Common marmosets (Callithrix jacchus) excel in a one-trial spatial memory test, yet perform poorly in a classical memory task.

When quantifying animal cognition, memory represents one of the most tested domains and is key to understanding cognitive evolution. Memory tests thus play an important role in comparative cognitive research, yet slight variations in the experimental settings can substantially change the outcome, questioning whether different memory tests tap into different memory systems or whether they test memory at all. Here, we first assessed memory performance of 16 common marmosets (Callithrix jacchus) in two distinct paradigms varying in their format and delay. First, we examined marmoset memory in a 24-h delay memory test (24 h-DMT) in which they could freely explore an environment with three novel objects of which one contained food. We examined their retention the day after, and the procedure was iterated cumulatively with previous objects remaining in the enclosure until the marmosets had to choose the correct out of 30 objects. Second, we administered a classical delayed response test (DRT) in the same animals with three objects and a maximum delay of 30 s. In the DRT, marmoset performance was poor and not better than chance after 15 s already. However, individuals excelled in the 24 h-DMT, performing above chance level after 24 h even with tenfold the number of objects to choose from compared to the DRT. Moreover, individual performances in the two tests were not correlated, and typical age effects on memory could not be detected in both experiments. Together, these results suggest that the two tests explore different domains, and that the 24 h-DMT examines long-term memory. The outcome of the DRT is more difficult to assign to memory since individuals performed only moderately even in the 0-s delay condition. This puts into question whether this task design indeed tests memory or other cognitive processes.