Sensory impairment is common after stroke though problems with the assessment of sensation have hindered research into sensation and its recovery. The revised Nottingham Sensory Assessment [NSA] (Lincoln et al, 1998) is a standardised assessment but there have been difficulties interpreting results, as it is not possible to calculate total scores. Therefore the purpose of this study was: - To investigate the extent of sensory impairment and recovery in stroke patients - To investigate if a total score for the NSA can be obtained - To explore the factors that are related to sensory impairment and outcome Method Patients with a first stroke were recruited on admission to two rehabilitation units in Nottingham. The NSA, which measures tactile sensations, proprioception and stereognostic ability, was administered on admission and at two, four and six months after stroke. Rasch analysis was used to examine if total scores of the NSA could be calculated. Results Seventy patients were recruited during a fifteen-month period. Mean age was 71 years (SD 10.00) and 36 were men. Sensory impairment was common in stroke patients and was significantly related to stroke severity. Stereognosis was the most frequently and severely impaired sensation. Rasch analysis enabled total scores of the NSA to be calculated. These totals showed significant recovery at six months post-stroke for upper limb tactile sensations, stereognosis and proprioception. Lower limb tactile sensations did not show significant recovery. The severity of the stroke, initial sensory impairment and activities of daily living ability were significantly related to sensory recovery, however they only accounted for 46-71% of the variance. Conclusion Sensation is a complex ability and a problem in its own right. Sensory impairment was a reflection of stroke severity but low variance indicates other factors were involved. Therefore there is a need to assess sensory impairment after stroke. Sensory outcome could not be accurately predicted, suggesting other potentially treatable factors such as cognitive and perceptual ability are involved. Rasch analysis allowed calculation of total scores, but also importantly allowed the scale to be shortened, making the NSA a more useable outcome measure.