Background: In everyday scenarios, interpersonal communication often occurs in challenging acoustic conditions. This is also the case when interlocutors converse with physical obstructions like face masks, which have been found to impair voice propagation and radiation. To ensure efficient message transmission, speakers modify their speaking style to overcome these obstacles and make themselves more understandable. This adjustment includes the use of Lombard speech. While such mechanisms can enhance speech intelligibility for listeners in noisy environments, they may not necessarily be advantageous in other cognitive contexts or situations characterized by minimal or no background noise. The alteration in voice quality associated with Lombard speech could affect how the speech signal is perceived, potentially making it more disruptive for specific cognitive tasks. Furthermore, the extent and appropriateness of such modifications might vary, depending on individual speaker characteristics.

Methods: In the present study we examined the effects of speaker adaptations in face-masked speech on working memory performance, voice quality perception and psychoacoustic-phonetic measures. For this purpose, 33 participants were presented with audio recordings of a native German speaker uttering matrix-type sentences with and without a face mask in conversational and Lombard speech style, contained in the BEMASK corpus. Listeners completed a self-paced cued-recall task in a quiet environment and subsequently rated perceived voice quality using unipolar semantic differential scales. Statistical analyses were carried out using GLMMs.

Results: Our findings align with previous research that indicates a decrease in recall ability while wearing a face mask, as opposed to not wearing a mask, but this effect was not statistically significant. However, adjusting one’s speech style while wearing a mask impacted recall performance notably. Specifically, Lombard speech led to a significant reduction in recall, resulting in an average decrease of 5%. Psychoacoustic and phonetic measures of speech correlating with annoyance and increased vocal load were found to be higher in face-masked Lombard speech and participants perceived a significant change in voice quality between speech conditions.

Conclusion: In summary, our study replicated that face masks can slightly impact recall ability, but this effect was negligible. However, the adoption of Lombard speech, a common strategy for improving speech intelligibility, significantly reduced recall performance. These findings suggest that speech adaptations in the context of face masks may have unintended consequences on cognitive tasks. They highlight the complexity of speech adjustments and the need for a nuanced understanding of their impact, especially in varying cognitive contexts and for individual speakers.