Facilitation of speech processing by both expected and unexpected talker continuity

Abstract

Speech processing is faster for one continuous talker than mixed talkers. However, it is unknown whether listeners’ expectations about talker continuity affect this facilitation. We measured response times during three speeded word identification experiments that manipulated listeners’ expectations about talker continuity. First, we manipulated expectations about talker continuity by presenting words in pairs where both words were frequently produced by the same talker (talker-repeat trials) and rarely by different talkers (talker-change trials), or vice-versa. Word identification was faster in talker-repeat trials than talker-change trials, with equal facilitation from both expected and unexpected talker continuity. Unexpected talker changes did not slow processing more than expected changes. Second, a control experiment demonstrated that listeners’ expectations about repetitions of the word itself did affect word identification speed. Third, listeners identified words in conditions with one talker, two talkers presented randomly, or two alternating talkers. Word identification was faster whenever the talker was repeated compared to when the talker switched between trials, even if listeners could perfectly predict the talker switch (i.e., alternating-talker condition); talker continuity also facilitated word identification in the random condition. These results provide converging evidence that talker continuity facilitates speech processing in an automatic, feed-forward way, irrespective of listeners’ expectations.