Contrastivity in island repair: an argument for parallelism

1
Contrastivity in island repair an argument for
parallelism

• James Griffiths (University of Groningen)
• Anikó Lipták (Leiden University)

2
Synopsis

• In this talk, we are going to show that
• Sluicing and fragments show identical behaviour
  with respect to island obviation (contra Merchant
  2004).
• Whether any instance of clausal ellipsis shows
  island obviation depends on its relation with its
  correlate
• ? contrastive remnants in clausal ellipsis
  cannot repair islands
• ? non-contrastive remnants in clausal ellipsis
  can repair islands
• (iii) Behaviour in island repair in both cases
  is due to obligatory parallelism between remnant
  and correlate (Fox and Lasnik 2003).

3
Types of clausal ellipsis

• Interrogative fragment (usually referred to as
  sluicing)
• (1) A Someone hit John last night. B
  Who? B Who lthit John last nightgt?

Declarative fragment (usually referred to as
fragments) (2) A MARY hit John last night. B
No, BILL. B No, BILL lthit John last nightgt.
3
4
Fragments island obviation

• Interrogative fragments obviate strong islands
  (Merchant 2001)
• (3) A They want to hire someone who speaks
  a Balkan language.
• B Which Balkan language?
• B' Which Balkan language do they want
  to hire someone who speaks?
  

Declarative fragments obey strong islands
(Merchant 2004) (4) A Does Abby speak the
same Balkan language that BEN speaks? B
No, CHARLIE. B No, Abby speaks the
same Balkan language that CHARLIE speaks.
5
PF-deletion approach to fragments

• Both interrogative and declarative fragments are
  derived by movement of the remnant and
  PF-deletion (Merchant 2001, 2004)

(5) A Someone hit John last night. B Really?
Who1 t1 hit John last night? Step 1 movement of
who B Really? Who1 t1 hit John last
night? Step 2 PF-deletion of TP

• (6) A MARY hit John last night.
• B No, BILL t1 hit John last night.
• B No, BILL t1 hit John last night.

5
6
PF-deletion theory of islands (1)

• Merchant (2004), building on Lasnik (2001)
• i. All copies of island-violating wh-movement
  are -marked, except the topmost copy
• ii. -marked copies are uninterpretable to the
  PF component they prevent convergence
• iii. However, if all -marked copies are
  contained within an ellipsis site (which is
  deleted at PF), convergence can still occur
• Required assumption the remnant moves
  successive-cyclically by adjunction to every
  functional projection

7
PF-deletion theory of islands (2)

• (7) A They want to hire someone who
  speaks a Balkan language.

B' Which Balkan language do t1 they t1 want
to t1 hire t1 someone t1 who t1 speaks t1?
B' Which Balkan language do t1 they t1 want to
t1 hire t1 someone t1 who t1 speaks t1?
8
PF-deletion theory of islands (3)
Interrogative fragment

• Declarative fragment

someone t1 who speaks t1
language t1 that t1 speaks.
A They want to hire someone who speaks a
Balkan language.B Which Balkan language do they
t1 want to hire someone t1 who t1
speaks t1?
A Abby speaks the same Balkan language
that BEN speaks. B' (No,) CHARLIE t1 Abby t1
speaks t1 the same Balkan language t1
that t1 speaks.
9
Problems with the PF-deletion theory of islands
(1)

• Empirical problem I

Not all interrogative fragments are
island-insensitive
Sprouting (Chung, Ladusaw McCloskey 1995) (8)
A Sandy is trying to work out which student
would speak. B To whom1 is Sandy trying
to work out which student would speak t1?
Contrast sluicing (Merchant 2008) (9) A
Abby wants to hire someone who speaks Greek.
B What OTHER languages does Abby want to hire
someone who speaks t1?
10
Problems with the PF-deletion theory of islands
(2)

• Empirical problem II
• Not all declarative fragments are island
  sensitive (cf. also Valmala 2007)
• A Does Abby speak the same Balkan language that
  someone in your syntax class speaks?
• B Yeah, Charlie1 Abby speaks the same Balkan
  language that t1 speaks.
• A I imagine John wants a detailed list.
• B Im afraid he does. Very detailed1
  John wants a t1 list.

