Variability in vP-Subject Island Violations
Roumyana Pancheva and Barbara Tomaszewicz, University of Southern California
We provide new empirical evidence in support of three generalizations: (i) subjects are not categorical
islands ([8,9,10,15]); (ii) the islandhood of subjects is not due to a freezing effect, (contra [15] and in line
with [2,5]); and (iii) ellipsis has variable effect on subject island violations ([12]). Our findings come
from two domains: degree questions and comparatives. Both involve wh-movement of non-referential
elements, which cannot escape even weak islands ([3,11]) and which are not subject to constraints on
discourse referents posited by processing theories of islands ([9]).
1. SUBJECT ISLANDS. Contrasts as in (1) have been traditionally captured by constraints that
categorically prohibit sub-extraction from subjects ([7]). [15], however, argues that only subjects that
have moved to Spec, TP are islands and that extraction from vP-subjects is allowed and observed in
languages which can leave subjects in Spec, vP, such as the Slavic languages. In contrast, [2] and [5]
propose that vP-subjects are also islands. (See also [4]). [8] and [10] offer experimental evidence from
German and Russian, respectively, that different types of subjects behave differently wrt sub-extraction;
[8] further shows that there is substantial speaker variability in acceptability of sub-extractions.
2. POLISH DEGREE CONSTRUCTIONS. Polish phrasal comparatives (PCs) (2a) exhibit island
effects when the more-NP is a transitive subject ([13]). [13]’s explanation, which we adopt, is that od
‘from’/‘than’ has a non-overt small clause complement, whose subject it ECMs. There is wh-movement in
the than clause from a position parallel to that of the matrix more, as in clausal comparatives (CCs) (2b),
but in the absence of a wh-probe, the wh-operator moves to the edge of the vP only. The movement
creates a degree predicate, as in [6]. Movement of the whole subject out of Spec, vP targeting vP, as in
(3a) is precluded as too-local ([1]). The alternative in (3b) involves sub-extraction of the degree wh-word
from the subject, hence the island effects. The CC (2b) is grammatical, as no sub-extraction is required,
given that movement of the wh-subject to Spec, CP is not precluded as too-local (3c).
3. SUB-EXTRACTION FROM SUBJECTS IN POLISH. We tested [13]’s account in 3 acceptabilityrating studies in Polish, comparing PCs and CCs with more-NP transitive subjects to PCs and CCs with
more-NP objects (Exp.1,2), adverbs (Exp.2), and degree questions with or without sub-extraction from
subjects (Exp.3) (see 4) on a 1-7 scale. In Exp.1&2 repeated measures ANOVAs yield significant main
effects of type of than (od vs. ni ) and position of more (subject vs. object (vs. adverb)), and, most
importantly, significant interactions (5a,b). This suggests that (4a)’s lowest mean is not just a cumulative
effect of the two main factors but an additional effect, which we attribute to the island violation.
Underscoring this point, the main effects remain significant when the subject conditions are not included
in an ANOVA but there is no interaction (Exp.2: F(1,25)=0.77, p=0.39); i.e., the lower mean of (4e)
relative to (4c,d,f) is entirely cumulative. In Exp.3 repeated measures ANOVA also yields significant
main effects of type (comparative vs. question) and of type of wh-movement (sub-extraction from subject
vs. movement of the whole subject) (F(1,55)=110.79, p <.0001), as well as an interaction (5c). The results
of Exp.3 confirm the analysis of (2a) in terms of sub-extraction. Overall, several aspects of our findings
are notable. Sub-extraction from vP-subjects (PCs with od) is significantly degraded, relative to whmovement of the whole subject (CCs with ni ), suggesting that vP-subjects are islands in support of [2,5]
and contrary to [15]. There is a significant variability among speakers in rating violations of vP-subject
islands (PCs with od), with individual mean averages ranging 1.17-7 (Exp.1), 1.5-5.75 (Exp.2), and 1-7
(Exp.3), as in [8]. Similar variability is observed with overt sub-extraction in questions, with individual
means in the range of 1.17-7 (Exp.3). In contrast, the ungrammatical fillers are rated uniformly low: e.g.,
mean 1.17, range 1-2 (Exp.3) and similarly for Exp. 1&2. Clearly, there are Polish speakers for whom vPsubjects (and possibly TP subjects as well, given the question data) are not strong islands.
4. THE VARIABLE ROLE OF ELLIPSIS. A fourth acceptability-rating study revealed that NPdeletion ameliorates the violation of the vP-subject island: (6a) received a mean rating of 5.33 (on a 1-7
scale) comparable to the mean of 5.52 for (6b). We analyze (6a) as involving a left-branch movement of
the adjective to a higher projection in the small clause, followed by clausal ellipsis. The results are in line
with the observation that whereas clausal (TP)-ellipsis ameliorates islands more generally, vP-ellipsis
doesn’t [12].
(1)
a. * Who1 did [a story about t1] cause a sensation?
(2)
a.
b.
b.
Who1 did you read [a story about t1]?
