Episodic-Like Memory in Dogs: What Dogs Retain and for How Long

1. The Question Behind the Question

1.1 What Owners Mean

"Does my dog remember that?" is usually asked about something specific: yesterday's walk, a veterinary visit, a former home or a person associated with a frightening event. Underneath sits a larger question — whether a dog can retrieve information about a particular past event, rather than simply behaving differently because previous experience changed learning and expectation.

Those are different claims. A dog that stiffens outside a clinic may be responding to learned predictive cues. That response does not by itself show conscious recollection of a previous consultation, even if an event memory could also exist.

1.2 What Researchers Mean

Human episodic memory includes conscious recollection of personally experienced events and the subjective sense of mentally returning to them (Tulving, 2002). That experiential component cannot be established directly in a non-verbal animal.

Comparative researchers therefore use the more limited term episodic-like memory for behavioral evidence that captures particular features associated with episodic memory without claiming conscious re-experiencing. There is no single definitive animal test. Major approaches include what-where-when paradigms, incidental encoding followed by an unexpected question, and source-memory designs; each isolates different components and has different alternative explanations (Davies & Clayton, 2024).

1.3 The Canine Evidence Is No Longer One Paradigm

Dog research now includes several related lines: unexpected recall of another individual's action, recall of the dog's own recent actions, odor-based what-where-when tasks, delayed imitation over longer intervals, and a clinical-style episodic-like memory test used across age groups. These paradigms are not interchangeable, and success in one does not prove that all of them measure a single underlying memory system.

That distinction matters because "memory" is not one capacity. Recognition, semantic knowledge, working memory, conditioned responding, action memory and episodic-like retention can all produce behavior that an owner describes simply as "he remembers" (how a construct is made measurable).

1.4 Why the Weaker Term Matters

Episodic-like is not a dismissal of animal cognition. It is a guardrail. Behavioral experiments can test what information was encoded, how flexibly it can be retrieved and whether the animal appeared to expect a later memory test. They cannot directly establish the phenomenology of recall — whether remembering feels to a dog like mentally revisiting an event.

Accordingly, absence of evidence for conscious recollection is not evidence that dogs lack it. The correct conclusion is narrower: current methods do not allow that subjective component to be demonstrated directly (why behavior is not an identified inner state).

1.5 The Popular Version of the Question

Questions such as "Does she remember her first family?" or "Does he remember what happened as a puppy?" remain much harder than the laboratory tasks. They concern remote autobiographical content, not merely the ability to retrieve event information after seconds, minutes or hours.

They are not completely beyond investigation: owner-report work has collected naturalistic descriptions of event memories in dogs and cats, including events said to date back years (Lewis & Berntsen, 2020). But retrospective owner reports cannot verify the original event, isolate the memory mechanism or demonstrate conscious recollection. They are therefore descriptive evidence, not an experimental answer to the autobiographical-memory question.

2. The Problem of the Expected Test

2.1 Why Trained Recall Is Not Enough

A dog trained to fetch a named object or repeat an action on cue is clearly using memory, but this does not by itself establish episodic-like recall. Repeated training gives the animal reason to prepare for the request and may support stable rules or associations that do not require retrieval of a particular event.

This is why the distinction between expected and unexpected retrieval has become central in comparative memory research.

2.2 The Unexpected Question

One strategy is to arrange encoding so that the animal has no obvious reason to expect that the information will later be tested. If performance remains above chance when the memory request is introduced unexpectedly, that is evidence consistent with incidental encoding.

The word unexpected still needs care. Researchers cannot ask a dog what it expected. Experimental designs instead try to make one response highly practiced and another request surprising, and can use behavior around the switch as additional evidence. The result strengthens the incidental-encoding interpretation; it does not provide direct access to the dog's expectation state (Fugazza, Pogány & Miklósi, 2016b).

2.3 What-Where-When Was One Starting Point

A major comparative tradition began with scrub jays that recovered different food caches according to what had been stored, where it was hidden and how much time had passed (Clayton & Dickinson, 1998). This made three event features behaviorally testable without requiring verbal report.

But what-where-when is not synonymous with episodic-like memory. Later work has emphasized other features — including incidental encoding, source information and flexible retrieval — because no single behavioral signature maps perfectly onto the human construct (Davies & Clayton, 2024).

