During dialogue, people reach mutual comprehension through the production of feedback markers such as "yeah" or "okay." The purpose of the current study was to determine if mental load affects feedback production, as there is currently no consensus as to how mental load constrains the way in which dialogue partners reach mutual comprehension. In two experiments, pairs of participants interacted in order to complete a collaborative puzzle game. We manipulated the amount of mental load experienced by each participant by giving them a series of digits to memorize (or no digits) before the beginning of the game. In Experiment 1, the participants were given no information about their partner's mental load. In Experiment 2, each participant was told whether their partner had received digits to memorize. We found that although some results were identical in both experiments (directors produced more words, longer utterances, and fewer feedback markers than matchers), the effect of mental load was different in both experiments. Indeed, whereas in Experiment 1, mental load mainly affected the number of words and speech turns produced, in Experiment 2, participants who had to follow the instructions of their partner and were under low mental load produced more feedback markers when their partner was under high mental load. Taken together, these results help disentangle the contribution of experienced and perceived mental load on collaboration in dialogue. They also highlight the importance of being explicitly aware of each other's mental load in interpersonal coordination.