<<

ISLAND CONSTRAINTS AND EXTRACTION OF LEXICAL CASE-MARKED DPS IN FINNISH AND TURKISH

REBECCA TOLLAN University of York

Abstract

This paper investigates the acceptability of types of wh-question formation which involve extraction out of strong syntactic islands in with rich lexical case marking systems. By considering these types of extraction, I aim to test Cinque’s (1990) diagnostic regarding the distinction between different island- hood strengths, namely that so-called ‘strong’ islands, whilst sometimes allowing DP extraction, do not allow extraction of a PP due to unavailability of PP resumption at the extraction site. I will claim, based upon evidence from extraction of DPs in Finnish and Turkish, that this hypothesis is certainly on the right track, but suggest an potential alternative explanation for the PP-DP extraction asymmetry observed in other languages.

1. Introduction

Cinque (1990) proposes that the distinction between strong and weak islands lies in the acceptability of PP extraction, whereby an island from which PP extraction is licensed must also allow DP extraction (a so-called weak island, involving genuine movement), whereas an island allowing only DP extraction is considered as ‘strong’. According to Cinque, extraction from strong islands involves a resumption strategy, whereby the site from which a DP has been extracted is occupied by a null resumptive pronoun; since PPs cannot be resumed, PP movement is not licensed. Extraction from an adjunct island in English, for example, may take place only when any preposition selecting the extracted DP is stranded, and hence only the DP is moved, as in (1); the preposition cannot be pied-piped, as in (2). (1) did leave [without talking to_]? (2) *To whom did you leave [without talking_]? The aforementioned hypothesis, however, is based only upon data from English and Italian, both of which have relatively few types of case marking. Cinque does not consider languages in which arguments expressed as a PP in English and/or Italian would be expressed as a lexically case-marked DP. This paper addresses behaviour of extraction in two such languages: Turkish and Finnish, as well as concurrently considering analyses of Finnish by Nikanne (1993) which claim that all lexically case-marked DPs are in fact PPs. Based upon data collected by means of two questionnaires, I will suggest that Cinque’s hypothesis is, on

York Papers in Linguistics Series 2 ISSN 1758-0315 Issue 12a © The Author, 2012

103 Island Constraints and Extraction of Lexical Case-marked DPs in Finnish and Turkish

the whole, correct; that DPs with lexical case cannot be analysed as PPs under the Cinquean diagnostic; and that the structure of certain spatial DPs impacts upon their extraction properties.

1.1. Structural and lexical case

Throughout this paper, a distinction is made between ‘structural’ and ‘lexical’ case marking. I define structural case here as (according to Davison (2004)) case which is theta independent and assigned as a function of structural position. Lexical case is defined as that which is theta-related and assigned due to lexical selection.

2. Background: weak and strong islands in English

A distinction is typically made in syntactic literature between strong and weak island types. Strong islands (also referred to as ‘absolute’ (Szabolcsi 2002) or ‘locked’ (Postal 1997/8)) are those which do not generally allow extraction, unless, in occasional circumstances, with an appropriate resumptive pronoun. Weak (also known as ‘selective’ (Szabolcsi 2002) or ‘unlocked’ (Postal 1997/8)) islands allow more extraction than strong islands, although certain selectional requirements can restrict what types of phrases may be extracted (this is usually based upon referentality and specificity). The types of islands listed by Cinque (1990:7) as strong and weak are given below in (3) and (4), with relevant examples. (3) Strong islands a. island *Which books did [talking about _] become difficult? *Which conference did [that Anne attended _] surprise you? b. Complex DP island *To whom have you found [someone who would speak_]? *Who did you read [a book that Peter gave to_]? c. Adjunct island *To whom did you leave [without speaking_]? *Which shop did you go to the farm [after buying groceries from _]? (4) Weak islands a. Wh-island ? To whom didn’t they know [when to give their present_]? *How did they ask [who behaved_]? b. Inner negative island To whom [didn’t you speak_]? *How [didn’t you behave_]? c. Factive island To whom do you regret [that you could not speak_]? *How do you regret [that you behaved_]? d. Extraposition island To whom [is time to speak_]? *How [is it time to behave_]? This study focuses specifically on extraction from strong islands. Cinque observes that some islands of this type can sometimes allow extraction of a DP, but never a PP (as was illustrated in the examples in (1) and (2)). therefore makes the claim in (5) regarding the strong/weak distinction.

Rebecca Tollan 104

(5) Cinque’s strong/weak diagnostic (as worded by Szabolcsi 2002:4) Among those domains that do not allow all standard extractions, those that allow a PP-gap are weak islands, and those that can at best contain a DP-gap are strong islands (and their DP-gap is an empty pronoun). On this account, a strong island with PP extraction such as in (2) is ungrammatical because a PP cannot be resumed. If Cinque’s diagnostic is taken to be true, then languages which convey the semantics of prepositions with lexical case marking (i.e. thereby expressing a would-be English PP such as ‘to whom’ as a DP with a case affix) should allow more extraction from strong (i.e. subject, complex DP and adjunct) islands than does English. In particular, constructions of the type in (2) ought to be acceptable, since unavailability of (PP) resumption should not be a factor in a which expresses a PP argument of this type as DP. An investigation into DP extraction in such a language ought therefore to serve as a true test of Cinque’s claim in (5).

