Sensitive experimental techniques for the study of sound segregation

Introduction Segregation phenomena are often studied using the "double-vowel" identification task: subjects are presented with stimuli containing two vowels, and requested to identify both of them [1]. The number of trials on which both vowels are correctly identified (combinationcorrect score) is used as a measure of segregation. This technique has several drawbacks: a) The task is often too easy and may lead to near-perfect identification in all conditions (ceiling effect); b) Since the subject must always answer two vowels, cues that signal the multiplicity of sources are ignored; c) The combination-correct score does not reveal eventual asymmetries between vowels within a pair. This paper reports several modifications of that paradigm: an inter-vowel level mismatch was introduced to reduce ceiling effects, subjects were allowed to answer one or two vowels, and identification was scored separately for each vowel in a pair (constituent-correct score) [3,4].