Abstract Background Prospective studies linking infant temperament, or behavioral style, to infant body composition are lacking. In this longitudinal study (3 to 18 months), we seek to examine the associations between two dimensions of infant temperament (distress to limitations and activity level) and two anthropometric indicators (weight-for-length z-scores (WLZ) and skin fold (SF) measures) in a population at high risk of overweight. Methods Data are from the Infant Care and Risk of Obesity Project, a longitudinal study of North Carolina low income African American mother-infant dyads (n = 206). Two temperament dimensions were assessed using the Infant Behavior Questionnaire-Revised. A high distress to limitations score denotes an infant whose mother perceives that s/he often cries or fusses, and a high activity level score one who moves his/her limbs and squirms frequently. Cross-sectional analyses were conducted using ordinary least squares regression. Fixed effects longitudinal models were used to estimate anthropometric outcomes as a function of time varying infant temperament. Results In longitudinal models, increased activity levels were associated with later decreased fatness and WLZ. In contrast, high levels of distress to limitations were associated with later increased fatness at all time points and later increased WLZ at 12 months. Conclusion Infant temperament dimensions contribute to our understanding of the role of behavior in the development of the risk of overweight in the formative months of life. Identification of modifiable risk factors early in life may help target strategies for establishing healthy lifestyles prior to the onset of overweight.