In a previous experiment, we showed that domestic pigs, Sus scrofa, unlike many other species, performed accurately in a spatial memory task, where visits to a previously baited food trough were rewarded (win-stay). We investigated whether pigs have a predisposition for this strategy, by comparing their performance in a radial arm maze under either win-stay (N=10) or win-shift (N=10) reward contingencies. Contrary to our earlier results, only one of the animals in the win-stay condition was able to reach the imposed criterion level of accuracy. The performances of the other win-stay pigs did not deviate from random. All pigs in the win-shift condition reached criterion by day 25 of the experiment, and performed better than expected by chance. Analysis of the types of errors made matched our a priori predictions that shift movements would occur more frequently, especially within visits to the maze. We suggest that the difference in learning rates may reflect the fact that win-stay pigs needed to use two different rules, stay between trials and shift within trials, while win-shift pigs only needed to use the shift rule. In our previous study, win-stay pigs did not experience a conflict of rules and this may have facilitated stay learning. We found evidence of a recency effect in win-shift animals and a primacy effect in the win-stay group. However, we discuss the unsuitability of these specific terms in this type of experiment, and propose an alternative interpretation of the results.