The Brazelton Neonatal Behavioral Assessment Scale was administered to 143 healthy newborns. Infants were divided into average-superior and worrisome groups based on the rating of their performance on the "a priori" dimensions and overall profile. Twelve risk variables were entered into a stepwise discriminant analysis for each dimension and the overall profile. Each dimension was best predicted by a different combination of risk variables whose weights also varied across dimensions. Birthweight contributed to four of the discriminant analyses and was the single most powerful predictor for the dimension of motoric processes and the overall profile. The results indicate that the Brazelton Scale is differentially sensitive to factors associated with risk in the newborn and that the prediction of neonatal behavior can be increased by looking at combinations of variables rather than at single variables. (Author/MS)