Specificity effects in spoken word recognition and the nature of lexical representations in memory

When we hear speech, besides the linguistic content, we may gain a great deal of information about the speaker from their voice, such as their identity, age, gender, or emotional state. No word is uttered the same way, even by the same talker, so one of the main challenges in spoken word recognition research is to understand the cognitive processes that underlie the processing of a complex signal like speech in the presence of high variability. Previous research has shown that listeners encode both linguistic and talker-related voice properties of the speech signal in their memory representations. Speaker variability is not the only variable we encounter; we frequently hear speech in varying auditory contexts as well. Recent evidence suggests that auditory background details, such as non-linguistic sounds co-occurring with spoken words, may be incorporated in our lexical memory. Here, I first test the hypothesis that the acoustic glimpses of words (left-overs) produced by masking from the associated sounds, rather than the sounds per se, are retained in memory. I then identify and examine the role of a novel element in the relationship between a spoken word and its associated sound, perceptual integrality, in the retention of sounds in memory. Last, I investigate the potential impact of the unique pairwise associations between words and sounds on the encoding of sounds in memory. My findings suggest that background sounds can be encoded in memory, but only in certain conditions. Specifically, this can happen when the auditory episode of the word(s) consists of highly contrasted acoustic glimpses of the same word(s), and when the sounds are made integral to, hence more difficult to perceptually segregate from the words, through intensity modulation