Episodic-like Memory and other Behavior in Scrub Jays. Lecture 7 Psych 1090. I’ve done things a bit differently in this lecture…. assigning only a review paper for the earlier material. and giving the details in the lecture. What is episodic memory?. unique, personal, past experience.
Episodic-like Memory and other Behavior in Scrub Jays
Ive done things a bit differently in this lecture
assigning only a review paper for the earlier material
and giving the details in the lecture
What is episodic memory?
Explanation in terms of other forms of human memory:
Declarative memory is subdivided:
According to Tulving and Marlowitsch, episodic memory is
Supposedly, animals remember facts (semantic memory) but not personal experiences (episodic memory)
plenty of evidence to show that animals remember that x indicates y, or if x, do y,
but not necessarily when I saw x, I then remember doing y
The point may seem trivial, but its not..
The difference is in knowledge that is assumed to represent the way the world works for everyone (semantic memory)
And knowledge that is understood to represent only what one has personally experienced (episodic memory)
Which brings up another issue:
Some researchers argue that to understand that an experience is personal, a being must have full consciousness.
an attribute that is generally denied to animals.
To get around the argument that a being must be conscious in order to have personal memoriesi.e., to have episodic memorywe can simply agree
to define episodic-like memory as retrieval of what, where AND when
Thus the issue of personalization and of self-projecting past to future
And of assuming that others experience life similarly
Such a definition puts emphasis on the episode aspect of episodic memory, and is one that can be tested in animals
Note that other types of memory that have been studied may seem to involve time, but are not truly episodic.
For example, animals trained on delayed match-to-sample or delayed nonmatch-to-sample
may seem to be recalling previous, personal events.
That is, being shown a red sample at time T and then, at T+20 seconds, being shown red and green samples,
they have learnedvia trial and errorthat they get rewarded
only for matching or not matching the original sample shown at time T
One could argue that the animals are responding based on personal, event-based memories..
They may, however, simply be choosing or avoiding the most familiar object.
Which really has nothing to do with episodes
There is a distinct difference between
The difference between knowing a face is familiar
remembering that you saw this person last Saturday night at the bar in Harvard Square
But many animals likely demonstrate some kind of episodic memory in nature.
Nest parasites, like cuckoos, must keep track of the location and state of nest-building and egg laying in their hosts
so as to know when to drop their eggs
And because recently researchers found that cuckoos will destroy the nests of hosts who dump their eggs
Some connections might be made in terms of personal experience for both hosts and cuckoos
But probably the best example are food-caching birds
As we learned last lecture
They store thousands of food items in the autumn in thousands of locations, and recover them over the course of several winter months
And even though the scrub jays that are used by Clayton et al.
dont cache nearly as much as the nutcrackers and pinyon jays
they do cache some food, and for times a bit longer than chickadees
To connect to episodic memory:
To get clear data on such behavior, Clayton and Dickinson performed a series of experiments with scrub jays
Birds were allowed to cache wax-worms (perishable and preferred) and peanuts (non-perishable and liked) in visually distinct sites in the laboratory
Sites were ice cube trays with Lego blocks in varied patterns
Birds were divided into two groups, Degrade and Replenish
Birds in the Degrade group were given the chance to cache peanuts and wax worms in two different trays
and then recover at both 4 hours and 124 hours, and to learn that the worms would be horrid after 124 hours
They then were tested by being allowed to cache and recover the two different foods at the two intervals
But now the experimenters removed all the food items before recovery,
so birds couldnt possible smell the degraded worms.
