Overimitation in Dogs: Copying Actions That Do Nothing

1. What Overimitation Is — and Is Not

1.1 The Core Definition

Overimitation is the reproduction of an observed action that makes no causal contribution to a demonstrated physical outcome. A demonstrator performs an unnecessary step and then the action that actually opens, moves or retrieves something; the observer later reproduces the unnecessary step as well.

The term describes a behavioral pattern. It does not, by itself, identify the mental process behind that pattern. The same outward response could in principle arise from action matching, local or stimulus enhancement, learned attention to human hands, uncertainty about causality, social motivation, exploration or more than one of these processes.

1.2 Why the Comparison With Children Became Important

Human children often reproduce causally irrelevant steps even when the task is transparent, and the behavior can be strengthened by social or conventional framing. This has made overimitation part of the literature on cultural learning and norm-sensitive copying (Hoehl et al., 2019).

Great apes usually omit transparently unnecessary actions when the causal structure is visible, although their performance depends on task transparency and design. It is therefore too absolute to say that great apes “never” reproduce irrelevant actions. The more defensible comparison is that robust overimitation of clearly unnecessary actions is characteristic of human children and has generally not been demonstrated in great apes to the same extent.

1.3 What Makes Dogs Interesting

Dogs live in a social environment saturated with human movement, cues and reinforcement histories. They also learn socially from humans and other dogs. Finding that some dogs reproduce irrelevant human actions therefore raises a useful comparative question: is this simply attention and learned action matching, or does the social relationship with the demonstrator change what gets copied?

The research does not yet provide one final answer. What it does provide is a sequence of increasingly specific tests of caregiver familiarity, relationship quality, social priming, causal knowledge, human attention and development (what dogs learn by watching).

2. The Dot-Touching Paradigm

2.1 The Basic Task

In the best-known canine paradigm, a wooden apparatus contains a sliding door that gives access to food. Before operating the door, the demonstrator touches one or two colored dots positioned away from the mechanism. Touching the dots has no physical role in opening the compartment. The dog then receives an opportunity to interact with the setup.

The irrelevant action and the goal-relevant action are scored separately. That separation is important because a dog can copy the functional action without touching the dots, touch the dots without solving the task, do both, or do neither.

2.2 Measurement Does Not Require a Theory of Mind

Researchers can objectively record whether a dog touches the demonstrated target and how accurately it reproduces an action. That measurement does not require knowing what the dog believes. Interpretation is a different matter: calling the behavior overimitation is strongest when the irrelevant nature of the action is transparent to the subject, because otherwise causal misunderstanding remains a competing explanation.

2.3 Why Baseline and Control Conditions Matter

A colored target that has just been touched by a human may become interesting even if the dog is not copying the action itself. Baseline exploration and no-demonstration controls are therefore crucial. They help distinguish socially induced action matching from spontaneous contact with the apparatus, but they still cannot reveal the dog's internal reason for touching it (how behavioral constructs are operationalized).

2.4 Frequency and Precision Are Different Outcomes

Several papers report both whether a dog ever touched an irrelevant target and how precisely or repeatedly it copied. A headline such as “half the dogs copied” can therefore coexist with a low overall frequency across trials. This distinction is important throughout the literature: occurrence, accuracy and repetition are not interchangeable measures.

2.5 The Demonstration Is Performed Dog-Style

A detail of the procedure is easy to miss and worth stating: the human demonstrator performs the sequence on all fours, touching the dots with the nose and operating the mechanism in a dog-like manner. The point is to make the demonstrated action one the dog could plausibly reproduce with its own body, rather than a hand movement it has no equivalent for.

That design decision also constrains who can be tested. In the at-home study, guide dogs and service dogs were excluded because their caregivers could not perform the crawling demonstration, which is why that sample consisted of signal dogs (Gerwisch et al., 2025). Sampling in this literature is shaped by the apparatus as much as by the question (how measurement decisions shape findings).

