How effective is training frogs to jump? This is perhaps the most frequent question in biology that is subjected to statistical analysis: does a treatment make a difference? One can examine whether there is indeed a training effect, by first assuming the opposite. That is, the authors assume that training has no effect on the mean distance jumped. If they propose that the mean values have been calculated from samples taken from the same population, any difference they do find between the mean values would be the consequence of chance alone. They calculate the probability (or P value) of finding the observed difference between the mean distances jumped by the two groups, or something more extreme, given that they are taken from the same population. The authors have ignored, in the last experiment, the fact that they have carried out several previous experiments. Each time they did an experiment, they accepted that there was a 5% chance of a false positive result (i.e. they would conclude there was a difference, when in fact there wasn't one). This may increase the risk of concluding, purely by chance, that a difference exists when there actually is none: the more tests, the greater the risk that a false positive will occur. (Contains 5 figures.)