Background: We present an experiment that explores the nature of repetition priming of picture naming in a group of semantic stroke aphasic patients. The study was designed to extend previous investigations of repetition priming effects among other stroke aphasic patients and patients with semantic dementia. This work builds on previous work with semantic aphasic patients that shows enhanced picture-naming performance due to correct phonemic cues. Aims: To assess the extent to which semantic control deficits observed during semantic aphasic patients' picture naming are resolved by prior exposure to an identical stimulus, and to determine the optimal lag between prime and target to maximise naming success. Methods & Procedures: The procedure was carried out with five stroke patients who had all failed verbal and picture versions of tests of semantic association, revealing difficulties with manipulation of semantic information, and their performance was compared to five age- and education-matched controls. A total of 180 pictures to be named were presented individually on a computer screen in two sessions at least a week apart, with half preceded by an identical item in session one and the other half preceded by an identical item in session two. Three lags (0, 1, and 7 items intervening) were embedded in the pseudo-random structure such that it was unpredictable whether the next trial would be a repeat or not. Outcomes & Results: Considerable repetition priming was observed in this semantic aphasic patient group, bringing their performance up to control level at lag 0. Priming with a very short lag between prime and target (0-1 item) significantly reduced latency. Accuracy was significantly increased and semantic errors decreased with up to seven intervening items. Controls also benefited from repetition priming, but showed little variation in latency, accuracy or errors over this range of short lags. Conclusions: For patients with problems manipulating semantic information, repetition priming was an effective way to boost naming performance, although increasing the number of intervening items had a progressively detrimental effect. The observed repetition priming effects are interpreted within a connectionist model of speech production. © 2012 Copyright 2012 Psychology Press, an imprint of the Taylor & Francis Group, an Informa business.