2.3a Why Odor Is the Right Material for the "What"

Building the content component out of odors plays to canine perception rather than against it, which matters when a negative result could otherwise reflect the modality instead of the memory (how olfactory information is processed).

It also creates the interpretive difficulty discussed later: an odor trace changes over time, so "when" and "how strong the trace is now" are not fully separable in the way they would be with visual stimuli.

2.4 There Is No Three-Item Gold Standard

It is therefore better to ask what a study demonstrates than whether it "passes" a universal episodic-memory checklist. Does the task show what information was retained? Was encoding incidental? Could a fixed learned rule explain performance? Were the event components integrated or merely available separately? Could the subject flexibly use the information in a novel probe?

Different dog studies answer different subsets of those questions.

2.5 A Failure Is Ambiguous

Failure on an unexpected test can reflect forgetting, weak encoding, distraction, motivation or task misunderstanding. That is one reason individual differences are difficult to interpret from a single trial. Memory is inferred from performance, and performance depends on more than memory alone.

3. Remembering What Someone Else Did

3.1 The Design

The best-known canine demonstration used the "Do as I Do" procedure. Dogs first learned a rule that allowed them to reproduce a demonstrated human action. They were then trained to lie down after watching demonstrations, making lying down the expected response. On critical trials they were unexpectedly asked to imitate the demonstration instead (Fugazza, Pogány & Miklósi, 2016b).

3.2 The Result

Seventeen dogs were tested. They reproduced previously observed actions after both one-minute and one-hour delays, although performance was lower after the longer interval. The authors interpreted the result as evidence that information about the observed event had been retained despite the absence of an expected imitation request at encoding.

3.3 What the Decline Does — and Does Not — Show

The decline across the one-hour delay is consistent with forgetting of incidentally encoded event information and supports the authors' episodic-like interpretation. It is not a diagnostic marker of episodic memory. Many kinds of memory become less accessible with time.

Nor would stable performance automatically imply a mere trained association. In a separate expected-recall study, dogs retained demonstrated actions for delays from 1 to 24 hours, with high imitative performance across the tested intervals (Fugazza, Pogány & Miklósi, 2016a). Delay effects therefore have to be interpreted within the design that produced them.

3.4 The Sampling Constraint

The unexpected imitation paradigm requires substantial prior "Do as I Do" training. These dogs are therefore a selected, highly trained population. The finding demonstrates the capacity in dogs that can perform the procedure; it does not tell us how often ordinary untrained dogs would succeed under an equivalent incidental-recall test.

3.5 Why One Hour Is Not a Memory Limit

One hour is the longest delay tested in this particular unexpected-recall paradigm. It is a boundary of that experiment, not a demonstrated upper limit on canine event memory.

Expected imitation has been shown after delays up to 24 hours, and other memory systems operate on still longer timescales. The correct statement is therefore specific: unexpected recall of incidentally encoded demonstrated actions has been demonstrated up to one hour in this paradigm.

4. Remembering What the Dog Itself Did

4.1 Own-Action Memory

Fugazza and colleagues extended the logic from observed actions to actions performed by the dog itself. Dogs were taught to repeat their own behavior on request and were then tested with actions that included spontaneous behaviors not deliberately elicited for the memory test (Fugazza et al., 2020).

4.2 The Result

Dogs could repeat recently performed actions after delays ranging from seconds to one hour, with performance decreasing as the delay increased. The study supports the idea that dogs can form retrievable representations of their own recent actions. The authors discuss this as evidence relevant to representations of the self, but the claim is narrow: memory for one's own behavior is not equivalent to human autobiographical self-awareness.

4.3 Independent Evidence From Another Laboratory

Scagel and Mercado (2022) independently trained dogs to respond to a "repeat" cue and found that they could reproduce recently performed actions, including actions that had not been specifically trained as repeatable responses. The paradigm was not an incidental unexpected-memory test, so it should not be treated as a replication of Fugazza et al. (2020).

Its value is different: memory for one's own recent actions is not confined to one research group or one exact procedure. The specifically episodic-like interpretation still depends on features such as incidental encoding and unexpected retrieval.

