OBJECTIVES: Affective neuroscience research that investigates core symptoms of pediatric bipolar disorder (PBD) may be effective in differentiating PBD phenotypes. The current study used affect-modulated startle to examine potential differences in reactivity to emotional stimuli (reward and punishment) in narrow and broad phenotype PBD and controls. METHODS: Thirty children meeting DSM-IV bipolar disorder criteria (i.e. narrow phenotype PBD with defined manic episodes with elevated/expansive mood), 19 children meeting criteria for severe mood dysregulation (i.e. broad phenotype with chronic irritability, hyper-reactivity, and hyperarousal), and 19 controls completed a lottery startle paradigm involving reward (money) and punishment (loud noise). Startle probes were presented during anticipation of the emotional stimulus, immediately following the presentation of the stimulus, or during return to baseline following the stimulus. RESULTS: By self-report, patients and controls found the putative punishment to be preferable to the neutral condition. In the reward condition, patient samples reported greater arousal than did controls, but no between-group differences were found on the magnitude of startle response during the reward, punishment, or neutral conditions. CONCLUSIONS: The failure to find differences in affect-modulated startle between control children and those with narrow or broad PBD phenotypes speaks to the methodological challenges associated with studying reward mechanisms in PBD. Alternative paradigms that focus on different aspects of reward mechanisms are discussed.