??/
*Wi cej uczniów zwiedziło Czechy od Słowacji.
more students visited
Czech R. from Slovakia-GEN
Wi cej uczniów zwiedziło Czechy ni
Słowacj .
more students visited
Czech R. than Slovakia-ACC
‘More students visited the Czech Republic than Slovakia.’
(3)
a. od [PredP Slovakia3 [vP wh-many students2 [vP x2 visit x3 ]]]
b. od [PredP Slovakia3 [vP wh2 [vP d2-many students visit x3]]]
c. ni [CP wh-many students2 [TP Slovakia3 [TP [vP x2 visit x3 ]]]]
(4)
a. Tego wieczoru
b. Tego wieczoru
This evening
c. Zespół Impresja
d. Zespół Impresja
group Impresia
par
zata czyło tango
par
zata czyło tango
couples danced
tango
pary
zata czyły tango
pary
zata czyły tango
couples danced
tango
g. Ile
how-many
h. Ile
how-many
tego wieczoru
par
this evening
couples
par
tego wieczoru
couples this evening
a. Exp1
b. Exp2
c. Exp3
Subj
od
(4a)
4.38
3.93
4.07
a. W Polsce
b. W Polsce
in Poland
Subj
ni
(4b)
5.48
5.53
5.67
ci gle
ci gle
still
(violation of Anti-locality)
(vP-subject island violation)
(no sub-extraction needed)
od
poloneza.
ni
poloneza.
than polonaise
zata czył wi cejla tynoskich ta ców
zata czył wi cej latynoskich ta ców
danced
more
Latin
dances
e. Wszystkie
f. Wszystkie
all
(5)
(6)
wi cej
wi cej
more
Polish, Pancheva (2009)
Obj
od
(4c)
5.18
5.38
na
Obj
ni
(4d)
5.78
6.34
na
wi cej
wi cej
more
Adv
od
(4e)
na
5.09
na
ludzi
ludzi
people
lepiej
lepiej
better
od
zespołu T cza.
ni
zespół T cza
than group
Techa
od
poloneza.
ni
poloneza.
than polonaise
zata czyło
danced
zata czyło
danced
Adv
ni
(4f)
na
5.73
na
Polish
poloneza?
polonaise
poloneza?
polonaise
Subextr.
Q (4g)
na
na
4.99
pije czarn
pije czarn
drink black
Subj
Q
(4f)
na
na
5.92
herbat
herbat
tea
interactions b/n main
effects
F(1,34)=6.26, p=0.017
F(2,25)=3.99, p=0.025
F(1,55)=7.08, p=0.010
od
zielonej.
ni
zielon .
than green
[1] Chomsky, N. (1995). “Bare Phrase Structure”, in G. Webelhuth (ed.) Government and Binding Theory and the
Minimalist Program. Blackwell. 383–439. [2] Chomsky, N. (2008). “On Phases”. In Foundational Issues in Linguistic
Theory: Essays in Honor of Jean-Roger Vergnaud, ed. by R. Freidin, C. P. Otero, M. L. Zubizarreta, 133-166. MIT Press.
[3] Cinque, G. (1990). Types of A’-Dependencies. MIT Press. [4] Corver, N. (2006). “Subextraction” In M. Everaert and
H. van Riemsdijk (eds.) The Blackwell Companion to Syntax, v. IV. Blackwell. [5] Gallego, Á. & J. Uriagereka (2007).
“Sub-extraction from Subjects: A Phase Theory Account”, In J. Camacho, N. Flores-Ferrán, L. Sánchez, V. Déprez & M.
J. Cabrera (eds.) Romance Linguistics 2006. John Benjamins, 149-162. [6] Heim, I. & A. Kratzer (1998). Semantics in
Generative Grammar. Blackwall. [7] Huang, J. (1982). “Logical Relations in Chinese and the Theory of Grammar.”
Ph.D. thesis, MIT. [8] Jurka, J. (2009). “Gradient Acceptability and Subject Islands in German” Ms. University of
Maryland. [9] Kluender, R. (2004). “Are Subject Islands Subject to a Processing Account” In B. Schmeiser, V. Chand, A.
Kelleher, A. Rodriguez (eds). WCCFL 23, 101-125. [10] Kravtchenko, E., M. Polinsky, & M. Xiang (2009) “Are All
Subject Islands Created Equal?” poster at CUNY 2009, UC Davis. [11] Kroch, A. (1989). “Amount Quantification,
Referentiality and Long Wh-Movement” Ms. Upenn. [12] Merchant, J. (2008). “Variable island repair under ellipsis” In
K. Johnson (ed.) Topics in Ellipsis, CUP, 132-153. [13] Pancheva, R. (2009). “More Students Attended FASL Than
CONSOLE” FASL 18. [14] Ross, J.R. (1967). “Constraints on Variables in Syntax” MIT thesis. [15] Stepanov, A (2007).
“The End of CED? Minimalism and Extraction Domains” Syntax 10 (1), 80-126.