Several studies have reported that aphasic patients may perform substantially better on lexical than sublexical perception tasks (e.g., Miceli et al., 1980). These findings challenge claims made by models of speech perception which assume obligatory sublexical processing (e.g., McClelland and Elman, 1986; Norris, 1994). However, prior studies have not closely matched the phonological similarity of targets and distractors or task demands of the sublexical and lexical perception tasks. The current study addressed shortcomings of these prior studies, testing 13 aphasic patients on sublexical and lexical tasks matched in phonological similarity of stimuli and task demands. When the lexical and sublexical tasks were not matched (Experiment 1a), as in prior studies (e.g., Miceli et al., 1980), several patients with impaired sublexical perception were within the control range on tasks tapping lexical perception. In contrast, when the lexical and sublexical tasks (sublexical: syllable discrimination, auditory-written syllable matching (AWSM); lexical: word discrimination, lexical decision, and picture-word matching (PWM)) were matched on these factors (Experiments 1b and 2), in most instances, patients were impaired on both sublexical and lexical tasks relative to controls and performance on the lexical tasks was not significantly greater than that on the sublexical tasks. For two patients, performance on one lexical task was statistically better than that on one sublexical task, but the advantage was not replicated across other task comparisons. The current study is consistent with models of speech perception which assume obligatory sublexical processing and fails to support models that do not require successful sublexical perception in order to access lexical levels (e.g., Goldinger, 1998; Hickok and Poeppel, 2000).