3. Case marking systems in Turkish and Finnish

Of those languages whose case marking systems allow for certain would-be English PPs to be expressed as DPs, Turkish and Finnish were chosen as test-languages for this study. Reasons for this choice were based upon three conditions: firstly, availability of native speakers who were able to give grammaticality judgements; secondly, licensing of certain types of extraction from islands in general (Polish and German, for example, were considered as test- languages for the study but were ruled out due to general dislike of speakers for extraction of any type, even where no lexical case marking is involved); thirdly, the specific types of case- marking found in Turkish and Finnish will, when studied together, allow for testing of a wide variety of different PP-equivalents. In particular, Finnish has a rich system, thereby allowing for investigation into extraction of different locative DPs. Finnish does not, however, have an equivalent to the Indo-European (which would be typically used to express the PP ‘to whom’ from example (2)). Turkish on the other hand, has only two types of locative case but also has a dative case marker, allowing for more effective testing of sentences of types (1) and (2). A background to the basic and case systems in these two languages follows in sections 3.1–2. 3.1 Case marking in Turkish

Turkish belongs to the Altaic language family. Basic word order is SOV. It is generally regarded as a wh-in situ language, although Akar (1990) proposes that Turkish wh-phrases undergo raising to [spec, CP] at LF. Turkish has a six-case system, as outlined in (6). The former three (nominative, accusative and genitive) are structural cases and the latter two (locative and ) are lexical. Dative case can have either structural or lexical use (von Heusinger & Kornfilt 2005). Turkish also has a number of postpositions. (6) Turkish case system Nominative: (zero ) expresses grammatical subject Accusative: (-i) expresses grammatical Genitive: (-in) expresses ownership/ Dative: (-e) expresses movement towards, as conveyed by the English preposition ‘to’ Locative: (-de) expresses static position Ablative: (-den) expresses movement away 105 Island Constraints and Extraction of Lexical Case-marked DPs in Finnish and Turkish

I aim in particular to collect acceptability judgements for extraction of dative- and locative- marked DPs from strong islands ( is addressed as part of the discussion of Finnish locative case in section 6).

3.2 Case marking in Finnish

Finnish is a member of the Uralic language family. Basic word order is SVO, although many types of scrambling are permitted. The Finnish case system comprises some sixteen different cases, as listed in (7). In addition, Finnish also has a small number of postpositions. (7) Finnish case system (from Holmberg & Nikanne 1993) Structural cases: Nominative: (zero inflection) expresses grammatical subject Genitive: (-n) expresses ownership Partitive: (-ä) expresses specific grammatical object Accusative: (-n/-Ø) expresses less-specific grammatical object General locative cases: Translative: (-kse) expresses change of state Essive: (-nä) expresses English ‘as’ Internal locative cases: Inessive: (-ssä) ‘in’ (stative) Illative: (-on) ‘into’ Elative: (-stä) ‘from (within)’ External locative cases: Adessive: (-llä) ‘on’ (stative) Allative: (-lle) ‘onto’ Ablative: (-ltä) ‘from (on)’ Marginal cases: Abessive: (-ttä) ‘without’ Comitative: (-ne) ‘together with’ Instructive: (-n) expresses instrument/method Prolative: (-tse) ‘via’ Due to the small scale of this study, my collection of acceptability judgments for Finnish island conditions focuses on extraction of DPs marked for internal locative case. The external locative cases are, however, an important part of the analysis in section 6. Research into marginal and more general locative cases awaits further research. Although case-marked in Finnish appear at first glance to be DPs, Nikanne (1993) claims that lexical (or ‘semantic’) cases (i.e. locative and marginal cases) must be analysed as PPs. Her reasons for preferring a PP analysis are limited: notes only that lexical case DPs have the same distribution as PPs (example in (8), from Nikanne 1993:75-6). (8) a. talo sijaitsee liki Toukolan kylää house is-located near Toukola+GEN village+PAR ‘The house is located near Toukola village’ b. talo sijaitsee liki Toukolan kylässä house is-located Toukola+GEN village+INE ‘The house is located in Toukola village’

Rebecca Tollan 106

Nikanne concludes that lexical case phrases are PPs, whose head is an empty postposition which selects D and assigns to it the relevant case, as illustrated in (9). (9) The lexical case phrase according to Nikanne ‘93 kylässä ‘in (the) village’

PP

DP Ø [P, uD, case: INESSIVE] kylä+ssä [D, case:INE]

If this analysis is correct, extraction of lexical case from a strong island should be impossible. In section 5 I will argue that Nikanne’s model is incompatible with Cinque’s claims regarding strong islands and instead propose an alternative, whereby lexical case phrases are analysed as DPs after all.