3. The Foundational Findings

3.1 Dogs Copied Their Caregiver

Huber et al. (2018) reported that roughly half of the tested pet dogs reproduced a causally irrelevant action demonstrated by their own caregiver at least once, although the overall frequency of irrelevant-action copying was low. The finding was notable because the action had no role in obtaining the reward.

Dogs copied the irrelevant component whether it had been demonstrated alone, before the functional action or after it. This shows that the functional action did not simply eliminate all copying of the unnecessary one. It does not, however, prove that dogs understood the causal structure of the apparatus.

3.2 A Stranger Produced Far Less Copying

Huber et al. (2020) repeated the dot-touching task with an unfamiliar experimenter as demonstrator. Only very few dogs reproduced the irrelevant action, markedly fewer than in the caregiver condition from the earlier study.

This is strong evidence that demonstrator identity or familiarity matters in this paradigm. It is not, on its own, proof that attachment quality or affiliation is the causal mechanism. A caregiver differs from a stranger in many ways at once: familiarity, reinforcement history, predictability, attention history, cue meaning and social relationship.

3.3 Relationship Quality Did Not Show the Predicted Correlation

Huber, Kubala and Cimarelli (2022) directly tested whether relationship quality predicted overimitation. Fifty-eight dogs entered the final analysis. Twelve touched one irrelevant dot and only three reached the highest irrelevant-action copying score by touching both.

The key result is often overstated: the overall correlation between relationship-test scores and overimitation was not significant. The few dogs with the highest copying accuracy showed high levels of behaviors such as gazing and synchronization, but this was a small descriptive subgroup rather than evidence for a general dose-response relationship between “better bond” and more overimitation.

The study therefore supports a hypothesis worth testing, not a demonstrated causal chain.

3.4 Familiarity Is Better Supported Than “Bond Strength”

By 2022 the safest summary was already narrower than many secondary accounts suggested: dogs behaved differently when the demonstrator was their caregiver versus an unfamiliar person, but individual measures of relationship quality did not reliably predict copying within the caregiver group.

That distinction remains important in 2026. Familiarity and relationship quality are not the same variable.

4. The Independent Canid Study

4.1 A Different Design Produced a Different Picture

Johnston, Holden and Santos (2017) investigated irrelevant-action copying in dogs and dingoes with a different apparatus and a standardized human experimenter. Their studies did not produce the same pattern later reported in the caregiver dot-touching work.

This cannot be reduced to “they used a stranger”. The studies differed in several ways at once: apparatus, procedure, species comparison, demonstrator arrangement and the question being asked. Demonstrator familiarity is one plausible contributor to the divergence, not an isolated causal explanation.

4.2 What the Comparison Can and Cannot Say About Domestication

A dog-dingo comparison addresses evolutionary history; caregiver-versus-stranger comparisons address individual experience with a particular human. These levels should not be collapsed. Evidence that a caregiver changes performance does not show that domestication is irrelevant, and a domestication comparison cannot tell us which feature of a caregiver relationship matters.

4.3 The Evidence Base Is Still Concentrated

Most direct positive studies of dog overimitation come from the same Vienna research program. The program has become methodologically broader over time, but independent replications using closely matched procedures remain limited. That concentration is an important limitation when generalizing to dogs as a species.

5. Social Motivation: Tested, but Not Demonstrated as a Single Cause

5.1 Social Priming Did Not Increase Overimitation

Mackie and Huber (2023) tested whether briefly priming the dog-caregiver relationship or attention to the caregiver would increase copying. There was no significant main effect of the priming conditions on irrelevant-action copying.

A null result does not prove that social motivation is irrelevant. It shows that these particular manipulations did not produce the predicted change. Without an independent measure showing how strongly the prime changed the targeted state, failure to change copying cannot distinguish “wrong mechanism” from “weak manipulation”.

5.2 Post-Goal Copying Became an Important Clue

In the 2023 study, irrelevant-action copying often occurred after the dog had already achieved the physical goal. This timing is difficult to reconcile with a simple account in which the dog touches the irrelevant target only because it believes that doing so is required to obtain the reward.

