Ana Gouvea, Colin Phillips & David Poeppel
University of Maryland, College Park
Subject center-embedded relatives (1) are easier to process than object center-embedded relatives (2) (Caplan & Waters, 1999; Gibson, 1998; King & Just, 1991). When comparing center-embedded (1,2) and right branching relatives (3,4), right branching relatives are harder to process than center-embedded in English (Hakes et al., 1976; Holmes, 1973). Based on cross-linguistic experimental and syntactic data from English and Brazilian Portuguese (BP), we argue that a particular syntactic property, the possibility of extraposition, conditions the manner in which relative clauses are processed.
In a previous RSVP experiment (200ms/word; n=16 BP and n=36 English), we examined subject and object center-embedded (1,2) and right-branching (3,4) relatives. Both languages showed the familiar subject/object contrast. However, a cross-linguistic difference was found: in BP but not in English, center-embedded were more difficult than right-branching relatives. Similarly, in a second experiment using eye-tracking and similar materials in English, the subject/object contrast was found (p=.0002), but the relative clause region in right branching relatives received longer first pass reading times than center-embedded relatives (p<.02).
To account for the cross-linguistic difference and the difficulty with right branching relatives in English, we examined the syntactic properties of both languages. We hypothesize that the difficulty with right branching relatives in English stems from the fact that English allows extraposition from relative clauses (5). This possibility creates a temporary ambiguity during the processing of right branching relatives in which the relative clause can modify the NP in object or in subject position. BP does not allow this kind of extraposition, so no ambiguity is created, and right branching relatives present no processing difficulty.
To evaluate this hypothesis, we examined different types of extraposition from relative clauses in BP and English in two grammaticality judgment survey experiments. In the first experiment (n=43 BP, n=48 English) several types of extraposition from relative clauses were compared to ambiguous, good and bad controls. English speakers accepted some types of extraposition, Brazilian speakers do not. In the second experiment (n=23 BP, n=27 English), a subset of extraposed relative clauses (6 and 7) were compared to good and bad controls. The results again show that English speakers accept examples of extraposition from relative clauses, Brazilian speakers do not (p<.0001).
These data, as well as converging Hungarian facts about relative clauses and extraposition, suggest that processing theories using working memory metrics must be sensitive to the syntactic properties of the languages while integrating information into structure.
Examples
(1) The man that _ is pinching the woman is talking to the child again. (Subject Center-embedded)
(2) The man that the woman is pinching _ is talking to the child again. (Object center-embedded)
(3) The child is talking to the man that _ is pinching the woman again. (Subject Right-branching)
(4) The child is talking to the man that the woman is pinching _ again. (Object Right-branching)
(5) Any girli could break the table easily thati takes karate lessons. (Frazier & Clifton, 1996, p.98)
(6) Did the journalist see the athlete running that hates to swim?
(7) Did the waitress see the baker cooking the bread that does not like cookies?
References
Caplan, D., & Waters, G. (1999). Behavioral and Brain Sciences, 22: 77-126.
Frazier, L., & Clifton, C. (1996). Construal. Cambridge, MIT Press
Gibson, E. (1998). Cognition, 20:22-34
Hakes, D., Evans, J., & Brannon, L.(1976). Memory and Cognition, 4 (3): 283-290
Holmes, V. M. (1973). Journal of Verbal Learning and Verbal Behavior, 12: 285-293
King, J., & Just, M.A. (1991). Journal of Memory and Language, 30: 580-602