This volume is a collection of eight articles, part of the outcome of a five-year project funded by the British Arts and Humanities Research Council and entitled ‘Null subjects and the structure of parametric theory’. The introduction by Ian Roberts & Anders Holmberg is a thorough literature review of the pro-drop parameter, taking us as far back as Perlmutter (Reference Perlmutter1971), who was the first to link null subjects to the presence of person morphology on the verb. Roberts & Holmberg show that parameters as defined in generative grammar are powerful tools, which need to be constrained while maintaining their descriptive power. Contra Newmeyer (Reference Newmeyer and Pica2004), Roberts & Holmberg (and Theresa Biberauer and Michelle Sheehan too) promise to restore faith in the Government and Binding notion of parameters by refining these through the Minimalist framework.
In my view, this is a very thought-provoking volume with exceptional theoretical rigour. The authors not only propose typologies worth pursuing, but they also open up avenues for future research. For this review, I have decided to provide a summary of every article in this volume for the reasons that (i) the contributions are very different from each other and do not always entertain a coherent set of theoretical assumptions, and hence (ii) a detailed account of each chapter should provide signposting and navigation points for readers.
Chapter 1, ‘A deletion analysis of null subjects’, by Ian Roberts, investigates the status of pro in rich agreement null subject languages (NSLs). Instead of thinking of pro as a null pronoun with inherently unspecified φ-features, Roberts returns to Holmberg's (Reference Holmberg2005) idea that pro is a deleted pronoun. Unlike Holmberg (Reference Holmberg2005), he takes pro to be a weak pronoun, not a φP but a D(eterminer) which occupies the specifier position of the Tense phrase (SpecTP). Thus, pro checks the EPP-feature. Evidence for this claim comes from object–verb NSLs like Turkish. Pro is defective, given that its D- and φ-features are included in those of T (a subset of the features of the probe), and as a defective goal, it deletes. Roberts further proposes that ‘consistent NSLs’ (which permit null subjects in all tenses and persons) are the ‘inverse of radical pro-drop languages’ (which allow null subjects as well as other null nominal arguments) (85):
a. Radical pro-drop is possible if φ-agreement is not obligatory.
b. Consistent null subjects are possible iff there is no impoverishment of T's φ-features [in the sense of Müller Reference Müller, Brandt and Fuss2005].
(85, ex. (34)).
In consistent NSLs, T has a D-feature and pro is required, but by virtue of being defective it remains null. English, on the other hand, has impoverished φ-features on T- it lacks both a D-feature and pro. Radical NSLs lack φ-features on T altogether (just like English), but the trigger for null subjects and null objects involves other deletion processes (83). In Roberts's analysis, pro is then no longer an empty category, but a pronoun that gets deleted. Just as Minimalism has dispensed with traces, other empty categories (such as pro and PRO) may also be redundant.
In Chapter 2, entitled ‘Null Subject Parameters’, Anders Holmberg puts forward a typology of null subjects for consistent and partial NSLs. Consistent NSLs like Italian and Greek have T with an unvalued D-feature (D=definiteness) among their other φ-features. T seeks a goal to value these features. All features get valued by the subject, which incorporates into T. In the case of an overt subject, the subject moves to SpecTP, satisfying the EPP-feature. In the case of a null third-person pronoun, the subject is D-less with valued φ-features. The uninterpretable D-feature of T is valued by an A(boutness-shift)-topic situated in the C-domain (following Frascarelli Reference Frascarelli2007); there is no movement of the subject pronoun, just a chain headed by T and spelled out as an inflection on the verb. The A-topic satisfies the EPP on T. In the case of first- and second-person null subjects, the D- feature of T is valued by the speaker/addressee feature, and the EPP-feature is checked by the null A-topic. In partial NSLs like Brazilian Portuguese and Finnish, finite T does not bear an uninterpretable D-feature. In these languages, a bare φP subject can still be incorporated into T as a result of Agree with T, but it will not be interpreted as definite but as generic or arbitrary. Embedded null subjects in partial NSLs lack a D-feature and do not incorporate into T. They can remain null because they have a local antecedent. As for partial NSLs with first- and second-person null subjects, they are derived just like definite null subjects in consistent NSLs. All in all, Holmberg postulates three parametric distinctions: [+/−] D in T; [+/−] P (PF EPP) in T, which, when present requires the overt presence of an element in SpecTP (as in English); and [+/−(φ-dependent)] EPP.