It is consistent with an additional non-instrumental motivation to reproduce an observed action. That interpretation remains an inference, not a directly measured “social goal”.

5.3 Being Watched Did Not Increase Irrelevant-Action Copying

Mackie, Trehorel and Huber tested 63 family dogs while the caregiver either watched the dog or turned away during the dog's trials. Caregiver attentional state had no significant effect on the accuracy of irrelevant-action copying (Mackie, Trehorel & Huber, 2025; published online in 2024).

The manipulation did affect another part of performance: dogs copied the relevant sliding-door action more accurately when watched. This is useful precisely because it shows that social attention can matter to task behavior without producing the predicted change in overimitation itself.

5.4 What the Social Account Looks Like After These Tests

Caregiver demonstrations reliably matter compared with unfamiliar or absent demonstrators in some paradigms, but no simple variable has yet explained the effect. Relationship scores, short social primes and the caregiver's immediate visual attention have not produced a consistent “more social connection = more irrelevant copying” pattern.

6. Causal Understanding: The 2024 Test

6.1 Why Causal Misunderstanding Matters

A dog may copy an irrelevant action because it does not realize that the action is irrelevant. If so, giving the dog prior experience with the task should reduce copying once the dog has learned which action actually produces the reward.

6.2 Prior Experience Changed Timing, Not Overall Frequency

Mackie and Huber (2024) tested 81 family dogs in a new task using a ball reward. Dogs with prior experience of the apparatus were compared with dogs without that experience and with a no-demonstration control group.

Prior experience did not simply eliminate irrelevant-action copying. Instead, it shifted when copying occurred. Dogs that already knew the task tended to obtain the reward first and interact with the irrelevant element afterwards; dogs without prior experience were more likely to copy the irrelevant action before achieving the goal.

6.3 What the Study Supports

The timing result is difficult to explain by causal misunderstanding alone. A dog that already possesses the ball and then reproduces the irrelevant demonstrated action no longer needs that action to explain how the reward was obtained.

The authors interpret this pattern as evidence compatible with a secondary social goal. A more conservative formulation is that task knowledge changes the organization of copying and leaves room for a non-instrumental motivation. The study does not directly identify what that motivation is.

6.4 Learning Across Trials Still Matters

In the same study, irrelevant-action scores decreased across trials. Dogs therefore did not behave as fixed “overimitators” or “non-overimitators”; experience with the task changed performance. This is another reason not to treat overimitation as a stable personality trait.

7. Relationship Quality Tested Again: Assistance Dogs in 2025

7.1 A Stronger Relationship Did Not Mean More Overimitation

Gerwisch et al. (2025) tested 74 dogs at home: 25 assistance dogs, 24 family dogs and 25 family dogs in a no-demonstration control condition. Caregivers of the two demonstration groups also completed the Monash Dog Owner Relationship Scale.

The relationship scores differed significantly between assistance- and family-dog groups, with the assistance-dog caregivers reporting the stronger bond, but assistance dogs did not overimitate significantly more than family dogs. Across the demonstrated groups, 17 of 49 dogs — 34.7 percent — touched at least one dot: 45.8 percent of the family dogs against 24 percent of the assistance dogs, the opposite of the predicted direction (Gerwisch, Mackie, Folkertsma & Huber, 2025).

The authors offer a specific explanation for the low assistance-dog figure. Their sample consisted of signal dogs, trained to monitor their caregiver for medical events, and such a dog may be occupied with its actual job while the caregiver stands still and ignores it for a minute. A second possibility they raise is that assistance-dog training teaches an animal which details in the environment to treat as irrelevant (how training history shapes what transfers).

7.2 Why This Matters for the Relationship Hypothesis

If stronger measured caregiver relationships directly produced more overimitation, assistance dogs should have shown a clear increase. They did not. The result does not make the dog-human relationship irrelevant, but it argues against a simple linear account in which relationship strength alone determines copying.

