The aim of this study was to determine functional connectivity among patients with pediatric bipolar disorder (PBD) who are responders to pharmacotherapy and those who are nonresponders, and learn how they differ from healthy controls (HC) while performing a task that engages affective and cognitive neural systems. PBD participants (n = 34; 13.4 ± 2.3 years) were defined as responders if there was ≥ 50% improvement in Young Mania Rating Scale (YMRS) scores (n = 22) versus nonresponders with < 50% improvement (n = 12) with one of three mood stabilizing medications (divalproex, risperidone, or lamotrigine). HC (n = 14; 14.2 ± 3.1 years) participants also were scanned at baseline and follow-up. During functional magnetic resonance imaging, participants performed a color-matching task in which they had to match the color of positive, negative, or neutral words with colored dots. Independent component analysis was used to identify functionally connected networks across the whole brain, which were subsequently interrogated using region-of-interest analyses to test for group differences. A frontolimbic network was identified that showed impaired functional integration in PBD relative to HC when participants viewed negatively valenced words. PBD medication responders showed greater connectivity of the amygdala into the network before and after treatment compared with nonresponders, with responders showing a pattern more similar to HC than to nonresponders. Regardless of medication type, the degree of amygdala functional connectivity predicted medication response as well as the improvement in YMRS scores across responders and nonresponders. These findings suggest that increased functional integration of the amygdala within the frontolimbic network might be a biomarker of general mood stabilizer medication responsivity in bipolar disorder.