Minimalist syntax relates to sound and meaning at two interfaces: one is the syntax–phonology interface, on which plenty of studies have been conducted within the Distributed Morphology framework (see e.g. Embick & Noyer Reference Embick, Noyer, Ramchand and Reiss2007); the other is the syntax–semantics interface, which has received relatively less attention, at least since the advent of Minimalism. Terje Lohndal’s Phrase Structure and Argument Structure: A Case Study of the Syntax–Semantics Interface provides a programmatic but thought-provoking approach to this intriguing aspect of the model. Lohndal argues that it is possible to establish a transparent mapping from syntax to Neo-Davidsonian logical forms such that each spelt-out unit corresponds to a conjunct at logical forms before they are conjoined and existentially closed. This proposal is justified by two seemingly unrelated reductionist hypotheses: (i) the elimination of the structural distinction between specifier and complement in favor of a simple head–complement relation, and (ii) a full thematic separation hypothesis proposing that all thematic arguments be severed from the lexical verb/root and introduced by functional projections instead. If this proposal is on the right track, then an equally simple mechanism for both syntax and semantics turns out to be tenable.
The first chapter offers a synopsis of the book, beginning with an outline of the theoretical framework adopted in the book. The main focus of this chapter is a survey of the two major approaches to argument structure and its alternations, the lexicalist approach (e.g. Chomsky Reference Chomsky1981, Levin & Rappaport Hovav Reference Levin and Rappaport Hovav2005) and the constructivist approach (e.g. Borer Reference Borer2005, Ramchand Reference Ramchand2008), developed in subsequent chapters. Lohndal successfully demonstrates that the practice of attributing argument structure flexibility to lexical specifications has been challenged in recent years by the constructivist approach, which questions the traditional wisdom of anchoring the verbal adicities/valencies in the lexicon, modulo an event variable obtained from categorizer v in the syntactic computation. In other words, verbs are always monadic since there is only an event argument.
In Chapter 2, ‘The size of a verb’, Lohndal further develops the full thematic separation hypothesis. He argues that verbs do not have thematic arguments at all and that the argument structure effect follows from the Neo-Davidsonian event semantics of a verb in that each argument is introduced by its own separate thematic predicate. In other words, full thematic separation can be realized in the syntax. This claim is justified in two ways: on the one hand, the positing of an entirely lexically-encoded adicity is ‘not a virtual conceptual necessity for formal semantics’ (24); on the other, it is challenged empirically by the properties exhibited by external and internal arguments. External argument separation is realized by positing a dedicated functional head Voice/v to introduce the Agent. Inspired by Schein’s (Reference Schein1993) initiative in full thematic separation, Lohndal discusses the possibility of dissociating the Theme from the verb in five types of construction, namely reciprocals, adjectival passives, focus, event-delimitation, and obligatoriness of the Theme. General labels like ‘external argument’ or ‘internal argument’ become theoretically insignificant in the face of the more specific and precise labels like Agent, Theme, etc. However, it has to be admitted that these arguments are not as persuasive as those for Agent-severing, especially the arguments based on the Theme’s role in measuring out the event described by the verb phrase, and the obligatory presence of the Theme. Finally, given the challenges that the lexicalist tradition faces in accounting for English lexical flexibility (e.g. Borer Reference Borer2005), Lohndal demonstrates that this complexity can be accommodated if a full thematic separation hypothesis is adopted. This is because the complexity is largely due to the interaction of conceptual meaning and language use, not the variant lexical adicities. If this perspective is adopted, some undesirable theoretical burdens resulting from the lexicalist approach can be circumvented, including unconstrained enlargement of subcategorization frames and redundant lexical specifications of context-dependent meaning. Consequently, the ‘size’ of the verb is reduced to its proper limit in that its lexical specification includes only an event argument, and the lexicon is considerably simplified in favor of an interface approach. This is one of the core contributions of this generative constructivist model to the architecture of minimalist syntax.
In Chapter 3, ‘Specifiers and phrase structure’, Lohndal argues for a specifier-less approach to phrase structure, which can be achieved either by relegating the roles that the specifier has assumed since its inception in 1970 to more fundamental technical mechanisms in Minimalism, or by proving that these primary roles assigned to specifiers are only apparent. The major functions of specifiers are (i) to host generalized subjects, (ii) to host phrasal modifiers, and (iii) to provide landing sites for movement. Lohndal argues that (i) specifier as generalized subject (or spec–head relations in general, including spec–head agreement and case valuation) can be successfully subsumed by the prominent mechanism Agree in Minimalism; (ii) the phrasal modifier located in specifier can be reshaped as the complement of a dedicated functional head following the tree-pruning theory of Spell-Out (see below); (iii) the landing site for movement is simply another phrasal position, thus, there is no need for a specifier qua specifier assumption; (iv) the constraint on movement (Relativized Minimality) can also be largely captured by Agree with the incorporation of featural specifications; and (v) the asymmetry between specifier and complement with respect to (sub)extraction can also be explained alternatively by a specifier-less account. Thus, one immediate virtue of eliminating specifier is to dispel the existing mysteries or problems concerning the notion per se. Developing this argument, Lohndal further argues that there is no relational difference between specifier and complement, a step that restricts all syntactic relations to the ‘most local’ one: the head–complement relation. Thus, Merge will be maximally simplified in the spirit of Occam’s Razor. This takes us a step closer to the elimination of the entire phrase structure component, a natural continuation of Chomsky’s consistent reductionist endeavor toward a simple Universal Grammar (UG; see Chomsky Reference Chomsky2013) in that the basic relational primitive in minimalist syntax has been reduced to the head–complement relation only. However, it should be emphasized that the nullification of the role of specifiers does not mean that they are not important at all, but rather that ‘their role may not be as significant as commonly assumed’ (54) in the generative tradition.
