A recently published article by Modford, Hogg, and Roberts (1997) raised concerns about the observation code used by Guess, Rues, and Roberts (1993) in a longitudinal research project to assess emerging behavior state patterns in infants and young children. Specifically, Mudford et al. found low within-state reliability scores when they used the code with 3 adults who had profound mental retardation. From these findings, Mudford et al. questioned whether the state code satisfies the natural science approach for measuring observable behavior and whether behavior state in only a hypothetical construct. In response, we have raised concerns about the thoroughness of the reliability data collected by Mudford et al. and the conclusions and interpretations they derived from these data.