11
Problems with the PF-deletion theory of islands
(3)

• Empirical problem III
• Languages where interrogative and declarative
  fragments are structurally identical (e.g.
  Hungarian) show the same pattern of island
  obviation as English.
• Hungarian wh- and focus-fronting both target the
  same structural position, FocP (Horváth 1986, É.
  Kiss 1987, Brody 1995)
• (12) a. Tegnap FocP kit TP
  hívott meg Mari ?
• yesterday who.A invited PV Mari
• Who did Mari invite yesterday?
• b. Tegnap FocP PÉTERT TP hívta meg Mari
  .
• yesterday Péter.A invited PV Mari
• Mari invited PÉTER yesterday.

11
12
Problems with the PF-deletion theory of islands
(4)

• The same position, FocP, is targeted by the
  remnants in clausal ellipsis
• (13) A Mari meghívott valakit.
• Mari PV.invited someone.A
• Mary invited somebody.
• B FocP Kit TP hívott meg Mari
  ? interrogative fragment
• who.A invited PV Mari
• Who?
• (14) A Kit hívott meg Mari?
• who.A invited PV Mari
• B FocP PÉTERT TP hívta meg
  Mari. declarative fragment
• Péter.A invited PV Mari
• Mari invited PÉTER yesterday.

13
Problems with the PF-deletion theory of islands
(5)

• FocP is one functional projection higher than TP.
  Consequently, no -marked traces should escape
  PF-deletion, either in the case of interrogative
  fragments or declarative ones.

Thus, the PF-island approach predicts that all
Hungarian fragments should obviate islands.
14
Problems with the PF-deletion theory of islands
(6)

• However, just like in English, island-obviation
  is permitted in interrogative fragments and
  prohibited in declarative fragments.
• (15) Keresnek valakit aki beszél egy
  bizonyos szláv nyelvet,
• search.PL somebody.A REL speaks a
  certain Slavic language.A
• de nem tudom, melyiket.
• but not know which.A
• They are looking for someone who speaks a
  certain Slavic language but I don't know which
  one.
• (16) A OLYAN KUTATÓT keresnek, aki az
  OROSZT beszéli?
• such researcher.A search.PL REL
  the Russian.A speaks
• Are they looking for a researcher who speaks
  RUSSIAN?
• B Nem, a KÍNAIT.
• no the Chinese.A
• No, Chinese.

15
Types of fragments (1)
Contrastive vs. non-contrastive remnants
contrast relation with respect to the
correlate in the antecedent ? Contrastive
fragments (17) A Did John eat A PIZZA for
dinner? B No, a SALAD. (salad ?
pizza) (18) A John ate A PIZZA for
dinner. B No, a SALAD. (salad ?
pizza) ? Non-contrastive fragments (19) A What
did John eat for dinner? B A salad.
(salad ? what) (20) A John ate something
for dinner. B Indeed, a salad. (salad ?
something)
16
Types of fragments (2)

• Important notes
• ? contrastivity of a remnant ? contrastive focus
  on the remnant
• (21) A What did John eat for dinner?
  (non-contrastive fragment)
• B A SALAD, and not a steak.
• ? contrastive remnants always have the
  information structural role of contrastive focus
• (22) A John eat a PIZZA for dinner.
  (contrastive fragment)
• B No, A SALAD.

