Case and PPs
5.4 Capturing the distribution via size
5.4.2 Potential counter-examples: can dressed PPs be separated from their complement?their complement?
iskola school
k¨ozel close
van be.3sg
hozz-´a.
allat-3sg
‘The school is close to it.’
not: ‘The school is close (by).’
Prodeic thus cannot replace DP, it can only replace KP. Consider now the PP-structure in (84), which is the structure of both dressed and naked Ps.
(84) PP
KP
DP K
P
Naked PPs spell out only P, therefore the replacement of KP by prodeic is unproblematic (and it also makes the case-marker disappear). Dressed Ps, on the other hand, spell out both K and P.
If prodeic replaces KP in a dressed PP, this forces the dressed P to Underassociate its K feature and spell out only P. The Superset Principle makes this possible in principle. I have argued, however, that dressed PPs belong to the class of lexical items to which the Superset Principle cannot apply: they cannot be compressed either upwards or downwards. This causes dressed Ps to be incompatible with prodeic.
In sum, the fact that dressed Ps cannot have a prodeiccomplement follows from the size of the lexical items involved. Prodeic can only replace KP, and dressed Ps must spell out K, therefore only one of them is possible in any given PP.
5.4.2 Potential counter-examples: can dressed PPs be separated from their complement?
Is there extraction from dressed PPs, after all?
I have emphasized that the relationship of dressed Ps and their complements is a very close one:
these postpositions cannot be separated from their complement either by way of extraction or an intervening degree modifier. Some data, however, seem to contradict this generalization. In (85), the P seems to have been extracted from a dressed PP. It appears in the the so-called verbal modifier position (it precedes the verb in neutral sentences and follows the verb in sentences with focus and negation). In this configuration an agreement-marker appears on the postposition even with R-expression complements (recall that Ps normally agree only with pronouns), and the complement
15The Superessive case -on/en/¨on appears in the formrajt- when its complement is a pronoun rather than a lexical noun. This is an allomorphic variation that should not distract the reader.
bears overt Dative case. (86) shows the version of this example with the P and its complement adjacent for comparison.16,17
(85) J´anos
‘John is standing next to the tree.’
(86) J´anos
‘John is standing next to the tree.’
Conversely, in (87-a) the complement of P seems to have been extracted from the PP. Again, the Ground is marked with Dative case and the P bears an agreement marker.
(87) a. A
‘It is John that is standing next to the tree.’
b. [F oc A
‘It is the tree that John stands next to.’
This means that as soon as the dressed P is separated from its complement, the complement must be overtly case-marked. This confirms my view that the lack of overt case has a deep connection to the adjacency effects.
While the discontinuous dressed PPs illustrated in (85) to (87-b) are definitely marked and statistically far less frequent than the contiguous dressed PP in (86), we still need to account for how they arise; and if they involve extraction from the PP, then we need to reconsider the adjacency effects seen in the previous section. It will be shown, however, that these examples do not involve subextraction from dressed PPs, and these data can be accommodated into the account of the adjacency effects developed above.
Plain extraction is explanatorily inadequate
The pattern exhibited by discontinuous dressed PPs is reminiscent of the case alternation exhibited by possessors. As we have seen in Chapter 4, Hungarian possessors can appear either in the Nominative or in the Dative case (88); and Dative (but not Nominative) possessors can be separated from the rest of the DP. Further, the agreement paradigm on the dressed P is identical to the paradigm of possessive agreement, too.
(88) a. a
‘the book of the boys’
b. a
‘the book of the boys’
(89) a. *a
‘the book of the boys is heavy’
16This is possible only with Place and Goal Ps but not Source Ps. It is also not possible with the temporal interpretation of PPs, even if they are formally identical to spatial PPs.
(i) *Kata
‘Kate comes from beside John.’
(ii) *Kata
‘Kate came before the lecture.’
This, however, is not a quirk of dressed Ps: unambiguously source adverbs likekint-r˝ol ‘from outside’ orinnen
‘from here’ never become verbal modifiers either (´E. Kiss, 2002). See ´E. Kiss (2002) for an account in terms of aspectual interpretation and Sur´anyi (2009b, 2011) for a recent explanation which builds on the different merge-in heights of different types of PPs.
17The agreement is obligatorily overt on Place Ps and optionally covert on Goal Ps. I have nothing insightful to say about why this should be so. It is important to emphasize, however, that when the complement has no overt case and the P follows it, agreement is not grammatical.
b. a
‘the book of the boys is heavy’
The similarities with the possessive construction have been repeatedly pointed out in the liter-ature. Many analyses hypothesize an actual, formal or frozen possessive structure in dressed PPs (c.f. Mar´acz, 1986; ´E. Kiss, 1999, 2002; Sur´anyi, 2009b; R´akosi, 2010 and Laczk´o and R´akosi, 2011).18 This possessive structure, in turn, is often exploited to explain the dative-marked putative extraction from dressed PPs.
