Reliability and Validity of Thin Slice Judgments of the Working Alliance: A Multilevel Investigation

Observational coding provides a means of measuring psychotherapy processes when self-report data is unavailable. The quality of the therapeutic alliance is a robust predictor of outcome as well as the most widely studied element of psychotherapy. The current study is the first to measure the alliance using thin slicing, or nonexpert judgments formed based on brief video clips. Four archival psychotherapy sessions from 53 adults were coded by undergraduates viewing either thin slices or entire sessions. Results indicated that four to five one-minute slices were sufficient to represent a session. We provide empirically derived coding guidelines specifying the number of raters and segments needed to optimize reliability for one-minute slices and entire sessions. A novel multilevel interrater reliability approach was applied. In addition, a single slice from three therapy sessions (3 minutes of coding total) was sufficient to predict therapy outcome (r = .28), with maximal predictive power when five slices from four sessions were aggregated (r = .33). Findings support the utility of thin slicing for efficiently capturing the alliance without compromising validity. More broadly, these data highlight the potential of thin slicing for tracking elements of psychotherapy across multiple timescales; this represents a possible avenue for addressing the time- and labour-intensive nature of existing coding methods, a current bottleneck in psychotherapy research. Further applications for research and clinical training are discussed.Draft version 1.0, 3/7/25. This paper has not been peer reviewed.