Voice Recognition in Dementia

Impaired recognition of familiar voices (phonagnosia) has been described in patients with focal brain lesions [7] and in behavioural variant frontotemporal lobar degeneration (FTLD) [5], but the mechanisms that underpin deficits are less well characterised than for defective face recognition (prosopagnosia). The syndrome of semantic dementia (SD) is characterised by progressive breakdown of semantic knowledge associated with selective atrophy of the anterior temporal lobes, and offers a useful disease model in which to study different aspects of person knowledge. SD may present with progressive prosopagnosia in association with predominant atrophy of the right temporal lobe [3, 6], and voice recognition may also become impaired in such cases [3]. However, voice recognition has seldom been assessed in detail and the cognitive relations between the processing of voices, faces and other aspects of person knowledge remain to be clarified. Emerging evidence in the normal brain suggests a framework for the organisation of voice processing that is somewhat analogous to the processing of faces [1,2]. In particular, parallels can be drawn between cases of “associative” phonagnosia and associative prosopagnosia [4,5], in which impairments of familiar voice recognition occur despite intact perception of voices, supporting a hierarchical cognitive model of voice recognition in which perceptual mechanisms precede and interact with semantic mechanisms. In this study we investigated perceptual and semantic dimensions of voice processing in patients with semantic dementia (SD) and in a comparison group of patients with typical Alzheimer’s disease (AD), in relation to healthy age-matched controls. Using a novel neuropsychological battery, voice recog-