Whereas evidence has accumulated that nonhuman animals have access to the strength of their memory trace, it is unclear whether such metamemory contains components, as proposed by Hampton (2005). We assessed whether capuchin monkeys could recognize details of memorized items using a delayed matching-to-sample task. We used compound stimuli separable into 2 dimensions, “what” and “where.” Two monkeys were trained to memorize both “what” and “where” a sample was and answer both/either “what” and/or “where” the sample was depending on each task after a delay. In Experiment 1, the monkeys were tested on a modified delayed matching-to-sample task with selectable task option. In some trials, the monkeys could choose which task they wanted to do. In other trials, they were forced to choose a specific task because the other icon had no effect. Results showed that both monkeys selected the task at which they were better in forced trials, but accuracy in chosen trials was not significantly higher than that in forced trials. In Experiment 2, we inserted stimuli that could individually interfere with “what” or “where” after the sample and before the matching test to manipulate task difficulty, and tested whether monkeys would choose the easier task, that is, with the noninterfered-with dimension. However, task choice was unaffected by the interfering stimuli. These results did not support the hypothesis that capuchin monkeys metacognitively monitor detailed contents of their memory traces. We suggest that working memory load might have affected performance, and advocate further research to test this possibility with simpler tasks.