In Chapter 3, ‘Control into finite clauses in partial null-subject languages’, Anders Holmberg & Michelle Sheehan shed light on the null subject properties of Brazilian Portuguese, Marathi and Finnish. Embedded subjects in these NSLs may be null, provided they are bound by some antecedent in a higher clause. This binding relation, however, does not observe the strict locality conditions discussed by Landau (Reference Landau2004) for Hebrew PRO and finite control. Thus, the null subject may be bound by an antecedent two clauses up, allow split antecedents or sloppy readings. Akin to the analysis in Chapter 2, Holmberg & Sheehan's proposal is that the uninterpretable D-feature of the null pronoun enters an Agree relation with a c-commanding antecedent in a higher clause (provided there is no intervening DP which could block the formation of the chain). The null pronoun moves to the edge of the CP-phase, thereby making it accessible to its antecedent. As the authors themselves point out, this analysis predicts stricter locality between the antecedent and the null subject than is found in these partial NSLs – an issue which is left for future investigation.
In Chapter 4, ‘Semi null-subject languages, expletives and expletive pro reconsidered’, Theresa Biberauer focuses on Germanic languages and tries to establish whether they require an expletive pro. She argues that within the Minimalist framework, an expletive pro may not be justified, and shows for Germanic that (i) SpecTP may not always accommodate subjects, and (ii) the EPP can be satisfied by a variety of categories. Biberauer proposes a typology according to which languages differ as to whether they exhibit V-to-T, VP-to-SpecTP or DP-to-SpecTP raising, deriving the different movement operations from different feature compositions of T (165). Some languages are D-oriented and satisfy the EPP nominally, whereas others are V-oriented and satisfy it via V-movement. Grammars of languages with T-specifications that do not require DP-raising to SpecTP will not require expletive elements. In Germanic, only English requires a DP to raise to SpecTP. Other Germanic languages do not require a subject in SpecTP, as evidenced by subject-raising and expletive insertion contexts in German, Dutch, and Afrikaans. In Afrikaans, ‘verb-early’ and verb-final word orders provide evidence that SpecTP is not exclusively reserved for subjects, but, instead, hosts the whole vP (including the subject). In German and Dutch, SpecTP is only optionally occupied by a subject (optional EPP requirement on T). The lack of expletive pro in these Germanic languages questions further the existence of NSLs with quasi-argumental and non-argumental pro.
In Chapter 5, Anders Holmberg looks into ‘The null generic subject pronoun in Finnish: A case of incorporation in T’. This type of G(eneric)-pronoun is characteristic of partial NSLs like Brazilian Portuguese, Hebrew, and Marathi. The G-pronoun receives Case, triggers agreement, binds anaphors and behaves like a subject – except that it cannot satisfy the EPP in Finnish, which requires checking by either an expletive or an XP with topic status. The G-pronoun lacks topic content so that an XP is required to appear clause-initially. G is probed by T, but is not a copy of T because T has a D-feature, whereas G does not. The features of the G-pronoun are a subset of the features of T, and T and G form a chain, but there is no movement of G-to-T. Since G does not head the chain, G cannot be spelled out and cannot check the EPP-feature of T.
