Effects of pitch accents in attachment ambiguity resolution

Abstract

Previous work has found that listeners prefer to attach ambiguous syntactic constituents to nouns produced with a pitch accent (Schafer, Carter, Clifton, & Frazier, 1996). This study examines what factors underlie previously established accent attachment effects by testing whether these effects are driven by a preference to attach syntactic constituents to new or important information (the Syntax Hypothesis) or whether there is a bias to respond to postsentence probe questions with an accented word (the Salience Hypothesis). One of the predictions of the Salience Hypothesis is that selection of accented words should be greater when a sentence is complex and processing resources are limited. The results from the experiments presented here show that the probability of listeners’ selecting accented words when asked about the interpretation of a relative clause varies with sentence type: listeners selected accented words more frequently in long sentences than in short sentences, consistent with the predictions of the Salience Hypothesis. Furthermore, Experiment 4 demonstrates that listeners are more likely to respond to postsentence questions with accented words than with nonaccented words, even when no ambiguity is present, and even when the response results in an incorrect answer. These findings suggest that accent-driven attachment effects found in earlier studies reflect a postsentence selection process rather than a syntactic processing mechanism.

Keywords:

• Pitch accents
• Prosody
• Relative clauses
• Syntactic processing

Acknowledgements

We would like to thank Keturah Bixby and Susan Smith for their assistance with recording stimuli and running participants. This project was supported by Grant Number R01DC008774 from the National Institute on Deafness and Other Communication Disorders. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institute on Deafness and Other Communication Disorders or the National Institutes of Health.

Notes

¹There are some exceptions to these claims. Boundaries may provide time for processing semantic information (Blodgett, 2004; Schafer, 1997) and tonal movement at edges of intonational phrases may signal information about discourse structure (Pierrehumbert & Hirschberg, 1990).

²Schafer et al. (1996) found that focus attraction effects were greater when critical items were focused with a contrastive accent than with a new information accent. Thus, we used an L + H* pitch accent to mark focused elements in this and the following experiments.

³The reported fixed effects are based on a model that includes the random by-participant slopes for accent and the random intercepts for both participants and items. The inclusion of random by-participant slopes for RC type was justified in a likelihood ratio test (p<.05) [In this model, the interaction between accent and RC type was marginally significant (p<.07)]. However, it yielded a perfect negative correlation between the random intercepts and random slopes for RC type, indicating that the model overfit the data. Thus, we report the model that does not include the random slopes for RC Type.

⁴We would like to thank an anonymous reviewer for pointing this out.

⁵The percentage of other responses closely matches that from Schafer et al. (1996)'s Experiment 1 (9%), which also used a recall task.