Processing Long Distance Dependencies in Spanish
Leticia Pablos & Colin Phillips
University of Maryland, College Park
Abstract
Design and Predictions
This study uses long distance topic-clitic dependencies in Spanish to demonstrate a
pre-verbal dependency formation effect that arises from the prediction of clitic
pronouns.
A controversial issue in research on the real-time construction of long distance
dependencies is the question of whether a fronted phrase receives its argument role
via direct association with a verb or mediated via argument positions. We investigate
whether long distance dependencies are formed before the verb, focusing on overt
elements in argument positions such as clitic pronouns in Spanish. This would
support the use of intermediate argument positions projected by the verb.
Background
There was a manipulation of the clitic pronoun at the matrix verb to control for the effects
expected at the embedded clitic.
Context:
Yendo a la escuela, mi hermana mayor y yo vimos a mis amigas Ana e Irene.
‘Going to school, my sister and I saw my friends Ana and Irene’
a. Non-clitic condition:
Target: Clitic is predicted
Topic _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ las
A estas chicas, mi hermana mayor más tarde
dijo que ya lo cree que las conoce
desde hace tiempo
‘These girls, my old sister later said that she indeed thinks she knows them for a
long time’
Predictions: In (a) condition, clitic iiss predicted and integrated naturally
b. Clitic condition:
Filled Gap Effects in Head-initial Languages.
In studies on head-initial languages, it has been shown that long distance dependencies
are created at the verb and that the parser slows down when there is a phrase filling the
position posited for the gap (Stowe 1986, Clifton & Frazier 1989).
Attempts have been made to explain whether the fronted phrase in these dependencies
receives its argument role via direct association with the verb or through its argument
positions but none of the theories have found decisive evidence for which of the two
approaches is correct, because the slowdown effect is found post-verbally in
head-initial languages.
(Pickering & Barry 1991, Sag & Fodor 1994, Bever & McElree 1988, MacDonald 1989,
Gibson & Hickok 1993).
What do Englishmen cook gap/trace/copy
Indirect Association
Transformational Grammar
(--> Projection Principle)
Direct Association
HPSG/GPSG
Categorial Grammar
Dependency Grammar
etc.
What do Englishmen cook
Pre-verbal Filled Gap Effects
The way the previous theories could be disentangled would be searching for effects of
dependency formation associated with argument positions that precede the verb. This
filled gap effect would not be expected under the direct association theory.
Control: Clitic is not predicted
Topic _ _ _ _ _ _ _ _ les _ _ _ _ _ _ _ _ _ las
A estas chicas, mi hermana mayor más tarde les dijo que ya lo cr ee que
conoce desde hace tiempo
las
‘These girls, my old sister later told them that she indeed thinks she knows them for
a long time’
Predictions: In (b) condition, a longer residual reading time is expected at “las ” if the
parser has compl eted the dependency at the first available clitic and has to check coreference
Procedure and Participants
56 native speakers of Spanish participated in the study, 16 students from the University
of Maryland,College Park and 40 from the University of Deusto. We used the self-paced
reading paradigm.
Materials
Experimental sentences consisted of 24 sets of sentences of the 2 conditions described
above and 72 filler sentences. A context sentence preceded its item and was identical in
both conditions. A yes/no question comprehension task followed each item.
Results
Results: 55 subjects (16 Latin-American & 40 Castilian)
150
F1 (1, 54) = 58.73, p< .01
F1 (1, 54) = 3.62, p= 0.06
F2 (1, 23) = 39.41, p< .01
100
F2 (1, 23) = 4.27, p= 0.05
Matrix verb
Clitic “las”
Clitic Prediction Paradigm
In this study, we focus on clitic pronouns in Spanish and use a clitic
prediction paradigm to search for pre-verbal dependency formation effects.
Reading Time
50
Clitico
0
No-clitico
-50
Why use Topic-clitic Dependencies?
Slower residual r eading times are shown
at the region of the main verb “decir” at
condition (a) and at the clitic “las” in
condition (b).
