Mechanisms of natural selection can be identified using experimental approaches. However, such experiments often yield nonsignificant effects and imprecise estimates of selection due to low power and small sample sizes. Combining results from multiple experimental studies might produce an aggregate estimate of selection that is more revealing than individual studies. For example, bony pelvic armour varies conspicuously among stickleback populations, and predation by vertebrate and insect predators has been hypothesized to be the main driver of this variation. Yet experimental selection studies testing these hypotheses frequently fail to find a significant effect. We experimentally manipulated length of threespine stickleback (Gasterosteus aculeatus) pelvic spines in a mesocosm experiment to test whether prickly sculpin (Cottus asper), an intraguild predator of stickleback, favours longer spines. The probability of survival was greater for stickleback with unclipped pelvic spines, but this effect was noisy and not significant. We used meta-analysis to combine the results of our mesocosm experiment with previously published experimental studies of selection on pelvic armour. We found evidence that fish predation indeed favours increased pelvic armour, with a moderate effect size. The same approach found little evidence that insect predation favours reduced pelvic armour. The causes of reduced pelvic armour in many stickleback populations remain uncertain.