The overall objective is to evaluate and compare the efficacy of two common treatments for CB in children and adolescents with autism using multiple outcome measures. Our central hypothesis is that both behavioral treatment and medication will produce reductions in CB, but behavioral treatments, including behavior treatment + medication, will produce a greater reduction in CB when measured using direct observation within the contexts most associated with challenging behavior. We have four aims: Aim 1: Evaluate the concordance between indirect and direct measures of challenging behavior. We hypothesize that behavior change (% change) will be discordant between indirect and direct measurement data, regardless of treatment type, with direct measurement resulting in significantly greater change at each measurement timepoint. Aim 2: Compare the efficacy of psychotropic medication, behavior therapy, and combination therapy (medication+behavior therapy) on CB using indirect measures (ABC-I, BPI, CGI, MOSES) and direct observation (antecedent analysis). We hypothesize that all treatment modalities will reduce CB on both outcome measures at each standard timepoint, but behavioral therapy and combination therapy will produce a significantly greater reduction on direct observation of CB in the contexts most associated with challenging behavior. Aim 3: Compare the social validity of psychotropic medication, behavior therapy, and combination therapy using the Treatment Acceptability Rating Form-Revised (Reimers et al., 1992). Aim 4: Compare the social validity of measurement methods (indirect vs. direct).