The Natural History of Late-stage “Pure” Semantic Dementia

Relatively little is known about the neuropsychological profile of late-stage semantic dementia. This article provides a detailed assessment of patient MK who, despite her very severe semantic impairments, remained cooperative to testing and, unusually, did not show additional behavioral/personality changes. Although MK’s initial presentation was typical of semantic dementia (SD), her performance began to deviate from the normal pattern. She developed impairments of single word repetition and regular word reading, and began to produce phonological errors in picture naming and spontaneous speech. These deficits might suggest that late-stage SD includes an independent disorder of phonology. An alternative possibility, however, is that phonological processing cannot proceed normally in the face of profound semantic degradation. A series of experiments supported the latter explanation of MK’s deficits. In picture naming, MK showed little effect of progressive phonological cueing, did not reveal an increased sensitivity to word length or phonological complexity and continued to show a high degree of item-specific consistency in both accuracy and errors: she tended to produce the same erroneous phonemes for each item. She remained sensitive to the effects of phonological similarity in immediate serial recall. Letter substitution errors in regular word reading were more common for lower frequency letters (e.g., Q, Z). These letters also produced more item errors in immediate serial recall, suggesting that a frequency-graded loss of letter knowledge, rather than separate orthographic and phonological deficits, accounted for the deficits in both of these tasks. These findings are discussed in terms of theories that posit strong interactivity between phonology and semantics. We would like to thank David Bateman at the Royal United Hospital, Bath, UK and Roy Jones at St. Martin’s Hospital, Bath, UK for referring the patients to us. This work was carried out in part while the first author was on a study visit to the MRC Cognition & Brain Sciences Unit, Cambridge, funded by the Experimental Psychology Society. The work was also supported by a grant from the NIMH (MH64445).