4.3a What the Replication Adds and What It Does Not

An independent laboratory reaching a comparable result with a different training route is the strongest thing that can happen to a small literature. Dogs that learned a repeat cue could reproduce recent actions, including ones never trained as repeatable and ones they innovated on cue, which the authors read as an abstract concept of repeating that generalizes across actions (Scagel & Mercado, 2022).

What it does not add is an unexpected-test design. The dogs knew the repeat request was coming, so the study speaks to flexible access to action memory rather than to incidental encoding (what canine cognition research can establish).

4.4 Spontaneous Actions Are Important — but Not Everyday Autobiography

In the 2020 study, some actions were produced spontaneously rather than requested by an experimenter. That reduces the concern that the dog encoded the behavior only because the laboratory explicitly marked it as important.

It still does not show that dogs later replay ordinary meaningful episodes, remember why they acted, or attach a narrative interpretation to the event.

4.5 Forgetting Is Consistent With the Interpretation

Performance decreased across longer delays in the own-action work. That pattern is compatible with episodic-like retention, but it cannot identify the underlying memory system on its own. Forgetting curves occur in many memory tasks.

5. What, Where and When

5.1 The Odor Paradigm

Lo and Roberts (2019) tested dogs with spatially and temporally unique sequences of odors. The tasks asked whether dogs could use information about which odor had appeared, where it had appeared and its temporal position in a sequence.

5.2 What the Study Shows

The dogs successfully used information corresponding to what, where and when across several experiments and responded to probe conditions designed to prevent a single simple spatial rule from solving every trial (Lo & Roberts, 2019).

The important limitation is explicit in the paper: the results do not establish whether what, where and when were bound into one integrated episodic representation or were stored and retrieved as separable pieces of information. The study therefore supports flexible use of all three components without settling the integration question.

5.3 What "When" Means Here

The temporal component concerned relative position and temporal information within the experimental sequence. It should not be translated into a claim that dogs remembered an absolute clock time or consciously experienced how long ago an event occurred.

Because odor was the principal stimulus material, the paradigm is particularly well matched to canine perception. That is a strength. It also means that conclusions should stay tied to the discrimination structure actually tested rather than being extended to autobiographical time.

5.4 Flexible Probes Reduce — but Do Not Eliminate — Alternatives

Changing the probe conditions makes rigid response rules less plausible and is an important feature of the study. It does not logically exclude every possible non-episodic strategy. This is one reason modern reviews argue for converging evidence across multiple paradigms rather than treating one task as decisive (Davies & Clayton, 2024).

5.5 Clinical Relevance: Aging

The translational motivation is no longer only theoretical. Sanches et al. (2022) tested 90 clinically healthy domestic dogs of different ages with a Canine Episodic-like Memory Evaluation Test. Older dogs were more likely to fail the task than younger adult dogs, which the authors interpret as age-related decline in episodic-like memory.

This does not turn the test into a stand-alone diagnosis of canine cognitive dysfunction. Performance can be influenced by many factors, and the study involved clinically healthy dogs. It does show that age differences have been tested and that episodic-like memory is being developed as a potential comparative and veterinary cognitive measure (what changes in aging dogs).

6. What This Evidence Does Not Establish

6.1 Not Direct Evidence of Mental Time Travel

Human episodic memory includes conscious re-experiencing. The dog studies provide behavioral evidence for event information and retrieval; they do not establish that subjective phenomenology. Equally, they do not establish its absence.

6.2 Not a General Measure of Time Perception

Remembering temporal order, retaining an event across a delay and estimating elapsed duration are related but different problems. The literature on interval timing and absence duration should therefore be kept separate (what is known about time perception in dogs).

6.3 Not a Necessary Explanation for Behavior Problems

A dog that becomes distressed near a veterinary clinic can be explained by associative learning from predictive cues without invoking episodic recall. That is a sufficiency argument, not proof that the dog has no event memory of the clinic.

For applied work, the useful principle is to start with mechanisms that can be observed and changed — triggers, predictions, reinforcement history and emotional state — rather than constructing an unverifiable autobiographical episode (why the mechanism changes the plan).

6.4 Recognition Is Not the Same as Episodic Recall

A dog that recognizes a person after a long interval demonstrates retention of information about that individual. Recognition does not tell us whether the dog also retrieves a specific past encounter as an event.

