Short- and long-term influences of repeated speech examples on segmentation in an unfamiliar language analog
Neyla Sfeir,Dominic Guitard,Nelson Cowan
DOI: https://doi.org/10.3758/s13421-024-01517-8
2024-01-21
Memory & Cognition
Abstract:Because segments in fluent speech (e.g., words and phrases) are not reliably separated by pauses, a key task when listening to an unfamiliar language is to parse the incoming speech into segments to be learned. We aim to understand how working memory contributes to that segmentation learning. One cue to segmentation occurs when a segment is repeated in varying contexts. Cowan ( Acta Psychologica , 77 (2), 121–135, 1991) explored a language analog to study how segmentation occurs during immediate memory of speech, and found effects of segment presentation frequency, stimulus length, and serial position. Here we ask whether those effects extend from working memory to long-term memory. Overlapping segments were presented (e.g., mah bar slo mi and slo mi geh ), varying numbers of times (presentation frequencies) to determine how varying the schedule of repetition patterns would affect perception of a unified test pattern formed from the two of them (e.g., mah bar slo mi geh ). These constructions provide an analogy to how segments occur in varying contexts in speech. Participants were to indicate where they heard the boundaries between syllables. In immediate memory, the perceived boundaries more often reflected the most frequently presented pattern, and often reflected both pattern boundaries (in this example, mah bar / slo mi / geh ). In a long-term memory follow-up, however, the original presentation frequencies only mattered for certain short test pattern configurations. We suggest that working memory for speech, without a semantic component, may be an incomplete basis to learn longer segments in an unfamiliar language.
psychology, experimental