BACKGROUND: Unilateral spatial neglect causes difficulty attending to one side of space. Various rehabilitation strategies have been used but evidence of their benefit is lacking. OBJECTIVES: To determine the persisting effects of cognitive rehabilitation specifically aimed at spatial neglect following stroke, as measured on impairment and disability level outcome assessments and on destination on discharge from hospital. SEARCH STRATEGY: We searched the Cochrane Stroke Group Trials Register (last searched 4 July 2005), MEDLINE (1966 to July 2005), EMBASE (1980 to July 2005), CINAHL (1983 to July 2005), PsycINFO (1974 to July 2005), UK National Research Register (July 2005). We handsearched relevant journals, screened reference lists, and tracked citations using SCISEARCH. SELECTION CRITERIA: We included randomised controlled trials of cognitive rehabilitation specifically aimed at spatial neglect. We excluded studies of general stroke rehabilitation and studies with mixed patient groups, unless more than 75% of their sample were stroke patients or separate stroke data were available. DATA COLLECTION AND ANALYSIS: Two review authors independently selected trials, extracted data, and assessed trial quality. MAIN RESULTS: We included 12 RCTs with 306 participants. Only four had adequate allocation concealment, that is a low risk of selection bias. A large number of outcome measures were reported. Only six studies measured disability and two investigated whether the effects persisted. The overall effect (standardised mean difference) on disability had a wide confidence interval that included zero and was not statistically significant. For discharge destination there were clinically significant effects but in both directions and the confidence interval of the odds ratio included one. In contrast, cognitive rehabilitation did improve performance on some, but not all, standardised neglect tests. The number of cancellation errors made was reduced and the ability to find the midpoint of a line improved immediately and persisted at follow up. These effects appeared likely to generalise from the samples studied to the target population, but were based on a small number of studies. AUTHORS' CONCLUSIONS: Several types of neglect specific approaches are now described but there is insufficient evidence to support or refute their effectiveness at reducing disability and improving independence. They can alter test performance and warrant further investigation in high quality randomised controlled trials. As we did not review whether patients with neglect benefit from rehabilitation input in general, such patients should continue to receive general stroke rehabilitation services.