Recent research suggests that when viewing a face two social categories (e.g., sex and race) can be activated simultaneously. However, multiple social categories – including age, race and sex – can be extracted from faces. In the present study we present a new method, motivated by previous research exploring the costs and benefits associated with language-switching, to examine whether performance on an attended social categorization task (e.g., sex classification) was impacted by changes – switches – in two unattended social category dimensions (e.g., race and age). We predicted an interaction between the effects of transition (switch versus repeat) on an attended social categorization task and transition on both of the two unattended social category dimensions. Specifically, we hypothesized that when, across two trials, the attended categorization repeated (e.g., male–male) people would be quicker and more accurate when the unattended social categories also repeated (e.g., younger face–younger face) relative to when they switched (e.g., younger face–older face). Conversely, when, across two trials, the attended categorization switched we expected that people would be quicker and more accurate when the unattended social categories also switched relative to when they repeated. These predictions were supported across three experiments, in which it was found that when unfamiliar face stimuli were categorized according to age (Expt. 1a), race (Expt. 1b) or sex (Expt. 1c) performance was impacted by the switch/repeat status of the unattended categories. These results suggest that, even when cognitively occupied, we automatically and simultaneously extract information from faces that pertain to two unattended, task-irrelevant social categories.