7.3 Overimitation Also Occurred at Home

The study is useful for another reason: it moved the task out of the laboratory. Family dogs that watched demonstrations performed the irrelevant actions more than the no-demonstration control group, at a marginally significant level, showing that the phenomenon is not restricted to the laboratory. The control condition earns its place here: none of the control dogs touched both dots and only four touched one, which rules out plain exploration of novel objects as the explanation.

Two further details belong in any careful summary. In 87.5 percent of the trials in which dogs copied both actions, they opened the bucket first and touched the dots afterwards, matching the post-goal pattern reported earlier. And dogs that gazed longer at their caregiver scored lower on the relevant action, which the authors read as gaze functioning as a request for help rather than as an index of affiliation (what human-directed gazing does).

8. Development: Puppies in 2026

8.1 Six- to Eight-Week-Old Puppies Touched the Dots

Gerwisch, Folkertsma and Huber (2026) adapted the at-home paradigm for six- to eight-week-old puppies. At least one dot was touched by 27.78% of puppies, compared with 24.5% in the adult comparison sample.

That superficial similarity does not establish early overimitation.

8.2 Baseline Exploration Changes the Interpretation

Puppies touched irrelevant dots during baseline exploration at a rate that did not differ clearly from their test trials, and they did not reproduce the demonstrated two-dot sequence precisely. The authors therefore describe the result as ambiguous and explicitly raise exploration as an alternative explanation.

8.3 Development Is No Longer Completely Untested

The 2026 study means that ontogeny is no longer a blank research gap. What remains unknown is how copying changes within the same dogs across development, how attachment and training history contribute over time, and whether the adult caregiver effect emerges gradually or through particular experiences.

A longitudinal design following puppies into adulthood would address those questions more directly than comparing separate age groups.

9. What Counts as Overimitation?

9.1 Competing Mechanisms Remain

Local enhancement, stimulus enhancement, social facilitation, prior reinforcement for interacting with human-indicated objects, exploration and action matching remain relevant alternatives or contributors. The 2024 prior-experience study weakens a purely causal-misunderstanding account, but it does not eliminate these other mechanisms.

9.1a Why the Control Conditions Decide the Question

Everything in this field depends on the no-demonstration baseline. Dogs investigate novel objects, and a board a human has just visited is interesting on its own; without a control group, dot-touching means nothing. Two studies make this concrete: in the at-home sample only four control dogs touched a single dot and none touched both, and in the puppy study the baseline was high enough to change the interpretation entirely (Gerwisch et al., 2025, 2026).

That is also the honest answer to the recurring objection that the dogs are merely curious. Curiosity is not excluded as a contributor, but it cannot by itself explain a difference between demonstration and control groups (why an observed action is not an identified motive).

9.2 The Normative Account Cannot Be Demonstrated Here

In children, researchers can test whether a copied sequence is treated as the “right” or conventional way to act. Dogs cannot verbally report such judgments, and the dot-touching tasks do not provide behavioral evidence strong enough to claim norm-following.

Calling canine dot-touching overimitation is therefore useful as an operational comparison, not as proof that dogs share the same normative cognition proposed for human children (why behavior alone does not identify an inner state).

9.3 Low Rates Still Matter — but Need Perspective

The fact that only a minority of dogs overimitate in many newer samples does not make the effect meaningless. It does mean the phenomenon is variable, task-dependent and unsuitable as a default expectation for an individual dog.

9.4 Great-Ape Comparisons Need Careful Wording

The strongest comparative claim is not “apes are at zero”. Great apes often omit clearly irrelevant steps in transparent tasks, but some studies with causally opaque tasks have reported copying of unnecessary-looking actions. Dog-human comparisons should therefore be framed around differences in robustness, context and task transparency rather than an absolute species boundary.

9.5 What the Term Is Still Good For

Used carefully, the label earns its place. It names a specific, measurable phenomenon — reproducing an action that demonstrably does not contribute to the outcome — and it connects the canine work to a large comparative literature with an established methodology. The claim that survives is narrower than the headline: some dogs reproduce causally irrelevant actions at low rates, mainly when the demonstrator is a familiar person, and mainly once the actual goal has been reached.

