Do You Really Know How Strong Some Bird’s Memory Are?

Do You Really Know How Strong Some Bird’s Memory Are?

Last Updated on by Mitch Rezman

Many people think that episodic memory is a trait only humans have because it lets us consciously reimagine memories.

There are big problems when trying to find out if animals have episodic memory because these phenomenological parts are part of what it means.

However, it’s important to note that when humans remember an event, they may remember details from that event that didn’t matter to their wants, thoughts, or desires at the time.

This “accidental” information is still stored instantly in the memory and is later retrieved as part of a complete picture of the event.

The incidental encoding and unexpected question paradigm shows this trait of human episodic memory and can be used to study how animals remember things.

This type of memory is called “episodic-like memory” because there is no proof for the associated phenomenology during recall.

We tested seven Eurasian jays (Garrulus glandarius) in this way to see how well they could use visual information from observer-made “caches” to solve a surprise memory test.

These birds did better than expected, which suggests that Eurasian jays can store, retrieve, and access visual information that happened during a remembered event.

This is similar to how people remember events in short bursts.

Editor: Miquel Llorente, School of Education and Psychology, University of Girona, Spain Facultat de Educacio i Psicologia, Universitat de Girona, SPAIN

Starting off

Mental time travel means being able to go through your own subjective time and go back to events you remember or forward to events you might think might happen in the future.

Episode-based memory is the type of memory we use when we think about and consciously re-create events from our own lives.

So, when people think about the past, they often distinguish between “knowing” and “remembering”, with episodic memory being the latter. Semantic memory, on the other hand, is the remembrance of facts that are not connected to conscious experience. This is also known as “knowing.”

This is the declarative, or clear, memory system that stores and retrieves long-term memories. It is made up of episodic memory and semantic memory.

Because episodic memory includes a conscious experience during recall, which is the main difference between it and semantic memory, its phenomenology is built into its definition.

There are big problems with looking into whether non-human animals have episodic memory because the only proof of conscious episodic recall in people comes from language-based reports, and there aren’t any agreed upon non-linguistic behavioral markers of consciousness yet.

Many psychologists think that episodic memory (and therefore mental time travel) is something that only humans can do.

They say that even though animals understand their environment well, they can’t remember and relive past events as clearly as humans can.

So, by this definition, it is impossible to tell if animals have episodic memory. As a result, researchers have instead focused on behavioral models that show how human episodic memory works in animals.

However because there is no proof of a conscious experience during recall, this kind of memory is called “episodic-like memory”.

Clayton and colleagues did a series of important studies with scrub-jays (Aphelocoma sp.) that were the first to look into episodic memory.

It is thought that these jays have developed episodic-like memories to help them remember the contents, location, and timing of their caches so that they can retrieve foods before they go bad.

This is because they store foods that last a long time (like nuts) and foods that go bad quickly (like insect larvae) for later use.

Using this natural event, these studies show that when jays retrieve trial-specific caches after different amounts of time, they remember combined representations of the “what” (food type), “where” (cache location), and “when” (time of caching compared to retrieval) information and use these memories to control their behavior.

Since these groundbreaking studies, many more have used the “what-where-when” method to test different types of animals, such as rats, great apes, other corvids, parids, and cuttlefish.

However, it has been suggested that the people who took part in these studies might have been able to remember the encoding event without necessarily remembering it in an episodic way.

Since this type of training is needed over and over to learn the rules of the experiment (like how fast certain foods break down), the animals may learn to anticipate a memory test after being shown the encoding situation.

In this way, they might learn that certain details about the event they were shown are needed for the memory test that is coming up soon. The animal may then be able to remember this information and store it in a memory trace that it can use later in the expected memory test, without actually remembering the original event when the memory test is given (which is needed for an episodic-like memory account).

Another way to test episodic memory in animals that are not humans, and one that might be the most convincing [37,38], is the incidental encoding and unexpected question approach.

This paradigm is based on the idea that when humans use episodic memory to remember a specific event, we may remember details from that event that didn’t matter to our wants, thoughts, or desires at the time. This unimportant (or “incidental”) information is still stored naturally in the memory and is later retrieved as part of a complete picture of the event.

When we remember going to the grocery store the day before, for example, we might remember the color of the cashier’s T-shirt or the pattern on the counter.

To do this, the “unexpectedly” asking subjects to remember incidental knowledge about a certain event is meant to show how incidental encoding works in human episodic memory.

