Capuchin monkeys learn to use information equally well from individual exploration and social demonstration

Abstract

The limited evidence of complex culture in non-human primates contrasts strikingly with human behaviour. This may be because non-human primates fail to use information acquired socially as effectively as they use information acquired individually. Here, monkeys were trained on a stimulus discrimination task with a win-stay, lose-shift (WSLS) reward structure. In a social learning condition, the experimenter performed an information trial by choosing between the available stimuli; in an individual condition, monkeys made this choice themselves. The monkeys' subsequent test trials displayed the same stimulus array. They were rewarded for repetition of rewarded ('win-stay') and avoidance of unrewarded ('lose-shift') information trial selections. Nine monkeys reached our pre-determined performance criterion on the initial two-stimulus stage. Their ability to generalise the WSLS strategy was then evaluated by transfer to a three-stimulus stage. Minimal differences were found in information use between the social and individual conditions on two-stimuli. However, a bias was found towards repetition of the information trial, regardless of information source condition or whether the information trial selection was rewarded. Proficient subjects were found to generalise the strategy to three-stimuli following rewarded information trials, but performed at chance on unrewarded. Again, this was not found to vary by source condition. Overall, results suggest no fundamental barrier to non-human primates' use of information from a social source. However, the apparent struggle to learn from the absence of rewards hints at a difficulty with using information acquired from unsuccessful attempts; this could be linked to the limited evidence for cumulative culture in non-human primates.

Keywords: Cumulative culture; Discrimination learning; Individual learning; Non-human primates; Social learning; Touchscreen.

Fig. 1

a Arbitrary visual reward cue used during training and task, and b Stage A example stimuli

Fig. 2

WSLS success over sessions separated by subject, whether criterion was met or not, and information source. Sessions were grouped into blocks of five. NB: includes the three sessions where criterion was met

Fig. 3

WSLS success on Stage A for all subjects by information source (top), information type (middle), and the interaction between information source and type (bottom). Dashed line indicates chance and the black point indicates the mean. ‘N’ indicates the number of data points (trials) for each individual subject

Fig. 4

Repeats on Stage A for all subjects by information source (top), information type (middle), and the interaction between source and type (bottom). Horizontal significance brackets indicate group differences. Exact binomial tests found that monkeys repeated above chance on both rewarded and unrewarded problems overall (both p < 0.001; middle panel, vertical significance brackets indicate difference to chance); when also separated by source, monkeys repeated significantly above chance (bottom panel) on all (p < 0.001) except unrewarded problems in the social condition (p = 0.19). NB: For the rewarded information type, successful and repeat responses are iso-directional; for the unrewarded information type, successful and repeat responses are in opposite directions. Dashed line indicates chance and the black point indicates the mean. ‘N’ indicates the number of data points (trials) for each individual subject

Fig. 5

Example stimuli from each stage. One selection was made, and one stimulus rewarded at both stages

Fig. 6

Repeats on all sessions of Stage B broken down by information type and by information source. The optimal response for rewarded trials was to repeat, and for unrewarded was to shift. Dashed line indicates chance

Fig. 7

Response (shift or repeat) to rewarded and unrewarded information trials by all criterion subjects (left) and separated by information source (middle and right). The last three sessions of Stage A before criterion was met (“pre-crit”), three sessions where criterion was met on Stage A (“criterion”, included for illustration only) and first three sessions of Stage B (“Stage B”) are shown. The optimal response for rewarded trials was to repeat (above-chance performance), and for unrewarded was to shift (below-chance performance). Dashed line indicates chance (Stage A = 0.50; Stage B = 0.33)