Listeners can recognize familiar speakers from their voices alone with remarkable accuracy and across a wide range of utterances. However, itis relatively unknown how, when, and under which conditions voices become familiar. Here, I will give an overview of recent EEG and fMRI studies from our lab using recognition memory paradigms to unravel the neural mechanisms of voice learning and subsequent recognition under conditions pertaining to the speech material, speaker and listener age as well as attentional factors. In two studies we showed that following learning of unfamiliar voices by means of brief sentences, voices can be recognized among novel voices, even from previously unheard sentences. This suggests the successful acquisition of speech-invariant voice representations. In the EEG, learned compared to novel voices elicited a suppression in beta-band oscillations from ~300 ms following voice onset independent of speech content, indicating the detection of newly-learned speaker identities at test. In fMRI, explicit voice recognition independent of speech content recruited both voice-sensitive cortex areas of the right superior temporal gyrus and extra-temporal areas including the right inferior frontal cortex. Furthermore, our research on the role of speaker and listener age shows that both young and old adult listeners are better at learning old compared to young voices an effect possibly related to the distinctiveness of old voices. Finally, I will present data suggesting that successful voice recognition is enhanced by intentional learning and is partly dissociable from non-intentional learning in ERP patterns during learning (from ~250 sm) and recognition (from ~500 ms). Overall, these studies mutually show that newly-learned voices can be recognized following only a few voice repetitions by means of brief sentence stimuli, and that voice learning is modulated by characteristics of the stimulus material and top-down mechanisms as reflected both in behavioral and neurophysiological measures.
[more]