Desktop version

Home arrow History

  • Increase font
  • Decrease font

<<   CONTENTS   >>

The rise of thematic relations

The issue of ^-assignment remains a perennial trouble-maker for the syntactic theory. The labeling-based framework as developed in Chomsky (2013c, 2015b) does not admit solutions available under earlier versions of the minimalist apparatus; in particular, neither phrase structural distinctions between complements and specifiers are admissible in the most recent formulations of the syntactic theory, which explicitly strives to get rid of X-bar theoretic notions, nor does a featural approach, assimilating 0-roles to (formal) features participating in the derivation, fit in the labeling framework. The latter possibility, taking 0-roles to be determined in the course of standard featural interactions during a derivational process, adds 0-features to the inventory of syntactically visible properties of syntactic objects (cp. e. g. entirely different in details, but agreeing as to the featural status of ^-relations, proposals in Manzini and Roussou (2000) and Hornstein (1999, 2003); see further Boskovic (1994), Lasnik (1995: n. 27), Boskovic and Takahashi (1998), Fanselow (2001), Otsuka (2010), a. o.); in the current theoretical setting such features would presumably belong to the set of unvalued features introduced by the phase head and subject to the mechanism of feature inheritance (if we take 0-features to be properties of the verb, with Hornstein (1999: 79 n. 17)). They would arguably lack valued counterparts on the side of argument DPs, hence, beside being unable to participate in labeling, they would be parasitic on some other valued-unvalued relationship between features of syntactic objects—this is standardly assumed for the abstract Case, checked/assigned in {(, ф) structures. Thus, after feature inheritance in the vP phase, Case feature borne by IA is deactivated concomitantly with the establishment of the {(, (^-relationship between R and IA (involving, as it must do for labeling purposes, a pair of features differing along the valued-unvalued dimension), as in (14).


The object a is labeled, on current assumptions, as in instance of the {XP, YP} structure with both complex syntactic objects—IA and в in (14)—sharing features relevant for labeling purposes, (-features in this case, resulting in a configuration which may be labeled {(, (): the configuration in which the Case feature of IA is deactivated. No such relationship obtains between R and the (IA)-copy (trace). The 0-role of IA cannot be determined/assigned in the non-tail position of the chain, leaving the trace position uninterpreted in this respect—after all, the whole concept of chains in the A-domain relies on the need to establish a link between a position relevant for thematic interpretation and the position in which ф and Case relations are established, the requirement that the foot of an A-chain be a 0-related position being enforced for all argumental DPs together with the assumption of the vP-internal EA position and the IA point of origin for subjects of unaccusatives and passives, a property of syntactic structures which is firmly established by now. It is furthermore a property of structures like в that they are labeled by LI’s, being instances of the {X, YP}-schema: proceeding by minimal search, the labeling algorithm cannot take an LI and a complex YP object to be equally distant. It is with such considerations in mind that Chomsky (2013b: 66) would take в in (15) to be labeled simply by V (the particular landing site of IA and its antilocality-violating movement are not proposed in Chomsky (2013b), but are irrelevant for the issue of the labeling of в there).


The copy of IA remaining in the complement position of V being invisible to the labeling algorithm, the structure is labeled by the LI, V in this case; similarly for the EA case, in which the structure may be labeled by the v*-LI after raising of the EA.


For structures in (15) and (16), Chomsky (2013b: 66) assumes labeling to be required for thematic purposes, taking still in a standard way V and v* to be relevant LI’s capable of determining the label.