4. Questionnaire and results

Two small-scale questionnaires (one for Turkish, one for Finnish) were put together in order to obtain judgements of acceptability for different types of strong island violations with extraction of different case-marked nouns. Constraints both on the scale of this project and on the availability of speakers imposed some limitations of the number of token sentences in each questionnaire. The results, therefore, are based only upon the judgements a small number of sentences by a small number of speakers. Further testing, with more speakers and more construction types, is most definitely required to substantiate all judgements and analyses given from here onwards.

4.1 Questions

The types of constructions used in the questionnaires are given in (10). For each island type, a question with extraction of a structural accusative/genitive (Turkish) or partitive (Finnish)1- marked DP was included as base-test. This was done on the basis that, if extraction of a DP marked with structural case is acceptable (i.e. where no lexical-case semantics are involved), then a benchmark is set for DP extraction from this island type to be permitted per se. (10) Adjunct island extraction Finnish: Mitä lähdit sanamatta? (Partitive extraction) what-PAR leave-2SG say-ABE-2SG ‘What did you leave without saying _?’

1 Extraction of in Finnish is much preferred to extraction of (Huhmarniemi 2009) due to referent specificity: partitive case marks a specific physical object whilst accusative marks a less-specific, often non-physical entity. Parititve case was therefore used as the base-test in Finnish. Extraction of is not permitted due to use of the genitive as a default case marker as default subject case; see Vainikka (1993) for further detail. 107 Island Constraints and Extraction of Lexical Case-marked DPs in Finnish and Turkish

Finnish: Kenen puutarhassa paloit istuessasi? (Inessive extraction) who garden-INE surburnt-2SG sit-whilst-2SG ‘*?In whose garden did you get sunburnt whilst sitting _?’ (also tested with extraction of elative and illative cased marked nouns) Turkish: Peter kimi gördükten sonra bıraktı? (Accusative extraction)

Peter who-ACC see after leave-3SG

‘Who did Peter leave after seeing_?’

Turkish: Kitabı verdikten sonra kime bıraktın? (Dative extraction)

book-ACC give after who-DAT leave-2SG

‘*To whom did you leave after giving the book _ ?’

(also tested with locative extraction)

Complex DP island extraction Finnish: Mitä Sara sai tilaisuuden tutkia? (Partitive extraction) what-PAR Sara receive-3SG opportunity explore-INF ‘What did Sara receive an opportunity to investigate_?’ Finnish: Missä talossa Sara sai tilaisuuden elää? (Inessive extraction) which-INE house-INE Sara receive-3SG opportunity live-INF ‘In which house did Sara receive an opportunity to live_?’ Sentential subject island extraction Turkish: Kimin evi Melike'nin gittiği seni şaşırttı? (Genitive extraction)

whose house-GEN Melike-GEN go-PAST 3SG you-ACC surprise-PAST 2SG

‘*To whose house did [that Melike went _] surprise you?’ Turkish: Kiminle Melike'nin gizli kodu verdiği Deniz'i çakıştırdı? (Dative extraction)

whom-DAT Melike-GEN secret code give-PAST 3SG Deniz-ACC make cross-PAST 3SG

‘*To whom did [that Melike gave the secret code _] make Deniz cross?’

(also tested with locative extraction) Each questionnaire was responded to by three native speakers of the relevant language. Full details of questionnaire construction (including filler sentence types) and all questionnaire tokens are given in the appendix. Judgement results from the questionnaires are discussed in section 4.2.

4.2 Results and analysis

Results for extraction from each island type are listed in the tables in (11) and (12). These are based upon judgments of questionnaire respondents, full details of which can be found in the appendix. Participants in the Finnish survey reported that some extraction types, though

Rebecca Tollan 108 grammatical, were unnatural and ‘awkward’-sounding; this is indicated in the table in (12) with a question mark. (11) Extraction in Turkish

Island Type Case of extracted English Extraction DP equivalent allowed? Adjunct Accusative DP Yes Adjunct Dative PP Yes Adjunct Locative PP Yes Adjunct2 Ablative PP No Sentential subject Genitive DP Yes Sentential subject Dative PP Yes Sentential Subject Locative PP Yes

(12) Extraction in Finnish

Island Type Case of extracted English Extraction allowed? DP equivalent Complex DP- Partitive DP Yes infinitival complement Complex DP Inessive PP Yes infinitival complement Adjunct Partitive DP Yes Adjunct Inessive PP Yes Adjunct Elative PP ? Adjunct Illative PP ?

Based upon the results in (11) and (12), it seems that Cinque’s diagnostic is correct (or at least, on the right track). Extraction of locative-marked DPs from within strong islands is acceptable in Finnish and in Turkish (as is dative extraction in Turkish), where its English equivalent, (expressed as a PP) is not. The discussion from here onwards will therefore be concerned with two principle issues: firstly, an alternative to Nikanne’s analysis of lexical case phrases as PPs (as discussed in section 3.2) and secondly, further investigation into the unnaturalness of elative and illative extraction in Finnish and the unacceptability of Ablative extraction in Turkish. This will in turn suggest that Cinque’s diagnostic in (5) may require some further attention.

