Social cognition has received growing interest in many conditions in recent years. However, this construct still suffers from a considerable lack of consensus, especially regarding the dimensions to be studied and the resulting methodology of clinical assessment. Our review aims to clarify the distinctiveness of the dimensions of social cognition. Based on Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statements, a systematic review was conducted to explore the factor structure of social cognition in the adult general and clinical populations. The initial search provided 441 articles published between January 1982 and March 2017. Eleven studies were included, all conducted in psychiatric populations and/or healthy participants. Most studies were in favor of a two-factor solution. Four studies drew a distinction between low-level (e.g., facial emotion/prosody recognition) and high-level (e.g., theory of mind) information processing. Four others reported a distinction between affective (e.g., facial emotion/prosody recognition) and cognitive (e.g., false beliefs) information processing. Interestingly, attributional style was frequently reported as an additional separate factor of social cognition. Results of factor analyses add further support for the relevance of models differentiating level of information processing (low- vs. high-level) from nature of processed information (affective vs. cognitive). These results add to a significant body of empirical evidence from developmental, clinical research and neuroimaging studies. We argue the relevance of integrating low- versus high-level processing with affective and cognitive processing in a two-dimensional model of social cognition that would be useful for future research and clinical practice. (JINS, 2018, 24, 391-404).