The researchers found that after 4 hours, the birds preferred to visit the sites where they had cached the worms
but went to the peanut sites after 124 hours
suggesting that they knew precisely what was where and the time delay
But maybe the birds just remembered which caches they had already emptied
Or maybe they just more rapidly forgot worm sites
i.e., maybe some evolutionary rather than memory process was at work
So the researchers worked with the Replenish group
who never learned that the worms degraded
because the researchers put in fresh ones before allowing recovery at 124 hours
When tested with all food removed, these birds preferentially went to the worm sites, at both 4 and 124 hours
Thus the behavior of prioritizing which food to recover was learned, and not some genetic instinct
And was nice preliminary evidence of episodic-like memory
But this design still didnt examine whether the birds in Degrade
understood more than time makes worms decay
i.e., just that something about the time at which they stored the various foods was important
So now the birds could store both foods (nuts and worms) in one tray at one time
then a few days later could again store both foods in another tray
and after a short interval after the last caching, were allowed to recover
So, in order to get worms and nuts appropriately (avoid yucky worms)
they had to remember which tray was cached when
And not just better memory for nicer food
And they succeeded on that task
To tease this out even further, the researchers designed another experiment
in which the birds got to store the different foods at different times,
and then recover them at the same later time
Test: predict worms
Peanut, then Worm
Worm, then Peanut
Test: predict nuts
Birds with experience with degraded worms
chose worms at a significantly higher rate than peanuts when the worms were cached last
and peanuts at over twice the rate when the worms were cached early and likely degraded
Interestingly.and critically.birds with no experience with degraded worms chose the worms in both cases
that is, they just chose on preference
not as to what was likely to have happened to the worms
But what if the birds were trained on something totally counter-intuitive.
That worms were yuckky after a couple of hours
But somehow were ok after a few days?
Test: predict nuts
Peanut, then Worm
Worm, then Peanut
Test: predict worm
And the birds acted as predicted
functioning on the basis of what they had learned about worms
And it wasnt what they learned during the test phase
because they acted OK from trial 1
Moreover, the birds behavior was not just a matter of familiarity of the tray
that is, somehow associating tray with the issues
because the tray exposure was the same for each type of caching
The switch by the birds in the Degrade group
requires the birds to recognize a particular cache site
in terms of both its content and the relative time that has elapsed between caching and recovery
The birds HAD to recall information about
Toothe information was acquired on the basis of a single, trial-unique personal experience
That is.something getting quite close to episodic memory
Note the result cannot be explained by the simple rule
search the side of the tray in which food was stored most recently, regardless of food type.
birds couldnt use recency, because each food was cached at the same time in different trials
Researchers next wanted to make sure that the birds remembered
not only which sites have been depleted (data from experiments by Kamil and Balda that we discussed)
but also exactly WHAT was recovered
not just the choice of spoiled vs. unspoiled,
but also what might be more appealing at a particular time.
which relies on a very personal memory for what has been consumed recently
not just always go for X if its fresh
So, they tested whether what they fed the birds just prior to recovery would affect what they recovered.
and also made them remember where the different foods were stored in two different trays.
The trials are quite complicated!
(1) cache P in one side of each tray
(3) 3 hrs later, allowed to recover P from one tray and K from the other
A few mn
(2) cache K in one side of each tray
(4) Prefeed bird P, see what it does
Assumption is that bird that is full of peanuts
will choose the kibble, and remember where the kibble still was
even after additional time delay
Of course, during test, all food was removed by the experimenter to prevent any odor or disturbance cues
So birds should not rely on a scent or preference for food
but on satiation
Note that birds have to INTEGRATE information as to
9 of 12 birds upheld the prediction of feeding from the tray
that held food different from what it had just eaten
Birds made twice as many searches to the place where there was intact different food
as to where there was intact same food
birds made very few searches to depleted sites
And this did not depend on training about good or bad types of food
Three main conclusions from these data
Now, those of you who have been mulling about the peanut/mealworm study may have realized a possible flaw in the experimental design
Maybe the birds just have a differential memory for remembering stuff that spoils versus stuff that does not.
rather than remember when stuff was cached
In other words.
Birds that experienced degradation not only learned something about what is good to eat when
But also that, in general, its best to remember where peanuts are stored because they are a more reliable source of food
So dont clog memory with information about degradable mealworms.
Thus researchers had to figure out a way around that problem..
They did so by replicating the experiments with crickets instead of worms
Crickets degrade more slowly than worms,
so the researchers could see if birds progressively recovered crickets more slowly as the days went by
Thus, if birds were just not bothering to remember where a degradable food was cached
and they didnt like the degradable crickets as much as the degradable worms
but they liked the crickets only as much as the peanuts
they should start skipping the crickets immediately
If, instead, they remembered crickets as well as peanuts,
and when they cached
they should go for the cricket for the first day or so after caching and stop after about 3 days
Interestingly, almost none of the 6 birds involved went for peanuts on the first three days
By day 5, all 6 birds went to peanuts first..
