spent further outlining and detailing the nonsentential analysis. As it stands, the reader gets the impression at several points in the volume that the two lines of analysis are simply discussing different sets of data, with the nonsententialists focusing on fragments displaying anti-connectivity effects, and the sententialists concentrating on the opposite pattern. This might indicate that both approaches are needed, but for different sets of data, and that the main challenge is to draw the dividing line between the two (as is suggested by Merchant in his paper); but the discussion of the Serbian case system in the epilogue seems to suggest that the editors want to extend the nonsentential analysis to all fragmentary utterances. In light of this, a more detailed and worked-out discussion of the central proposal seems indispensable.
In the second half of the volume, the empirical range of Progovac's nonsentential analysis is extended by applying it to data from different registers (chapter 6, ‘Extending the nonsentential analysis: The case of special registers’ by Kate Paesani), child language (chapter 7, ‘The narrowing acquisition path: From expressive small clauses to declaratives’ by Christopher Potts & Tom Roeper), second language acquisition (chapter 8, ‘Nonsententials in second language acquisition’ by Nicola Work), aphasia (chapters 9, ‘How language adapts to the brain: An analysis of agrammatic aphasia’ by Herman Kolk, and 10, ‘Nonsententials and agrammatism’ by Patricia Siple), pidgins (chapter 11, ‘Reduced syntax in (prototypical) pidgins’ by Donald Winford) and creoles (chapter 12, ‘Copula variation in Guyanese Creole and AAVE: Implications for nonsentential grammar’ by Walter F. Edwards). The breadth that is thus achieved is simply impressive. It is the first time that a volume on fragments includes data from so many subareas of linguistics. This means that to the extent that Progovac's account can be successfully extended to cover these data, it receives strong additional support. As always, however, there is a trade-off between breadth and depth, that is, the more (types of) data one discusses, the less detailed the discussion will be. Specifically, while it is well-known that the types of data just mentioned display certain superficial similarities with non-impaired adult fragmentary speech (e.g. use of non-finite verb forms, omission of determiners, etc.), the question remains how ‘deep’ or telling these similarities are. That question can only be answered by an in-depth comparison between the two data sets, and it is precisely such a comparison that is lacking in most of the papers. There are exceptions, of course, such as Siple's detailed discussion of the correlation between the lack of finiteness and subject omission in agrammatics (274f.), but on the whole the reader is struck not only by the similarities but also by the differences between the various sets of data; for example, recipes allow object drop much more easily than adult fragments (156f.), and in pidgins, copula omission and determiner drop might be due to first-language influence, not default case (299). Accordingly, it remains to be seen to what extent the small clause analysis can be successfully extended to these other areas.
All in all, though, The syntax of nonsententials is an interesting and innovative volume that will undoubtedly play a central role in the literature on ellipsis from now on. Not only does it represent the most substantial defense of the nonsentential analysis so far, it also provides a new stimulus for the debate by bringing in considerable amounts of new data.