5. The lexical case phrase as a DP: an analysis of case assignment

Nikanne’s (1993) analysis of lexical case phrases as DPs (adopted by Vainikka 1993 and Huhmarniemi 2009) needs to be revised in light of the results in (11) and (12), if are to take Cinque’s strong-weak diagnostic as correct. If Nikanne’s view of a lexical case phrase (as shown in (9) and repeated below for convenience) were correct, we would expect for them to resist extraction from strong islands due to unavailability of PP resumption.

2 Due to time constraints, the judgement for this construction type was only tested with one participant 109 Island Constraints and Extraction of Lexical Case-marked DPs in Finnish and Turkish

(9, repeated; kylässä ‘in (the) village’)

PP

DP Ø [P, uD, case: INESSIVE]

kylä+ssä [D, case:INE]

One possible way to reconcile Cinque’s and Nikanne’s analyses is to query whether the null P in Nikanne’s model is stranded in situ when the DP is moved to [spec, matrix CP] in the formation of these long-distance questions. A simple test to tell whether this could be the case here is to check whether overt postpositions may be stranded. We cannot tell for Turkish questions, since the wh-element remains in situ. However, in Finnish, all postpositions must be pied-piped; stranding is not an option (as shown in (13)). (13) Postposition movement in Finnish Pied-piped: Kenen Kanssa soit illallisen? who with eat-2SG dinner ‘With whom did you eat dinner?’ Stranded: *Kenen soit illallisen kanssa? who eat-2SG dinner with ‘Who did you eat dinner with?’ Nikanne’s analysis in (9) must therefore be rejected. Instead, we must assume a DP structure for a lexical case phrase, where case is a feature on the which selects it. The structure of the Finnish ‘istua puutarhassa’ (to sit in the garden) is as in (14) (14) VP

istua [V, uD, DP case: ILLATIVE,

Θ: location] puutarhassa [D,

case:ILL, uΘ: location]

By contrast, the structure of the English equivalent involves different selectional requirement on the verb to sit. Sit selects P instead of D, and it is P which assigns case3 (see (15)).

3 I assume here that English P assigns , as according to Hornstein & Weinberg (1981)

Rebecca Tollan 110

(15) VP

sit [V, uP , PP Θ:location ]

in [P, uD, DP case: OBL]

the garden [D, ucase: OBL, uΘ: location]

As such, we can make the following distinctions (in 16) between Finnish and English. There is no reason at this point not to assume that the analyses in (14) for Finnish does not also hold true for Turkish. (16) Certain in English (e.g. sit) select prepositional complements. These prepositions select DPs. Case is assigned to the DP by the preposition. The Finnish equivalents to such verbs select complements. Case is assigned to the DP by the verb. From this point onwards, I consider all lexical case phrases in Finnish and Turkish to be DPs.

6. Locative case in Finnish

The following analysis focuses specifically on internal locative case in Finnish and in particular the issue surrounding the unnaturalness of extraction of illative and - marked nouns. As was reported by participants in the questionnaire in section 4, questions in which such nouns are extracted from a strong island sound less natural than those from which is extracted. They are, however, by no means ungrammatical by comparison with other examples given in the literature (e.g. Huhmarniemi 2009) where extraction is most certainly not acceptable. Sentences and judgements are given in (17). The sentence in (d), where illegal extraction of a broad (zero) accusative interrogative has taken place, is given as a point of comparison for the judgements of sentences in (a-c). This is purely in order to emphasize the fact that extraction in (b-c), whilst denoted with a question mark, is still far better than what we could consider ‘true’ illegal extraction ((d)4). Judgements from (a-c) are in accordance with feedback from the questionnaire.

4 Extraction in (d) is unacceptable because the accusative interrogative ‘mikä’ has no specific referent. By contrast, its partitive counterpart ‘mitä’, which has more specific reference, may be extracted, as below (from Huhmarniemi 2009:30)

Mitä Pekka kertoi että tontille rakennetaan _ ? what-PAR Pekka told that site-ALL build-3PL ‘What did Pekka tell that they are building on the site_?’

111 Island Constraints and Extraction of Lexical Case-marked DPs in Finnish and Turkish

(17) a. Inessive case extraction Kenen puutarhassa paloit istuessasi? who garden-INE sunburnt-2SG sit-whilst-2SG ‘*?In whose garden did you get sunburnt whilst sitting _?’ b. extraction ? Kenen taloon kiljuit nimeni käveltäysi? whose house-ILL shout-2SG my name walking-after-2SG ‘*Into whose house did you shout my name after walking _?’ c. Elative case extraction ? Kenen talosta huutelit nimeäni tultuasi pelastetuksi? whose house-ELA call my name be rescued-after-2SG ‘*From whose house did you call my name after being rescued _?’ d. Broad accusative extraction (from Huhmarniemi 2009:30) *Mikä Pekka kertoi että tontille rakennetaan? what-ACC-Ø Pekka told that site-ALL build-3PL ‘What did Pekka tell that they are building on the site?’ Hence we are able to say that inessive extraction is better than illative/elative extraction, which is by far better than zero-accusative extraction. So the question to be addressed now is this: why does extraction of illative or elative case not sound as natural as extraction of inessive case? The first step to solving the puzzle is to consider the semantics of the case markers. Here, a noticeable distinction comes to light: inessive case (equivalent to the English preposition ‘in’ (see (7) in section 3.2)) carries stative meaning. Both illative (equivalent to ‘into’) and elative (equivalent to ‘away from’) convey directional meaning, which involves motion. So a potential hypothesis is that, for some reason, extracting a DP with directional locative case (e.g. illative) from within a strong island is more laboured than extracting a DP with stative locative case (i.e. inessive case), and that it is this which causes sentences such as (17 b-c) to sound slightly degraded compared with (a). A convenient means of testing this idea is to consider the corresponding external locative cases (adessive, allative and ablative). The latter two are the external equivalents of illative and elative cases and hence also convey directional meaning. Adessive (equivalent to the English preposition ‘on’) corresponds to inessive (‘in’). This is summarized in (18) (18) Internal and external locative cases in Finnish