17
Types of fragments (3)
Fragment contrastive non-contrastive
IS role of fragment contrastive focus information focus or contrastive focus
some illustrative examples ? corrections A The PIZZA was cold. B No, the STEAK. ? answers to yes/no questions A Was the PIZZA cold? B No, the STEAK. ? contrast sluices I know how many PIZZAS John ate, but I dont know how many SALADS. ? elaborative fragments A You have eaten something. B Yeah, a pizza. ? answers to wh-questions A What did you eat? B A pizza. ? tag-questions What did you eat, a pizza?
18
The correct generalization on island obviation
(1)
English non-contrastive remnants show island
obviation (23) A I heard that a biography of one
of the Marx brothers is going to be published
this year. B Yeah, of Groucho. / B'
Excellent. Of which / whom? (24) A I imagine
John wants a detailed list. B I'm afraid he
does. Very detailed. / B' How detailed? (25) A
I heard that Irv and a certain someone from your
syntax class were dancing together last
night. B Yeah, Bill. / B' Really?
Who? (26) A I heard they hired someone who
speaks a Balkan language fluently. B Yeah,
Serbo-Croatian. / B' Really? Which? (27) A I
hear that Abby is likely to get mad if Ben speaks
to one of the guys from your syntax class. B
Yeah, John. / B' Really? Who?
19
The correct generalization on island obviation
(2)
English contrastive remnants show no island
obviation (Merchant 2004) (28) A I heard that a
biography of the youngest Marx brothers is going
to be published this year. B No,
of the OLDEST. (29) A I imagine John wants a
short list. B No, LONG. (30) A I heard
that Irv and John were dancing together last
night. B No, BILL. (31) A I heard they
hired someone who speaks Bulgarian fluently.
B No, SERBO-CROATIAN. (32) A I hear that
Abby is likely to get mad if Ben speaks to Mary.
B No, BILL. (33) A Abby wants to hire
someone who speaks Greek. B What OTHER
languages?
20
The correct generalization on island obviation
(3)
Chinese (34) A ni renshi yi-ge jiang
shenme wen de ren? you know one-cl
speak what language de person lit. You know
someone who speaks what language?
B E-wen. Russian Russian (35) A ta
renshi yi-ge jiang e-wen de ren he know
one-cl speak Russian de person He knows
someone who speaks Russian. B bushi,
ri-wen not.be Japanese No, Japanese.
21
The correct generalization on island obviation
(4)
Turkish (36) A Hasan kim-i görecegiz diye
bir ekmek daha almis? Hasan who-A will.see
for one bread more bought lit. Hasan
bought another loaf of bread because he will see
who? B  Mehmed-i. Mehmed-A Mehmed. (37)
A Hasan Mehmed-i görecegiz diye mi bir
ekmek daha almis? Hasan Mehmed-A will.see
for Q one bread more bought  lit. Hasan
bought another loaf of bread because hell see
Mehmed? B ? Hayir, Ali-yi.
no Ali-A No, Ali.
22
The correct generalization on island obviation
(5)
? empirically correct generalization on island
obviation (38) Contrastive fragments cannot
obviate islands. Non-contrastive fragments
can.
23
Parallelism in fragments

• Hint behaviour of non-contrastive remnants in
  wh-in-situ languages (Chinese, Turkish)
• (39) A ni renshi yi-ge jiang shenme wen
  de ren?
• you know one-cl speak what
  language de person
• lit. You know someone who speaks what
  language?
• B E-wen.
• Russian
• Russian
• (40) A CP whi island node ti ? LF
  of question
• B CP fragmenti island node ti
  LF/PF of answer
• ? Scopal parallelism (Fox 2000, Fox and Lasnik
  2003)
• Variable in the antecedent and the elided clause
  are bound from parallel position.