Used loosely, the term imports everything that has been shown in children and nothing that has been shown in dogs. That is the version worth resisting, and the reason this article keeps repeating what each study actually measured (what canine cognition research can and cannot establish).

10. Where This Fits in Canine Social Learning

10.1 Trained Imitation Is a Different Phenomenon

In Do as I Do, dogs are first trained to match their own behavior to demonstrated human actions after a specific cue. Dogs can later generalize that rule to novel actions, and they retain demonstrated actions over delays: in a version of the paradigm combined with a two-action control, 12 dogs reproduced human actions after intervals of one to 24 hours, with performance staying high as the delay grew (Fugazza, Pogány & Miklósi, 2016). The authors read that as long-term declarative memory of imitative actions (how memory is consolidated in dogs).

That is fundamentally different from spontaneous overimitation. In one case the dog has learned an explicit “copy this” rule; in the other, researchers measure whether an unrequested irrelevant action appears after a demonstration.

10.2 Selective Imitation Is Not a Settled Result

Range, Virányi and Huber (2007) reported that dogs selectively copied a conspecific's unusual paw action depending on whether the demonstrator appeared to have a reason for using the paw. The result was interpreted as sensitivity to action efficiency.

Kaminski et al. (2011) failed to replicate that conclusion under a modified design and argued against a rational-imitation interpretation. Huber, Huber, Range and Virányi (2012) replied that methodological and baseline differences made the replication non-equivalent. The appropriate summary is therefore that selective imitation has supporting evidence, but its interpretation and robustness have been debated.

10.3 Dogs Learn Socially From Both Humans and Dogs

The broader social-learning literature is much larger than the overimitation literature. Dogs can learn locations, actions and object-related solutions by observation under some conditions. Those findings should not be used as evidence that every matching response is imitation, and overimitation should not be used as a shortcut for all social learning.

10.3a Household Reports of Dogs Copying Dogs

Owners of multi-dog households routinely report that a new dog picked something up from the resident one. What that reflects is usually unknown: social facilitation, local enhancement, independent learning in the same environment under the same contingencies, or genuine copying. The everyday observation is real; the mechanism behind it is not established (how dogs influence each other).

10.4 Overimitation of Other Dogs Remains Open

The caregiver dot-touching program has focused on human demonstrators. Whether dogs reproduce causally irrelevant actions demonstrated by a familiar canine partner in a matched paradigm remains an open question.

11. What This Means for Training

11.1 Do Not Turn a Laboratory Effect Into a Training Method

Spontaneous overimitation is not a validated teaching technique. The dot-touching studies were designed to investigate social cognition, not to establish the most effective way to teach recall, loose-leash walking, targeting or cooperative-care behaviors.

11.1a The Handler Is Part of the Setup

The most transferable result is not about copying at all. It is that the same demonstration produces different behavior depending on who performs it, and that a dog attends closely to what a familiar person does around a task. For a trainer that argues for a habit rather than a technique: record sessions, watch what the hands and body do before the cue, and be able to say which of those movements are meant as signals (what dogs read from posture and movement).

The corresponding household observation is ordinary. A behavior that runs with the person who taught it and falls apart with someone else is not defiance; it is a change in a variable this literature identifies as relevant. The fix is that the other person teaches a few repetitions themselves rather than simply issuing the cue.

11.2 Demonstration Has Been Compared With Shaping — in a Different Paradigm

It is incorrect to say that demonstration-based dog training has never been compared with shaping. Fugazza and Miklósi directly compared the trained Do as I Do method with shaping/clicker training. In their 2014 study, Do as I Do was more efficient for the tested complex object-related tasks within the study's time limits. A 2015 follow-up found advantages for Do as I Do in training consistency, memory and generalization for the tested object-related action, with more mixed results for a body movement (Fugazza & Miklósi, 2014b, 2015).

Those experiments show that social learning can be deliberately incorporated into dog training. They do not show that spontaneous overimitation is useful in training, because the dogs in Do as I Do were explicitly trained to imitate and reinforced for doing so.

