Poor context processing is related to multiple cognitive dysfunctions in schizophrenia and may constitute a core deficit of the disorder. We examined attention to social contextual information when interpreting the mental state of characters within naturalistic social scenes. Groups of healthy (n = 26) and schizophrenia (n = 24) participants viewed a series of image-pairs depicting target characters presented in isolation (Series 1: no-context) and within a social context (Series 2: context). Gaze position was recorded using the SR Eye-Link binocular eye tracker while participants performed a mental state inference task. Mean eye movement variables were calculated for each image series (context versus no-context) to examine group differences in social context processing, while detailed analysis of attention to specific contextual items was conducted using one image-pair. Overall, the schizophrenia group demonstrated significantly fewer eye movements (saccades) when viewing faces alone (Series 1). When viewing faces within social scenes (Series 2), healthy individuals significantly shortened the duration of fixations (versus Series 1), but this was not as pronounced in schizophrenia. Analysis of attention to contextual details (secondary faces and objects) in a single image-pair revealed that schizophrenia patients spent relatively more time than controls viewing the target face under both conditions (i.e., at the expense of attending to contextual information in Series 2) and showed a significant delay in directing attention to secondary faces in the scene. These findings demonstrate abnormal attention to social contextual information in schizophrenia when viewing faces in naturalistic contexts. Implications for face processing in real world situations are discussed.