Sources of variation in masking by competing speech

Abstract

The literature on speech recognition within a competing speech environment shows great variation with respect to the amount of masking produced, as well as variation in the improvement realized from spatial separation of the target and masking speech. Presumably, the sources of variation include differences in methodology and stimuli used by different researchers. In the current study the target speech stimuli were nonsense sentences spoken by a female talker. The competing speech was created from similar nonsense sentences recorded by 10 other female speakers. Five different maskers were created, each a combination of two of the 10 talkers who had similar fundamental frequencies. The results showed that the variation in masking produced by the different two-talker combinations was substantial only when target and masker were presented from the same loudspeaker. As a consequence, the observed benefit from spatial separation varied widely among the different maskers, and was ordered according to fundamental frequency in an unexpected way. The results are interpreted as showing greater variation in informational masking than in energetic masking. Also tested was the influence of different ways of blocking the trials, which effectively manipulated the uncertainty of the masking stimulus from trial to trial. [Work supported by NIH DC01625.]