Previous accounts of asymmetries in passivisation

Under Government and Binding (GB), asymmetries in passives were taken to stem from the violation of the Case Filter (Baker, 1988b; Larson, 1988, e.g.).

Thus, in GB, the reason why (27c) is ungrammatical is the lack of case valuation on the recipient argument.

(27) a. John gave Mary a book.

b. Mary was given a book.

c. *A book was given Mary.

Recently, locality-based accounts emerged, explaining the ungrammaticality of (27c) with a locality constraint on A-movement, (28). Under (28), the theme cannot move to [Spec;TP] because a higher argument (goal/recipient), placed closer to T, blocks the movement of the theme.

(28) Shortest Move/Closest Attract

K attracts F if F is the closest feature that can enter into a checking relation with sublabel of K.

(Chomsky, 1995, 297) However, while locality restrictions explain why themes cannot move over re-cipients/goals, they do not explain how symmetric passives are allowed. Thus, locality restrictions do not account for cases where themes can move over a higher argument. Therefore, a new explanation has been proposed where argu-ments move successive cyclically, stopping at an intermediate position. In this position, the moved theme and the recipient/goal become equidistant from T (Anagnostopoulou, 2003; Holmberg and Platzack, 1995; McGinnis, 2001, a.o.). Due to this equidistance fromT, either of the objects can passivise.

Various proposals have been made as to what allows such movement to an in-termediate position. For example, for passives of ditransitives, Anagnostopoulou (2003) proposed the Specifier to vAPPL parameter, in (29).

(29) The Specifier to vAPPL parameter

Symmetric movement languages license movement of DO to a specifier of vAPPL. In languages with asymmetric movement, movement of DO may not proceed via vAPPL.

(Anagnostopoulou, 2003, 157) Anagnostopoulou (2003), following Marantz (1993), assumes the universal struc-ture of double object constructions represented in (30).³

(30) [_vP Agent v [v_{AP P LP} Ben/GoalvAP P L[_{V P} V Theme ]]]

Under (29), in languages which allow for DO passivisation, the DO is successive-cyclically moved to the outer specifier ofApplP, as in (31).

(31) [vAP P LP DO [vAP P LP IO [_{vAP P L}^′ vAP P L [V P VtDO ]]]]

Such movement to the outer [Spec;ApplP] position, allowed in languages that show the specifier to vAPPL parameter, makes the theme and goal/recipient equidistant fromT. Following Chomsky (1995), Anagnostopoulou (2003) takes multiple specifiers of the same head to be equidistant from a c-commanding head that triggers movement. This equidistance in (31) allows either of the arguments to move further to [Spec;TP]. Asymmetric languages cannot move the theme through [Spec;ApplP]; hence, these languages can passivise only the higher argument, i.e. the goal/recipient.

A very similar, although conceptually superior, analysis is proposed in McGin-nis (2001), where the specifier of vAPPL parameter is reduced to an independent property of a given language, namely the nature of phases (Chomsky, 1999, 2000, 2001). Under Phase Theory, syntactic derivations proceed in chunks called phases. Once a given phase is complete, it is sent to spell-out for phono-logical and semantic interpretation. It is commonly assumed that vP and CP constitute phases, and thusv and C are taken to be phase heads. When these phase heads project, whatever is in their complement position is sent to spell out. Crucially, due to thePhase Impenetrability Condition (PIC), in (32), the complement of a phase head is no longer accessible to operations above the head at phase spell-out.

(32) Phase Impenetrability Condition (PIC)

In a phaseα with head H, the domain H is not accessible to operations outside α, only H and its edge are accessible to such operations.

(Chomsky, 2000, 108)

3Note that both authors assume that the applicative head is a light applicative verb, i.e.

vAP P L. This is in contrast to, e.g. Jeong (2007); McGinnis (2001); Pylkkänen (2002, 2008) who take the applicative head to be of theAppltype, i.e. resulting inApplPas its maximal projection, notvP.

What is at theedge of the phase, i.e. the specifier of the phase head, together with the head itself are visible to the structure above the phase head. Movement to the edge of a phase is taken to be triggered by an optional EPP feature, similar to the EPP on T, which causes movement of external arguments to [Spec;TP]. A non-phase EPP, like the one onT, is taken to be obligatory, while a phase EPP is commonly assumed to be optional. Whenever the EPP feature is present on v, it will trigger movement of a lower DP to v’s outer specifier position. This movement, in turn, will allow the lower DP to avoid being frozen in the complement position under the PIC.

In what follows, we discuss how Phase Theory has been integrated into the theory of applicatives (Citko, 2011, 2014; Jeong, 2007; Lee, 2005; McGinnis, 2001, a.o.). Recently, it has been proposed that, in contrast to low applicatives, high applicatives constitute phases (McGinnis, 2001). Under the low applicative structure of Pylkkänen (2002, 2008) where the IO is a co-argument of DO, the lack of phasehood of low applicatives has been attributed to anti-locality constraints (e.g. Jeong, 2007; Lee, 2005). Under anti-locality, all applicative phrases can be taken to be phases. However, while the phasehood of vP and CP are common theoretical assumptions in the literature, the phasehood of ApplP is stipulative. What is more, the phasehood ofApplP can be done away with under the hypothesis as to two applicative types -vP-internal and ApplP-internal - proposed in Chapter 2 of this thesis. Under our hypothesis, only vP-internal applicatives are taken to constitute phases, as expected ofv heads.

ApplP-internal/vP-external applicatives are not phases.

In the discussion to follow, we abstract away from the problem ofweak and strong phases, namely the differences between PIC, in (32), and PIC2, in (33).

(33) Phase Impenetrability Condition (weak PIC, or PIC2)

Given the structure [ZP Z ... [HP α [_H^′ H Y P ]]], where H and Z are phase heads, the domain of H is not accessible to operations at ZP; only H and its edge are accessible to such operations.

(Chomsky, 2001, 14)

Both in Polish and Icelandic, ExperiencerDAT-ThemeN OM constructions show evidence that the PIC is too strong, and that the PIC2 is more accurate. This is because, in both languages,T can agree with a lower phase-contained theme, which indicates that untilC merges,T has access to the lower phase. However, for the discussion in this chapter, it suffices to assume the traditional, strong, understanding of PIC.