Is There Any Intermediate Position between the Criterial and S Selection Positions?
‘‘On Wh-Movement’’ forcefully argued for a stepwise application of movement, hence, in present terms, for a positive answer to question. Over the last 25 years or so, quite diverse and compelling evidence has accumulated in favor of this conclusion: in addition to the two positions relevant at the interface, chains must involve intermediate positions.
Intermediate positions can trigger morphosyntactic phenomena, making them detectable in different languages. For instance, intermediate wh-movement to an embedded C system triggers embedded I-to-C movement in Belfast English (Henry 1995).
(1)
French stylistic inversion is triggered by a wh-element in the local C system: intermediate movement triggers it, as in (2d) with representation (2f ), while a nonlocal wh-element per se is unable to trigger it, as in (2e) (Kayne and Pollock 1978).
(2)
Other examples are past participle agreement in French (see Kayne 1989) and the C agreement of the Celtic (and many other) languages (Irish ex ample from McCloskey 1996).
(3)
(4)
Intermediate positions are not assigned special interpretive properties; nevertheless, they are visible at the interpretive interface, as shown by reconstruction effects (Lebeaux 1988). A sentence like (5d) is therefore ambiguous, admitting both Bill and John as antecedents of the anaphor himself. The latter interpretation must be made available by the inter mediate trace in the embedded C system. Under the copy theory of traces, the relevant representation is (5e).
(5)
The evidence for intermediate positions in movement chains, between s-selection and criterial positions, is thus strong and diverse, ranging from purely syntactic, to morphosyntactic, to interpretive. See Chung 1998 for a detailed review.
Why is it that chains involve such intermediate positions, positions that are not motivated by interface needs? The standard answer is that intermediate positions are enforced by the theory of locality: movement cannot land too far away; linguistic representations can be indefinitely extended owing to the recursive nature of the system, but each step of the linguistic computation is local and can cover only a limited amount of structural ground. Chains can be indefinitely long because the operations creating them can iterate indefinitely, but each chain link itself is local. The fundamental intuition of ‘‘On Wh-Movement’’ (and previous work within the Extended Standard Theory tradition, starting from Chomsky 1973) has retained all its force here.
Putting things in slightly more general terms, we can say that two competing forces determine the form of chains, both being construed as economy guidelines:
Movement as a last resort: Movement must be motivated; it can only be performed to achieve an interface effect; there is no ‘‘free movement.’’
Locality: Movement cannot land too far away.
The movement-as-a-last-resort guideline would favor chains with just two positions, the positions motivated by interface effects; locality would favor a proliferation of positions very close to one another. Permissible chains therefore represent a point of equilibrium between these two desiderata.
