Use of pitch accents and pitch range in processing and production

Abstract:
Author: Affiliation:

Perceivers are more likely to report a gapping reading of (1) when the subject is accented, and thus parallel to the remnant (Sharon), than when the object is accented (Carlson, 2001).

However, the critical pitch accents on the first clause subject or object were realized in the same pitch range as the remnant.  Thus the parallelism of either the accents or the pitch range (or both) might explain the result.

To tease apart the effects of accent vs. pitch range, ambiguous replacive sentences (as in (2)) were examined in an auditory questionnaire.  The pitch range of the accent on the remnant (Dr. Green) was varied, along with the placement of accents in the first clause:

With the Subject Accent prosody and matching pitch range, subject interpretations of the remnant reached 85%, vs. 80% with the non-matching prosody; with the matching Object Accent prosody, subject interpretations were only 47%, vs. 57% with the non-matching prosody.  Clearly, the accent placement had the strongest effect (p's < .001).  However, the conditions in which the pitch accent on the remnant was in the same pitch range as an earlier accent were also significantly more effective than the non-matching conditions in producing interpretations with the intended contrast (significant interaction, p's <.05).

A production study including conjoined and gapping sentences (as in (1)) in disambiguating contexts found that naive speakers used pitch range (as well as pitch accent placement and type) to vary the relative prominence of first-clause arguments.  (3) shows the average peak F0 in Hz for the first-clause subject, object, and the post-conjunction DP:

In the subject context, the first-clause subject contrasted with the post-conjunction DP, and thus the subject was produced in a higher range than it was in the object context (t(9) = 4.01, p=.003).  In the object context, the object DP contrasted with the later DP, so the object was realized in a higher range than in the subject context, (t(9) = 9.14, p<.001).

These studies support a view of prosody in which gradient features such as pitch range can optionally be used by speakers along with pitch accent type and location to mark intended contrasts.  Similarly, perceivers can use pitch range information along with pitch accent type and placement to help determine an interpretation during sentence processing, though they are not required to do so.