This study examined the validity and reliability of the Neuropsychiatric Inventory Questionnaire version (NPI-Q), a proxy-reported format of the interview-based NPI, in assessing neuropsychiatric symptoms in 173 patients with stroke or transient ischemic attack (TIA) having cognitive impairment. The NPI-Q was validated against the NPI as a gold standard. Informants took approximately 7 minutes to complete the NPI-Q. Bland-Altman analysis revealed a bias of 0.7 points, with 95% limits of agreement between –8.6 and 10.0 between the total symptom scores of the NPI and NPI-Q. The NPI-Q correlated significantly with the NPI in individual and total symptom scores and caregiver distress scores. In predicting presence of symptoms on the NPI, the NPI-Q yielded, on average, sensitivity of 74.1% and specificity of 79.5%. On the NPI-Q, informants tended to overreport symptoms in patients with less severe symptoms but underreport with increasing symptom severity. Internal consistency of the NPI-Q was acceptable (Cronbach's α = 0.756). One-week test–retest reliability of the NPI-Q was excellent (intraclass correlation coefficient = .990). The NPI-Q is a valid and reliable instrument for screening neuropsychiatric symptoms in patients with stroke and TIA.