Episodic memory for words enhances the language familiarity effect in talker identification

Speech perception and talker identification are intertwined. Speech from a single talker is recognized more efficiently than speech from multiple talkers; likewise, in the “language familiarity effect”, talker identification is more accurate in one's native language than a foreign one. Models of speech perception implicate episodic memory to explain effects of phonetic variability. We investigated whether these models can also account for the language familiarity effect. Listeners learned to identify voices speaking English and Mandarin in conditions differentially favoring episodic memory: (1) all talkers repeated the same sentences and (2) each talker said completely unique sentences with no repeated words. The language familiarity effect was stronger when talkers' speech had identical content, suggesting that episodic lexical access enhances talker identification in a native language. Foreign-language talker identification did not differ between the conditions, suggesting that episodic memory for voices is filtered by lexical abstraction possible only in a familiar language.