The birds seemed to make a categorical decision as to change their behavior around day 4
Thus, it could still be that the birds were not directing a lot of resources to remembering crickets
or had some kind of intermediate range memory in which they stored information about degradable food
compared to long-range memory for nondegradable items.
So, the idea was to give the Degrade birds staggered caching and then see what the birds did on progressive days.
Cache day 1
Cache day 2
Cache day 3
Recover day 4
Recover day 5
Recover day 6
This time, tho, some birds were givendegraded crickets on Day 3
Before, remember, crickets were still pretty good on Day 3
If they really remembered what they cached and when
they would go for peanuts on day 6 in the last tray that was cached
only if they experienced degraded crickets on day 3
But if they were the group that had not had rummy crickets on Day 3 for the first two trays,
they should still go for crickets on day 3
because crickets were usually good until day 4
And, to make sure that the birds just now didnt decide that any cached crickets were worthless,
they were given good crickets on a Day 1 recovery trial
The idea was to see if the researchers could tweak the encoded information.
Showing that the information was indeed encoded
and not just that crickets were relegated to a specific memory bin
Most of the birds that got degraded crickets on Day 3 for trays 1 and 2 did search for peanuts on Day 3 for tray 3.
In general, the birds had encoded some kind of semantic memory about the state of crickets over time (i.e., that they degrade in a few days)
And then integrated this information with respect to where and when they had cached the crickets..
And they were able to update their general semantic knowledge of the world based on new information.
These data fit with the ethology of the scrub jays, who cache items under varying conditions.
and thus must determine the effects of, e.g., a hot spell on their caches
So, these data suggest that birds fulfill the criteria for episodic memory:
What, When, Where
But is this behavior identical to that of humans, as Suddendorf and Busby question?
We still do not know.
We dont know if the birds are simply reacting on the basis of semantic memory.
That is, on some version of this is just how the world works
versus some understanding that the situations are specific to themselves or allow them to project ahead.
Specifically, one can argue that the birds have each remembered a setalbeit a very complex setof facts and integrated these facts.
Thus demonstrating an extremely high level of complex cognitive processing
What is needed is to show that the animal remembers a specific episode
that is, where I was and what I was doing on September 11, 2001.
And not just a collection of facts that allow me to solve a problem
because what you were doing was dissociated from the events
Is there any way to get around that problem?
Possibly, by designing experiments in which a bird has to separate out its own experiences from that of other birds..
This strategy isnt perfect
because it still cant separate out the bird as a dispassionate observer from what is going on around it
But it is a start in the right direction
Researchers realized that scrub jays will steal from one anothers caches..
And that as a precaution, some jays will re-cache items
if their initial caching is observed by other jays.
So maybe researchers could work with that knowledge to see if they could personalize the birds memories
And also see if the birds could project their knowledge to the future
So, one group of birds watched another group cache, and then was allowed to pilfer.
Another group of birds never had this experience of being allowed to pilfer.
although they could watch caching
Both sets of birds were then allowed to cache in the presence of a competitor.
Both sets of birds were then given the opportunity to re-cache their hoards in private.
Pretty much only those that had pilfered other birds stores did re-cache
These data begin to suggest that birds do have memories that, if not specific to themselves,
at least suggest some sense of attribution to others.
Specifically.if I pilfered another birds hoard whom I watched, someone watching me may pilfer mine
Of course, one can always come up with alternative explanations.
Something a bit less mentalistic.any ideas???
Wellit seems that birds are acting as tho if they do something, all other birds will do the same thing.
that is, generalization
Such behavior does not really impute intentionality or personal experience.
What one would really have to show is that birds recognize specific individuals
and that these individuals are indeed pilferers
and react only in the presence of a bird that they have reason to expect will pilfer their food.
not simply react on a generalized basis
Testing this with respect to actual knowledge is difficult
But possibly testing it with respect to the birds having some idea
of what a potential pilferer may or may not be able to see might work
Some evidence exists for ravens
If other ravens are around
the storer will wait until those birds are distracted or have moved away
or will make false caches and false recaches
or store behind barriers
But, again, are the birds acting as if they really have a Sally-Ann sense of theory of mind
or are they also working on a sort of generalization of this is what should be done in this circumstance
without fully knowing why?
