An investigation of locus equations as a source of relational invariance for stop place categorization

Abstract

Locus equations were investigated as a potential higher-order metric capable of illustrating relational invariance for place of articulation in voiced initial stop consonants independently of vowel context. Locus equations are straight-line regression fits to data points formed by plotting onsets of F2 transitions along the y axis and their corresponding midvowel nuclei along the x axis. Twenty subjects, 10 male and 10 female, produced /b/v/t/, /d/v/t/, and /g/v/t/ tokens for 10 vowel contexts. Each CVC token was repeated in a carrier phrase five times yielding 150 tokens per subject. Formant measures were obtained using the MacSpeech Lab II speech analysis system. Locus equation scatter-plots revealed extremely tight clustering of points around the regression line that were consistent across speakers and gender. Derived slope and y-intercept parameters were significantly different across stop place categories. The relative value of F2onset as it linearly changes in relation to the coarticulatorily produced vowel reflects an acoustic correlate of relational invariance for stop place. A discriminant analysis using F2onset & vowel as predictors showed 82%, 78%, and 67% classification rates for labial, alveolar, and velar place. Using derived slope and y-intercept values as predictors led to 100% classification into stop place categories. A neurobiologically oriented perspective on the invariance issue is explored and a brain-based recognition algorithm for stop place integrating burst and F2 cues is offered. [Work supported by NSF.]

Putting phonetic context effects into context: A commentary on Fowler (2006)

“…It is almost certainly the case that there are regularities in acoustics that result from coarticulation (see, e.g., Sussman, McCaffrey, & Matthews, 1991) and that listeners are sensitive to these regularities in speech perception (Holt, Lotto, & Kluender, 1998;Nearey, 1997;Smits, 2001). Nevertheless, spectral contrast may play a supporting role even in these cases.…”

Non-isomorphism in efficient coding of complex sound properties

“…To the extent that patterns of covariance among acoustic attributes in natural sounds are efficiently coded, these non-isomorphic representations may inform how the auditory system exploits different patterns of redundancy to learn to distinguish different speech sounds. For example, extraction of relational properties across variations consequent to coarticulation (e.g., locus equations, Sussman et al, 1991Sussman et al, , 1998 or anatomy (scaling of formant frequencies across changes in vocal tract length across talkers, Kluender et al, 2011) are the most direct speech analogs to non-isomorphism demonstrated here. In related studies employing fMRI, Okada et al (2010) report that responses in bilateral posterior superior temporal sulcus were sensitive to phonemic variability (intelligibility) of speech sounds, but not to acoustic variability.…”

“…While the novel sounds employed here varied only along three acoustic dimensions (one of which varied randomly), patterns of covariance naturally scale to high-dimensional feature spaces. In complex natural stimuli such as speech, multiple forms of stimulus attribute redundancy exist concurrently and successively (e.g., Delattre et al, 1955;Kluender et al, 2011;Lisker, 1978;Repp, 1982;Sussman et al, 1991Sussman et al, , 1998. To the extent that patterns of covariance among acoustic attributes in natural sounds are efficiently coded, these non-isomorphic representations may inform how the auditory system exploits different patterns of redundancy to learn to distinguish different speech sounds.…”

Cochlea-scaled entropy, not consonants, vowels, or time, best predicts speech intelligibility

“…Production of every consonant and vowel overlaps in time (and in the vocal tract) with speech sounds that precede and follow. In addition, acoustic characteristics within vowels provide rich evidence concerning neighboring consonants and vice versa (8)(9)(10)(11).…”