Objective: Clinical fMRI is routinely used to lateralize language dominance in surgical planning, but it is not possible to reliably localize areas of language cortex. This study evaluated a novel clinician-based approach to language localization in which analysis is based on a suite of three fMRI tasks, flexibly thresholded in analysis, and a novel neurocognitive model of language extending beyond Broca's and Wernicke's areas. Initial validity and reliability data were derived. Method: Data from 19 temporal lobe patients (15 left/4 right; 7 female) undergoing evaluation for possible epilepsy surgery at UCLA received routine clinical fMRI (object naming; auditory naming; reading tasks) and Wada testing (16 left dominant, 2 right, 1 mixed). Two neuropsychologists independently selected and analyzed 1–3 of 6 T2* sequences to identify language regions including Broca's and Wernicke's Areas; Angular gyrus; Basal Temporal Language Area; Exner's Area; and Supplementary Speech Area. Results: Across the 38 maps generated, language laterality matched Wada dominance closely (left dominant: 29/32 cases; right 4/4; mixed 1/2). Analysis revealed reasonable correspondence between maps generated by separate raters, with an average of 62% of clinicians' maps overlapping and DICE coefficient of 0.58. Raters typically selected similar or identical runs (63%) with differences between generated maps more typically reflecting variation in extent of activation, and higher correspondence observed in identification of specific language regions. Conclusion This study provides initial data for a novel approach to clinical fMRI that capitalizes on clinicians' knowledge of brain-behavior relationships to map a series of regions, damage to which has been associated with language impairment.