Research Article: Does opportunistic testing bias cognitive performance in primates? Learning from drop-outs

Date Published: March 20, 2019

Publisher: Public Library of Science

Author(s): Michèle N. Schubiger, Alexandra Kissling, Judith M. Burkart, Lesley Joy Rogers.


Dropouts are a common issue in cognitive tests with non-human primates. One main reason for dropouts is that researchers often face a trade-off between obtaining a sufficiently large sample size and logistic restrictions, such as limited access to testing facilities. The commonly-used opportunistic testing approach deals with this trade-off by only testing those individuals who readily participate and complete the cognitive tasks within a given time frame. All other individuals are excluded from further testing and data analysis. However, it is unknown if this approach merely excludes subjects who are not consistently motivated to participate, or if these dropouts systematically differ in cognitive ability. If the latter holds, the selection bias resulting from opportunistic testing would systematically affect performance scores and thus comparisons between individuals and species. We assessed the potential effects of opportunistic testing on cognitive performance in common marmosets (Callithrix jacchus) and squirrel monkeys (Saimiri sciureus) with a test battery consisting of six cognitive tests: two inhibition tasks (Detour Reaching and A-not-B), one cognitive flexibility task (Reversal Learning), one quantity discrimination task, and two memory tasks. Importantly, we used a full testing approach in which subjects were given as much time as they required to complete each task. For each task, we then compared the performance of subjects who completed the task within the expected number of testing days with those subjects who needed more testing time. We found that the two groups did not differ in task performance, and therefore opportunistic testing would have been justified without risking biased results. If our findings generalise to other species, maximising sample sizes by only testing consistently motivated subjects will be a valid alternative whenever full testing is not feasible.

Partial Text

Cognitive performance in animals can be assessed as pure proof of principle, i.e. to test whether a specific ability can be found in a given species. However, cognitive performance is increasingly assessed also for comparative purposes, both between and within species. At the inter-specific level, two or more species are tested with the same cognitive task(s) in order to explore species differences and similarities in cognitive abilities. These differences are usually hypothesised to emerge because the species have faced different selection pressures in their evolutionary past, such as challenges posed by their social or ecological environment. Examples of such studies suggest that enhanced inhibitory control is found in species with a fission-fusion social group structure [1] or a feeding ecology that requires more patience [2], that more tolerant species show increased socio-cognitive performance [3], and that species relying on a diet rich in fruit [4] or food caching species [5] show enhanced spatial memory. Encountering specific social and ecological challenges can thus, over evolutionary time, lead to domain-specific cognitive adaptations.

For each of the six cognitive tasks, we analysed whether individuals who finished the task in the expected time (i.e. would have been included in opportunistic testing) differed in performance from individuals who needed longer (i.e. would have been excluded in opportunistic testing but were included in our full testing). We ran Generalised Linear Mixed Models (GLMMs), with the exception of the A-not-B task (task 2) for which we computed Fisher exact tests owing to the small number of trials in this task. The outcome variables for cognitive performance were the response variables, and subject was included as a random factor in all models. Testing time needed to complete a given test was included as a binary fixed factor: expected amount of time vs. longer. Individual test session, species (where applicable) and all two-way interactions were also included as fixed factors in the models. We calculated all models with biologically meaningful factor combinations and identified the best model using the Akaike criterion corrected for small sample sizes (AICc, [31]).

The main question was whether the amount of testing time a subject required to complete a task affected its cognitive performance in that task. Our results show that this was not the case and that in each task, the performance of subjects who needed longer than expected to complete the task did not differ from those who completed the task in time.

This study was performed in accordance with the Swiss legislation and licensed by the Veterinary Office of the Canton of Zurich (Licence number 183/13, 24826, degree of severity: 0, i.e. no harm). Thus before, during and after this study, the monkeys were never constrained or subjected to any pain, suffering or injury and their general state of health was not impaired. All cognitive tasks were conducted non-invasively between the monkeys’ regular feeding times. The monkeys could freely enter and leave the test enclosure without being handled by humans at any time and were never isolated from their social groups. After the completion of this study, the monkeys continued living at the Primate Station, eventually participating in other non-invasive studies.




Leave a Reply

Your email address will not be published.