Long recognition intervals therefore cannot be used to infer long autobiographical recollection, just as the one-hour incidental-recall experiment cannot be used to rule it out.

6.5 Fear Conditioning and Episodic Memory Are Not Either-Or

A frightening event can produce durable associative fear learning without requiring episodic recollection. That is sufficient to explain many later responses to cues associated with the event (how fear memories are formed and changed).

But different memory systems can operate in parallel. A conditioned response does not prove the absence of event memory; it simply means episodic recollection is not needed to explain the observable behavior.

6.5a Not a Claim About Understanding Why

Retaining what happened is not the same as grasping why it happened. A dog that can reproduce an action it watched has stored the action, not an account of the causal structure behind it (what dogs understand about causes).

That distinction matters in applied writing, where "he remembers" often slides into "he understands", and from there into "he knows better". Each step adds a claim the evidence does not carry.

6.6 Anticipation Is Not Evidence of Episodic Foresight

A dog waiting near the door before a familiar walk can be explained by learned temporal and environmental cues. Such anticipation does not demonstrate episodic future thinking. It also does not prove that future-oriented representation is impossible; the ordinary observation simply does not distinguish the alternatives.

7. Other Memory Systems in Dogs

7.1 Deferred Imitation and Long-Term Declarative Memory

Expected recall can last substantially longer than the one-hour incidental test. In a deferred-imitation study, 12 dogs retained demonstrated actions over delays of 1, 2, 12 and 24 hours, with consistently high performance across the tested intervals (Fugazza, Pogány & Miklósi, 2016a).

That line of work goes back to the demonstration that dogs reproduce an action after a delay at all, interpreted at the time as evidence of declarative memory (Fugazza & Miklósi, 2014). All of it is deferred-imitation evidence, not an incidental episodic-like test. Its main relevance here is methodological: one hour must never be presented as a general upper limit on canine memory.

7.1a Imitation as a Trained Skill

Both main paradigms rest on a dog having learned an imitation or repetition rule first, which is itself a substantial piece of learning and part of the wider literature on what dogs acquire by watching (what dogs learn by watching).

That dependency has a consequence worth stating: any claim about memory drawn from these studies is a claim about dogs who can be trained to this standard, and the training itself may change what they attend to.

7.2 Working and Spatial Memory

Short-term retention of a baited location or recently seen object answers a different question. Strong working or spatial memory does not by itself establish memory for a unique personal event.

7.3 Associative Learning Remains Central

Associative learning can explain a very large share of everyday behavior and often provides the most directly testable account of why a dog reacts in a familiar context. Episodic-like findings do not invalidate associative mechanisms; the two can coexist and answer different questions (how learning is arranged in practice).

7.4 Semantic Knowledge Is Different Again

Dogs that retain object labels over long periods demonstrate semantic-like knowledge about names and objects, not necessarily event memory. Retrieving a named toy, repeating one's own previous action and choosing the odor encountered earlier are different tasks and should not be collapsed into a general claim about a "good memory".

7.5 Consolidation and Sleep

Sleep-dependent consolidation has also been studied in dogs with electrophysiological and behavioral methods. That work concerns stabilization of learned information and sits alongside the episodic-like literature rather than proving the same construct (how sleep and memory interact in dogs).

8. What Follows for Practice

8.1 Delayed Punishment Is Not Rescued by Better Memory

Even if a dog retains information about an earlier action, that does not solve the learning problem created by a delayed consequence. Memory for an event and understanding that a later aversive event is contingent on one specific earlier behavior are separate requirements.

That is the stronger reason not to use episodic-like memory as an argument for punishment "after the fact". The memory studies did not test delayed correction, causal attribution of punishment or moral evaluation (what punishment costs).

8.2 The Guilty Look Is Still Not Evidence of Guilt

Nothing in the episodic-like memory literature establishes that the familiar "guilty look" reflects recollection of a misdeed plus an evaluation that it was wrong. Studies of that expression point instead to the immediate social context and the owner's behavior (what the guilty look tracks).

8.3 What the Memory Findings Do Support

Dogs can retain event information that was not necessarily encoded for an expected future memory test. They can also remember recent self-performed actions and use what-where-when information flexibly in an olfactory task.

Those are substantive cognitive findings. They do not require the stronger claims that dogs review their life history, consciously relive specific events or remember only for the short intervals used in these experiments.