-100
The same way as English speakers predict a gap after processing a wh-phrase,we predict
that once they have processed a fronted phrase, Spanish speakers could predict an
upcoming clitic and interpret the dependency bef ore encountering the verb that
subcategorizes this fronted phrase.
-150
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
Region
A estas chicas, mi hermana mas tarde (les) dijo que ya lo cree que las conoce desde hace…
1 of 56 subjects was excluded for scoring below 80% accuracy on the comprehension
questions (accuracy of the comprehension questions was high in general average 92%).
English: What
What
do Englismen cook gap ?
gap?
In the non-clitic conditions, there was a large slowdown at the main verb (region 9). In
the clitic conditions, the effect at the most embedded clitic “las” (region 15) was
significantly larger than at the no-clitic condition in (a).
Spanish: A estas chicas, mis amigos las adoran.
Topicalization
clitic verb.
Conclusion
Off-line Tasks
We ran an off-line rating task to verify the predictability of the clitic after processing
a fronted phrase. The test was given after the subjects did the self-paced reading task.
5
Context:
Caminando por la playa, mi madre y yo nos encontramos
a dos señoras mayores.
Grammaticality Rating
4.5
4
a. Non-clitic condition:
A esas señoras, mi madre conoce.
3.5
b. Clitic condition:
A esas señoras, mi madre las conoce.
3
2.5
2
1.5
The results from our experiment show effects of dependency completion at the most
embedded clitic pronouns( las =region 15) in the clitic conditions. This supports the use
of clitics as arguments that help to form a dependency before encountering the
subcategorizing verb. Moreover, this effect shows the parser may not expect a clitic
when it has already completed the dependency at the first available clitic (les =region
8) although it has recently activated the same ref erent.
An effect of clitic absence/presence was found at the main verb that could either mean
the verb is highly predicted in (b) when preceded by “les” or that subjects are surprised
when not encountering a verb with this verb in particular.
The goal of this study was to look for a pre-verbal dependency completion effect in a
head-initial language by taking advantage of clitic pronouns in Spanish. In the case of
topic-clitic dependencies in Spanish, the parser can interpret the dependency at a
position prior to the verb through an argument position (e.g: clitic pronoun) projected
by this verb. Finally, whether the referent for the clitic has been recently accessed
makes a difference in clitic processing. The contrast shown at the clitic “las” for both
conditions reflects the expectancy for a clitic once the fronted phrase has been
processed.
Acknowledgements
1
CLITIC
NO-CLITIC
We would like to thank Jon Franco and Jon Ortiz de Urbina at the University of Deusto
for their help in running the experiment. This work was supported in part by grants from
the Basque Government-Dept. of Education, NSF BCS-0196004 (to Colin Phillips) and
from Human Frontiers Science Program RGY-0134 (to K. Sakai & CP).
References
Bever, T.G. & McElree, B. (1988). ‘Empty categories access their antecedent duri ng c omprehension’. Linguistic Inquiry, 19, 35-43.
Clifton, C. JR & Frazier, L. (1989). ‘Comprehending sentences with long distance dependencies’. In G. Carlson & M. Tanenhaus (Eds.), Linguistic
Structur e in Language Pr oc essing (pp. 273-317). Dordrecht, The Netherlands, Kluwer.
Gibs on, E. and Hickok, G. (1993) ‘Sentence processing with empty categori es’, Language and Cognitive Processes 8, 147- 161.
MacDonald, M. C. (1989) ‘Priming effects from gaps to antecedents’, Language and Cognitive Processes, 4, 35-56.
Pi ckering, M . & Barry, G. (1991). ‘Se ntence processing without empty categori es’. Language and C ognitive Processes, 6, 229-259.
Sa g, I .A., & Fodor, J .D. (1994). Extraction without traces. In R. Aranovich, W. Byrne, S. Preuss, & M. Senturi a (Eds.), Proceedings of the 13th Annual
Meeting of the West Coast Conference on Formal Linguistics (pp. 365-384). Stanford, CA:CSLI.
Stowe, L. A. ( 1986). ‘Plausibility and the processing of unbounded dependencies: an eye-tracking study’. Journal of Memory and Language, 35, 454-475.