24
Parallelism in non-contrastive fragments
Scopal parallelism at work in English interrogati
ve fragments (see also Merchant 2001) (40) A
Mary kissed someone last night. B Who1
ltMary kissed t1 last nightgt? LF A someone1
?x (TP Mary kissed x1 last night)
B who1 ?x (TP Mary kissed x1 last
night) declarative fragments (41) A Mary
kissed someone last night. B Yeah, Bill1
ltMary kissed t1 last nightgt. LF A someone1
?x (TP Mary kissed x1 last night) B
Bill1 ?x (TP Mary kissed x1 last night)
25
Parallelism in contrastive fragments (1)
Observation 1 A well-formedness condition on
contrastive fragments (42) A John was tired
and hungry. He had a pizza in his favourite
restaurant on his way home. B No, a
SALAD. (43) A Of all the people who were
hungry, JOHN had a pizza in his
favourite restaurant on his way home. B No,
a SALAD. (44) A Of all the things he likes,
John had a PIZZA in his favourite restaurant
on his way home. B No, a SALAD. ? Felicity
condition on contrastive fragments Contrastive
fragments are only felicitous if their correlate
is contrastively focused.
26
Parallelism in contrastive fragments (2)
Observation 2 Contrastive focus is
island-sensitive (Drübig 1994, Rooth 1997, Krifka
2006) (45) John was only happy when JILL
arrived. there is a set of times when someone
arrived (when a arrived, when ß arrived, etc.),
and John was happy when one of these
arrived'. LF John (only) Jill arrived1 ?x
(vP was happy at (time point) x1). John
(only) Jill1 ?x (vP was happy at (time point)
x1 arrived).
27
Parallelism in contrastive fragments (3)
The island sensitivity of contrastive focus can
be evidenced in Hungarian when an island
contains contrastive focus, the entire island
must be pied-piped to the preverbal FocP (46)
János was (only) happy when JULI
arrived. overt focus movement of JULI bad
result JULI1 örült János isl amikor t1
megjött. Juli was.happy János
when arrived island in-situ (the English
pattern) bad result Örült János isl amikor
JULI jött meg. was.happy
János when Juli arrived PV island
pied-piped to focus position fine result
örült János was.happy János
isl amikor JULI jött meg
when Juli arrived PV
28
Parallelism in contrastive fragments (4)
Taking stock so far we have established two
things (1) Felicity condition on contrastive
fragments Contrastive fragments are only
felicitous if their correlate is contrastively
focused. (2) Contrastive focus is island
sensitive. When contrastive focus finds itself
in an island, the entire island is interpreted
contrastively. ? If a contrastive fragment has
a correlate inside an island, the whole island is
interpreted contrastively.
28
29
Parallelism in contrastive fragments (5)
(47) A isl Amikor JULI jött
meg1 örült János t1 ? when Juli
arrived PV was.happy János Was János happy
when JULI arrived? Long answer B Nem, isl
amikor BEA jött meg1 TP örült János
t1 . no when Bea arrived PV
was.happy János No, when BEA
arrived. Short answer B Nem, BEA.
no Bea No, Bea.
29
30
Parallelism in contrastive fragments (6)
Long answer (47) A isl Amikor JULI jött
meg1 örült János t1? when Juli
arrived PV happy.was János Was János happy
when JULI arrived? B Nem, isl amikor BEA
jött meg1 TP örült János t1 . no
when Bea arrived PV was.happy
János No, when BEA arrived. ? Overt
movement of the island in As question is
parallel to the overt movement of the island in
Bs answer.
31
Parallelism in contrastive fragments (7)
Short answer (47) A isl Amikor JULI jött
meg örült János t1 ? when Juli
arrived PV happy.was János Was János happy
when JULI arrived? B Nem, BEA1 isl
örült János amikor t1 megjött
no Bea was.happy János when
arrived No, BEA. ? Overt movement of the
island in As question is not parallel to the
island obviating short answer in Bs
answer. These data show that obligatory
parallelism with the correlate requires overt
movement of the island in the answer.
32
Parallelism in contrastive fragments (8)
? Contrastive remnants also observe scopal
parallelism. The variable in the antecedent and
the elided clause are bound from parallel
position. (i) Overt focus-movement languages
(Hungarian) parallelism holds at PF/LF. (ii)
Covert focus-movement languages (English)
parallelism holds at LF. English Long answer
pattern (48) A Was John happy when JILL
arrived? B No, when BEATA arrived. A when
JILL arrived1 ?x (vP John was happy at
(time point) x1)? B when BEATA arrived1 ?x
(vP John was happy at (time point) x1)? ?
in this scenario, LF parallelism is satisfied
33
Parallelism in contrastive fragments (9)
Short answer pattern (49) A Was John happy
when JILL arrived? B No, BEATA. A when
JILL arrived1?x (vP John was happy at (time
point) x1)? B BEATA1 ?x (vP John
was happy at (time point) x1 arrived ). ? in
this scenario, LF parallelism not satisfied
34
Parallelism in contrastive fragments (10)
Conclusion Island sensitivity in contrastive
fragments is solely determined by scopal
parallelism. Short answers Parallelism
violated ? island repair does not obtain (50)
A island node correlate1 t1
(LF) B fragment1 island node
t1 Long answers Parallelism
satisfied (51) A island node correlate1
t1 (LF) B island node
fragment1 t1
35
Consequences for the PF-theory of islands

• The concept of PF-interpretable traces is no
  longer required to explain strong
  island-sensitivity in fragments
• A simpler description of strong islands
• Chomsky (1973) Subjacency
• XP1 BN YP BN t1 .
• NB
• (53) is a PF constraint it can be obviated if
  the BNs are rendered unpronounced.
• LF constraints (such as parallelism) must also
  be obeyed to obviate strong islands.

