Nonhuman primates perform poorly in trap tasks, a benchmark test of causal knowledge in nonhuman animals. However, recent evidence suggests that when the confound of tool use is avoided, great apes' performance improves dramatically. In the present study, we examined the cognitive underpinnings of tool use that contribute to apes' poor performance in trap tasks. We presented chimpanzees (Pan troglodytes), bonobos (Pan paniscus), and orangutans (Pongo abelii) with different versions of a maze-like multilevel trap task. We manipulated whether the apes had to use their fingers or a stick to negotiate a reward through the maze. Furthermore, we varied whether the apes obtained visual information about the functionality of the traps (i.e., blockage of free passage) or only arbitrary color stimuli indicating the location of the traps. We found that (a) apes in the finger-maze task outperformed apes in the tool-use-maze task (and partially planned their moves multiple steps ahead), and (b) tool-using apes failed to learn to avoid the traps and performed similar to apes that did not obtain functional information about the traps. Follow-up experiments with apes that already learned to avoid the traps showed that tool use or the color cues per se did not pose a problem for experienced apes. These results suggest that simultaneously monitoring 2 spatial relations (the tool-reward and reward-surface relation) might overstrain apes' cognitive system. Thus, trap tasks involving tool use might constitute a dual task loading on the same cognitive resources; a candidate for these shared resources is the attentional system.