MR diffusion/perfusion mismatch may help identify patients for acute stroke treatment, but mixed results from clinical trials suggest that further evaluation of the mismatch concept is required. To work effectively, mismatch should predict prognosis on arrival at hospital. We assessed mismatch duration and associations with functional outcome in acute stroke. We recruited consecutive patients with acute stroke, recorded baseline clinical variables, performed MR diffusion and perfusion imaging and assessed 3-month functional outcome. We assessed practicalities, agreement between mismatch on mean transit time (MTT) or cerebral blood flow (CBF) maps, visually and with lesion volume, and the relationship of each to functional outcome. Of 82 patients starting imaging, 14 (17%) failed perfusion imaging. Overall, 42% had mismatch (56% at <6 h; 41% at 12-24 h; 23% at 24-48 h). Agreement for mismatch by visual versus volume assessment was fair using MTT (kappa 0.59, 95% CI 0.34-0.84) but poor using CBF (kappa 0.24, 95% CI 0.01-0.48). Mismatch by either definition was not associated with functional outcome, even when the analysis was restricted to just those with mismatch. Visual estimation is a reasonable proxy for mismatch volume on MTT but not CBF. Perfusion is more difficult for acute stroke patients than diffusion imaging. Mismatch is present in many patients beyond 12 h after stroke. Mismatch alone does not distinguish patients with good and poor prognosis; both can do well or poorly. Other factors, e.g. reperfusion, may influence outcome more strongly, even in patients without mismatch.