While the assumption that thematic interpretation requires labels to be established before Transfer to the C-I interface occurs still holds in the approach of Chomsky (2013c, 2015b), the details become considerably more convoluted given the properties of roots, replacing V/N-heads at the bottom of the structure, and the mechanics of head movement, as discussed in particular in Chomsky (2015b). The elimination of phrase structural notions makes it impossible to mantain a configurational characterization of 0-roles along the lines pursued in the GB period, ‘complement’ and ‘specifier’ being at best purely descriptive terms, used, if at all, for reasons of backward compatibility with earlier analyses, but possessing no official status in current theory. An analysis of label-theoretic behaviour following the lines indicated above suggests that ^-properties are best understood as arising in the interpretive component, being dependent upon properties of syntactic objects which, although established in narrow syntax, are not exclusively directed at such aims as providing thematic interpretation. In the Spartan framework of Chomsky (2013c, 2015b), whereas IA undergoes set- merge with the root, hence appears in a direct syntactic configuration with a LI, EA does not, its only syntactic relationship being with a vP, a complex phrasal structure, never with the v* itself. Chomsky (2015b) partly evades the problem, partly hints at possible solutions, stating that ‘other questions arise about thetamarking, a matter not considered here. Note that theta-marking involves the entire relation v*-R-complement, not just R-complement, and accordingly takes place at the phase level’ (Chomsky 2015b: 13). The crucial point is that—in contrast to structures with V as a head—current representations of the verbal phrase do not involve a ^-assignment opportunity for IA at the point of its external merge: head movement of R to v*, determining its verbal nature, is required. A label-theoretic perspective together with properties of head adj unction structures supports taking ^-relevant configuration to ‘involve the entire relation v*-R-complement, not just R-complement’, and an analogous view on the EA case—configurations not syntactic in the sense of being theoretically relevant for an account of the workings of narrow syntax. This remains in conformity with the hypothesis that ^-properties and their interpretation belong to the postsyntactic, C-I component. The original formulation of UTAH was cast with the D-structure level as relevant for determining ^-properties (Baker (1988: 46): ‘Identical thematic relationships between items are represented by identical structural relationships between those items at the level of D-structure.’), but with D-structure gone in early minimalism, ^-properties had to be moved to LF:

.. .realizing that conceptually the UTAH is part of the theory of the relationship between language and the C-I system tells us where it should fit in the grammar: like the Theta Criterion, it should be understood as an output condition on LF, the interface level related to C-I. (...) There is one slight complication: to apply the UTAH at LF one must say that it looks not at simple NPs, but rather at the tails of chains, in order to factor out the effects of movement. (Baker 1997: 122)

Once LF is also eliminated, it was reasonable to assume that relevant configurations are interpreted in the C-I component, differentiated on the account above due to differences between structures in which IA and EA enter the derivation. This non-uniformity of ^-assignment should not be particularly worrying even if ^-assignment were seen as a syntactic issue—after all, even the most restrictive understanding of UTAH (Baker 1988, 1997) does not require that ^-assignment occur in identical relationships with 0-assigners, only that identical 0-roles be assigned in identical configurations—but it is perceived as an anomaly, witness Chomsky’s remarks in Piattelli-Palmarini, Uriagereka, and Etxeberria (2009: 53-54) on unclarities about the proper characterization of the place of external arguments (Shim and Epstein (2015) discuss the issue at length, putting forward proposals of a unification of ^-relevant configurations as always involving a head-XP relationship either resulting from a radical transfer of the VP or a doubly-peaked structure familiar from Epstein, Kitahara, and Seely (2012)). It is a property of syntactic module as currently conceived that it does not avail itself of operations which would exhibit characteristics of substitution, a theme to which we return in sections 2.2.1-2.2.2, 2.3.3 and 3.5 at more length; it is a property which distinguishes it from C-I components—(sub)components responsible for recognition and use of patterns across structures/expressions and (subcomponents constituting the repository of prelinguistic concepts as well. As a consequence, the transition to the lexicon requires that a concept qua a root have some properties imposed at this stage, and it may be supposed that it is the order of participants; on the other hand, concepts in their prelinguistic sense, while being unable to enter into inferential relations in the strict sense of the word—as brought into focus in the quote from Reinhart (2006) in section 1.1.1—are usable in their original environment (allowing substitution and, if they are neutral in the sense of Fine (2000), anchoring constraints on their applicability in there being distinguished conceptual complexes in the Finean sense, with equivalence classes under substitution of participants being concepts more properly so called) and may serve as providing the anchoring point for constraints on admissible interpretation functions in the syntax-related component(s) of the C-I system, restraining properties of those concepts qua roots which correspond most closely to original ones: the transition from non-linguistic concepts to roots involves not merely imposing an order on participants (slots for participants, more accurately), but also a possibility of manipulating their number—the intricate relationship between the lexicon and being available as a syntactic object and non-linguistic conceptual modules need not consist only in acquiring a property of being subject to merge. Thematic roles may be seen as in this context as abstract interpretive concepts, analogous to concepts properly so called in nonlinguistic modules, equivalence classes in this case arising in the C-I component.

<<   CONTENTS   >>

Related topics