Most models of speech perception assume that attentional mechanisms are critical to speech processing and representation. In addition, recent work has shown that listeners encode social and linguistic information cued by phonetic variation in spoken words early and robustly. On this basis, we might assume that attentional mechanisms modulate encoding, resulting in the differential weighting of linguistic experiences, with social underpinnings. Here, we explore the role of attention in speech perception, with an eye toward socially-meaningful variation, within the dichotic listening paradigm.
Dichotic listening was introduced by Cherry (1953), who found that listeners can focus on one auditory stream while suppressing another, and that aspects of the suppressed stream may nevertheless attract attention. Listeners shadowed speech presented in one ear and ignored speech presented in the other. The ignored speech switched partway through, either from English to German, from forwards to backwards, or from a man’s to a woman’s voice. Listeners only reported explicit awareness of the switch when the sex of the unattended talker changed. This result has been interpreted to imply that, if the ignored stream captures attention, only its physical properties will be retained.
Two issues arise from this paradigm: (1) it ignores numerous social and linguistic differences between male and female voices; and (2) it uses gross measures of explicit recall, which may provoke null effects. Given current theory, the recent plentiful literature illuminating talker-specific effects, and the multi-tasking nature of being human, we might expect that listeners retain some degree of implicit memory for richly-detailed aspects of speech even in the absence of focused attention. Furthermore, early-activated social representations may modulate attention and the encoding of talker-specific episodic traces.
We extend Cherry’s paradigm to allow for fine-grained sensitivity to talker identity. We present a proof-of-concept pilot study establishing that detail of ignored speech is retained even when minimally attended and discuss the potential of this methodology to address issues of talker-specific attention and the weighting of linguistic events more broadly.
[joint work with Meghan Sumner and Grant Shillington]
