The effects of high variability training on voice identity learning

Abstract

High-variability training has been shown to benefit the learning of phonetic contrasts and new face identities. Here, we investigated whether high-variability training also aids voice identity learning. In Experiment 1, we contrasted high variability training (including stimuli extracted from several different recording sessions and speaking styles) with low variability training (including stimuli extracted from one recording session of read sentences; i.e., one speaking style). Listeners learned to recognise 4 voice identities (2 through high-variability training, 2 through low-variability training) and were subsequently tested on an old/new recognition task using novel read sentences. We found no high-variability training advantage in Experiment 1 – instead we found a high-variability disadvantage. However, in this experiment there was full overlap of speaking style in the low variability voices across the two task phases. We therefore ran a second experiment (Experiment 2), in which test stimuli were sourced from a previously unheard speaking style: The manipulation of high versus low-variability training was otherwise achieved in the same way as before (many speaking styles versus one speaking style). In this experiment, we observed a high-variability advantage. Findings are discussed in the context of the mechanisms thought to underpin advantages for high-variability training.