BACKGROUND: Neuropsychological investigations can help untangle the aetiological and phenomenological heterogeneity of schizophrenia but have scarcely been employed in the context of treatment-resistant (TR) schizophrenia. No population-based study has examined neuropsychological function in the first-episode of TR psychosis. METHODS: We report baseline neuropsychological findings from a longitudinal, population-based study of first-episode psychosis, which followed up cases from index admission to 10 years. At the 10-year follow up patients were classified as treatment responsive or TR after reconstructing their entire case histories. Of 145 cases with neuropsychological data at baseline, 113 were classified as treatment responsive, and 32 as TR at the 10-year follow-up. RESULTS: Compared with 257 community controls, both case groups showed baseline deficits in three composite neuropsychological scores, derived from principal component analysis: verbal intelligence and fluency, visuospatial ability and executive function, and verbal memory and learning (p values⩽0.001). Compared with treatment responders, TR cases showed deficits in verbal intelligence and fluency, both in the extended psychosis sample (t = -2.32; p = 0.022) and in the schizophrenia diagnostic subgroup (t = -2.49; p = 0.017). Similar relative deficits in the TR cases emerged in sub-/sensitivity analyses excluding patients with delayed-onset treatment resistance (p values<0.01-0.001) and those born outside the UK (p values<0.05). CONCLUSIONS: Verbal intelligence and fluency are impaired in patients with TR psychosis compared with those who respond to treatment. This differential is already detectable - at a group level - at the first illness episode, supporting the conceptualisation of TR psychosis as a severe, pathogenically distinct variant, embedded in aberrant neurodevelopmental processes.