Weak Crossover
An analysis that treats the movements as EPP-driven, rather than semantically motivated scrambling, predicts that the landing site of such movement should exhibit A-position properties. One test that identifies Amovement is the ability to override Weak Crossover (WCO) effects. In English, for example, when a nominal expression is moved to a non-argument position across a pronoun, the pronoun cannot be construed as anaphoric on the moved phrase, even though the anaphoric interpretation is possible when no movement occurs. Thus, compare the overt wh-movement in (1a) and the covert LF movement of the quantified NP in (1b) with (1c), where the nominal expression does not move.
(1)

In contrast, movement to an argument position over a pronoun does not induce a WCO effect, as a comparison of (2a) and (2b) demonstrates (see Mahajan 1990 for further discussion).
(2)


We have already established that movement to Spec-TP does not affect focus projection in simple cases where no anaphoric relations are involved. The focus projection facts generalize to the weak crossover paradigms for Russian in (3–4).
(3)

(4)


In the non-deviant (b) examples, the focus projection can include the entire clause. Note that even though the antecedent of the pronoun crosses over the pronoun overtly, no crossover effect occurs (i.e., the anaphoric relation with the pronoun is not blocked), suggesting that the moved constituent lands in an A-position before undergoing QR to a non-argument position. This reinforces the analysis that movement of the internal argument in accusative unaccusative constructions is to Spec-TP.
We are assuming here that the weak crossover configuration, which is generally prohibited, involves a pronoun whose antecedent is construed as a variable to its right (cf. Chomsky 1976). The sentence (5a) violates weak crossover at LF because the trace of každuju deυočku “every girl,” which is construed as a variable, functions as the antecedent of the pronoun to its left.
(5)

In contrast, in (37b), when každuju deυočku “every girl” is scrambled overtly across the subject ee sobaka “her dog” the anaphoric relation is not prohibited. This suggests that at LF, the variable bound by the quantified expression každuju deυočku “every girl” occupies the scrambled position rather than the verb complement position in which accusative NP is assigned its thematic function. The idea is that covert QR creates a weak crossover configuration whereas overt scrambling does not.
There is, however, a crucial difference between the examples in (5) and the accusative unaccusative constructions in (3–4). In the former, scrambling in the (b) example disrupts focus projection so that only the nominative NP can be focused, whereas in the latter focus can project to the entire clause. The difference concerns whether the dislocated phrase moves to Spec-TP (EPP-movement) or to some higher projection (scrambling).
The weak crossover facts demonstrate that both Spec-TP and the position to which an argument may scramble are A-positions. The difference in focus projection construal now strongly suggests that the position to which an internal argument moves in an accusative unaccusative construction must be Spec-TP.