The frequency of divorce in pair-bonding species should vary not only with the fitness costs of separating from suboptimal mates, but also with the opportunities for finding alternatives of higher quality or compatibility. Such opportunities should be more commonplace in group-living species, especially the fission–fusion groups that typify many social vertebrates. Here we tested the hypothesis that greater fluidity of the social environment results in higher divorce rates by altering flock composition of captive zebra finches, Taeniopygia guttata, soon after birds paired. Twelve flocks (10 males and 10 females per flock) were established in breeding aviaries and observed until 80% of birds in each flock paired. After removing birds that did not pair, flock composition was manipulated to achieve either a more fluid environment (by exchanging half the pairs between flocks) or a less fluid one (all pairs remained in their original flock); flock size was then restored to 20 birds by adding four new, unmated individuals. Following manipulation of group composition, some birds divorced and remated and, as predicted, most such instances (9 of 10 divorces) occurred in the higher fluidity treatment. Remated pairs took longer to initiate clutches and produced fewer offspring than stable pairs, supporting the idea that divorce has short-term costs. In the premanipulation phase of the experiment, members of pairs that subsequently divorced engaged in high levels of courtship/pairing activity, but they appeared not to resolve nest site selection. These findings support the behavioural incompatibility hypothesis as a driver of divorce. Results suggest that incidence of divorce in natural populations could be significantly higher than typically reported, given that shifts in mate affiliation relatively soon after pairing are seldom investigated. Findings also indicate that caution should be exercised regarding use of frequency of affiliative behaviours to quantify pair bond formation and strength.