Exploring the role of working memory for speech perception in energetic and informational masking: a dual-task study

Working Memory, a limited-capacity process, which supports simultaneous storage and manipulation of information, is thought to be important for any task which unfolds in time. Speech perception is one example, particularly when the speech is presented in a background noise. The nature of the background noise is often described in terms of energetic or informational masking: energetic masking occurs due to a physically overlap between target and masker signals while informational masking is mainly due to distractions and intrusions of masker information. We hypothesize that working memory is differently engaged depending on masking characteristics. Particularly, energetic masking may necessitate the engagement of the manipulation aspect of working memory in order to restore speech fragments into an intelligible signal. In contrast, informational masking may necessitate the engagement of both storage (of an intelligible but distracting masker) and manipulation (involved in the restauration of speech fragments).

We investigated these hypotheses with a dual-task paradigm. The primary task was a Speech-in-Noise task where low predictability sentences were presented in two types of noise, speech-modulated noise (energetic) and 3-talker babble (informational). The secondary task was one of four secondary working memory tasks, which required for successful completion either storage processes alone or storage and manipulation processes, and which engaged either the verbal or non-verbal domain. The different working memory aspects were operationalised by the following tasks: digit span forwards - verbal storage; digit span backwards - verbal storage and manipulation, Corsi span forwards - non-verbal storage; Corsi span backwards - non-verbal storage and manipulation.

In concordance with our hypotheses we predicted that reverse span measures would cause disruption of speech intelligibility in both types of masking whereas forward span measures would disrupt intelligibility disproportionally to informational versus energetic masking. Moreover, only verbal storage measures would affect intelligibility in information masking due to the modality specificity of the storage component.