11.2a What the Findings Do Not License

Three sentences go beyond the data. That dogs learn best by watching: copying rates are low, variable and not compared against reinforcement-based teaching in the dot paradigm. That a dog copying its person understands the purpose of the action: the action has no purpose, which is the point of the design. And that overimitation demonstrates the strength of a bond: the study built to test exactly that found no significant correlation (Huber et al., 2022), and the assistance-dog comparison went in the opposite direction to the prediction (Gerwisch et al., 2025).

The reason this matters is that the finding is genuinely interesting and therefore attractive to overstate. A phenomenon can be real, replicable in one paradigm and still useless as a diagnostic for an individual animal (why group findings rarely transfer to individuals).

11.3 Clean Handler Mechanics Remain Good Advice — for Other Reasons

Keeping hand movements, posture and cues consistent is good training practice because dogs readily learn compound cues, contextual cues and stimulus-control relationships. Overimitation provides an interesting example of dogs attending to irrelevant human actions, but it does not by itself prove that an accidental hand movement will become a “superstitious behavior chain”.

Do not use the overimitation literature as evidence for a mechanism it did not test (how reinforcement changes behavior).

11.4 Caregiver Effects Should Not Be Generalized to Every Household Difference

A dog performing a cue for one person but not another can reflect differences in reinforcement history, stimulus control, posture, timing, context and generalization. The caregiver-versus-stranger overimitation result shows that demonstrator identity can matter in a specific copying task; it does not establish that ordinary cue failures across household members are caused by overimitation.

11.5 Showing Is Not the Same as Teaching an Imitation Rule

Simply demonstrating a behavior once is not equivalent to using Do as I Do, and neither is equivalent to shaping. Trainers should describe which learning procedure they are actually using rather than grouping all visually guided learning under “the dog copies me”.

11.6 What the Research Does Support for Practice

The most defensible practical message is modest: dogs pay close attention to familiar humans, socially demonstrated actions can influence what some dogs do, and the identity and history of the demonstrator can alter performance. How much this contributes to ordinary training depends on the procedure and has to be tested rather than assumed.

11.7 Three Situations Where This Shows Up

The dog that works for one person. Demonstrator identity changed copying in every study reviewed here, and the same variable plausibly affects ordinary cue-following. Having each household member teach rather than only cue is the practical response.

The behavior with a mysterious extra step. A dog that circles before lying down or paws the mat before settling has often been reinforced in a sequence that contained that element, sometimes because the handler was doing something at the same moment. Video usually finds it, and removing it later is harder than not building it in (why elements of a chain persist).

The class where nothing works. A group setting changes who demonstrates, who watches and what else is happening. Performing at home and not in class is a change in exactly those variables rather than a failure of obedience.

12. Summary at a Glance

Dogs can reproduce causally irrelevant human actions — but the frequency is low and varies strongly between individuals and studies (Huber et al., 2018; Gerwisch et al., 2025).

Caregiver versus stranger matters — dogs copied the irrelevant dot-touching action far less from an unfamiliar experimenter than from their caregiver in the matched Vienna paradigm (Huber et al., 2020).

Relationship quality did not show the predicted overall correlation — in 58 dogs, the 2022 study found no significant relationship between overall relationship-test scores and overimitation; only three dogs reached the highest irrelevant-copying category (Huber et al., 2022).

Social priming did not increase copying — the 2023 manipulation produced no significant main effect on irrelevant-action copying (Mackie & Huber, 2023).

Being watched did not increase irrelevant-action copying — caregiver attention did not change overimitation accuracy in 63 dogs, although relevant-action copying was affected (Mackie, Trehorel & Huber, 2025).

Knowing the task changed when dogs copied — prior experience shifted irrelevant action copying toward the post-goal period without simply eliminating it (Mackie & Huber, 2024).

Assistance dogs did not overimitate more — despite group differences in caregiver-relationship scores, assistance dogs and family dogs did not differ significantly in overimitation (Gerwisch et al., 2025).

