Nearly all theories of spoken word perception presume a lexicon with singular entries corresponding to each word. In turn, the perceptual system is presumed to operate by matching such entries (e.g., nodes, spectral templates, distributed patterns) to the variable signals that speakers produce, requiring either a process of normalization or sophisticated guessing. In contrast, episodic theories assume that people store multiple entries, in the form of detailed perceptual traces, for each in the lexicon. Such multiple-trace theories are robust to variation, and they provide a natural explanation for extra-linguistic learning, such as learning voices.
In this presentation, I will review selected evidence for detailed perceptual episodes, including assessment of their specificity, longevity, and the factors that shape their memorial forms. Following this, I will address the continuity of memory and perception, emphasizing the role of stored traces in perceptual tasks.
Finally, I will review my applications of a specific model (Hintzman's MINERVA 2) to data in speech perception and production. Key issues and challenges, including the wide-ranging roles of selective attention, will be discussed.