The role of working memory in selective attention: Integrality of two speech dimensions across and within a category boundary

When two dimensions act together as a single factor to characterize a stimulus, they are said to be integral. However, one dimension may interfere with discrimination of the other to a greater degree, representative of an asymmetrically integral relationship. The current study investigated the nature of integrality between initial stop consonant (/d/, /t/) and voice-onset time (VOT), two linguistic dimensions, and fundamental frequency (fo), a nonlinguistic dimension and the extent to which working memory is influential in determining these relationships. In a set of four experiments, participants engaged in selective attention tasks (dual-channel oddball tasks) and item-recognition tasks to examine the integrality of (1) fo (high, low) and initial stop consonant (/d/, /t/), with consonant variations occurring across a perceptual category (/da/-/ta/), and (2) fo (high, low) and VOT variations occurring within one perceptual category (/ta/ only). In addition, item-recognition tasks that contained shorter delay periods (500 ms) and longer delay periods (3000 ms) between the initially-presented stimulus and probe presentation were tested. Garner Interference (decline in behavioral performance in reference to baseline performance) was observed in the filtering tasks when the target was characterized by fo or VOT, both when VOT varied across and within a category boundary. Experiments 1 and 2 revealed symmetric integrality between fo and VOT, when VOT varied across and within a category boundary. This was consistent with previous results (Melara & Marks, 1990; Zdanczyk, 2005) showing interference when a nonlinguistic dimension or linguistic dimension is varied irrelevantly. Experiments 3 and 4 revealed asymmetric integrality in short and long delay conditions, but the direction of this asymmetry was dependent on whether participants were holding linguistic information characterized by across- or within-category distinctions. Experiment 3 illustrated interference by irrelevant consonant differences (across-category VOT differences), while Experiment 4 illustrated interference by irrelevant fo differences. Thus, when selection of a relevant dimension occurred within the context of working memory, an asymmetry was revealed. Potential reasons are discussed concerning the basis of this effect, including the influence of intrinsic differences between the two dimensions, interactions between attention and working memory, and differences within the tasks themselves