Reconsideration of the use of peer sociometrics for evaluating social-skills training. Implications of an idiographic assessment of temporal stability.

Social-skills training studies using sociometric procedures as dependent measures have often yielded mixed results as to the improvement of the subjects. Failure to document improvement in peer acceptance subsequent to behavior change has led some to question the validity of social-skills interventions, whereas others have questioned the psychometric properties of the measures themselves. This study examined the temporal stability of the two major types of peer measures used in social-skills interventions studies; peer-nomination measures of liking and peer-rating measures of liking. Subjects were 87 children in three fourth-grade and two fifth-grade classrooms. Temporal stability was assessed across time intervals of 2,6, and 8 weeks. Temporal stability was examined as it traditionally has been at the group level (using Pearson product-moment correlations), and at the level at which data are normally examined for change in social-skills interventions, at the level of the individual child (using phi and Cramer's V coefficients). Assessed at the group level, the three types of peer measures were generally moderately to highly stable. Stability coefficients for individual children's scores on the peer measures, however, indicated instability at the level of the individual child. These problems regarding stability at the individual, idiographic level may be especially relevant when sociometric procedures are used as dependent measures in individual subject design studies. Conceptual and practical implications of the findings for the assessment of social-skills interventions are discussed.