67
class of behaviors at a time. For example, one observer could code the verbalizations of a dyad, after which another observer would code the physical behaviors. This process might prove to be more efficient in that it would save individual observers time and also promote the development of "expertise" in the various behavior classes. In general, more frequent reliability checks should occur both during training and during coding of research tapes to ensure observer accuracy and prevent drift. Confusion matrices are particularly helpful, as noted earlier, in helping observers to identify which codes they are mistaking for one another.
This study took a step toward simplifying the DPICS II by combining the categories of Criticism and Smart Talk as well as the categories of Labeled Praise, Contingent Praise, and Unlabeled Praise. Although this step clearly expedited the coding process by requiring less discriminations, combining Criticism and Smart Talk did not improve the reliability of these categories for either parents or children. Combining the praise categories, however, significantly improved the reliability estimates, particularly for the child categories which occur less often.
Validity
Differences in Self-report Measures Between the Clinic-referred and Non-referred Fathers
Significant differences were found between the groups on the ECBI, PSI, and PLOC-SF. In addition, for the clinic-referred group, the scores on the ECBI (both Problem and Intensity scores) and the PSI (Child Domain, Parent Domain, and Total scores) were in the clinical range while the scores for the comparison group were in the normal range. These findings suggest that the ECBI and the PSI are good tools for