This thesis aims to further our understanding of social attention, and its manifestation in adults with autism spectrum disorders (ASD) and their typically-developed (TD) peers. Atypicalities in social attention have been proposed to play a crucial role in the development of autism. If social attention difficulties persist across the life-span, we would also expect them to impair ongoing social interactions in adolescents and adults with ASD. However, social attention in adulthood has been little examined. Instead, research tends to focus on more complex social cognitive difficulties, or to investigate attention to social stimuli presented in isolation. Our understanding of the role of social attention in autism is further inhibited by conflicting evidence on the influence of high-level input and low-level stimulus properties on selecting the focus of attention in TD individuals. The studies presented here tackle these issues by assessing social attention in adults using stimuli which: present social information in a realistic context; measure spontaneous attentional processes; and provide control over the low-level properties of stimuli. Three studies each employed a method newly applied to the study of social attention. These were: a free- description task that coded verbal accounts of social scenes; a social change detection task that recorded change detection speed and accuracy for alterations to social and non-social aspects of a person; and a preferential-looking task that presented social and non-social scenes side-by-side, while recording eye-movements. It was predicted that findings from each study would indicate a social attention bias in TD adults, while people with ASD would have either a weaker social attention bias or no bias at all. In contrast to predictions, these results showed that people with ASD spontaneously attend to social stimuli, as revealed by the social content of their verbal descriptions and their rapid and accurate detection of changes to eye-gaze direction. However, eye-tracking data in the preferential-looking task indicate that the social attention bias is subtly different in people with ASD, who show a reduced attentional priority for social information, and less persistence in looking at social stimuli over time, compared to TD participants. A series of cross-task analyses examining relationships between tasks indicated that a single social attention construct which operates across tasks and scenarios may not exist. These studies also emphasise the need to make distinctions between different types of social information and the idea of a hierarchy of social stimuli available in the real world is proposed. Taken together, the studies reported in this thesis provide new data indicating that social attentional difficulties found in children with autism do not continue in adulthood. Strong attentional preferences for social information, which override the influence of low-levelstimulus properties, are found in both TO and ASD groups. The findings also contribute a new way of thinking about the construct of social attention, in particular indicating that different types of social information may interact with individual attentional preferences. These data are interpreted in the context of recent findings of perceptual atypicalities in people with ASD, which may interact with their social difficulties. The motion and multisensory properties of real-life social interaction may present specific processing difficulties for people with ASD. If so, the mild group differences found in our studies could translate into profound problems for people with ASD in the real world, and this is an area ripe for futureresearch.