A species' intelligence may reliably predict its invasive potential. If this is true, then we might expect invasive species to be better at learning novel tasks than non-invasive congeners. To test this hypothesis, we exposed two sympatric species of Australian scincid lizards, Lampropholis delicata (invasive) and L. guichenoti (non-invasive) to standardized maze-learning tasks. Both species rapidly decreased the time they needed to find a food reward, but latencies were always higher for L. delicata than L. guichenoti. More detailed analysis showed that neither species actually learned the position of the food reward; they were as likely to turn the wrong way at the end of the study as at the beginning. Instead, their times decreased because they spent less time immobile in later trials; and L. guichenoti arrived at the reward sooner because they exhibited "freezing" (immobility) less than L. delicata. Hence, our data confirm that the species differ in their performance in this standardized test, but neither the decreasing time to find the reward, nor the interspecific disparity in those times, are reflective of cognitive abilities. Behavioural differences may well explain why one species is invasive and one is not, but those differences do not necessarily involve cognitive ability.