Compelling as the above argumentation is, Lohndal does not explore the feasibility of specifier-less approach in DP, AP, PP, etc., which is essential for a full exploration of the proposal. In addition, the distinction between adjunct and complement is not discussed at all in this approach. Furthermore, this argumentation is essentially focused on English, which risks the proposal of cross-linguistic counterevidence.
A significant added value of this new approach, according to Lohndal, is the availability of transparent mapping from syntax to Neo-Davidsonian logical forms featured by the conjunction of each independent Spell-Out of the predicate as well as every single thematic argument. This is the focus of Chapter 4. According to Chomsky’s bare phrase structure, the merger of two non-heads should be filtered out for the sake of labeling (Chomsky Reference Chomsky2013: 43). However, Lohndal argues that this problem can be solved by exploiting the dynamic nature of Spell-Out, a mechanism that can be triggered and applied to the complement of a functional head whenever the maximal projection of this functional head is about to be merged with another phrase. In this way, the complement of this functional head is pruned away from the ongoing syntactic derivation, reducing the newly-built functional phrase under consideration to a bare head again. The syntactic computation can proceed to the end in this fashion. Put simply, the solution to the problem lies in assuming both complement and the traditional specifier to be merged with the same functional head, with Spell-Out/pruning reducing its maximal projection to the head itself in the dynamic derivation. In addition, the relative smallness of Spell-Out domain echoes Hornstein’s (Reference Hornstein2009) claim that no long-distance agreement is allowed, only strictly local agreement is possible.
Turning to logical form, the transferred domains are stacked in a sequence, and then conjoined in line with compositionality principle conjunction (see Pietroski Reference Pietroski2005) once the syntactic derivation comes to an end. Semantic computing then stops insofar as each spelt-out conjunct is existentially closed. One of the ensuing virtues of this conjunctive approach to meaning lies in a systematic mapping between syntax and Neo-Davidsonian logical forms, which is a step closer to Montague’s ideal of homomorphism between syntactic and semantic rules, manifested by the straightforward reversal procedure from logical forms back to the hierarchical syntactic structure. Hence, Lohndal argues that – if not better – conjunctive semantics is no weaker than the more standard type-driven semantics, with a typed lambda calculus, proposed in Heim & Kratzer (Reference Heim and Kratzer1998). The significance of this full transparency between syntax and semantics, argues Lohndal, lies in bestowing a parallel mechanism on these two components when speech production and parsing are taken into consideration. However, Lohndal’s approach is based on the assumption that both verbal predicate and thematic predicate are sentential, although, as he admits, ‘it is rather unclear what exactly makes a predicate a sentential predicate’ (106). Moreover, the separate existential closure proposal still needs further justification when it comes to the closure of event variables in the thematic predicates.
Another important contribution of this theory lies in proposing Thematic Integration, a syntactic mapping principle that translates syntactic argument into thematic relations. This rule links thematic arguments with the thematic predicates, which is reminiscent of Uniformity of Theta Assignment Hypothesis in the minimalist setting. Moreover, thematic uniqueness is essentially captured thanks to the strict argument–functional-head correspondence.
With respect to linearization, since VP is below all arguments, a VSO order is derived. Therefore, syntactic movements are needed. In this book, movement can be achieved via the creation of multiple copies in parallel ‘workspaces’, followed by a remerge into the spinal ‘workspace’, akin to Nunes’s (Reference Nunes2004) sideward movement. However, Lohndal’s movement offers a simpler mechanism, since pruning renders chain formation and chain reduction unnecessary. Thus, the simplified movement is more consistent with a simpler syntax.
Chapter 5, ‘Semantic competence and compositionality’, concludes the book with a discussion of the theoretical consequences of this new theory of the syntax–semantics interface, focusing on compositionality (as discussed above) and the distinction between grammatical meaning and conceptual meaning. Lohndal notes that logical forms are somewhat underspecified in that many facts about argument structure are ‘outside the syntax proper’ (132). More specifically, what ‘traditionally ha(s) been analyzed by way of theta roles should rather be analyzed by way of conceptual/situational meaning and language use’ (133). Thus, Lohndal has linked grammatical meaning with conceptual meaning. The latter is required to have a thematic grid committed to adicities. Consequently, ‘no semantic information is available in the syntax’ (16), which crucially dissociates semantics from syntax. As for compositionality, it should be acknowledged that the mapping from logical forms to situational meaning is holistic, not compositional. Therefore, it is safe to say that the Neo-Davidsonian approach to semantics provides a near-perfect match between syntax and semantics.
To conclude, this novel theory of mapping from syntax to semantics is extremely interesting and thought-provoking, and will surely benefit future minimalist study of this area, despite the fact that some of the details of this new approach remain programmatic, for instance, the cross-linguistic viability of a specifier-less phrase structure, the robust testification of existential closure of event variables in thematic predicates.