The people being tested don’t know that any of the knowledge they learn in an encoding trial will be useful in a later memory test because they are not expecting it. Subjects can use information that wasn’t needed to solve an encoding task to do well on a later memory test. This shows that they can store, recall, and access information that isn’t directly related to a remembered event, which is a feature of human episodic memory. Scientists have looked into episodic memory in a number of nonhuman species using incidental encoding and surprising question paradigms.

A few species of corvids have been studied in depth for episodic memory using the what-where-when memory paradigm.

However, as far as we know, no researchers have tested their skills using the incidental encoding and unexpected question paradigm.

As a result, we tested Eurasian jays (Garrulus glandarius), which are in the Corvidae family and are related to the scrub jay, on their ability to use knowledge that was encoded by accident in a memory task that was not what we expected.

Like scrub jays, Eurasian jays often store both perishable and non-perishable foods for later use.

They rely on these caches to keep them going when food is scarce [40]. There is also evidence that they have complex cognitive skills, like object permanence, observational spatial memory, and the ability to use a variety of cache-protection strategies, though these results were not repeated in later work.

Most importantly, though, Eurasian jays have shown that they can plan for the future by giving up what they want right now to meet their expected needs in the future.

People think that episodic memory helps them think about and make predictions about the future. Based on this, we thought that Eurasian jays would be able to store random information about an event and then recall it to complete an unexpected memory task.

Methods – People and houses

This study used seven Eurasian jays, three females, and four males. All the birds that were available and willing to be tested made up the group.

In 2015, all of the jays were raised by hand and lived in a large outdoor aviary that was about 20m long, 10m wide, and 3m high.

They were kept in a social environment in this aviary, which was located at the Sub-Department of Animal Behaviour, University of Cambridge, Cambridgeshire, UK.

At one end of the aviary, smaller sections measuring about 6m by 2m by 3m were connected to testing areas inside the aviary that were each 2m by 1m by 2m.

For food, the jays were given soaked cat food biscuits, eggs, vegetables, seeds, and fresh fruit. They could also drink water whenever they wanted, even during tests.

One hour before the tests, the jays’ regular food was taken away from the cage to get them more excited about taking part.

The subjects chose to take part, which made them more motivated and were split into different areas for testing. The tester talked to the jays through an open window next to the rooms inside.

How to Do It

The Animal Welfare Ethical Review Body at the University of Cambridge looked over the studies and gave their approval.

They were done with a non-regulated procedure license (NR2021/49).

The treatments were all non-invasive and only looked at behavior. No one had to be put to sleep or killed during any of them.

In earlier studies, the jays had been trained to use cups (e.g., [51]). The cups in this study and other unpublished studies were all the same: they were all red and had no other noticeable features.

This is because the cups in these studies were just placeholders for food that could be kept underneath them. As a result, the jays had never been taught to pay attention to any traits, unique or not, on the cups before this experiment.

A spatial memory training step was done to make sure they could find the food that was hidden under the cups.

During this time, the testing area was set up with a platform next to the window for the experimenter and a perch in the middle.

Four identical cups were set up in front of the perch, with a uniform distance between them and the experimenter’s window.

When the bird went into the compartment, it was cut off from the rest of the aviary.

It could only get to the test compartment and a closed-off area of the aviary next to it.

The jays watched as the researcher turned over one of the cups, put a mealworm inside it, and then put it back where it was.

They did this while staying on the perch and facing the researcher.

The birds then chose by pulling on a string on top of the cup (S1 Fig) to see what was inside (Fig 1B).

After a decision was made, the trial was over, and the bird was told to go back to its perch before the process started all over again for the next trial.

After baiting a cup, the researcher kept their arms by their sides and turned their head and eyes straight ahead to make sure that the bird wouldn’t pick the wrong thing by chance.

It was not truly random where the baited cup was placed; the mealworm never ended up in the same cup more than twice in a row.

In a single run of 10 trials, all 10 birds met the training standard of 8/10 successful trials.


All analysis was conducted using RStudio.

We conducted binomial generalized linear mixed models (GLMMs) to investigate whether the jays chose the correct cup above chance levels (0.25), with ‘individual’ as a random effect, for both the main and control test data.

To test against the null value of 0.25, we included an offset model (the logit transform of the null value) as a fixed effect.

A second binomial generalized linear model (GLM) was used to see if the experimental factors had an effect on how well the birds did in the main test.

