Pure linguistic interference during comprehension of competing speech signals

Abstract

Speech-in-speech perception can be challenging because the processing of competing acoustic and linguistic information leads to informational masking. Here, a method is proposed to isolate the linguistic component of informational masking while keeping the distractor's acoustic information unchanged. Participants performed a dichotic listening cocktail-party task before and after training on 4-band noise-vocoded sentences that became intelligible through the training. Distracting noise-vocoded speech interfered more with target speech comprehension after training (i.e., when intelligible) than before training (i.e., when unintelligible) at −3 dB SNR. These findings confirm that linguistic and acoustic information have distinct masking effects during speech-in-speech comprehension.

Distracting Linguistic Information Impairs Neural Tracking of Attended Speech

“…To make sure the training was efficient in improving the intelligibility of 4-band NV speech, we compared the participants' comprehension of the NV signals before and after training. Consistent with previous findings (9), and as shown in Fig. 2A, the training significantly improved the perception of 4-band NV speech.…”

“…In our study, when the distractor 8 was displayed on the right ear, it is primarily processed in the left language-dominant hemisphere, and may have its processing facilitated thus offering stronger interference when it is more intelligible (12). Effects of ear of presentation were not reported in Dai et al (2017), but additional analysis of those data revealed a similar (but not statistically significant) asymmetric pattern.…”

“…In line with this prediction, the comprehension of target speech is known to depend on the acoustic complexity of the distracting sounds (7,8). On top of acoustic effects, the intelligibility of distracting speech alone is also a source of interference, such that the same noise-vocoded speech background impairs more strongly the comprehension of target speech when it is intelligible as compared to when it is unintelligible (9).…”

“…3A-B). Overall, entrainment to target speech was stronger than that to distracting speech (main effect of speech type: (F(1, 24) = 62.15, p < 0.0001, interaction of data 9 and speech type (F(1, 24) = 63.72, p < 0.0001, post-hoc tests: data, target > distractor, p < 0.0001; surrogate, target vs. distractor, p = 0.176, Bonferroni corrected), and this specifically for delta oscillations (interaction between frequency and speech type (F(1, 24) = 8.39, p < 0.01, posthoc tests: delta, target > distractor, p < 0.0001; theta, target > distractor, p < 0.0001, Bonferroni corrected, Fig. 3A-B).…”

“…Based on previous findings, we hypothesized that linguistic masking should impact delta entrainment. In order to isolate the linguistic effect from the effect of acoustic competition between the target and distracting speech, we used a novel A-B-A training paradigm in which the linguistic content of the distracting stimulus was manipulated while its acoustic properties were kept constant (9).…”