Recent research in the use of eye-gaze and joint attention behaviour shows inconsistent evidence of social orientation impairments in children with ASD. These inconsistencies have been linked to both differences in levels of functioning between children with autism and limitations in design as well as to methodological factors. This study explored individual differences in use of eye-gaze among 3-4 years olds with high and low functioning autism spectrum disorders (HF- and LF-ASD). The design of this study enabled comparison with children with mental retardation without autism (MR) and typical development (TD). Methods: Frequency, duration and targets of gaze were coded based on video recordings of two ADOS-G activities. The results of these measures were compared to parental and clinical reports of eye-gaze and joint attention behaviour in the two autism diagnostic instruments, ADI-R and ADOS-G. Results: Mean use of eye-gaze was lower in the ASD groups compared to control groups. No difference in within-group variance was found between children with ASD and TD. Levels of functioning were not associated with the amount of eye-gaze initiated by children with ASD. However, higher levels of both cognitive and social functioning were related to better responsive joint attention (RJA), while only social functioning was related to initiating joint attention abilities (IJA). Conclusion: Children with ASD are as heterogeneous as children with typical development in use of eye-gaze, and this diversity across groups might be related to differences in levels of functioning on more advanced joint attention behaviour. The findings are discussed in relation to methodological factors, causal theories and the importance of early identification and intervention.