UMass Amherst
UMass Amherst | Library | Umail | Spire | People Finder 

Search

Match case Regex

Current issue

Categories

WHISC Back Issues

Other Ling Newsletters

Subscribe to this blog's feed
[What is this?]
Creative Commons License
This weblog is licensed under a Creative Commons License.

« Craige Roberts Visit | Main | Syntax Reading Group»

Rene Kager Talk

René Kager
University of Utrecht

Phonological constraints in speech processing

Tuesday, March 27, 4:00 pm, Dickinson 212

Abstract

René Kager and Keren Shatzman

A growing body of literature addresses the notion of gradient well-formedness (GWF). Gradience is the phenomenon that native intuitions about the well- formedness of forms fall on a scale with multiple values, rather than into a binary division of well-formed versus ill-formed. Perhaps the best studied cases of GWF involve OCP-PLACE, the constraint against adjacent consonants with identical place of articulation. Gradient OCP-PLACE effects occur in languages such as Arabic (Greenberg 1950; McCarthy 1986; Frisch, Pierrehumbert & Broe 2004), Hebrew (Berent & Shimron 1997, 2003), English (Berkley 2000), Japanese (Kawahara, Odo & Suno 2005), and Muna (Coetzee & Pater 2006). Although gradient judgments have been shown to correlate with lexical factors such as neighbourhood density (Ohala & Ohala 1986; Bailey & Hahn 2001) and with low-level phonotactic factors such as onset and rhyme frequency (Coleman & Pierrehumbert 1997; Bailey & Hahn 2001), many studies assume an additional involvement of grammatical constraints. Yet, constraints are often motivated on the basis of lexical evidence alone (Frisch, Pierrehumbert & Broe 2004; Coetzee & Pater 2006). That is, lexical under-representation of forms that violate a constraint is interpreted as direct evidence for this constraint. This raises the issue to what extent GWF constraints have grammatical significance beyond the lexical data. Several studies support the mental reality of GWF constraints by means of native well-formedness judgments on non-words (Berent & Shimron 1997, 2003; Frisch & Zawaydeh 2001; Coetzee to appear). Still, eliciting judgments involves a meta-linguistic task, which runs the risk of allowing semi-conscious factors to influence judgments which are absent in speech processing. This problem is aggravated when no time limits are imposed on subjects' responses. Hence, stronger evidence for the mental reality of GWF constraints would require an online task, eliciting subjects' responses to non-words under more realistic conditions of speech processing and establishing correlations between response latencies and the degree of well-formedness of non-words, as predicted by phonological constraints. Here we address the role of GWF constraints in speech processing in Dutch, using the online task of lexical decision. In this task, subjects classify stimuli as either words or non-words, and their response latencies are measured. The task is known to be strongly influenced by lexical processing factors. Hence, if an independent effect of constraints were to emerge, this would constitute strong evidence in favour of the psychological reality of constraints. We will present lexical statistics on avoidance of identical place in Dutch, which reveals that similarity avoidance is influenced not only by degree of similarity, and distance, but also by initial position in the stem. We then nominate three candidate constraints against labial co-occurrence: classical OCP-LAB, self-conjoined *LAB2, and alignment ALIGN-LAB. Our experimental study shows that phonological constraints indeed influence speech processing in a way that is independent of lexical factors. We will compare the relative merits of the three constraints for speech processing.

March 15, 2007 12:11 AM | 05:07 (2007-03-15) , visiting scholars