One can set up an experiment in which the storer sees who is stealing some of its cache
And determine if the storer will then respond in a way that suggests it knows something about this specific competitor
compared to how it treats food stored in private
So, storers had three trials where they could store in private
And three trials where they knew that one potential pilferer was watching
In each case, two trays were available, at different ends of their cage
The same potential pilferer saw all three trials
Interestingly, storers also sometimes cached outside of the trays, but experimenters removed those caches
Experimenters also noted how much was eaten
Then, in the presence of the storer, the pilferer was given one tray
while the other tray was out of reach of both birds (trays marked w/ Legos)
And pilferers could eat a cache or re-cache it in same or different site
In contrast, after the storer had stored in private
the trays were observed by both birds, but no one could reach either tray
thus observation time for all four trays was equal
Then the storer, now in private, got back two trays
What did storer do?
For whatever reason, birds cached a bit more when being observed
Pilferers got about 50% of what had been stored when they could access a tray
But overall, about the same number of items were left for recovery remember, 3 unpilfered trays
The original storers didnt differ in how much they ate or checked their caches when they got the trays back in the two conditions
But they did re-cache more items in the trays that the pilferers had observed than in trays stored in private
They discriminated the particular trays that the pilferer had observed
And, interestingly, they took food out of the observed trays more often
Stuffing it in sites in the cage that could not be accessed by the pilferer at all
They also ate a lot more after being in trials in which the observer ate some of their caches
Here they werent just observed, but actively saw their caches being pilfered,
which is a fairly strong condition
In a separate study on other cachers, the Grey jay
This species would cache in front of a nutcracker because nutcrackers dont pilfer
but wouldnt cache in front of Stellar jays that do
None of these observational studies, however, deal with social effects
In the wild, many of these birds have social hierarchies that affect who has access to food
And may affect caching if good sites are limited
In a 2005 paper (Dally, Clayton, and Emery, Behaviour) looked at caching in an aviary in which there existed a social hierarchy
They found that the alpha pair would cache more than the beta pair
And that the alpha pair knew where each other cached and didnt care who recovered the cache
But the alpha pair seemed to defend their caches from the other birds
And when the alphas were removed, the betas then became the alphas
And took over the caching
Probably the subordinate birds knew they could not defend cache sites from the domiants
And basically just gave up on it
The effect might, however, have to do with being in an aviary
In the wild, the subordinate might need to cache more.
Here, the birds were basically well-fed and limited only slightly
Might be a totally different story in the wild w/ limited food available and the need to steal and cache
In yet another 2005 study (Animal Behaviour, Dally, Emery and Clayton) looked at positional factors in social settings
Here birds had a choice of where to store under conditions of being watched or not
sites were either close or far
If the bird had a general theory of just storing as far away from any possible observer
it would always use the far tray
Even if, in this case, the bird saw a screen being pulled down between it and a neighbor
If, however, it was sensitive to the condition of not being viewed
It should cache randomly in the two possible trays
The expt would show whether the bird was really aware of what the other bird could see
Now, this time the birds did not see anyone eat their food before the trays were removed
And they had successive trials so that they could learn what was happening to their hoardes
Which could all affect results
Again, for whatever reason (maybe fear of pilfering?) more food was cached when the birds were observed
And more food was put in the far tray when the birds were observed
suggesting that being observed had an effect
More food from the near trays were re-cached during recovery
Note that in the three hours in between caching and recovery, the screen was installed
So the storers had no idea what had happened to their trays
This condition is quite different from the one in which the storers saw the pilferers go to work
But the fact that the storers re-cached in private suggests that they knew that by caching all the stuff initially in the far space wouldnt be good either
Although this might suggest a sort of awareness, it may not
Could just be a simple idea of trying to avoid being near a conspecific during caching
And moving stuff away from the site it was at
What if you put non-cacheable food near the observer?
Would the bird go near the observer?
What if you played with the details of where the observer could be?
