Testing social learning theory using reinforcement's residue: A multilevel analysis of self-reported theft and marijuana use in the national youth survey

Critics have expressed concerns regarding measurement strategies or analytic techniques often used in social learning research (Horan and Phillips, 2003; Krohn, 1999; Sampson, 1999; Tittle, 2004). In response to these concerns, this study tests the hypothesized causal relationships among reinforcement, general definitions, and self-reported crime (theft and marijuana use) using a multilevel modeling approach with longitudinal data from the first five waves of the National Youth Survey (NYS), as well as with indirect parent and friend reinforcement measures that incorporate both the assumed products of reinforcement (expected consequences of behavior) and the efficacy of reinforcement (expected influence of the reinforcement source). Within-subject analyses present a challenge to the theory as social learning variables do not covary significantly over time with criminal offending rates. Between-subject analyses offer support for the theory as across-person differences in average parent and friend reinforcement are significantly related to offending rates, and these reinforcement-crime relationships are mediated partially or fully by learned definitions. Implications of these findings are discussed.