Behavioral studies of facial emotion recognition (FER) in autism spectrum disorders (ASD) have yielded mixed results. Here we address demographic and experiment-related factors that may account for these inconsistent findings. We also discuss the possibility that compensatory mechanisms might enable some individuals with ASD to perform well on certain types of FER tasks in spite of atypical processing of the stimuli, and difficulties with real-life emotion recognition. Evidence for such mechanisms comes in part from eye-tracking, electrophysiological, and brain imaging studies, which often show abnormal eye gaze patterns, delayed event-related-potential components in response to face stimuli, and anomalous activity in emotion-processing circuitry in ASD, in spite of intact behavioral performance during FER tasks. We suggest that future studies of FER in ASD: 1) incorporate longitudinal (or cross-sectional) designs to examine the developmental trajectory of (or age-related changes in) FER in ASD and 2) employ behavioral and brain imaging paradigms that can identify and characterize compensatory mechanisms or atypical processing styles in these individuals.