The turn to 'evidence-based education' in the past three decades favours one type of evidence: experiment. Knowledge brokers ground recommendations for classroom practice on reports of experimental research. This paper distinguishes "field" and "laboratory" experiments, on the basis of control and precision of causal ascription. Briefly noting problems with knowledge brokers' extrapolating from field experiments, the paper's main focus is on extrapolating from laboratory experiments, using the case of 'interleaving'. It argues that knowledge brokers often extrapolate from laboratory experiments as if they are field experiments. By considering both laboratory and 'extra-lab' interleaving studies, it suggests that an alternative extrapolation--creating laboratory effects in the classroom--has little pedagogical value. The conclusion suggests focussing on mechanisms, contexts and outcomes as a more useful basis for brokering pedagogical knowledge from laboratory experiments.