"Jigsaw" is a peer learning procedure derived from social interdependence theory, which suggests that individuals positively linked by a common goal can benefit from positive and promotive social interactions (Aronson & Patnoe, 2011). Although jigsaw has often been presented as an efficient way to promote learning, empirical research testing its effect on learning remains relatively scarce. The goal of the present research is to test the hypothesis that a jigsaw intervention would yield a meaningful effect size (d = 0.40) on learning outcomes, in 5 randomized experiments conducted among sixth-graders. The jigsaw intervention was compared to an "individualistic" (N[subscript Exp 1]= 252; N[subscript Exp 2] = 313) or a "teaching as usual" (N[subscript Exp 3A] = 110; N[subscript Exp 3B] = 74; N[subscript Exp 3C] = 101) approach on the same pedagogical content. Across the 5 experiments, we did not find empirical support for this hypothesis. Internal meta-analytic estimates (ES = 0.00, 95% CI [-0.10, 0.09]) showed that, overall, the jigsaw intervention did not produce the expected positive effects on learning. The reasons why jigsaw classrooms may not always prove beneficial for learning are discussed.