Pynte & Colonna (2000) showed that in silent reading of French complex noun phrases (N1 de N2), long relative clauses (RCs) preferentially attach high (to N1), and short RCs low (to N2). A similar length effect occurs in English, where long RCs tend to attach low and short RCs even lower (Fernandez, 2000). The implicit prosody hypothesis (IPH; Fodor, 1998) proposed that length effects arise because readers project default prosody onto sentences, as determined by language-specific phonological principles which are sensitive to both syntactic structure and constituent length (Selkirk, 2000).
Quinn et al. (2000) reported evidence that default prosody differs between long and short French RCs. The present study, along similar lines, was designed to put the IPH to a more substantial test.
Eight of Pynte & Colonna's experimental sentences, in 4 versions each (long/short, high/low attachment disambiguated by number agreement), were read aloud by 10 native speakers of Canadian French. (A standard ambiguity resolution questionnaire, with different subjects, established that silent reading attachment preferences for Canadian French resemble those noted above for European French.) Translation-equivalent English materials were read by 10 native speakers of American English. Measurements of fundamental frequency (F0) and constituent-final lengthening (including pauses) were taken for N1 and N2. Lengthening and/or F0 re-set are characteristic acoustic properties associated with prosodic breaks. Maynell (1999) showed for English that a pre-RC break favors high RC-attachment.
For short RCs, results were similar for French/English, and forced-high/-low attachment. F0 dropped between N2 and RC, and there was no pre-RC lengthening, suggesting that the default prosody includes no break before short RCs in either language.
For long RCs, only English forced-low attachment exhibited no break. In the other three cases (English forced-high, French forced-high and -low), there was significant pre-RC lengthening and no F0 fall between N2 and RC. This is indicative of a prosodic boundary before long RCs, which in French is indifferent to the height of the RC in the syntactic structure.
For ambiguous RC-attachment in silent reading, the IPH predicts correctly that low attachment is preferred for short RCs in both languages. For long RCs, English prosodic principles are flexible, letting prosodic phrasing mirror the syntactic structure; but French apparently has a strong default, requiring a break between N2 and a long RC regardless of the attachment intended. The IPH therefore predicts that French readers project a break before long RCs, favoring the higher site as Pynte & Colonna observed.
This data pattern strongly supports the claim that prosodic phrasing is the cause of ambiguity resolution preferences, not vice versa.
Examples (material in parentheses was in LONG-RC version only)
References
Fernandez, E. (2000) Bilingual Sentence Processing: Relative Clause Attachment in English and Spanish. Unpublished Ph.D. dissertation, City University of New York.
Fodor, J. D. (1998) Learning to parse? Journal of Psycholinguistic Research 27.2, 285-319.
Maynell, L. (1999) Effect of pitch accent placement on resolving relative clause ambiguity in English. Poster presented at the 12th Annual CUNY Conference on Human Sentence Processing.
Pynte, J. and S. Colonna (2000) Decoupling syntactic parsing from visual inspection: the case of relative clause attachment in French. In A. Kennedy,
R. Radach, D. Heller and J. Pynte (eds.) Reading as a Perceptual Process. Oxford: Elsevier.
Quinn, D., H. Abdelghany and J. D. Fodor (2000) More evidence of implicit prosody in silent reading: French, English and Arabic relative clauses. Poster presented at the 13th Annual CUNY Conference on Human Sentence Processing.
Selkirk, E. (2000) The interaction of constraints on prosodic phrasing. In M. Horne (ed.) Prosody: Theory and Experiment. Amsterdam: Kluwer.