Recently, researchers have started to measure the working alliance repeatedly across sessions of psychotherapy, relating the working alliance to symptom change session-by-session. Responding to questionnaires after each session can become tedious, leading to careless responses and/or increasing levels of missing data. Therefore, assessment with the briefest possible instrument is desirable. Because previous research on the Working Alliance Inventory has found the separation of the Goal and Task factors problematic, the present study examined the psychometric properties of a 2-factor, 6-item working alliance measure, adapted from the Working Alliance Inventory, in three patient samples (N = 1095, 235, and 234). Results showed that a bifactor model fit the data well across the three samples, and the factor structure was stable across ten sessions of primary care counseling/psychotherapy. Although the bifactor model with one general and two specific factors outperformed the one-factor model in terms of model fit, dimensionality analyses based on the bifactor model results indicated that in practice the instrument is best treated as unidimensional. Results support the use of composite scores of all six items. The instrument was validated by replicating previous findings of session-by-session prediction of symptom reduction using the Autoregressive Latent Trajectory model. The 6-item working alliance scale, called the Session Alliance Inventory, is a promising alternative for researchers in search for a brief alliance measure to administer after every session. The working alliance concerns the quality of the collaborative relationship between therapist and patient in the process of psychotherapy, and has been studied extensively as a predictor of psychotherapy outcome (e.g. Horvath, Del Re, Fluckiger, & Symonds, 2011). Most studies to date have relied upon a single alliance measurement early in treatment to predict outcome in the form of symptom relief or improvement in functioning at the end of treatment.Lately, however, more complex models of the alliance-outcome relationship have been used: in a small number of studies, the alliance has been measured after each psychotherapy session and this measure was used to predict symptom change to the next session (Crits-Christoph, Gibbons, Hamilton, Ring-Kurtz, & Gallop, 2011;Falkenström, Granström, & Holmqvist, 2013;Hoffart, Øktedalen, Langkaas, & Wampold, 2013;Tasca & Lampard, 2012).The working alliance is usually measured using self-report instruments, most commonly filled out by the patient. In contrast to observer methods for measuring the working alliance, questionnaires tend to focus not just on the current session, but on the quality of the alliance so far in treatment. This means that when the alliance is measured after session three, questions focus not just on the quality of the alliance in session three, but on the quality of the alliance from treatment start up to and including session three. This makes sense in studies where the alliance is measured ju...