M. W. Eysenck has proposed that the performance deficits often associated with elevated levels of anxiety reflect an underlying restriction in the functional capacity of working memory. The current experiment was designed to test this account, by contrasting the performance of 24 high anxiety and 24 low anxiety students on a grammatical reasoning task performed under two simultaneous memory load conditions. As anticipated, the high anxiety group did indeed show disproportionately long decision latencies on this reasoning task. Furthermore, the effects of the simultaneous memory load manipulation were fully consistent with predictions derived from Eysenck's theoretical account of this performance deficit. Specifically, although all subjects were slower to perform the reasoning task under the high simultaneous memory load condition, this slowing was disproportionately severe for the high anxiety subjects. It is argued that these results provide support for Eysenck's proposal, that a restriction in working memory capacity underlies the performance deficits commonly shown by highly anxious individuals.