Working memory for sign language: a window into the architecture of the working memory system.

Traditionally, working memory has been divided into two major domains: verbal and visuo-spatial. The verbal domain of working memory can be characterized either by its relationship to language or by its grounding in auditory processing. Because of this ambiguity, languages that are not auditory and vocal (i.e., signed languages) pose a challenge to this conception of working memory. We describe several experiments with deaf users of American Sign Language (ASL) that explore the extent to which the architecture of working memory is determined by the constraints of auditory and visual processing and the extent to which it is determined by the characteristics of language. Various working memory effects were investigated: phonological similarity, word length, and articulatory suppression. The pattern of evidence strongly supports the existence of a sign-based 'rehearsal loop' mechanism parallel to the speech-based rehearsal loop. However, we also discuss evidence pointing to differences between the speech loop and the sign loop from forward and backward digit span tasks with deaf and hearing subjects. Despite their similarities based on linguistic properties, the speech loop and the sign loop appear to diverge due to the differing processing demands of audition and vision. Overall, the results suggest that the architecture of working memory is shaped both by the properties of language structure and by the constraints imposed by sensorimotor modality.