This paper discusses issues connected with the reliability of the Neonatal Behavioral Assessment Scale (NBAS) in terms of behavior prediction, neonatal behavioral organization and stability, and consequent implications for study of newborns. Discussion focuses on: (1) reliability, and (2) prediction and neonatal assessment. The NBAS is seen as a basically reliable instrument, yet its predictive reliability for stability of behavior is questioned. The view that environmental influence on developmental outcome is minimal or uniform is contested. Need is seen for specifying the functional interaction between variables which determine developmental outcomes. Development of improved measures of the range and variability of neonatal behavior is discussed. It is noted that the NBAS tester is trained to elicit and score the best observable behavior, rather than the characteristic (or modal) behavior, particularly on orientation items. Possibilities for measuring discrepancies in scores between best and characteristic behavior are discussed. It is contended that the central issue in neonatal assessment is not the identification of individual temperamental characteristics, but rather the identification of environmental features which interact to produce developmental outcomes. A number of tables are included. (Author/BF)