Interpretation biases have long been theorized to play a central role in depression. Yet, the strength of the empirical evidence for this bias remains a topic of debate. This meta-analysis aimed to estimate the overall effect size and to identify moderators relevant to theory and methodology. PsycINFO, Embase, Web of Science, Scopus, PubMed, and dissertation databases were searched. A random-effects meta-analysis was performed on 87 studies (N=9443). Results revealed a medium overall effect size (g=0.72, 95%-CI:[0.62;0.82]). Equivalent effect sizes were observed for patients diagnosed with clinical depression (g=0.60, 95%-CI:[0.37;0.75]), patients remitted from depression (g=0.59, 95%-CI:[0.33;0.86]), and undiagnosed individuals reporting elevated depressive symptoms (g=0.66, 95%-CI:[0.47;0.84]). The effect size was larger for self-referential stimuli (g=0.90, 95%-CI[0.78;1.01]), but was not modified by the presence (g=0.74, 95%-CI[0.59;0.90]) or absence (g=0.72, 95%-CI[0.58;0.85]) of mental imagery instructions. Similar effect sizes were observed for a negative interpretation bias (g=0.58, 95%-CI:[0.40;0.75]) and lack of a positive interpretation bias (g=0.60, 95%-CI:[0.36;0.85]). The effect size was only significant when interpretation bias was measured directly (g=0.88, 95%-CI[0.77;0.99]), but not when measured indirectly (g=0.04, 95%-CI[-0.14;0.22]). It is concluded that depression is associated with interpretation biases, but caution is necessary because methodological factors shape conclusions. Implications and recommendations for future research are outlined.