This study aims to evaluate reliability of clinical functional magnetic resonance imaging (fMRI) in identifying language lateralisation index (LI), verified with Edinburgh handedness inventory (EHI), in brain tumour patients. Methods In this retrospective study, 31 of a single surgeon’s brain tumour patients over a 12 year period have been selected. Lesion type varied, 12 (39 percent) were high grade gliomas, 10 (32 percent) low grade gliomas, 3 (10 percent) meningiomas, and 6 (19 percent) other types. Patients underwent language fMRI paradigms for preoperative assessment, and a neuroimaging analyst was able to identify an LI value for at least one Brodmann area (BA). For each paradigm, a neuroimaging analyst attempted to calculate LI for Wernicke’s area (BA 22) and Broca’s area (BA 44 and 45). Results Of 113 total LI values, 66 (58 percent) were concordant to EHI-predicted hemispheric dominance. Reliability of language LI appears dependent upon the type of language task performed. Verb generation correctly identified Broca’s area in 18 patients (64 percent) and Wernicke’s area in 11 patients (61 percent), sentence completion correctly identified Broca’s area in 18 patients (72 percent) and Wernicke’s area in 9 patients (60 percent), and naming correctly identified Broca’s area in 7 patients (47 percent) and Wernicke’s area in 3 patients (27 percent). Conclusions Results show limited correlation between language LI determined by fMRI and EHI. The main limitation of this study is that language LI is being compared to EHI, rather than gold standard measure of hemispheric dominance (e.g. Wada).