Assessment in social skills training often utilizes procedures such as partial-interval recording (PIR) and momentary time sampling (MTS) to estimate changes in duration in social engagements due to intervention. Although previous research suggests PIR to be more inaccurate than MTS in estimating levels of behavior, treatment analysis decisions have been found to be similar across procedures. To addresses limitations of previous studies that did not find substantial differences in treatment analysis decisions, PIR and MTS estimates were compared to those of continuous duration recording of social engagements to determine the effect on treatment interpretation. Five previously conducted social skills training data sets were coded using PIR, MTS, and duration recording. Treatment analysis interpretations using MTS were found to closely match those made using duration recording, but interpretations using PIR significantly overestimated the effectiveness of the intervention. Implications of findings are discussed in terms of assessment for social skills training in both research and applied settings.