Semantic Interference From First- and Second-Language Maskers During Speech-on-Speech Listening.

Abstract

Purpose: Broad linguistic properties of masker speech (e.g., phonological and prosodic familiarity) have been shown to affect target speech transcription during speech-on-speech listening. However, it is unclear whether the semantic content of the masker impacts listening accuracy as well. If it does, it would suggest that the semantic content of the masker is co-activated with that of the target rather than being suppressed by a semantic filter mechanism. Also unclear is whether semantic interference from the masker, if found, is attenuated when the masker is in a second language (L2). These questions were addressed in two preregistered experiments.

Method: We used a speech-on-speech task in which participants transcribed a target sentence in their first language (L1), presented simultaneously with a masker sentence that was either semantically related (semantically competing [SC]) or semantically unrelated (neutral [N]) to the target sentence and was either in L1 or L2. Transcription accuracy and intrusion errors (i.e., reports of masker words) were measured as a function of masker semantic relatedness (SC, N) and masker language (L1, L2).

Results: In Experiment 1, we found that the semantic relationship between target and masker speech (either in L1 or L2) did not impact target transcription performance, even though overall accuracy was poorer when the masker was in L1 than L2. This suggests that the masker speech was not semantically processed, at least not to the extent that it affected transcription accuracy. However, SC maskers elicited more intrusion errors than N maskers. These findings were replicated in Experiment 2. A surprise recognition task (Experiment 2) suggested that this semantic intrusion effect likely resulted from listeners misstreaming targets and maskers, rather than from target-masker co-activation.

Conclusions: The results show that listeners do not process the semantic content of masker speech, which provides evidence for an effective semantic filter mechanism. This contrasts with findings indicating that other broad linguistic features of a masker can interfere with target transcription.