An aim of many students of human development is to penetrate behavioral variability in such a way as to reveal stable phenomena. This broad aim is challenged on numerous fronts when researchers study infant behavior, for students of infants are everywhere confronted with variability. But, in the context of research, variability of subjects' performance is often viewed as undesirable. Major errors in science occur when researchers look for results in the wrong place or use inappropriate measures in search of lawful relationships. Such errors may be committed when ubiquitous phenomena are classified as "noise" which obscures the revelation of lawful events. It is the thesis of this paper that variability of behavior is one such phenomenon. Evidence presented in support of this thesis focuses on issues related to (1) the development of the Neonatal Behavioral Assessment Scale, (2) Linn's (1983) findings of behaviorally stable and behaviorally variable groups of infants, which lead researchers to take the fact of variability across test examinations as an informative indicator in and of itself, and (3) subsequent studies forcing investigators to seriously consider variability as a dimension of behavior which may facilitate development. It is pointed out that nearly 15 years had passed before researchers were able to see variability data as important information relevant to processes terminate adequacy of the development of normal and at-risk infants. (RH)