Demand characteristics are a fundamental methodological concern in experimental psychology. Yet, little is known about the direction, magnitude, consistency, and mechanisms underlying their effects. We conducted a three-level meta-analysis of 195 effect sizes from 40 studies that provided experimental tests of demand effects by manipulating the hypothesis communicated to participants. Results indicated that demand characteristics tend to produce small overall increases in hypothesis-consistent responding (d = 0.22, 95% CI [0.11, 0.33]). However, these effects were extremely heterogeneous (between-study τ = 0.31; within-study σ = 0.20), with the estimated distribution of true effects ranging from d = 1.98 (a massive increase in hypothesis-consistent responding) to d = -1.44 (a massive increase in hypothesis-inconsistent responding). Contrary to conventional motivation accounts, we did not find evidence that demand effects were driven by post-hoc measures of participants’ motivation or opportunity to adjust their responses. We did, however, find robust evidence for accounts that emphasize the role of participants’ expectancies about the hypothesized effects. Similar findings emerged in a direct replication of one recent study included in the meta-analysis. Taken together, results underscore the importance – and challenges – of understanding and controlling for demand characteristics in experimental design.