Nicola Clayton's Jay Planning Study and the Limits of Caching Evidence
Nicola Clayton and colleagues’ jay planning study found that neither future planning nor compensatory caching best explained the birds’ caching patterns. A model with equal caching rates across locations performed best in both experiments. Six birds reached the final test in each experiment. The result limits what these tasks establish about planning, without showing that jays lack conscious experience or all forms of foresight.
Piero Amodio, Johanni Brea, Benjamin G. Farrar, Ljerka Ostojić and Nicola S. Clayton published Testing two competing hypotheses for Eurasian jays’ caching for the future in Scientific Reports on 12 January 2021. Amodio and Brea share equal contribution credit. The paper tests explanations of food storage behaviour. It does not directly measure consciousness.
Nicola Clayton’s jay planning test
Earlier studies reported that jays cached food where it would later be absent. The 2021 paper asks whether birds anticipate future availability or respond to past availability. Those explanations can predict the same action in a stable environment.
The Future Planning Hypothesis proposes that birds recall what happened, where and when, then use those events to predict future food availability. The Compensatory Caching Hypothesis proposes location-specific preferences that change with past food availability. A bird would cache more of a food where it had encountered that food less often. That rule can distribute resources without requiring a representation of a future event.
The authors changed access to three compartments on a three day cycle. Food conditions changed on a two day cycle. After nine daily experience trials, the birds received one caching test with access to all compartments. The schedules made predictions based on past scarcity diverge from predictions based on the next expected event.
Experiment 1 used peanuts and macadamia nuts. Experiment 2 used food availability versus absence, with one food type. The planning account had two variants, provisioning for the next relevant event or for several future events. The authors compared constrained versions of these accounts with models allowing other caching patterns.
Which birds reached the analysis
The study began with a colony of nine Eurasian jays. That number is different from the final sample available to explain caching choices.
| Sample or analysis | Experiment 1 | Experiment 2 |
|---|---|---|
| Birds available at the start | Nine | Eight |
| Birds reaching the final caching test | Six | Six |
| Models compared | 13 | 9 |
| Highest posterior model probability | 0.997 | 0.72 |
| Best supported model | Rates vary by bird and food, with equal rates across compartments | Equal rates across compartments |
In Experiment 1, Rome and Lima did not meet the food pre-test criteria. Testing stopped for Dublin on welfare grounds. The other six birds completed the final test.
Jerusalem was euthanized due to illness unrelated to testing before Experiment 2. Dublin did not complete familiarisation. Lima entered an unintended compartment during the last experience trial and did not receive the final test. Rome reached this experiment’s final test despite not reaching the first. Thus, five birds contributed final-test data to both experiments. These were overlapping samples from one colony.
What the uniform model establishes
Spatially uniform rates do not mean that every bird placed exactly the same number of items everywhere. Nor do they mean that every bird preferred the same food. In Experiment 1, the favored model retained differences between birds and food types while removing dependence on compartment.
In Experiment 2, the reported posterior probability for the compensatory model was 0.16, below the uniform model’s 0.72. Some individuals showed patterns compatible with compensatory caching. The authors nevertheless found that equal rates across locations best explained the combined data.
These probabilities compare the specified models under the analysis’s assumptions. They are not probabilities that jays are conscious or that all possible planning accounts are false. The uniform model describes the observed distribution. It does not identify a complete cognitive mechanism that caused it.
Task demands constrain the inference
The authors discuss several limits. Their schedule may impose greater cognitive demands than earlier designs. Experience trials lasted 15 minutes, compared with two hours in the earlier Raby and colleagues protocol. The jays may have needed more experience to learn the food-location relationships.
Each bird received only one final caching test per experiment. The authors acknowledge that behavioural variability could conceal some effects compatible with either hypothesis. They regard their results as inconsistent with very large effects under this design. This is a narrower conclusion than a general rejection of corvid planning.
The experiment also addresses a particular compensatory account. The paper separately discusses mnemonic association, in which learning occurs when a bird retrieves a previous cache. One test of two explanations does not exhaust every alternative mechanism.
Planning evidence and conscious experience
Planning and episodic-like memory are relevant to questions about the temporal reach of experience. A successful caching response would still require an argument connecting the inferred mechanism to subjective experience. Conversely, failure on one schedule cannot establish the absence of experience.
The site’s Timescapes analysis of nonhuman temporal experience asks how to characterize time from another animal’s perspective. Its Lane and Clayton discussion of dimensional welfare asks which capacities should inform care. The jay experiment places a useful limit between those questions. Uncertainty about one temporal capacity should remain attached to that capacity and task, rather than becoming a judgment about all experience or welfare.
For the evolutionary origins of consciousness section, this is a reason to distinguish observed behaviour, inferred mechanism and consciousness attribution. Animal criteria need tests that can distinguish plausible explanations. They also need limits on what a negative result excludes.
What remains unresolved
Nicola Clayton’s jay planning study leaves both tested mechanisms unsupported as the best account of these data. It also leaves open whether different schedules or more experience would reveal planning or compensatory effects. Its contribution is a test with distinct predictions and an explicit model comparison, rather than a ranking of which species possess experience.
That distinction carries into the broader review of consciousness evidence. A behavioural measure becomes useful when its alternatives and failure conditions are stated. A single positive or negative result cannot take the place of those conditions.
Source
Amodio, P., Brea, J., Farrar, B. G., Ostojić, L., and Clayton, N. S. Testing two competing hypotheses for Eurasian jays’ caching for the future. Scientific Reports 11, article 835, 12 January 2021. Sample counts follow General Methods and the two Results sections. Model results appear in Figures 4 and 7. Limitations follow the Discussion.
Researchers covered here
- Piero AmodioUniversity of Cambridge and Stazione Zoologica Anton Dohrn, as listed in the 2021 paperJoint first author of the 2021 Eurasian jay study comparing future planning and compensatory caching
- Johanni BreaEPFL, as listed in the 2021 paperJoint first author and data analyst of the 2021 Eurasian jay caching model comparison
- Benjamin G. FarrarUniversity of Cambridge and IGDORE, as listed in the 2021 paperCo-author and Experiment 1 data collector in the 2021 Eurasian jay caching study
- Ljerka OstojićUniversity of Cambridge, IGDORE and University of Rijeka, as listed in the 2021 paperCo-designer and co-author of the 2021 Eurasian jay study testing future planning against compensatory caching
- Nicola S. ClaytonDepartment of Psychology, University of CambridgeComparative cognition, episodic-like memory in corvids, mental time travel in animals, Fellow of the Royal Society (FRS)