Background: Schizophrenia is accompanied by significant impairment in psychosocial functioning, which is only partially explained by clinical symptom severity. Recently, these impairments have been strongly associated with deficits in neurocognition and social cognition. Although the Global Assessment of Function (GAF) scale remains the most widely used measure of psychosocial function in clinical practice, it is unclear whether this instrument is sensitive to changes in cognition, or merely provides a snapshot of symptom severity. To investigate this, we assessed whether variation in GAF score was explained by performance on measures of neurocognitive and social cognition, particularly after variation associated with symptom severity had been accounted for. Methods: 216 patients with schizophrenia were assessed using the GAF scale, two theory of mind tasks (the 'Hinting' task and 'Reading the Eyes in the Mind' task), and a neuropsychological battery sensitive to the areas of deficit typically seen in schizophrenia -IQ, episodic memory, working memory and attentional control. Results: Using linear regression analysis, symptom severity explained 24% of the variance in GAF scores (F(3, 188) = 21.14, p b .001). While neuropsychological performance explained a further 4.7% of variation (r 2 change = .047, F change (1, 187) = 12.63, p b .001), social cognition did not explain any further variance in functioning (r 2 change =.006, F change (1, 186)=1.63, p=.20). Conclusion: These data indicate that GAF scores are primarily sensitive to variation in clinical symptoms severity and not at all sensitive to variation in social cognition, an important determinant of real world outcome. Doing so highlights the need to supplement the measurement of psychosocial function using the GAF in clinical practice with functional measures that are more sensitive to deficits in social cognition.