Academic journal article Perception and Psychophysics

Listeners Judge Talker Sex More Efficiently from Male Than from Female Vowels

Academic journal article Perception and Psychophysics

Listeners Judge Talker Sex More Efficiently from Male Than from Female Vowels

Article excerpt

Speech routinely provides cues as to the sex of the talker; in voiced sounds, these cues mainly reflect dimorphism in vocal anatomy. This dimorphism is not symmetrical, however, since during adolescent development, males specifically diverge from a previously shared trajectory with females. We therefore predicted that listeners would show a corresponding perceptual advantage for male sounds in talker-sex discrimination, a hypothesis tested using very brief, one- to eight-cycle vowel segments. The expected performance asymmetry was observed in threshold-like tests of multiple different vowels in Experiments 1-3, and a signal detection design in Experiment 4 helped rule out possible response bias effects. In confirming our counterintuitive prediction, the present study illustrates that a biological and evolutionary perspective can be helpful in understanding indexical cuing in speech.

Researchers have long noted that in addition to its linguistic value, speech provides salient acoustic cues as to talker characteristics, including biological sex, individual identity, emotional state, and more. Although they are widely acknowledged to be important, these "personal" or "indexical" cues (see, e.g., Abercrombie, 1967; Ladefoged, 1967) have received less empirical attention than have the linguistic components (see, e.g., Goldinger, Pisoni, & Luce, 1996; Johnson & Mullennix, 1997; but see Hollien, 2002; Rose, 2002). Understandably, investigators have emphasized the question of how listeners recover linguistic information hi the face of variable speech acoustics, rather than focusing on that variability itself. Work on indexical cuing has therefore tended to be pragmatic rather than theoretically inspired-for instance, in its orientation toward improving speech synthesis or automatic speech recognition (e.g., Childers & Wu, 1991; Wu & Childers, 1991; but see Fitch, 2000). However, indexical aspects of speech can also play a crucial role hi understanding phonetic content (e.g., Lachs, McMichael, & Pisoni, 2000; Pisoni, 1997), and it is important both to document such cues and to develop a stronger theoretical understanding of then- occurrence and form.

Bachorowski and Owren (1999) suggested adopting an explicitly biological approach to indexical cuing of talker sex and individual identity, arguing that naturally occurring anatomical variation is the most basic and stable source of cues to these traits. If so, perception of talker sex in particular should first and foremost mirror the audible correlates of anatomical dimorphism, which occur species-wide in humans. Aspects of this approach are implicit in previous investigations of talker-sex perception (e.g., Bennett & Montero-Diaz, 1982; Coleman, 1976; Lass, Hughes, Bowyer, Waters, & Bourne, 1976) and can be elaborated more explicitly by combining principles of sexual selection with the source-filter theory of speech production (e.g., Fant, 1960).

Vocal Production, Sexual Selection, and Anatomical Dimorphism

Voiced or phonated sounds are of particular interest in this context-that is, those sounds whose source energy is created through regular vocal-fold vibration that gives rise to a rich set of harmonics. The fundamental frequency of vocal-fold vibration (/T)) largely determines the perceived pitch of such sounds, whereas vocal-tract cavities above the larynx also leave a characteristic "imprint" on the sound's frequency spectrum through their shape- and size-specific filtering properties. This supralaryngeal vocal-tract transfer function is modeled as a series of resonances (formants) and produces characteristic patterns of high- and low-energy harmonics across the frequency spectrum.

Investigators interested in indexical features have shown that unambiguous acoustic correlates of an adult talker's sex occur in both f^sub 0^ and formant frequencies (e.g., Bachorowski & Owren, 1999; Childers & Wu, 1991; Hillenbrand, Getty, Clark, & Wheeler, 1995; Peterson & Barney, 1952; Rendall, Kollias, Ney, & Lloyd, 2005; Wu & Childers, 1991) and that both adults (e. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.