We describe our investigations of MNA, who had a progressive, severe and gl
obal loss of semantic knowledge (semantic dementia), Her verbal vocabulary
was restricted to a few common words and she was also unable to recognize c
ommon objects from sight, By contrast, she had a well-preserved digit span
(7-8 digits). In this series of experiments, we focused on her ability to r
epeat lists and sentences in which familiarity, meaningfulness, morphology
and syntactic structure were manipulated. In list repetition tasks, we foun
d that MNA showed a reliable effect of phonological similarity, word freque
ncy and stimulus lexicality, but was unaffected by linguistic complexity, w
ord length, semantic coherence or the status of individual stimuli as 'know
n' or 'unknown'. In sentence repetition, her performance was not influenced
by any semantic variables. However, there was a substantial effect of the
frequency of the constituent vocabulary, even for words outside the range o
f her retained vocabulary. The influence of syntax was restricted to minor
effects of morphology, The phonemes of syllables and the syllables of words
are bound by their co-occurrence rather than their meaning. We conclude th
at the phonological representation of words is functionally independent of
the semantic system.