Meaning (internal/external) Internal case External case in/on-STATIVE Inessive Adessive into/onto-MOTION Illative Allative away from (in/on)-MOTION Elative Ablative

If it is directional case semantics which impedes ease of extraction of illative and elative case-marked DPs, the same types of judgements as in (17 b-c) ought to be observed for similar questions with extraction of an allative or an ablative case-marked DP. I turn to this investigation in section 6. , on the other hand, should be as easily extracted as inessive case.

Rebecca Tollan 112

6.1. Extraction of external locative case

Examples of extraction of external locative case-marked DPs are given in (19). Adjunct extraction is tested as a means of fairest possible comparison with (17 a-c). Please note also that these further grammaticality readings from here onwards are based only upon the judgement of one native speaker. Clearly, further formal testing is much-needed, but my subsequent analysis will be based solely upon these judgements. (19) a. Adessive extraction Kenen kirjoituspöydällä nukahdit istuessasi? whose desk-ADE fall asleep-2SG sitting-whilst-2SG ‘*On whose desk did you fall asleep whilst sitting _?’ b. Allative extraction ? Mille helikopteri laskeutumisalustalle tarkistit ajan laskeuduttuasi? which-ALL helicopter pad check-2SG the time landing-after-2SG? ‘*Onto which helicopter pad did you check the time after landing _?’ c. Ablative extraction ? Kenen katolta huutelit nimeäni tultuasi pelastetuksi? whose roof-ABL shout-2SG my name rescued-after-2SG ‘*From whose roof did you shout my name after being rescued _?’ The judgements for (19) (i.e. allative and ablative extraction results in a slightly less well- formed question than adessive extraction) mirror those for (17) and support the hypothesis that the semantics of motion and/or direction seem to be responsible for the awkwardness of allative/illative/ablative/elative extraction compared with the two stative locative cases. The issue to address now is how this distinction can be encoded syntactically such that extraction of directional case DPs results in a slightly degraded-sounding question by comparison with stative case DPs. One possibility is to appeal to Nikanne’s view of lexical (for my purposes here, locative) case phrases as PPs with a null postposition, as discussed in sections 3.2 and 5. Suppose that, whilst stative case phrases (inessive, adessive) are DPs, directional locative case phrases (illative, elative, allative, ablative) are PPs with the structure in (9). This analysis is immediately problematic though, since it would predict ungrammaticality (comparable to that for the question in (17d)) of questions with directional case extraction. So instead, what is needed is an analysis which explains in particular why the structures in (17 b-c) and (19 b- c) are awkward-sounding, but not ‘ungrammatical’.

6.2 Place and Path

Peter Svenonius (2004) also observes the distinction between ‘stative location’ and ‘directional location’, and considers them as a separate class of syntactic entities, which he calls Place (stative) and Path (directional) (see 20).

113 Island Constraints and Extraction of Lexical Case-marked DPs in Finnish and Turkish

(20) Place and Path projections Place elements give information about the physical configuration of the relationship between a Figure and a Ground. Path elements give information specifying whether a place is a goal or a source, and may specify the orientation of the trajectory. (Svenonius 2004) Using evidence from the Chadic language Zina Kotoko, Svenonius argues in favour of a projection hierarchy in which Path occurs outside of Place, as in (21). (21) Path > Place > DP

PathP

Path PlaceP

Place DP

In English, Path is selected by a by a preposition with directional semantics, as in (22a). (22) a. E.g. ‘The child jumped onto the bench5’

VP

jump [V, PP

uP]

onto [P, upath, PathP

case:OBL]

(to) [Path, PlaceP

uplace ]

Ø [Place, uD] DP

the bench [ucase: OBL] As Svenonius concedes, it is unclear whether case is assigned to the DP from the preposition, but for the sake of arguments, I have assumed that the preposition assigns (oblique) case. When no direction is involved, a Place head only is selected (22b).

5 I have assumed here that the PP ‘onto’ is selected by ‘jump’, as opposed to being a VP adjunct; whether or not the PP is a complement or an adjunct is irrelevant to this discussion.

