Predation is a near ubiquitous factor of nature and a powerful selective force on prey. Moreover, it has recently emerged as an important driver in the evolution of brain anatomy, though population comparisons show ambiguous results with considerable unexplained variation. Here, we test the reproducibility of reduced predation on evolutionary trajectories on brain evolution. We make use of an introduction experiment, whereby guppies (<i>Poecilia reticulata</i>) from a single high predation stream were introduced to four low predation streams. After 8–9 years of natural selection in the wild and two generations of common garden conditions in the laboratory, we quantified brain anatomy. Relative brain region sizes did not differ between populations. However, we found a general increase and striking variation in relative brain size of introduced populations, which varied from no change to a 12.5% increase in relative brain weight, relative to the ancestral high predation population. We interpret this as evidence for non-parallel evolution, which implies a weak or inconsistent association of relative brain size with fitness in low predation sites. The evolution of brain anatomy appears sensitive to unknown environmental factors, or contingent on either chance events or historical legacies of environmental change.