In Chapter 6, ‘“Free” inversion in Romance and the Null Subject Parameter’, Michelle Sheehan analyses cases of inversion in Spanish, Italian and European Portuguese. Evidence from constructions with unaccusative/unergative verb–subject, XP–verb–subject, and verb-first word order in Spanish reveal that V-to-I movement for EPP-satisfaction does not always apply (contra Alexiadou & Anagnostopoulou Reference Alexiadou and Anagnostopoulou1998). Following Chomsky (Reference Chomsky, Freidin, Otero and Zubizarreta2008), Sheehan assumes that T inherits its agreement features from C. Adopting Rizzi's (Reference Rizzi and Haegeman1997) split-CP system, she further assumes that the EPP-feature occurs in different projections in different languages (FinP for Spanish and TP for Italian and European Portuguese, although European Portuguese exhibits additional complexities). Sheehan views EPP-satisfaction as a spectrum. One end of the spectrum features languages like English, Italian, and European Portuguese, in which the EPP is divorced from discourse features, which are ‘scattered’ on FinP. Spanish occupies a place in the middle of the spectrum: here, the EPP is associated with discourse features on FinP whenever these are present, otherwise it triggers subject movement to SpecTP. Discourse-configurational languages like Hungarian occupy the other end of the spectrum, where the EPP is always associated with discourse features. Effectively, word order variation in the Romance languages stems from the feature requirements of CP. Sheehan provides an excellent overview of inverted word orders, but it is not entirely clear why verb–object–subject word order in Italian and Spanish cannot be derived in a similar manner, and I would have welcomed more detail concerning the technicalities of her analysis.
In Chapter 7, ‘Subjects, Tense and verb-movement’, Theresa Biberauer & Ian Roberts argue that the triggers for V-to-T movement and null subjects should be kept distinct, exploring the Tense feature(s) (aspect and mood included) of T. They first draw a distinction between agreement and tense inflection and arrive at the following typology:
a. Rich agreement and rich tense inflection: hence V-to-T and null subjects, e.g. Italian, Greek, Spanish, etc.
b. Poor agreement but rich tense: hence V-to-T, but no null subjects, e.g. French. …
c. Poor tense and poor agreement: hence no V-to-T and no null subjects, e.g. Modern English.
(267, ex. (5a–c))
The remaining sections of the article are dedicated to establishing which languages instantiate which parametric option, with the authors considering in particular tense agreement and the availability of V-to-T movement. Biberauer & Roberts retain the assumption that rich agreement and the EPP constitute the trigger for null subjects.
In the final chapter, Ian Roberts investigates ‘Varieties of French and the Null Subject Parameter’, and proposes that in certain registers, Modern French displays null subjects. Focusing mostly on subject clitics, he argues that there is some special ‘interrogative conjugation’ (motivated by the existence of person distinctions). This explains why in subject–clitic inversion environments in Modern French, subjects consistently remain null. The φ-features of C are not passed to T, since they are realized in C by the subject–clitic. Given that T has no agreement features, the subject does not need to raise to SpecTP. Accordingly, no subject is tolerated in these constructions unless it is attracted to SpecCP (giving rise to complex inversion). Roberts assumes that in these environments of subject–clitic inversion, there is a null subject marked as pro which occupies SpecCP for EPP-satisfaction. Understanding these cases of subject–clitic inversion to instantiate residual verb-second (akin to English root interrogatives), he takes cliticisation in subject–clitic inversion constructions to be a process of affixation in which the inflected verb always undergoes movement. Unlike Northern Italian dialects, Modern French behaves like a consistent NSL due to the availability of interrogative conjugation.
While space restrictions prevent me from giving an evaluation of all of the individual chapters, I offer here some general observations. Despite the richness and variety of the contributions, none of the chapters (except in passing notes) deals with discourse pro-drop languages like Chinese and Japanese. In the various contributions to this volume, the availability of null subjects appears to be dependent on the featural specification of T (and C). It would be intriguing to see how null subjects in discourse pro-drop languages which lack person and agreement marking, tie in with this T-dependency. Overall, the authors restore faith in the notion of parameters within a Minimalist perspective, but some of the typologies introduced here struck me as computationally complex and as involving too many variables (this is particularly true for Holmberg's contribution). All the same, I look forward to future work and further empirical data from these scholars, as it is only through further investigation of the kind represented in this volume that syntactic theory can be advanced.