Rebecca Tollan 114

b. E.g. ‘The child sat on the bench’

VP

sit [V, uP] PP

on [P, uplace, PlaceP case: DAT]

Ø [place, uD] DP

the bench [ucase: DAT]

The important suggestion here, as far as an analysis of Finnish locative case is concerned, is that more structure is involved when the notion of direction is conveyed. In addition, it is questionable as to whether Place has any distinct semantics to P at all. In (b), Path conveys direction (‘to’), and if we assume that ‘on’, which refers to a Ground, is indeed a preposition generated in [head, P], then Place plays no role (syntactic or semantic) in the structure. In other words, it would make sense to assume that the semantics conveyed by Place are akin to those conveyed by P, and as such, the Place projection is, in effect, obsolete. Going forward then, I will assume a slightly modified version of Svenonius’ analysis, as in (23). (23) P is, by default, stative. The semantics of P give information about the relationship between a figure and a ground point. P selects D. In order for any directional, non-stative semantics to be conveyed, P must select Path, which selects D. The semantic role of a P which selects Path is to give information about the beginning or end point of the trajectory of motion. Whilst we can assume that (23) is true of English syntax, a further modification is needed for Finnish lexical (locative) case, since, as was established in section 5, lexical case is assigned to the DP by the selecting verb, such that no postposition is involved. As such, the following (24) can be said for Finnish. (24) Finnish verbs (e.g. ‘istua’ sit) which assign stative locative case (inessive, adessive) select D. Those with directional semantics (e.g. ‘pelastus’ rescue) must select Path. Path selects D. Since I have already assumed that, in English, case is assigned to the relevant by the preposition, I shall assume for consistency with prior argumentation that case in Finnish is assigned by the verb. So, whilst the analyses of the verb ‘sit (in)’ in (14) and (15) in section 5 (repeated for convenience) still hold, the structure of the VP ‘rescue (from)’ involves an extra projection, as shown in (14’) for Finnish and (15’) for English.

115 Island Constraints and Extraction of Lexical Case-marked DPs in Finnish and Turkish

(14, repeated; Finnish istua puutarhassa ‘to sit in the garden’)

VP

istua [V, uD, DP case: ILLATIVE,

Θ: location] puutarhassa [D,

ucase:ILL, uΘ: location]

(14’)6 Finnish pelastus ‘to rescue’

VP

pelastus [V, PathP uPath, case: ELATIVE, Θ: Ø [Path, uD] DP location ]

D, [uΘ: location, ucase:ELA]

(15, repeated; English sit)

VP

sit [V, uP, PP Θ:location]

in [P, uD, DP case: OBL,]

the garden [D, ucase: OBL, uΘ: location]

6 Since no distinction in extractability of external and internal cases has been observed, I assume that Finnish locative verbs can assign either external or internal case.

Rebecca Tollan 116

(15’) English rescue

VP

rescue [V, uP, PP Θ: Location]

from [P, upath, PathP case: OBL]

Ø [uD] DP

D [ucase: OBL, uΘ: location]

Somehow, the addition of an extra projection when directional case is assigned (i.e. illative, elative, allative and ablative cases) impedes extraction of a DP from a strong (e.g. adjunct) island, whereas this extra structure is not present when stative case (inessive or adessive) is assigned. So how does it impede extraction? There are two possibilities. Firstly, consider that the null head of Path is stranded in situ, and the resulting question sounds slightly degraded due to the need for the syntactic processor to make a long distance semantic link, across an island boundary, between DP and PathP (this account assumes that some element of the directional semantics is left behind in Path). This possibility seems slightly unfavourable, however, since we have already established that Finnish does not allow preposition stranding and it therefore seems unlikely that it should allow null Path stranding. The other, more likely option, is that null Path is pied-piped with the DP. As such, the need to pied-pipe (albeit a null element) over an island boundary causes the awkwardness of the resulting construction7.

7 The distinction between directional and stative locative case and the effects of directional semantics on extraction can by substantiated with data from Turkish Whilst the questionnaire results suggest that Turkish (stative) locative case-marked DPs are able to be extracted from strong islands, the same is not true of Turkish ablative case. In fact, the effect of directional case is more severe in Turkish than in Finnish: the example in (25), in which an ablative case-marked DP has been extracted from an adjunct island, is judged to be a completely unacceptable construction (see 11).

(25) *Kimin catisindan kurtarildiktan sonra rahatlamis hissediyor muydun?

whose roof-ABL rescue-ABL after relieved feel-2SG

‘*From whose roof were you feeling relieved after being rescued _?’

Why this directional case extraction should be less acceptable in Turkish than in Finnish is not clear, but we can only assume that some other syntactic or semantic factor is at work in Turkish which means that the presence of Path impedes upon extraction to a greater degree than in Finnish.