Puppy dot-touching may be exploration — six- to eight-week-old puppies touched dots, but did so similarly during baseline and did not precisely reproduce the two-dot demonstration (Gerwisch et al., 2026).

Do as I Do is not overimitation — it is a trained imitation procedure and has been experimentally compared with shaping/clicker training (Fugazza & Miklósi, 2014b, 2015).

At-home testing produced comparable rates — Family dogs copied irrelevant actions in their own homes at rates similar to the laboratory studies, with 45.8 percent touching at least one dot (Gerwisch et al., 2025).

Copying mostly happened after the goal — In 87.5 percent of the trials with both actions copied, the dogs opened the container first (Gerwisch et al., 2025).

13. Research Gaps and Critical Appraisal

The evidence remains concentrated in one research program. The Vienna group has now tested multiple mechanisms and settings, which strengthens internal development of the field, but closely matched independent replication is still limited.

The mechanism behind caregiver effects remains unresolved. Familiarity matters in the classic comparison, yet relationship scores, priming, being watched and assistance-dog status do not produce one coherent affiliative dose-response pattern.

Training history is still difficult to separate from relationship history. Pet dogs spend years being reinforced for attending to human hands, targets, gestures and objects. More direct comparisons across training backgrounds would help.

Causal misunderstanding is no longer a sufficient single explanation. Prior task knowledge changed timing rather than simply abolishing irrelevant copying, but that does not reveal which non-instrumental process remains.

Development has been sampled, not mapped. The puppy study provides an important first developmental test, but longitudinal data from puppyhood to adulthood are still missing.

Exploration needs stronger controls. The 2026 puppy findings show why baseline interaction with irrelevant targets matters, especially in young animals.

Normative interpretation remains unsupported. No current canine paradigm demonstrates that dogs treat the copied action as a convention or as the “correct” way to act.

Practical transfer of spontaneous overimitation remains unknown. Social-learning training methods have been studied, including direct comparisons of Do as I Do and shaping, but no evidence shows that spontaneous copying of irrelevant actions improves or disrupts ordinary dog training in a predictable way.

Conspecific overimitation remains a useful open question. Dogs learn socially from other dogs, but a matched test of causally irrelevant copying from a familiar canine partner would help distinguish human-specific relationship effects from more general social-learning processes.

Post-goal copying needs its own explanation. That dogs mostly touch the dots after obtaining the reward is consistent across studies (Mackie & Huber, 2023, 2024; Gerwisch et al., 2025), and the proposed reading — a secondary social goal — has not been tested directly against simpler accounts such as returning to a place the caregiver made salient (how reward timing shapes what is learned).

Gazing is ambiguous as a measure. It has been used as an index of affiliation and, in the same literature, read as a request for help; dogs that gazed longer performed the relevant action less well (Gerwisch et al., 2025).

14. Conclusion

Dogs do sometimes reproduce human actions that are unnecessary for achieving a physical goal. The strongest early result was the contrast between a familiar caregiver and an unfamiliar experimenter, which established that the social identity or history of the demonstrator matters in the classic dot-touching task. What subsequent research has not supported is a simple equation of stronger bond with more overimitation. Overall relationship scores did not significantly predict copying in 2022; social priming did not increase it in 2023; prior causal knowledge changed the timing rather than the frequency of copying in 2024; caregiver attention did not increase irrelevant copying in the 2025 attentional-state study; and assistance dogs with different relationship scores did not overimitate more than family dogs in 2025. The 2026 puppy study adds a further warning: touching an irrelevant target can also reflect exploration.

The current evidence therefore supports a measured conclusion. Canine overimitation is a real, low-frequency and context-sensitive behavioral phenomenon, but its mechanism is unresolved. Caregiver familiarity, social learning, causal knowledge, exploration and prior learning history all remain relevant pieces of the explanation. For training, the most important distinction is between spontaneous overimitation and deliberately trained imitation. Do as I Do has been studied as a training procedure and can be useful for some tasks; spontaneous irrelevant-action copying has not been validated as a training tool.

