Item

Abstract

Four experiments examined whether Dutch listeners can learn to interpret a nonnative phoneme (English [\phontheta]) as an instance of a native category (Dutch [f] or [s]). During exposure in Experiment 1, two listener groups made lexical decisions to words and nonwords. Listeners heard [\phontheta] replacing [f] in 20 [f]-final words (Group 1), or [s] in 20 [s]-final words (Group 2). At test, participants heard e.g. [do\phontheta], based on the minimal pair doof/doos (deaf/box), and made visual lexical decisions to e.g. doof or doos. Group 1 were faster on doof decisions after [do\phontheta] than after an unrelated prime; Group 2 were faster on doos decisions. The groups had thus learned that [\phontheta] was, respectively, [f] or [s]. This learning was thorough: effects were just as large when the exposure sound was an ambiguous [fs]-mixture (Experiment 2) and when the test primes contained unambiguous [f] or [s] (Experiment 3). In Experiment 4, signal-correlated noise was used as the exposure sound. Listeners learned that the noise was an [f], irrespective of [f]- or [s]-biased exposure, showing that learning is determined by the new sound’s spectral characteristics. Perceptual learning in a native language is thorough, and can override years of second-language phonetic learning.