117 Island Constraints and Extraction of Lexical Case-marked DPs in Finnish and Turkish

7. PathP and its implications for Cinque’s strong/weak diagnostic

Whilst the results of the questionnaire have shown Cinque’s diagnostic in (5) to be, on the whole, correct, further analysis is required in order to take into account the distinction between stative and directional locative case movement. The original claim about DP extraction and resumption does not necessarily predict that any particular from of case marking should impede movement of a DP. The movement of a directional locative case DP is unacceptable in Turkish, and slightly degraded (compared with stative case DP movement) in Finnish. I have claimed that this is due to the addition of null Path head which must be pied-piped to [spec, matrix C] when its selected D is extracted from a strong island. It seems that pied-piping an extra element across and island boundary leads to ungrammaticality in Turkish, and a slightly awkward-sounding question in Finnish. As far as Cinque’s diagnostic is concerned, if ungrammaticality of PP extraction is due to unavailability of PP resumption, then acceptability of Path + DP extraction in should depend upon whether nor languages such as Turkish or Finnish allow PathP resumption. Based upon the small amount of data in this study, we could claim that ablative extraction from strong islands in Turkish is ungrammatical because Turkish does not have a resumption mechanism for Path. The data for Finnish is a little less clear because extraction of elements with a Path head was judged as neither wholly acceptable nor wholly unacceptable. Clearly, further grammaticality judgements, of a wider range of questions and from more speakers, are needed to allow for any firm hypothesis to be made.

8. Conclusion

I have shown through analysis of Turkish and Finnish long-distance question formation that Cinque’s strong/weak diagnostic for syntactic islands is, on the whole, correct. Extraction from a strong island of a DP marked with lexical case is generally acceptable in these languages, whilst the equivalent constructions in English, which involve PP extraction, are not. The extractability of Turkish and Finnish lexical case phrases does not favour Nikanne’s (1993) view that such structures are headed by a null postposition; instead, I adopt a ‘DP’ analysis in which case is assigned by the lexical and semantic properties of the verb which selects the DP in question. Based upon observations that extraction of a DP marked with directional locative case is somewhat degraded (although to a far greater degree in Turkish than in Finnish), I have built up a theory, based upon Svenonius’ (2004) analyses of spatial prepositions, which suggests that a null Path head is present in the syntax where directional semantics are involved. The need to pied-pipe this extra (albeit null) material is what causes extraction from strong islands of directional case-marked DPs to falter. The degree of unacceptability which this causes may be due to whether or not Path resumption is licensed. It is to be stressed, though, that the analysis in this paper is a hypothesis only, which requires substantial testing with more example questions and more native speakers.

Rebecca Tollan 118

References

AKAR, D. 1990. Wh-questions in Turkish. Unpublished M.A. thesis, Boğaziçi University, İstanbul, Turkey.

CINQUE, C. 1990. Types of A-bar Dependencies. Cambridge: MIT Press

DAVISON, A. 2004. ‘Structural case, Lexical case and the verbal projection.’ In Dayal, V. and Mahajan, A. (eds), Clause Structure in South Asian Languages. Dordrecht: Kluwer, pp. 199-225.

HOLMBERG, A. AND NIKANNE, U. 1993. Case and Other Functional Categories in Finnish Syntax, Studies in Generative Grammar 39: Mouton de Gruyter.

HORNSTEIN, N. AND WEINBERG, A. 1981. ‘Case Theory and Preposition Stranding’ in Linguistic Inquiry 12 (1), p.55-91.

HUHMARNIEMI, S. 2009. ‘Extraction islands in Finnish’ in Biolinguistica Fennica Working Papers 1 21–78.

NIKANNE, U. 1993. ‘On Assigning Semantic Cases in Finnish’. In Holmberg, A. & Nikanne, U. (eds.) Case and Other Functional Categories in Finnish Syntax, Studies in Generative Grammar 39: Mouton de Gruyter.

POSTAL, P. 1997. Islands. Manuscript, New York University.

POSTAL, P. 1998. Three investigations of extraction. Cambridge. Mass: MIT Press.

SVENONIUS, P. 2004. Spatial P in English University of Tromsø, available online at http://www.hum.uit.no/a/svenonius/papers/PEnglish3.pdf .

SZABOLCSI, A. 2002. ‘Strong vs. Weak Islands’ in The Syntax Companion, available at http://www-uilots.let.uu.nl/syncom/

VAINIKKA, A. 1993. ‘The Three Structural Cases in Finnish’. In Holmberg, A. & Nikanne, U. (eds.) Case and Other Functional Categories in Finnish Syntax, Studies in Generative Grammar 39: Mouton de Gruyter.

VON HEUSINGER, K. AND KORNFILT, J. 2005. ‘The case of the direct object in Turkish: Semantics, syntax and morphology’. In 9, 3-44.

119 Island Constraints and Extraction of Lexical Case-marked DPs in Finnish and Turkish

Appendix: Questionnaire constriction and distribution

All questionnaire respondents were native speakers of the language in question. Each question was judged on the basis of acceptability. All questions in the Turkish questionnaire are given in (1). (1) Turkish questions Adjunct island 1. Peter kimi gördükten sonra bıraktı?

Peter who-ACC see after leave-PAST

= Who did Peter leave after seeing?

2. Merve' verdikten sonra ne bıraktın?

Merve'ye-DAT to give after what leave

(=*? What did you leave after giving _ to Merve?)