8.4 Training Timing Comes From Learning Research

Fast, clear feedback is important in training, but that recommendation comes primarily from learning, contingency and reinforcement research — not from the episodic-like memory experiments. The fact that incidental event information can be retained does not generate a rule that training must build on an event "within minutes".

8.5 What to Say About a Single Bad Experience

A single aversive event can produce lasting behavioral change through conditioning and other learning processes. It is therefore unnecessary to assume that the dog consciously replays the episode in order to take the resulting fear seriously.

At the same time, the current evidence cannot justify the opposite claim that dogs definitely do not retain an episodic-like representation of such an event. The mechanism behind an observed fear response and the full contents of memory are not the same question.

8.5a Working With What a Dog Predicts

Most of what households attribute to memory is predictive: the clinic route, the shoes, the bowl being lifted. Those chains are built by association and can be changed by systematic work on the predictive cues themselves (how predictive cues are worked with).

The practical value of keeping this separate from episodic memory is that predictive chains are treatable and a recollection is not. Naming the mechanism decides whether there is anything to do.

8.6 What to Tell a Household

Two popular extremes are unsupported. "Dogs live only in the present" is too simple: they can retrieve information about past events and their own recent actions. "Dogs remember their life story like humans" goes beyond the available evidence because conscious autobiographical recollection has not been demonstrated.

For a particular old home, previous family or puppyhood event, the honest answer is that current experimental paradigms do not establish whether that specific remote event is consciously recollected.

8.6a One Habit Worth Adopting

If incidental retention is real, the minutes around a training session are not neutral: what the handler does before the first repetition and after the last one is available to the dog whether or not it was meant as part of the lesson.

Nothing dramatic follows, and one cheap habit does. Film a session occasionally and watch the parts before and after the training rather than the training itself (what else the dog is learning meanwhile).

8.7 Three Claims Worth Correcting

"He knows exactly what he did wrong." Dogs can remember actions. That does not show moral evaluation, and it does not show that a later scolding is understood as contingent on one earlier behavior.

"She cannot remember her old home because experiments only reach one hour." Wrong inference. One hour is the maximum delay in a particular incidental-recall paradigm, not a measured limit of canine memory. Expected imitation has been demonstrated after 24 hours, and naturalistic owner reports describe much longer event memories, although those reports cannot verify the underlying mechanism.

"Dogs live entirely in the present." Also unsupported. The experimental record includes memory for observed actions, self-performed actions, temporal order and what-where-when information.

9. Aging and Individual Differences

9.1 Age Has Been Tested

Sanches et al. (2022) examined 90 clinically healthy domestic dogs with an episodic-like memory evaluation. Older dogs performed worse as a group, providing direct evidence that age-related differences in this kind of task exist.

This corrects an older gap in the literature. What remains open is how much of the decline is specific to episodic-like memory versus sensory, motivational, attentional or broader cognitive changes, and how well such tasks distinguish normal aging from canine cognitive dysfunction in clinical populations.

9.1a Why the Aging Result Matters Beyond Dogs

A behavioral test that detects age-related decline in a companion species is interesting for two reasons at once. It gives canine medicine a candidate measure that does not depend on owner questionnaires, and it gives human research a model species that ages in the same households and environments as people (what changes in aging dogs).

The caution is the same one that applies to every promising test: a group difference between older and younger dogs is not yet a diagnostic instrument for an individual animal, and the step from one to the other requires validation work that has not been done (Sanches et al., 2022).

9.2 Individual Performance Is Not a Personality Label

A dog that succeeds or fails on one memory task should not be labeled as generally having "good" or "poor" memory. Different paradigms impose different demands, and there is little evidence yet showing how strongly performance correlates across unexpected imitation, what-where-when, own-action and clinical-style tests.

9.3 What Owner Reports Add

Owner-report work has collected naturalistic accounts of what people believe their animals remember, including events said to lie years back (Lewis & Berntsen, 2020). Such material is useful for mapping what people claim and for generating hypotheses, and it cannot verify the original event or identify the mechanism.

Read that way it is a legitimate part of the literature. Read as evidence that dogs recall their distant past, it is exactly the move this article argues against (why the measure decides what a claim can say).

