Helpful and hurtful behavior are commonly viewed as antithetical, with little overlap and are commonly found to be negatively related (e.g., Anderson et al., 2010). Anecdotally, however, there are many instances of behaviors that can be generally considered as representative of both help and harm (CBS, 2015; Karimi, 2015). Such behaviors, which I refer to as Prosocial Aggression (PA), have received relatively little attention in the social psychological literature. Two factors are identified as potential sources of this inattention. First, there is no current theoretical framework integrates social psychological theories of prosocial and aggressive behavior. Second, practical limitations of studying PA make such work difficult. The work presented here was designed to address these two limitations by developing a coherent theoretical account of PA behavior and validating a novel measure of PA. To this end, two studies were conducted. Study 1 provided an initial test of the PA task by manipulating the presence of victimization followed by measuring aggression toward the victimizer. In addition to the manipulation, long-term predictions of the PA model were tested in a cross-sectional manner. Personality factors that are theoretically relevant to PA (i.e., empathic anger, anti-bullying attitudes) were tested as likely predictors of PA behavior. Finally, an experiment (Study 2) tested the short-term predictions of the PA model in which empathy toward the (soon-to-be) victim, and vulnerability of the victim were manipulated and PA was subsequently measured. Results indicated no effect of either manipulation on PA. In addition, neither empathic anger, nor anti-bullying attitudes predicted PA. Discussion focuses on potential theoretical and methodological factors that may have limited the PA task’s effectiveness with an emphasis placed on directing future investigations into PA.