Skip to main content
Open Access Publications from the University of California


UCLA Electronic Theses and Dissertations bannerUCLA

Cue Integration and Contrast Shifts: Experimental and Typological Studies

  • Author(s): Yang, Meng;
  • Advisor(s): Sundara, Megha;
  • Keating, Patricia A
  • et al.

Auditory Enhancement has been put forth as an explanation for why certain acoustic phonetic cues co-vary to signal phonological contrasts more often than others. Under this account, listeners more readily associate two cues if they produce the same auditory effect, making the cues perceptually inseparable. Traditionally, evidence for enhancement has come from studies showing perceptual integration between enhancing cues, but even cues that do not share the same auditory effect have been shown to perceptually integrate. Further, language experience with co-variation between cues is often a confound in these studies.

In this dissertation, I present new evidence in favour of auditory enhancement from four experiments and one typological study.

In the first set of experiments, I use a modified cue weighting paradigm that mimics diachronic contrast shifts. Listeners categorizing synthesized speech stimuli were forced to shift their attention between a pair of acoustic cues based on how informative each cue was to the contrast. This was done for a pair of enhancing cues, pitch and breathiness, and a pair of non-enhancing cues, pitch and vowel duration, both of which have been shown to perceptually integrate. For each pair of cues, I tested two groups of listeners – English listeners, who had no phonemic experience with either cue pair, and Hani (Tibeto-Burman) listeners who had experience with both pairs of cues co-varying in the same contrast. The extent to which listeners were able to shift attention between non-enhancing cues was predicted to reflect their language experience. For enhancing cues, attentional shift was predicted to also be conditioned by whether the cues were in an enhancing relationship. These predictions were borne out, but there was an unpredicted finding that shifting between the enhancing cues was asymmetric.

This asymmetry was further explored in two experiments. The first of these investigated whether the asymmetry could be caused by both listener groups having more linguistic experience with pitch than with breathiness. Two additional groups of listeners were thus tested using the same paradigm: Tone listeners, who used pitch phonemically, and Phonation listeners, who used breathiness phonemically. Both of these groups also exhibited the same asymmetry, showing that the phenomenon is language-general.

In the final experiment, I tested the hypothesis that the asymmetry in attentional shift was caused by an asymmetric perceptual dependency between pitch and breathiness. Listeners categorized stimuli for which one cue was informative but the other was completely neutralized. The amount of attention listeners paid to the uninformative cue was predicted to differ if the percept of one cue was dependent on the other but not vice versa. Results from this experiment provided weak evidence in favour of the hypothesis.

Finally, I conducted a cross-linguistic typological survey of the synchronic co-variation and diachronic contrast transfer between the cue pairs I tested experimentally. While the cues in both pairs co-vary synchronically, only the enhancing cues participate in contrast transfer. Furthermore, the transfer of phonological contrast between the enhancing cues occurs overwhelmingly in the direction that matches the asymmetry in attentional shift observed in the lab.

The experimental and typological studies in this dissertation provide support for Auditory Enhancement, demonstrating that cues that converge on the same auditory effect are treated differently by listeners compared to cues that do not. Based on the results, I argue that i) auditory enhancement and perceptual integration should remain separate notions, and ii) perceptual associations that are not learned through experience may be asymmetric, but learned associations are necessarily symmetric.

Main Content
For improved accessibility of PDF content, download the file to your device.
Current View