OBJECTIVES: To find determining factors for persistent infarction signs in patients with transient ischaemic attack (TIA), herein initial diffusion lesion size, visibility on apparent diffusion coefficient (ADC) or fluid-attenuated inversion recovery (FLAIR) and location. DESIGN: Prospective cohort study of patients with clinical TIA receiving 3T-MRI within 72 hours of symptom onset and at 8-week follow-up. SETTING: Clinical workflow in a single tertiary stroke centre between February 2012 and June 2014. PARTICIPANTS: 199 candidate patients were recruited, 64 patients were excluded due to non-TIA discharge diagnosis or no 8-week MRI. 122 patients completed the study. PRIMARY OUTCOME MEASURES: The primary outcome was visible persistent infarction defined as 8-week FLAIR hyperintensity or atrophy corresponding to the initial diffusion-weighted imaging (DWI) lesion. RESULTS: 50 patients showed 84 initial DWI lesions. 29 (35%) DWI lesions did not result in infarction signs on 8-week FLAIR. 26 (90%, P<0.0001) reversing lesions were located in the cortical grey matter (cGM). cGM location (vs any other location) strongly predicted no 8-week infarction sign development (OR 0.02, 95% CI 0.001 to 0.17) or partial lesion area decrease (>30% of initial DWI-area, OR 14.10, 95% CI 3.61 to 54.72), adjusted for FLAIR-visibility, DWI-area, ADC-confirmation and time to scan (TTS) from symptom onset to baseline MRI. Acute FLAIR-visibility was a strong associated factor for persistent infarction signs (OR 33.06, 95% CI 2.94 to 1432.34). For cGM lesions area size was sole associated factor for persistent infarction signs with a 0.31 cm2 (area under the curve (AUC), 0.97) threshold. In eight (16%) DWI-positive patients, all lesions reversed fully. CONCLUSIONS: 16% of DWI-positive patients and one-third of acute DWI lesions caused no persistent infarction signs, especially small cGM lesions were not followed by development of persistent infarction signs. Late MRI after TIA is likely to be less useful in the clinical setting, and it is dubious if the absence of old vascular lesions can be taken as evidence of no prior ischaemic attacks. TRIAL REGISTRATION NUMBER: NCT01531946; Results.