10. Summary at a Glance

Episodic-like memory is not one single test. Comparative research uses multiple behavioral approaches because conscious re-experiencing cannot be established directly in non-verbal animals (Davies & Clayton, 2024).

Dogs can unexpectedly recall observed actions. In the Do-as-I-do paradigm, incidentally encoded demonstrations were reproduced after one minute and one hour, with lower performance after the longer delay (Fugazza, Pogány & Miklósi, 2016b).

Dogs can remember their own recent actions. This has been shown with more than one experimental approach and more than one research group (Fugazza et al., 2020; Scagel & Mercado, 2022).

Dogs can use what, where and when information. Lo and Roberts (2019) demonstrated flexible use of those components, but did not establish that they were bound into one integrated episodic representation.

One hour is not a canine memory limit. It is the longest delay used in the unexpected incidental-action paradigm. Expected imitation has been demonstrated after up to 24 hours (Fugazza, Pogány & Miklósi, 2016a).

Age differences have been measured. A study of 90 clinically healthy dogs found poorer episodic-like task performance in older dogs (Sanches et al., 2022).

Conscious recollection remains unresolved. The experiments establish behaviorally accessible event information, not the subjective feeling of mentally returning to the past.

10.1 Situations in Practice

The dog that "knew" it had done something. Memory for an action is not equivalent to knowing that the action violated a human rule. The guilty-look literature should be used to interpret the expression, not episodic-memory findings (what the guilty look tracks).

The rescue dog with an unknown history. Current behavior can justify a fear-treatment plan without inventing a specific past episode. A historical event may or may not have occurred; the dog's present triggers, responses and learning contingencies are the evidence available for intervention.

The dog that reacts before reaching the clinic. Predictive environmental cues and associative learning are sufficient to explain the pattern. Whether the dog also retrieves a particular prior visit is a separate, currently unresolved question.

10.2 What Would Move the Field

Three studies would settle most of the open questions. Testing pet dogs without months of prior imitation training, in a paradigm that does not depend on it. Extending delays beyond an hour to find where retention actually fails rather than where the method stops. And running two paradigms on the same animals to see whether they measure one capacity or several.

Reviews of the field make a related point at the conceptual level: no single task is decisive, and confidence should come from converging evidence across designs that fail in different ways (Davies & Clayton, 2024).

11. Research Gaps and Critical Appraisal

The paradigms measure overlapping but not identical things. Unexpected imitation, own-action repetition, what-where-when and the CEMET task have not been shown to load on a single common canine episodic-memory factor.

Integration remains unresolved in the odor work. Dogs used what, where and when information, but the study did not establish that these features were bound into one integrated event representation (Lo & Roberts, 2019).

Unexpected-recall samples remain small and highly trained. The strongest incidental-action paradigm depends on prior Do-as-I-do training, limiting generalization to ordinary untrained pets.

Longer incidental delays are largely untested. The one-hour result is informative but does not locate the point at which event information becomes unavailable.

Remote autobiographical content remains unresolved. Owner-report studies suggest that households observe behavior they interpret as memory for events years earlier, but these reports cannot verify the event, exclude cue-based retrieval or establish conscious recollection (Lewis & Berntsen, 2020).

Aging has been studied, but clinical discrimination remains incomplete. Older healthy dogs performed worse in Sanches et al. (2022), yet further work is needed to separate normal aging, canine cognitive dysfunction and non-memory contributors to task performance.

Neural evidence specific to canine episodic-like retrieval remains limited. The dog literature reviewed here is primarily behavioral; it does not yet map these paradigms onto a validated canine neural signature of episodic recollection.

Emotion and salience are major open questions. The experiments mostly use neutral or structured tasks. Whether emotionally important events are encoded, retained and retrieved differently is much closer to what households ask and remains poorly resolved.

Conscious re-experiencing remains outside the reach of these methods. No amount of success on a behavioral proxy should be silently converted into evidence that a dog experiences remembering in the same way a human reports it.

12. Conclusion

The strongest conclusion is neither "dogs live only in the present" nor "dogs remember their lives like humans." Dogs have demonstrated several behavioral capacities that overlap with components of episodic memory: unexpected recall of incidentally encoded observed actions, memory for their own recent actions, flexible use of what-where-when information and age-related differences in an episodic-like task. At the same time, the paradigms do not establish conscious re-experiencing, and the odor study does not show that what, where and when were necessarily integrated into one single memory representation.