If the parallels are due to a shared possessive syntax indeed, then extraction from dressed PPs could proceed along the same lines as extraction from possessive constructions. That is, the Ground, like the possessor, can move to a peripheral escape hatch position in the containing phrase (DP for possessives, PP for dressed Ps), and in this position it gets dative case (c.f. Mar´acz, 1984).
In Section 5 I will discuss the fine-grained decomposition of my K and P and I will argue that PPs do indeed have an underlying possessive structure.19 However, a ‘movement to and dative assignment in an escape hatch’ analysis of (85) leaves unexplained why the agreement marker appears on the P once the complement is separated from it. Dressed Ps agree only with pronominal complements but not with full DP-complements.
(90) (´en)
‘next to the house’
This means that a simple extraction analysis of (85) and (87-a) fails to provide a descriptively adequate analysis of the facts.
A resumptiveproanalysis of discontinuous dressed PPs
Mar´acz (1984) and in part ´E. Kiss (2002) suggest that apparent dative-marked extraction from dressed PPs involves a PP-internal resumptivepro. Mar´acz (1984) suggests that in the pertinent cases the subject of the PP is apro, which is bound by the dative marked constituent (the latter is extracted from the escape hatch of the PP). ´E. Kiss (2002), on the other hand, suggests that resumptive pro is involved only in a subset of the relevant cases. She observes that with the extraction of a plural complement the agreement on the adposition can be either singular or plural. This is similar to the agreement possibilities of extracted possessors.
(92) (A
‘A beautiful future is ahead of the boys.’
(93) A
‘A beautiful future is ahead of the boys.’
(´E. Kiss, 2002, pg. 190, the glosses have been modified)
E. Kiss (2002) proposes that dressed PPs have the structure of possessive phrases. The version´ of (93) with singular agreement on the postposition involves plain possessor extraction, while the version with the plural agreement involves apro possessor. In the latter case the dative possessor is generated outside the PP in a hanging-topic-like construction.
I agree with Mar´acz (1984) that all instances of apparent extraction from dressed PPs, that is, both the variant with the singular and the one with the plural agreement, involve a PP-internal
18Adpositions expressing an axial part in the sense of Svenonius (2006) are all dressed Ps in Hungarian. For axial part denoting Gungbe adpositions, Aboh (2005, 2010b) also suggest an underlying possessive structure. A possessive structure for locative PPs in more general was proposed in Terzi (2005, 2008); Botwinik-Rotem and Terzi (2008).
19E. Kiss (1999) and R´´ akosi (2010) point out some differences between possessive constructions and dressed PPs.
Heged˝us (2010b) argues that a possessive analysis of dative Grounds is on the right track for earlier stages in the language but not in contemporary Hungarian. I refer reader to these works for details.
resumptivepro.20 The reason for this is the already mentioned fact that PPs don’t agree with lexical complements, and in a plain extraction analysis of the singular agreeing variant the source of the agreement remains mysterious. (´E. Kiss does not give an explanatory account of why the extraction is obligatory with the overt P-suffix.) A resumptiveproanalysis of apparent extraction from dressed PPs enables us to keep two robust generalizations that we see again and again: i) these Ps agree only with pronouns and ii) they are inseparable from their complement.21
This analysis is supported by several considerations. First, the existence of null resumptive pros has been argued for independently in den Dikken (1999); ´E. Kiss (2002); Gervain (2002, 2003, 2004) and Gervain (2005) (in connection with possessive structures and sentences with long operator movement). Second, null resumptive pros are always invoked in Hungarian to account for otherwise inexplicable, mysterious agreement facts, and apparent extraction from dressed PPs involves precisely such a mysterious agreement pattern. Finally, cross-linguistically we often see that in non-P-stranding languages or with non-P-stranding Ps adjacency violations between the adposition and its complement are obviated by the inclusion of a resumptive pronoun in the clause.
For specific examples I refer the reader to Aboh’s (2004a; 2005; 2010b) discussion of Gungbe (Gbe) and Muriungi’s (2006) observations about Kˆıˆıtharaka (Bantu, Kenya). Both languages feature strandable and non-strandable Ps. Their non-strandable Ps become separable from their complement if the DP is resumed by an overt resumptive pronoun adjacent to the adposition. The claim here is that Hungarian uses exactly the same strategy with dressed Ps, only the resumptive pronoun used in Hungarian PPs has no phonetic form.