Previous studies in learning set formation have shown that most animal species can learn to learn with subsequent novel presentations being solved in fewer presentations than when they first encounter a task. Gibbons (Hylobatidae) have generally struggled with these tasks and do not show the learning to learn pattern found in other species. This is surprising given their phylogenetic position and level of cortical development. However, there have been conflicting results with some studies demonstrating higher level learning abilities in these small apes. This study attempts to clarify whether gibbons can in fact use knowledge gained during one learning task to facilitate performance on a similar, but novel problem that would be a precursor to development of a learning set. We tested 16 captive gibbons' ability to associate color cues with provisioned food items in two experiments where they experienced a period of learning followed by experimental trials during which they could potentially use knowledge gained in their first learning experience to facilitate solution I subsequent novel tasks. Our results are similar to most previous studies in that there was no evidence of gibbons being able to use previously acquired knowledge to solve a novel task. However, once the learning association was made, the gibbons performed well above chance. We found no differences across color associations, indicating learning was not affected by the particular color / reward association. However, there were variations in learning performance with regard to genera. The hoolock (Hoolock leuconedys) and siamang (Symphalangus syndactylus) learned the fastest and the lar group (Hylobates sp.) learned the slowest. We caution these results could be due to the small sample size and because of the captive environment in which these gibbons were raised. However, it is likely that environmental variability in the native habitats of the subjects tested could facilitate the evolution of flexible learning in some genera. Further comparative study is necessary in order to incorporate realistic cognitive variables into foraging models.