The time limits must also be read correctly. One hour is the longest delay in the unexpected incidental-action paradigm, not an upper boundary on canine memory; expected imitation has been demonstrated after 24 hours. For remote personal events such as a former home or puppyhood experience, current experimental evidence is insufficient to say whether a dog consciously recollects that specific episode. In practice, this distinction matters because present behavior can usually be assessed through current triggers, learning history and emotional responses without inventing an unverifiable narrative about the dog's past.

Key Insights (Takeaways)

• Episodic-like memory is a behavioral research category, not a direct demonstration of conscious recollection.

• Dogs have shown unexpected recall of incidentally encoded observed actions after delays up to one hour (Fugazza, Pogány & Miklósi, 2016b).

• Memory for the dog's own recent actions has been shown with different methods and research groups (Fugazza et al., 2020; Scagel & Mercado, 2022).

• Dogs can use what, where and when information, but the available odor study did not establish that all three were integrated into one episodic representation (Lo & Roberts, 2019).

• One hour is not a general memory limit: expected imitation has been retained for up to 24 hours (Fugazza, Pogány & Miklósi, 2016a).

• Age-related differences have been demonstrated in a 90-dog episodic-like memory study (Sanches et al., 2022).

• Remembering an action does not mean understanding a later punishment as contingent on that action, and it does not establish guilt or moral evaluation.

• For remote life events, the scientifically accurate answer is usually that current methods do not establish whether a specific event is consciously recollected.

References

Clayton, N. S., & Dickinson, A. (1998). Episodic-like memory during cache recovery by scrub jays. Nature, 395(6699), 272–274. https://doi.org/10.1038/26216

Davies, J. R., & Clayton, N. S. (2024). Is episodic-like memory like episodic memory? Philosophical Transactions of the Royal Society B: Biological Sciences, 379(1913), 20230397. https://doi.org/10.1098/rstb.2023.0397

Fugazza, C., & Miklósi, Á. (2014). Deferred imitation and declarative memory in domestic dogs. Animal Cognition, 17(2), 237–247. https://doi.org/10.1007/s10071-013-0656-5

Fugazza, C., Pogány, Á., & Miklósi, Á. (2016a). Do as I … Did! Long-term memory of imitative actions in dogs (Canis familiaris). Animal Cognition, 19(2), 263–269. https://doi.org/10.1007/s10071-015-0931-8

Fugazza, C., Pogány, Á., & Miklósi, Á. (2016b). Recall of others' actions after incidental encoding reveals episodic-like memory in dogs. Current Biology, 26(23), 3209–3213. https://doi.org/10.1016/j.cub.2016.09.057

Fugazza, C., Pongrácz, P., Pogány, Á., Lenkei, R., & Miklósi, Á. (2020). Mental representation and episodic-like memory of own actions in dogs. Scientific Reports, 10, 10449. https://doi.org/10.1038/s41598-020-67302-0

Lewis, A., & Berntsen, D. (2020). Pet memoirs: The characteristics of event memories in cats and dogs, as reported by their owners. Applied Animal Behaviour Science, 222, 104885. https://doi.org/10.1016/j.applanim.2019.104885

Lo, K. H., & Roberts, W. A. (2019). Dogs (Canis familiaris) use odor cues to show episodic-like memory for what, where, and when. Journal of Comparative Psychology, 133(4), 428–441. https://doi.org/10.1037/com0000174

Sanches, F. J., de Melo, J. C., Ferreira, S. P., Trinca, L. A., Arambula, O. G., Padovani, F. H. P., Schelp, A. O., & Amorim, R. M. (2022). Aging-related episodic-like memory decline in dogs. Behavioural Brain Research, 422, 113762. https://doi.org/10.1016/j.bbr.2022.113762

Scagel, A., & Mercado, E. (2022). Do that again! Memory for self-performed actions in dogs (Canis familiaris). Journal of Comparative Psychology, 136(2), 121–139. https://doi.org/10.1037/com0000310

Tulving, E. (2002). Episodic memory: From mind to brain. Annual Review of Psychology, 53, 1–25. https://doi.org/10.1146/annurev.psych.53.100901.135114