Birds happily ate in view of the observer
So they werent simply avoiding another bird
And their behavior during caching wasnt just due to avoiding the other bird
In terms of playing with the spacing, the experimenters set up the cages so that the birds could see one another in a limited way
and sometimes the observer was stuck on one side or the other
but the storer could move
Up close, pilferer can see only one site
No matter which side the potential pilferer was on
The storer cached in the far and unseen location
Interestingly, the caches were moved more often when the potential pilferer could move about
And sometimes the birds would cache in view and then recache several times
As if the confuse the pilferer as to where the item was.
Of course, the cacher could just be dithering here
Another problem is that the unseen site is always the further one away.
So we still dont know if the bird is acting upon what it thinks is the sight of the potential pilferer
Or it just wants to store in a far location.
Hence, another setup:
Seen and unseen are now equi-distant
Here jays could also cache totally in private
But, for whatever reason, they did not!
But when they were viewed, they put more in the out-of-sight tray than the one in-sight
But, this still could be some kind of out of sight, out of mind behavior.
What is needed is a system in which the pilferer can be seen by the storer
But the storer knows it is unseen
In yet another study, jays seemed to understand which bird was the observer and react appropriately
So, for example, it didnt care if its partner saw a cache (all in the family)
But did react if a dominant saw it cache
And the experimenters had problems because some birds stopped caching at all in the presence of observers
And birds would respond differently when birds that observed recovery were not birds that observed caching
So the birds seemed to attribute some knowledge to the observer
Or at least some memory of the observer and some memory of what might have happened in a related situation.
We are getting closer and closer to some kind of TOM
But its always difficult to know for sure
In terms of planning, however, a recent paper (Nature, 2007, 445: 919-921) does show clever behavior on the part of the jays.
First experiment involved planning for breakfast to see if birds would provision themselves depending upon their knowledge of the future
Later in day, learned that they could eat in B, too, but just powdered nuts
Then they were given whole nuts in B in the evening.
They put more nuts in C than A
Ostensibly to compensate for lack thereof
But to make sure the birds werent simply associating caching and hunger
The experiment was repeated with nuts in A and kibble in C
And then they were given both nuts and kibble in B
And they put kibble in A and nuts in C
Showing some level of planning
Overall, the data demonstrate incredible abilities that were not thought to exist in any nonprimate other than humans and apes, much less in a bird!
Which, of course, brings up the monkey paper..what about other nonprimates?
Now, of course, rhesus monkeys do not normally cache
It does not make ecological sense
Most of their food is perishable
And most of the time it is not limited in any real way
So that the idea of worrying about caches is not likely
What may be true for monkeys, however,
is when a particular tree is fruiting
with respect to a particular location in the forest
Maybe this is less precise, or is seasonalbut could be episodic
So, although monkeys might not be the best possible subjects
There might be some reason why they might have something like episodic memory.
But will this task show it?
Monkeys had two different foods, one yummy and one less so
They could then go back after either a long or a short delay
Both foods were available after short delay; only less yummy food after a long delay
Would they learn, over training,
which food would be available after different delays?
Note how this differs from the jays
in what is available
Learning that something just disappears over time can be seen as just someone else eating it up if you dont get to it first
which is a subtle difference from something that has to do with time per se
And this task used some monkeys with hippocampal lesions
Now we know that the hippocampus involves spatial memory
But some areas of the hippocampus also seem to be involved with episodic memory
or possibly, as we saw in the last lecture, the connections between the hippocampus and other parts of the brain
One idea is that the where and when of the what-where-when may be closely tied together.
which might explain corvid success on both sets of tasks
So, monkeys had three foraging sites, one empty and two w/ foods of differing appeal
In training, the placement of the sites vary between, but not within, sessions
So monkeys had to forget between sessions (time not given) and remember within session placement
Although there is no report of within-session timing, it sounds as tho it was just a few minutes between runs.
Enough to pull monkeys to cage, cover w/ tarp, and re-bait the sites
We are not told the time delay between sessions
But this time delay is crucial
Because it sets up what delay the monkeys expect during which the expected sites have been altered!
The assumption is 24 hrs, given that food preference tests were given on a 12 trial/day basis
Monkeys were considered trained if they went to preferred food on the last 3 of 4 runs in a session.
