Comparing two forms of a childhood perspective-taking measure using CFA and IRT.

Deficits in perspective-taking ability have been linked to social problems associated with disorders such as autism spectrum disorders (ASD) and conduct disorder. Even subtle deficits in perspective-taking are related to social adjustment and moral development. A common measure of perspective-taking abilities in children is the "Reading the Mind in the Eyes Task" ("Eyes task"; Baron-Cohen, Wheelwright, Spong, Scahill, & Lawson, 2001). The Eyes task was primarily developed for use in identifying individuals with ASDs, while its function with nonclinical populations has not been clearly addressed. Additionally, it is unknown whether the Eyes task can be used to measure specific deficits or abilities in the cognitive or emotional components of perspective-taking. In this article we assessed the structure and function of the Eyes task and an open ended or generative format of the same task (Generative Eyes Task; GET) found to measure emotional perspective-taking specifically. Confirmatory factor analyses found the traditional Eyes task to have the assumed single factor structure, while the GET has a clear 2-factor structure corresponding to emotionally valenced or neutral items. The Eyes task and the GET were also compared using item response theory. The Eyes task provided the most measurement accuracy at 2 standard deviations below the mean making it most accurate for populations with severe deficits, while the GET was most accurate at the mean level of perspective-taking. Based on these analyses, we conclude that the GET is more appropriate for use in nonclinical populations and when emotional perspective-taking abilities are of interest.