This article reports the findings of a known-groups validity study of the Empathy Assessment Index (EAI), which is a 20-item self-report instrument based on an emerging social cognitive neuroscience definition of empathy. A convenience sample formed 2 groups of study participants: social service providers (n = 197), and a group (n =453) of offender service recipients from community treatment agencies, receiving services for (a) sexual offenses (n = 251), (b) anger management or misdemeanor domestic violence (n = 155), and (c), domestic violence (n = 47). The literature suggests providers and offenders should have significantly different levels of empathy. To test this hypothesis, we compare the group scores on the overall EAI and 4 component scores (affective response, perspective taking, self-other awareness, emotion regulation). Differential item functioning analysis is used to determine whether the EAI items measure the same construct within each group, and highlights 5 items. Results include findings from a multilevel regression analysis, using the 20-and 15-item versions of the EAI. For both versions, controlling for gender, age, education, income, class, and race variables, we find treatment groups have lower average empathy scores than service providers; this difference is statistically significant for 2 of the 3 treatment groups. The emotion regulation and self-other awareness EAI components demonstrate known-groups validity in 1 of the 3 treatment groups. Researchers can likely increase both known-groups and content validity of the EAI by using perceptionaction items of greater specificity in the affective response component and by adding an affective mentalizing component.