They acquired what is known as a learning set
where they learn a task, and small variations on the task are learned successively more rapidly
Then they were tested.
They went into the room, had, again, a unique set of placements
had the chance to learn where the good food had been hidden
Both the controls and lesioned animals quickly learned to find treat
They also learned to avoid the less tasty food
So at least for these very short delays (a few minutes) within sessions
lesions had no effect on how well the animals remembered the two sites
Then they were tested after each of the study trials
Now with first 1 hr and then 25 hr delays
After a 1 hr delay, the monkeys were allowed into the site
Nothing had changed
Researchers noted which food was chosen first, but monkeys had to visit both sites to end a trial
After choosing the good food, the marker was removed to encourage choice of the second site
After 25 hrs, the monkeys went back into the experimental room
Nothing had changed in the placement, UNLIKE the previous training phase
Now, however, the good food was degraded to be yuckky
Now remember, each test trial consisted of 8 runs for the animals to learn where the foods were
Then one run an hour later and another run 25 hrs later
Each monkey had 30 of these test trials; not sure of the time between test trials
The experimenters expected that, with time, the monkeys would stop going to the preferred food after the 25 hr delay.
Remember, the jays learned that worms got yuckky after 4 days
But the monkey data didnt quite work out that way.
First of all, the monkeys didnt bait the areas themselves
Second, the sites changed initially from one day to the next
So after a 25 hr break, they may have expected the room to look different.
But now, sometimes, instead of seeing a different setup
which they had learned meant some trial-and-error choice
they saw a familiar pattern
How easy would it be to learn to ignore that????
We know that reversals are tough to acquire
And thats what the monkeys had to learn
So its likely there was a lot of interference going on
The lesioned monkeys more often examined unbaited sites
suggesting that they indeed had a poorer memory for the location of any kind of foodgood or bad
Toward the end of the paper, the authors admit that their training might have skewed the results
But they then discount the possibilityfor what seem to be odd reasons
The point is that the continued shifting of the sites for the animals during training blurs the situation
What IS an episode for them?
Unlike the Clayton studies, the monkeys were trained in a particular way, then tested with a glitch
and then were expected to figure out a correlation that was obvious to the researchers
but one that would not necessarily be obvious to an individual experiencing the task
Too, although Clayton couldnt put every control in every experiment
each experiment was a consistent whole, with no claims as to having controlled for every variable
Although one could see the work as really on large paper instead of several individual ones
These studies, of course, get into the realm of questioning what it is that the animals actually do know
and, of course, what they know that they knowi.e., metacognition
The suggestion is that metacognition is necessary for episodic memory
But do you need to know that you know that worms degrade?
Metacognition might be more important in the pilfering study
in terms of getting to a kind of theory of mind behavior
But none of these cases are clear
An animal might know what food is hidden where and whether it is still good
but not know HOW or WHY they know it
No reasoning would be involved, just some memory returning when visiting the area
Humans use language to declare this information
How can we get at it in animals that have no, or only very limited, interspecies communication skills?
Some studies purportedly allow monkeys to state if they do or do not remember the answer to a task
But in reality the animals reporting of uncertainty just allows them a third choice when a first choice doesnt surface quickly
It may not be any more conscious than any other type of choice they are given
Some researchersSuddendorfargue that episodic memory is really a means of future planning
using the past to project to the future
How would that help us understand animals?
Clearly the jays store for the future and monkeys do not
But that doesnt get to the idea of whether the jays behavior is a mixture of instinct and knowledge
or self-awareness of what is individually needed
Would seem that the jays actions in preparing for the contingency of another animal pilfering
might actually tell us more than the caching behavior itself
Wouldnt it be great to ask Jay A where he thought Jay B would search after A re-cached?
The implication is that Jay A thinks Jay B will look in the old site and be fooled
Otherwise why would Jay A re-cache?
But Jay A re-caches in sight of Jay B;
Such would seem to be a silly behavior, and thus researchers make an argument for confusion
But is Jay A really pondering such an outcome, or just kinda freaking out and maybe trying to find a slightly better hiding place?
or hoping B will lose interest?
The take-home message is basically the cleverness of the birds and the need for truly complicated experiments in order to uncover their cleverness