| Meta-analytic techniques were used to empirically review the social skill training literature. An extensive literature survey located both published and nonpublished research. This exhaustive search strategy resulted in three hundred and fifty-one social skill training studies. Sixty-three studies were identified consistent with current definitions of social skill training, and obtained from research designs that yielded difference score statistics, and which had outcome measures reported in sufficient detail to be converted to a common metric.; Relevant study characteristics were classified by three independent raters. Inter-rater agreement for these coded characteristics exceeded 90%. The descriptive results from this investigation suggested the following: The majority of the social skill training studies came from published sources. These studies were conducted primarily by trained professionals. Their subjects were frequently elementary and high-school aged male and female students. The majority of children were selected for participation on subjective grounds, for example, class assignment and volunteering. The average training program consisted of sixteen-sixty minute sessions, which occurred two times per week. In general, assessment procedures involved multiple measures across several informed sources. However, few programs attempted to program and/or assess for generalization.; The "Meta-Analysis Effect Size Calculator" was used to convert study outcome to a common metric. This conversion resulted in five hundred and eighty-six frequency weighted effect size estimates, the mean of which was.652 with a variance of.697 after correction for sampling error. This result indicated that the average participant in a training program was more socially skillful than 74% of those not treated.; Multiple regression analyses were conducted on coded study characteristics in search of moderator variables to account for portions of the non zero variance. The dependent variables for these analyses were classified according to Gresham's (1983) three social validity types (Type I, II, III). These results suggested that dependent measures from behavior observation (Type II), self report/role-play (Type III), and teacher ratings (Type I) produced effect size estimations in descending order of magnitude.; Future research implications were discussed. Specifically, the regression analyses suggested combinations of moderator variables which should produce larger effect size estimations. This future "best package" of social skill technology would involve: The sociometric selection of subjects who are trained by professionals. This training would involve frequent meetings (3-5 per/wk), where the total training time is greater than 360 minutes, and assessment devices sample from the three different types of socially valid dependent measures. This package would use a combination of training methodologies to include coaching and modeling. Effective training methods would also program and assess for generalization. |
