Listening to a talker is more challenging when their language matches that of a competing (masker) talker (single-language context) than when target and masker speech are in different languages (dual-language context) (Brouwer et al., 2012, doi:10.1121/1.3675943). The Target-Masker Similarity Hypothesis attributes this to acoustical overlaps between languages, where greater similarity makes streams harder to separate. For bilinguals, when the target is the first language (L1), ignoring a second language (L2) masker may also be easier due to the L2’s lower activation levels, causing less interference than an L1 masker. This raises questions about whether it is easier to ignore L2 speech primarily due to acoustical differences or also differences in activation levels.

The masker’s semantic content might also influence interference (i.e., informational masking). Research in single-language contexts suggests that maskers undergo some semantic processing. For example, perception of a semantically meaningful target is more impaired by a meaningful (e.g., “Rice is often served in round bowls”) than anomalous (e.g., “The great car met the milk”) masker. Most research compares interference caused by semantically meaningful vs. anomalous maskers; however, it is also possible to investigate whether the semantic relatedness of a meaningful masker to a target interferes with perception. If there is a semantic interference effect when the target is L1 and the masker is L1 or L2, it would suggest that maskers in any known language undergo semantic processing. However, if this semantic effect is absent with an L2 masker, it would suggest that bilinguals activate meaning only for L1 maskers due to the L1’s higher activation levels.

We investigated this by presenting Spanish-English bilinguals (target N = 120, current N = 88) with pairs of simultaneous sentences. In each pair, the target sentence was Spanish (L1), while the masker was either Spanish (L1) or English (L2). The masker was either semantically related or unrelated to the target.

We hypothesise that transcription accuracy will be higher with an L2 than L1 masker, in line with the Target-Masker Similarity Hypothesis. If this effect results only from acoustical differences between the languages, then the difference between L1 and L2 masker conditions should not vary across semantic conditions. However, if different activation levels influence the masker language effect, then an effect of semantic relatedness will be apparent with an L1 masker, but not L2. This would indicate that L1, but not L2, background speech is semantically processed. Preliminary data from this study will be presented.