Chains and occurrences: silent partners
The descent of chains
The assumption that intermediate links of syntactic chains are interpretively inert, invisible for the C-I component, is taken for granted especially with regard to A-movement chains. In the most recent version of the minimalist apparatus, such intermediate links are created exclusively due to labeling requirements; more exactly, only structures in which they appear are not filtered out under the labeling algorithm, which cannot label structures in which such movements have not occurred. They cannot be said to be triggered by any interpretive requirement, encoded as feature on the lexical item heading structures hosting such intermediate movement sites—on the most strict account, as in Chomsky (2013c, 2015b), ‘spurious’ features are eliminated on the hypothesis of the free nature of displacement. A possibility of inducing interpretive effects as a by-product of checking features at intermediate landing sites—a mechanism assumed to be operative in the framework in which every application of a syntactic operation had to be triggered, and abandoned ultimately only as late as Chomsky (2013c, 2015b)—does not even arise in this situation. Given that displacement is thus seen as a side effect of labeling requirements, copies left under the application of internal merge being invisible for the labeling algorithm and being crucially only parts of a discontinuous object, created successively because of the cyclic nature of the derivation by phases, there is little room for a suspicion that the links of the chain may have any significant role to fulfill during interpretation in the C-I component in particular. It is no wonder that, even when it is considered as a reasonable conjecture that every syntactic operation should induce an interpretive effect, it is felt that ‘not every position counts,’ intermediate chain links in A-movement particularly being vulnerable to syntactic annihilation. Discussing the requirement that every instance of merge be linked to interpretive effects at either side of the transfer procedure combined with the hypothesis of the free nature of merge, Yang (2015), developing the ideas proposed in Yang (2013, 2014), finds it necessary to include a ‘null effect’ among ‘the exhaustive list of interpretive effects of IM for the theory of Merge under non-determinism’ (Yang 2015: 417), and explicitly claims that ‘the intermediate link(s) in a syntactic chain should be without phonological effects or interpretations’ (Yang 2015: 419). Similarly, Gal- lego (2016), discussing similarities between syntactic chains created by internal merge and lexical items, considers the intermediate occurrence in an A-raising structure as in (1) as entirely interpretively inert (take (1) to involve A-raising with an occurrence in the embedded Spec-TP, the structure otherwise not involving intermediate landing sites for the derived subject of the main clause with a raising verb, there being no phasal boundaries on the way to the main clause Spec-TP): ‘not every member of the chain feeds the interfaces: the topmost one feeds PF, the bottommost does LF, and the intermediate ones are ignored’ (Gallego 2016: 149).
‘If the complex chain (...) is chunked down into CH1 and CH2 (...), then notice that only all the occurrences of CH1 have an interpretive role’, (Gallego 2016: 149) remarks:
This division of the chain in (1) is only done for illustrative purposes, as Gallego (2016) himself notes, since the chain is uniform (created under A-movement exclusively), hence it creates no potential problem for the principle of chain uniformity (as in the case of A-bar movement following A-movement, see various takes in Chomsky (2001: 49 n. 71), Chomsky (2008: 149)); chunking the chain in this way merely highlights the putative inertness of intermediate copies with regard to either A-P or C-I interface: ‘Although only CH1 is interpreted, CH2 cannot be ignored or deleted—unless we tamper with the structure. CH2 is thus “there" although invisible for computational processes’ (Gallego 2016: 150). In a similar vein, Sportiche (2016) starts with the Principle of Full Interpretation of Chomsky (1986b, 1995), Chomsky and Lasnik (1993), to capitalize on chains being syntactic objects, and interpretive processes applying to them, not to occurrences which constitute them separately:
The key general proposal we make is based on the observation that FI is already rightly, and routinely albeit often implicitly, assumed to apply to syntactic objects rather than their occurrences (for example to a chain—a set of movement copies—rather [than] individual copies). Suppose a syntactic object SO has multiple occurrences O1..., On. If anyone of these O, is semantically interpreted, SO satisfies FI. From this alone, it follows that all other O, j Ф i can be (but need not be) semantically ignored without violating FI: this conception derives the existence of interpretive Neglect, that is the possibility of interpretively ignoring up to all occurrences of an interpretable syntactic object but one. (Sportiche 2016: 1)
The general line of Sportiche (2016) is therefore to push towards the interpretive component the mechanism which used to be associated with syntactic operations in the GB period—all differences between various proposals put aside, they all may be traced back (pun intended) to the idea of trace deletion, originally developed to handle cases of islandhood violations under A-bar extraction which nevertheless do not seem to exhibit the degree of unacceptability expected in such cases, analyzed with the assumption that not properly governed traces in a nonuniform chain may be deleted to obtain a legitimate LF object—a syntactic way of repair in a restricted class of environments (see Lasnik and Saito (1984, 1992), Chomsky (1991), Chomsky and Lasnik (1993) and the discussion in Ki- tahara (1999) and Lasnik and Uriagereka (2005); one may compare the proposal of Baltin (2011) to delete traces before binding relations are established). The approach came to be generalized once shadows had been cast upon the very notion of chain, which had begun its life as a device to link D-structure properties of expressions and those to which applications of Move a give rise, thereby making S-structures ‘enriched D-structures’:
Note that there is now a good sense in which the S-structure (...) represents both GF-d and GF-в; namely, John bears the grammatical relation [NP, SJ by virtue of its actual position (...), and bears the relations [NP, S2] and [NP, VP2] by virtue of the positions of its traces. Suppose we associated with each lexical NP in S-structure a sequence (pj,...,pn) which, in an obvious sense, represents the derivational history of this NP by application of “Move a”; thus, p is the position of the lexical NP itself; p2 is the position (filled by a trace) from which it was moved to its final position; etc., pn being the position (filled by a trace) occupied by the NP in D-structure. Correspondingly, let us associate with each lexical NP in S-structure the sequence of grammatical functions (GFj,..., GFn), where GF. is the grammatical function of the element filling position p. (the lexical NP itself for i = 1, a trace in each other case), as determined by the S-structure configuration. Then GFn is the grammatical function of the NP at D-structure, which, by the d-criterion, is assigned a d-role in LF. (Chomsky 1981b: 24-25)