Acoustic-perceptual salience and developmental speech perception [pdf]   

    *winner of the Horace H. Rackham Distinguished Dissertation Award

Dissertation co-chairs: Patrice Speeter Beddor (Michigan Linguistics) and Janet F. Werker (UBC Psychology)

Inspired by the notion that some typologically less common contrasts may be perceptually less salient than ubiquitous contrasts, this dissertation investigates the perception, by infants and adults, of a relatively uncommon nasal place contrast (onset /na-ŋa/) against a more common nasal contrast (/ma-na/) in order to assess the role of acoustic-perceptual salience in the development of speech perception.  Do perceptually less salient contrasts show a pattern of development different from the well-known tendency for infants to successfully discriminate native and non-native contrasts in young infancy?  It is argued that phonetic contrasts that are perceptually less salient than others may require language experience to be discriminated in infancy. 

An acoustic analysis (Experiment 1) of onset /m n ŋ/ in Filipino showed that, in the perceptually relevant F2xF3 space, [na] was closer to [ŋa] than to [ma].  When presented with these same stimuli in a discrimination task (Experiment 2), English- and Filipino-speaking adults accurately discriminated [ma]-[na], native to both language groups.  The [na]-[ŋa] contrast, native to Filipino but not English speakers, was discriminated at chance level by the English listeners and was well discriminated by the Filipino listeners, although slightly but significantly less accurately than [ma]-[na].  When Filipino listeners were presented with the same contrasts in two noisy listening conditions (Experiment 7), discrimination of [na]-[ŋa] fell to near chance levels in the noisier condition (-5dB SNR), while accuracy on [ma]-[na] remained above 90% in both noisy conditions.  These findings suggest that the [ma]-[na] contrast is perceptually more salient than [na]-[ŋa] for adult listeners regardless of language experience.

When English-hearing infants, aged 4-12 months, were presented with the two Filipino contrasts (Experiments 3-5), they successfully discriminated the [ma]-[na] contrast but not [na]-[ŋa].  In Experiment 6, Filipino-hearing infants successfully discriminated native [na]-[ŋa] at 10-12 months, but not at 6-8 months.  Taken together, the results suggest that acoustic-perceptual salience affects the discrimination of nasal place contrasts in infancy, with the less salient [na]-[ŋa] contrast being more difficult to discriminate in infancy than more salient [ma]-[na].  Native language experience is required for the infant to perceptually segregate acoustically similar categories.

I've presented these ideas (in parts) at the  ASA (Fall 2004)[Results from a pilot study with Cantonese nasals], and BUCLD 30 [pdf], ASA (Spring 2006-Providence), ICIS (2006-Kyoto)[Coda nasal perception in infancy], LabPhon10 (2006-Paris), and the LSA (2007).

The dissertation results are reported in:

"The acoustic-perceptual salience of nasal place contrasts,"  (see above) Journal of Phonetics. [This paper extends the dissertation results with a third perception study and an additional acoustic analysis that considers the dynamic cues to nasal place.]

"Acoustic salience affects speech perception in infancy: Evidence from nasal place discrimination," (under review, with Janet Werker and Pam Beddor)