The top two graphs of depict the percentage of problem behaviors recorded by individual DCIs in each of the two shifts, for each of the two treatment groups, during primary observation sessions. Data points represent DCI performance on days when target problem behaviors occurred, providing the opportunity to make frequency recordings. When 2 DCIs were present, data are shown for each. The mean number of target behaviors per session ranged from 1 to 4.3 across conditions for all shifts except Group 1 (second shift). This shift encountered problem behavior more often, with means ranging from 4.5 to 15 across conditions.
The top two graphs depict percentage of target problem behaviors recorded by direct-care instructors for Groups 1 and 2 across first and second shifts in primary observations.
The results show that the in-service training had no effect on 5 of the 8 DCIs' frequency-recording performance (there was no opportunity to assess AB's performance in this condition). Supervisor presence and feedback led to immediate and marked improvements for those DCIs who did not respond to the initial training. The improvement in JA's frequency recording following the in-service training was maintained in the condition with supervisor presence and feedback, and AW showed additional improvement over that obtained with the in-service training. Improvements appeared to be maintained and perhaps increased during supervisor presence without feedback, although additional data are needed to confirm this finding. Group 1 (second shift) showed the greatest fluctuation in performance in the supervision conditions, perhaps due to the higher rate of problem behaviors that the group had to record.
The bottom graph in displays generalization data for those DCIs who had opportunities to respond in baseline and at least two experimental conditions. The number of sessions when problem behavior occurred is depicted above each bar. The data suggest that staff generalized accurate data collection to settings in which supervisors were absent, but this conclusion must be made with caution because of the small number of data points.
One implication of the findings is that the supervisor presence already inherent in the treatment environment did not lead to adequate frequency recording. Hence, the introduction of staff management techniques was apparently necessary. Also, consistent with other research (e.g., Gardner, 1972
), in-service training by itself was ineffective in improving performance, suggesting that skill deficits alone were not the major factor that contributed to DCI baseline or post-in-service performance.
DCI performance improved substantially with a very brief supervisory period (6 min) followed by feedback. Thus, most frequency recording occurred when supervisors were not present. Further, improvements were maintained even after feedback was abruptly withdrawn and may have generalized to sessions when supervisory techniques were never applied. Moreover, because target problem behaviors occurred in a minority of observation intervals, supervisor feedback regarding the nonoccurrence of problem behavior seems to have been enough to increase DCI frequency recording. Finally, it appears that high rates of problem behaviors (Group 1, second shift) resulted in more variable staff performance, suggesting the need for enhanced supervision (e.g., longer monitoring periods) or alternative methods of measuring problem behavior in such cases.
Supervisory techniques in other settings in which behavioral data are collected (e.g., classrooms) and with different types of data collection warrant further investigation. Another topic for research is whether in-service training is necessary or whether supervision and feedback alone are effective in improving staff performance. Nevertheless, the supervision methods in this study may be an efficient means of improving and maintaining staff recording of problem behavior, which may, in turn, support sounder evaluations of services for people with developmental disabilities.