The violence graduation hypothesis assumes a direct causal link between animal cruelty and interpersonal aggression. This hypothesis is based on two postulates: 1. that animal cruelty precedes interpersonal aggression, and 2. that the effect is specific to violent forms of antisocial behavior. The current study was designed to test the second of these two postulates: i.e., specificity. Comparing reports of prior animal cruelty in violent and non-violent prisoners and patients, a 14-study meta-analysis revealed that the violent group was significantly more likely to have a history of animal cruelty than the non-violent group. Although this seemed to support the violence graduation hypothesis, uncontrolled differences between the violent and non-violent groups provide an alternative explanation of these results. In a second meta-analysis using a different set of studies k = 5, animal cruelty was found to correlate as well with non-violent offending as it did with violent offending in male and mixed gender samples. These latter results suggest that at least in males, the animal cruelty-offending relationship is not specific to violence and that theories other than the violence graduation hypothesis may be required to explain the modest association that exists between animal cruelty and violent offending.