Voiding phasehood

There are several consequences of such modeling of head movement as a syntactic operation, even if its applicability is restricted in the manner suggested above. Head movement deeply affects label-theoretic behaviour of syntactic objects, being an operation which involves crucially establishment of relationships at the featural level, the labeling algorithm being understood as searching for features and configurations in which they appear (possibly searching exclusively for features, as noted in Chomsky (2013c: 45)). Adopting the hypothesis that {{X}, Y} structures are not labeled in precisely the same way that structures not involving self-merged constituents—in other words, that adjunction structures need merely to be identified as adjunctions and are not further subject to the requirement that there be either a LI identifiable by minimal search as the label or a featural interdependence between LI’s identifiable by minimal search in constituents of a structure—leads to the postulated outcome that the adjoined head (the head previously occupying the landing site) together with its possible featural specification becomes invisible for the labeling algorithm, whether the copying approach to feature inheritance is adopted or the inheritance-with- deletion is hypothesized to be in force. An immediate consequence of this view is invisibility of adjoined heads (=heads moved into) for labeling purposes: head movement makes them inaccessible for the syntactic preparation of interpretive procedures in the C-I component. A change in the order of operations in (10) substantially affects properties and behaviour of the structure:

(20) a. form R-IA by EM;

b. Merge v*, reaching the phase level;

c. R raises to v* forming R with v* affixed, hence invisible.

Head raising which precedes feature inheritance makes the latter process impossible, relevant features being now buried inside the adjunction structure, as in (21).


No feature inheritance means no labeling, no transfer of phasehood—the phase becomes weak in the classical minimalist sense, with no PIC effects and raising of IA required only once the C-phase has been reached, if at this level the sequence is parallel to (10), hence with feature inheritance and concomitant processes.


Both unaccusatives and passives may be assumed to have (22) as the common core, the results of R-v* movement voiding phasehood at the vP stage of the derivation and hiding features of v* from the view of the labeling algorithm (the ordering of operations has been proposed as the source of such effects with bridge verbs in Nomura (2015) as cited in Epstein, Kitahara, and Seely (2016), who themselves hypothesize that derivations may begin with external pair-merge of heads, with much the same effect for the vP case). This does not yet settle the issue of IA moving in (21) to a position analogous to the one in transitive cases (with a ‘strong’ phase instead of a ‘weak’ one), according to the recipe in (23) rather than to the one in (20).

(23) a. form R-IA by EM;

b. IM of IA;

c. Merge v*, reaching the phase level;

d. R raises to v* forming R with v* affixed, hence invisible.

Were it the ultimate landing for IA, the structure would not pass the labeling test, for obvious reasons—R has not inherited ^-features, and thus there is nothing to share between (the head of) IA and R; but when IA raises further—as it can, without incurring label-related penalties, since phasehood is voided immediately by R-v* raising—it leaves only a copy, by hypothesis invisible for the labeling algorithm, hence causing no problems with the structure:


The current theory of the interplay between merge, both external and internal, and operations leading to transfer to the interfaces opens a space of certain under- determinacy. The crucial point concerns the freedom of (internal) merge, which ceased to be an operation requiring to be triggered and taking place only once the phase level has been reached. Chomsky (2015b) ultimately abandons the idea that internal merge stands in need of being ‘justified,’ concluding simultaneously that ‘the simplest conclusion, then, would be that Merge applies freely, including IM. Hence IM can apply prior to merge of C and inheritance, and the problems are resolved’ (Chomsky 2015b: 14), and the unification of phases makes this conclusion valid also for the vP case. The assumption that phase-level memory suffices for distinguishing instances of feature valuation and of external and internal merge frees the latter from taking place obligatorily immediately before transfer, and only after external merge of the phase head has occurred, as it was the case as long as internal merge was seen as internal merge and features valued derivationally were seen as such only immediately after valuation—which was, to recall, the main reason for the rise of the ‘all phrases are phases’ alternative to Chomskyan formulation of phase theory in Epstein and Seely (2002) (developed subsequently in Epstein and Seely (2006), Lahne (2008), Muller (2011)). Operations are not driven by phase heads as the process was understood earlier: feature inheritance relates phase-heads and their nonphasal companions, but the phase head is no longer conceived of as ‘probing, neither for purposes of feature valuation nor for purposes of labeling—both falling under the umbrella of minimal search:

There isn’t any identifiable Probe. There’s just a search procedure (...) It takes a look at a syntactic object and it’s asking the question, “What are you?” (...) I don’t think we should take Probe-Goal too literally. In the case of tense and subject, you can think of the tense as searching for the subject, but that’s kind of anthropomorphic. It’s just that a relation exists, and it should be a minimal computational relation, and that minimal relation values unvalued features—actually, in both ways, like the tense will get valuation of ^-features and the subject will get Case. It’s just a relation established by Minimal Computation, which values unvalued features. (Chomsky 2015a: 81)

Complex internal merge of heads may be assumed to enjoy freedom of application no less than it is the case in simple cases of internal or external merge, affecting the course of the derivation depending upon its timing relative to other processes. As long as X0 in (25) remains in the workspace as a visible object, it is bound to transmit features to Y0, with labeling taking place within its complement until phasehood is activated on Y0, which happens simultaneously with X-Y raising.


In such cases, a must be labeled, being a complement of the phase head X0, as it happens in standard RTO or IA short movement cases; otherwise, when X-Y movement occurs before feature inheritance and the phasehood of X is voided (without its activation on Y, which would require prior feature inheritance), it is not a derivational point at which the labeling filter would apply to a—if it is formed, i. e. if (23) was executed; it seems therefore possible, but not necessary, to have an intermediate occurrence in a chain of IA leading across a ‘weak’ v to a higher position in which {f, ip)-label is established, and the lesson generalizes to the CP case (although the featural content poses here more questions) and to successive cyclic movement in general, with the possibility of intermediate landing sites at the edges of VPs and TPs (to use the traditional terminology; see Biskup (2014) on successive cyclic movement in the infinitival case). The basic line supports the view that so-called ‘weak’ phases are to be treated as phases tout court, with the crucial difference between ‘weak’ and ‘strong’ phases—there being a suspension of operations leading to and including transfer, and of opacity for further syntactic computation—resulting from the timing of head movement (see already Legate (2003), Sauerland (2003) for early arguments for the phasehood of unaccusative and passive vPs; Dikken (2006) proposes to reanalyze the evidence in other ways). It must be noted, though, that availability of voiding phasehood by head raising as outlined above depends to much extent on a proper elucidation of details of feature inheritance, in C-T cases in particular, where the featural content possibly affected by feature inheritance is richer than in the v-R case, an issue noted in section 1.3.4, with much open issues left unresolved in the current theory of phases. Elimination of pair-merge as a kind of merge distinct from set-merge and unification of phrasal adjunction and head adjunction, insofar as it is possible given differences between syntactic objects participating in both, not only reduces the number of primitive operations and eliminates the need to postulate substitutional operations in narrow syntax and the C-I component as it directly operates on structures delivered from the syntactic component, in accordance with general properties of narrow syntax as currently conceived, but also provides syntactic foundations for the interpretive import of labeling as discussed in sections 1.3.3, 1.3.4, 1.3.5 and 1.3.6. Mechanisms of head movement interact also closely with the rise of discontinuous syntactic objects consisting of multiple occurrences of complex syntactic objects—phrasal chains—via establishment of labeling configurations or voiding such possibilities. Phrasal chains, in turn, are yet another phenomenon in which ‘weakness’ of objects participating in the derivation plays an important role, which stands in need of elucidation with respect to its interpretive consequences.

< Prev   CONTENTS   Source   Next >