Objective: Even though the early alliance has been shown to robustly predict posttreatment outcomes, the question whether alliance leads to symptom reduction or symptom reduction leads to a better alliance remains unresolved. To better understand the relation between alliance and symptoms early in therapy, we meta-analyzed the lagged session-by-session within-patient effects of alliance and symptoms from Sessions I to 7. Method: We applied a 2-stage individual participant data meta-analytic approach. Based on the data sets of 17 primary studies from 9 countries that comprised 5,350 participants, we first calculated standardized session-by-session within-patient coefficients. Second, we meta-analyzed these coefficients by using random-effects models to calculate omnibus effects across the studies. Results: In line with previous meta-analyses, we found that early alliance predicted posttreatment outcome. We identified significant reciprocal within-patient effects between alliance and symptoms within the first 7 sessions. Cross-level interactions indicated that higher alliances and lower symptoms positively impacted the relation between alliance and symptoms in the subsequent session. Conclusion: The findings provide empirical evidence that in the early phase of therapy. symptoms and alliance were reciprocally related to one other, often resulting in a positive upward spiral of higher alliance/lower symptoms that predicted higher alliances/lower symptoms in the subsequent sessions. Two-stage individual participant data meta-analyses have the potential to move the field forward by generating and interlinking well-replicable process-based knowledge.