The relative roles of vowels and consonants in discriminating talker identity versus word meaning

Abstract

Three experiments tested the hypothesis that vowels play a disproportionate role in hearing talker identity, while consonants are more important in perceiving word meaning. In each study, listeners heard 128 stimuli consisting of two different words. Stimuli were balanced for same/different meaning, same/different talker, and male/female talker. The first word in each was intact, while the second was either intact (Experiment 1), or had vowels ("Consonants-Only") or consonants wels-Only") replaced by silence (Experiments 2, 3). Different listeners performed a same/ different judgment of either talker identity (Talker) or word meaning (Meaning). Baseline testing in Experiment 1 showed above-chance performance in both, with greater accuracy for Meaning. In Experiment 2, Talker identity was more accurately judged from Vowels-Only stimuli, with modestly better overall Meaning performance with Consonants-Only stimuli. However, performance with vowel-initial Vowels-Only stimuli in particular was most accurate of all. Editing Vowels-Only stimuli further in Experiment 3 had no effect on Talker discrimination, while dramatically reducing accuracy in the Meaning condition, including both vowel-initial and consonant-initial Vowels-Only stimuli. Overall, results confirmed a priori predictions, but are largely inconsistent with recent tests of vowels and consonants in sentence comprehension. These discrepancies and possible implications for the evolutionary origins of speech are discussed.

Assessing the perceptual contributions of vowels and consonants to Mandarin sentence intelligibility

“…Bonatti et al (2005) suggested that consonants and vowels were more tied to word identification and grammar, respectively, in continuous speech. In addition, speech materials and language-specific vowel-consonant ratio would also impact the relative contribution of vowels and consonants in speech perception (e.g., Owren and Cardillo, 2006). One of the reasons for the difference between isolated-word (e.g., Owren and Cardillo, 2006) and sentence perception is that the sentence context itself strongly constrains lexical competition.…”

“…In addition, speech materials and language-specific vowel-consonant ratio would also impact the relative contribution of vowels and consonants in speech perception (e.g., Owren and Cardillo, 2006). One of the reasons for the difference between isolated-word (e.g., Owren and Cardillo, 2006) and sentence perception is that the sentence context itself strongly constrains lexical competition. Hence listeners may retrieve or predict sentences' meaning by using a priori knowledge, language experience and contextual cues involved in a top-down processing.…”

“…Vowels are characterized by a relatively open vocal tract with sustained voicing in production and low frequency energy and long duration, whereas consonants are characterized by complete or partial vocal tract constriction in production and high frequency energy and short duration. A number of studies have attempted to investigate the perceptual contributions of vowels and consonants to speech (e.g., word and sentence) intelligibility (e.g., Owens et al, 1968;Cole et al, 1996;Cutler et al, 2000;Bonatti et al, 2005;Owren and Cardillo, 2006;Kewley-Port et al, 2007;Fogerty and Kewley-Port, 2009). These studies have produced evidence of greater contributions by consonants under some conditions (e.g., Owens et al, 1968;Cutler et al, 2000;Bonatti et al, 2005;Owren and Cardillo, 2006) and greater contributions by vowels under other conditions (e.g., Cole et al, 1996;Kewley-Port et al, 2007;Fogerty and Kewley-Port, 2009).…”

A unified account of categorical effects in phonetic perception

“…Vowel inherent spectral change has been found to be central for differentiating regional variations of American English (Jacewicz and Fox, 2013), and vowel spectra also effectively differentiate American, British, and Australian English (Ghorshi et al, 2008). Listeners are most sensitive to indexical information in vowels rather than consonants: they are more effective at discriminating between speakers based on vowels than consonants (Owren & Cardillo, 2006), and are more likely to notice when vowels, rather than consonants, are replaced with tokens from other speakers (Hertz et al, 2004;Hertz, 2006). If vowels are naturally more prone to slight differences across accents and speakers than consonants are, and listeners are sensitive to such differences, this could lead listeners to learn to treat vowels as though they have high meaningful variance.…”

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

“…4-6; but see ref. 7). Across multiple studies, investigators have deleted parts of waveforms of sentences corresponding roughly to consonants or vowels and replaced them with noise.…”