35
36
Summary of findings
Clausal ellipsis (TP-deletion) can leave behind
two types of remnants contrastive and
non-contrastive ones. Both types obey the
parallelism requirement on ellipsis. Contrastive
remnants can never give rise to island obviation
because - their correlate is contrastively
focused - when the correlate finds itself in an
island, the island is contrastively focused
as well - parallelism with the correlate forces
that the entire island is spelled out in the
remnant, obligatorily giving rise to long
answers/remnants
36
37
Results

• Due to parallelism, contrastive remnants do not
  obviate islands.
• ? We do not need a syntactic theory to account
  for lack of island obviation with contrastive
  remnants.
• The syntactic theory of island obviation can be
  hugely simplified.
• (ii) Sluicing and fragment (answers) are the same
  phenomena TP ellipsis.
• ? We move away from a construction-specific
  study of ellipsis.
• (iii) Contrastivity plays a role in determining
  properties of ellipsis. (see also results of
  Winkler 2005)

37
38
References
Bródy, M. 1995. Focus and Checking Theory. In I.
Kenesei (ed), Approaches to Hungarian 5, Szeged
JATE Press, 29-44. Chung. S, W. A. Ladusaw and
J. McCloskey. 1995. Sluicing and logical form.
Natural Language Semantics 3 239-282. Drübig,
H.B. 1994. Islands Constraints and the Nature of
Focus and Association with Focus, Technical
Report, Arbeitspapiere des Sonderforschunsbereich
s, 340, IMS Stuttgart. É. Kiss, K. 1987.
Configurationality in Hungarian. Budapest
Akadémiai Kiadó. Fox, D. 2000. Economy and
semantic interpretation. Cambridge, MA The MIT
Press. Fox, D. and H. Lasnik. 2003.
Successive-cyclic movement and island repair the
difference between sluicing and VP-ellipsis.
Linguistic Inquiry 34 143-154 Horvath, J. 1986.
Focus in the Theory of Grammar and the Structure
of Hungarian. Dordrecht Foris. Krifka, M. 1996.
Frameworks for the representation of focus. In
G.-Jan M. Kruijff, R. T. Oehrle, and G. Morrill
(eds), ESSLLI 96 Formal grammar. Universitat
des Saarlandes, Department of Computational
Linguistics and Phonetics. Lasnik, H. 2001. When
can you save a structure by destroying it? In
Kim, M. and Strauss, U. (eds.) Proceedings of
the North East Linguistic Society 31. Amherst,
MA GLSA. 301-320. Merchant, J. 2001. The
syntax of silence. Oxford studies in theoretical
linguistics 1. Oxford OUP. Merchant, J. 2004.
Fragments and ellipsis. Linguistics and
Philosophy 27, 661-738. Merchant, J. 2008.
Variable island repair under ellipsis. In K.
Johnson (ed.), Topics in ellipsis. Cambridge
CUP. Rooth, M. 1997. Focus. In S. Lappin (ed)
Handbook of contemporary semantic theory. Oxford,
Blackwell 271 297. Ross, J.R. 1969. Guess who?
Chicago Linguistics Society 5. Chicago, Illinois.
252-286. Valmala, V. 2007. The syntax of little
things. Paper presented at the 17th Colloquium on
Generative Grammar. Girona, June 2007. Winkler,
S. 2005. Ellipsis and focus. Mouton de Gruyter.
39
Appendix fronting in clefts and copula clauses
(1)
A problem the reparative effect of ellipsis on
LF violations? (54) A Who does every
syntactician admire? B Nobody.
B Nobody1 does every syntactician admires
t1. (Valmala 2007) (55) Every syntactician
admires nobody. (Valmala 2007)
Solution (54B) derived from a copula
clause. (56) Nobody1 there is t1 that every
syntactician admires. (57) There is nobody that
every syntactician admires. (i.e. there
is not anybody that every syntactician admires)
39
40
Appendix fronting in clefts and copula clauses
(2)

• (56) A Which lawyeri did even hisi clients
  hate?
• B Bob Andersoni.
• B Bob Andersoni even hisi clients
  hated t1. (WCO)
• B Even hisi clients hated Bob
  Andersoni. (Valmala 200711)
• Solution (56B) is derived from a copula clause.
• Bob Anderson1 its t1 that even hisi clients
  hate.
• Its Bob Andersoni that even hisi clients hate.

40