3. Kitabı verdikten sonra kime bıraktın?

book-ACC give after who-DAT leave

(= *To whom did you leave after giving the book _ ?)

4. Esra oturuken kimin bahçesinde güneşten yandı?

Esra sitting-while whose-ACC garden-LOC sun burn

=*?In whose garden did Esra get sunburnt whilst sitting _?

5. Kimin catisindan kurtarildiktan sonra rahatlamis hissediyor muydun? whose roof-ABL rescue-ABL after relieved feel-2SG =*From whose roof were you feeling relieved after being rescued _?’

Sentential subject island 1. [Melike Katıldığı] hangi partide Sam'i kıskandırdı?

Melike attend-PAST 3SG which party-LOC Sam jealous-make-PAST 3SG

= *Which party did [that Melike attended _ ] make Sam jealous?

2. Kimin evi [Melike'nin gittiği] seni şaşırttı?

whose house-GEN Melike-GEN go-PAST 3SG you-ACC surprise-PAST 2SG

=*To whose house did [that Melike went _] surprise you?

Rebecca Tollan 120

3. Kiminle [Melike'nin gizli kodu verdiği] Deniz'i çakıştırdı? whom-DAT Melike-GEN secret code give-PAST 3SG Deniz-ACC make cross-PAST 3SG

=*To whom did [that Melike gave the secret code _] make Deniz cross?

Finnish questions are given in (2).

(2) Finnish questions

Complex DP with infinitival complement 1. Mitä Sara sai tilaisuuden tutkia? what-PAR Sara receive-3SG opportunity explore-INF (What did Sara receive an opportunity to investigate?) 2. Kenen kanssa Sara sai tilaisuuden työskennellä? Who-GEN with-INE Sara receive-3SG opportunity work-INF (With whom did Sara receive an opportunity to work?) 3. Missä talossa Sara sai tilaisuuden elää? Which-INE house-INE Sara receive-3SG opportunity live-INF (In which house did Sara receive an opportunity to live_?) Complex DP with relative clause 4. Kenen kanssa Emmi näki miehen, dei pakeni muurin yli? Who-GEN with-INE Emmi see-ESG man that escaped-3SG over wall (*With whom did Emmi see the man that escaped over the wall_?) Adjunct 1. Mitä lähdit sanamatta? what-PAR leave-2SG saying without-ABE (What did you leave without saying?) 2. Kenen puutarhassa paloit istuessasi? who-GEN garden-INE sunburt-2SG sitting-whilst-2SG (*?In whose garden did you get sunburnt whilst sitting_?) 3. ? Kenen talosta huutelit nimeäni tultuasi pelastetuksi? who-GEN house-ELA shout-2SG my name after rescued-2SG (*from whose house were you shouting my name after being rescued?) 4. ? Kenen taloon kiljuit nimeni käveltäysi? Whose house-ILL shout-2SG my name walking-after-2SG (*Into whose house did you shout my name after walking_?)

Two types of filler were included: relative clause fillers (which are comparable on length to the types of questions in (1) and (2)) and questions with no island extraction. Fillers are given in (3). All filler types were judged as fully acceptable. (3) Fillers (a) Turkish 1. Anna kiminle seyahat etmeyi sever? (=with whom does Anna like to travel) 2. Neden Sophie kitabi Peter'a verdi? (=why did Sophie give the book to Peter) 3. Sophie nin tatilde kaldigi ev yakin zamanda yikildi (=the house that Sophie stayed in whilst on holiday was knocked down recently) 4. partide cocuklari eglendiren kisiyi herkes sevdi. 121 Island Constraints and Extraction of Lexical Case-marked DPs in Finnish and Turkish

(=everyone liked the person who entertained the children during the party) (b) Finnish 1. Missä Anna asui viime vuonna? (Where did Anna live last year?) 2. Sara puhui miehen kanssa joka voitti pääpalkinnon (Sara spoke to the man who won the top prize) 3. Mitä Paul antoi Saralle? (What did Paul give to Sara?) 4. Pyörä jonka Anna osti Peterille oli jo rikki. (The bike that Peter bought for Anna was already broken) Questionnaires were constructed such that no tokens that were of the same island or filler type occurred in succession. Participants said whether each sentence was acceptable (marked as /), unacceptable (marked as -) or acceptable but unnatural (marked as ?). A breakdown of results (excluding filler sentences) is given in (4) and (5). (4) Results (Turkish)

Island Type Case of Subject 1 Subject 2 Subject 3 extracted DP Adjunct Accusative / / ? Adjunct Dative / / / Adjunct Locative / / ? Adjunct Ablative - NA NA Sentential Locative / / / subject Sentential Genitive / / / subject Sentential Dative / / / Subject

(5) Results (Finnish)

Island Type Case of extracted Subject 1 Subject 2 Subject 3 DP Complex DP- Partitive / / / infinitival complement Complex DP Inessive / / / infinitival complement Adjunct Partitive / / / Adjunct Inessive / / / Adjunct Elative ? ? ? Adjunct Illative ? / ?

Rebecca Tollan 122

Rebecca Tollan Department of Language and Linguistic Science University of York Heslington York email: [email protected]