This model included an interaction between the fixed effects of “condition” (string, shape, or card) and “trial number” (the position in the test sequence, i.e., 1, 2, or 3).

To check our models’ assumptions, we used the DHARMa package. The models did not fail to converge and model assumption checks showed no deviation from expected distributions.

Whilst multiple studies have provided evidence in support of other corvids possessing an episodic-like memory system [11–14], this is the first such evidence in Eurasian jays.

Like other corvids, Eurasian jays habitually cache food in order to sustain them in periods of lower food availability [40].

Therefore, the temporal characteristics involved in this behavior may have been selected for the evolution of an episodic-like memory ability [27]. Amongst other corvids, Eurasian jays are especially reliant on cached food for survival [40], suggesting that this selection pressure may have been comparatively strong in their lineage.

Furthermore, Eurasian jays frequently pilfer (steal) conspecific caches, which most closely resembles the experimental situations, as in testing the jays did not cache themselves but instead watched an experimenter cache.

To facilitate pilfering, they possess the ability for observational spatial memory [42] and it has been suggested that a species dependency on stored food correlates with their skill at observational spatial memory.

Moreover, Eurasian jays have been demonstrated to use visual information to locate conspecific caches and seem to, at least in some instances, limit visual information available to competitors when caching themselves.

It is important to note, however, that as the birds in the current study were trained to observe the cup bating event and subsequently use this information to select a cup, an immediate test after baiting likely became expected over training. Whilst, at test, a substantial delay occurred between the baiting event and cup selection, the birds still learned to expect a test at the time of baiting, and therefore the encoding of the relevant information (spatial locations) was likely to be explicit. What is important here, however, is that even if the memory test after the delay was somewhat expected, the nature of the test, i.e., the relevant information and long-term storage required to solve it, was unknown to the jays.

Therefore, in order to be successful in this task the jays had to use visual information related to the cups that, at the time of encoding, held no value or relevance to the events unfolding at that time.

In the training phase (and the encoding trials), the only information necessary to solve the task and find the food was the spatial position of each cup.

Furthermore, this information only needed to be retained for a very short interval as, in this phase, they were allowed to choose a cup almost immediately after the food was hidden. Therefore, only short-term spatial working memory was needed to be successful at this stage of the experiment.

In the memory phase, however, the jays had to remember the original event, including how the cups looked, after a delay during which they could no longer see the cups and spatial information was lost because the cups had been moved to different places.

These features were present during the short retraining period, which means they might have been paid attention to.

However, without repeated training to learn that these visual features are useful as cues for food (especially since the highly reinforced spatial cues were still present), and without anticipating a memory test for these details, it’s likely that this information was stored by accident.

Indeed, across the same number of trials in the control task the birds did not learn to use visual information when the spatial cues were still present, but were now irrelevant, further suggesting that the visual information of the cups was not explicitly encoded in the main task.

The birds’ success in the main task, and failure in the control task, therefore demonstrates evidence for the use of episodic-like memory.

Whilst tests of episodic-like memory are classically designed to ignore the question of consciousness associated with memory recall, success in this paradigm may shed some light on the contents of a non-human animal’s subjective experience.

To be successful in the current task, the individual must recall a holistic representation of an event and then target specific incidental information within it in order to solve the unexpected test.

Furthermore, recent evidence demonstrates that humans have conscious access to incidentally encoded information within memories and can target this information in order to solve memory tasks [55–58].

However, it should be noted that while this study strongly suggests that Eurasian jays can store, retrieve, and access visual information that happens during a remembered event, it’s possible that this ability is limited to food-related information and doesn’t necessarily show the broad-based flexibility that is typical of human episodic memory.

So far as we know, all studies that look into episodic memory in corvids use some kind of food caching paradigm. Future studies should build on this work by testing these birds’ ability to remember other things, like social cues, like similar studies have done with other taxa.

Eurasian jays (Garrulus glandarius) show episodic-like memory through the incidental encoding of information James R. Davies, Elias Garcia-Pelegrin, and Nicola S. Clayton.

Depiction of the unique visual markers: in A) ‘string’ trials (colored card around the string attached to the top of the cup); B) ‘shape’ trials (a laminated colored shape attached to the front of the cup); and C) ‘card’ trials (a laminated colored and/or patterned card underneath the cup).


We thank the staff at the University of Cambridge Sub-Department of Animal Behaviour facility for their assistance with the birds and to the University of Cambridge and all the donors who have contributed to saving the corvid lab from closure for their financial support.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Close Menu