Great study. It is calling for a follow-up with neuronal recording, especially from mirror neurons in F5, to see whether neurons modulate their responses depending on observing intentional vs. accidental displays of reaching actions.

Minor comments : It would be interesting to dissociate the reward component of the task from the intention-direction component by placing the reward to the non targeted container.

Possible confounds:
1) Tamarins in the study were presented by the action in multiple trials and rewarded after each intentional grasp condition. This reward might have conditioned a preference toward the targeted container.
2) Moreover, the cue to the targeted container was more available from the reaching than from the hand flop trials. Therefore, reaching was a more reliable cue than a hand flop.

Interestingly, the main surprising findings, the existence of intentionality attribution and rational analysis in tamarins contradict with the major finding of the Herrmann-Tomasello study, which reported a significant advantage of human social cognition relative to nonhuman apes in theory-of-mind related tasks, like the one presented here.

Experimental ideas:
It would be worth to rethink the possible adaptation of the Gergely experiment to monkeys and apes.
Conversely, it would be nice to replicate this study on 14 mo old babies.