Key Insights (Takeaways)

• Dogs can copy causally irrelevant actions, but the effect is variable and usually low in frequency.

• Caregiver familiarity matters in the classic dot-touching paradigm; “relationship quality causes overimitation” is not supported as a simple general rule.

• The 2022 relationship study found no significant overall correlation between relationship quality and overimitation.

• Social priming and immediate caregiver attention did not reliably increase irrelevant-action copying.

• Prior causal knowledge changes when dogs copy an irrelevant action and argues against causal misunderstanding as the sole explanation.

• Assistance dogs did not overimitate more than family dogs despite group differences in measured caregiver relationship.

• At six to eight weeks, puppy dot-touching is not clear evidence of precise overimitation because baseline exploration can produce the same response.

• Spontaneous overimitation and trained Do as I Do imitation are different phenomena and should not be used as evidence for each other.

• For trainers, clean cueing and mechanics remain sensible, but ordinary cue dependence or accidental behavior chains should not be attributed to overimitation without evidence.

References

Fugazza, C., & Miklósi, Á. (2014a). 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., & Miklósi, Á. (2014b). Should old dog trainers learn new tricks? The efficiency of the Do as I do method and shaping/clicker training method to train dogs. Applied Animal Behaviour Science, 153, 53–61. https://doi.org/10.1016/j.applanim.2014.01.009

Fugazza, C., & Miklósi, Á. (2015). Social learning in dog training: The effectiveness of the Do as I do method compared to shaping/clicker training. Applied Animal Behaviour Science, 171, 146–151. https://doi.org/10.1016/j.applanim.2015.08.033

Fugazza, C., Pogány, Á., & Miklósi, Á. (2016). 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

Gerwisch, K., Folkertsma, R., & Huber, L. (2026). Early overimitators or eager explorers? Ontogenetic influences on overimitation in dog puppies. Applied Animal Behaviour Science, 304, 107133. https://doi.org/10.1016/j.applanim.2026.107133

Gerwisch, K., Mackie, L., Folkertsma, R., & Huber, L. (2025). Comparing overimitation between assistance dogs and family dogs at home. Applied Animal Behaviour Science, 292, 106839. https://doi.org/10.1016/j.applanim.2025.106839

Hoehl, S., Keupp, S., Schleihauf, H., McGuigan, N., Buttelmann, D., & Whiten, A. (2019). 'Over-imitation': A review and appraisal of a decade of research. Developmental Review, 51, 90–108.

Huber, L., Kubala, D., & Cimarelli, G. (2022). Overimitation in dogs: Is there a link to the quality of the relationship with the caregiver? Animals, 12(3), 326. https://doi.org/10.3390/ani12030326

Huber, L., Popovová, N., Riener, S., Salobir, K., & Cimarelli, G. (2018). Would dogs copy irrelevant actions from their human caregiver? Learning & Behavior, 46(4), 387–397. https://doi.org/10.3758/s13420-018-0336-z

Huber, L., Range, F., & Virányi, Z. (2012). Dogs imitate selectively, not necessarily rationally: Reply to Kaminski et al. (2011). Animal Behaviour, 83(6), e1–e3. https://doi.org/10.1016/j.anbehav.2012.03.020

Huber, L., Salobir, K., Mundry, R., & Cimarelli, G. (2020). Selective overimitation in dogs. Learning & Behavior, 48(1), 113–123. https://doi.org/10.3758/s13420-019-00400-w

Johnston, A. M., Holden, P. C., & Santos, L. R. (2017). Exploring the evolutionary origins of overimitation: A comparison across domesticated and non-domesticated canids. Developmental Science, 20(4), e12460. https://doi.org/10.1111/desc.12460

Kaminski, J., Nitzschner, M., Wobber, V., Tennie, C., Bräuer, J., Call, J., & Tomasello, M. (2011). Do dogs distinguish rational from irrational acts? Animal Behaviour, 81(1), 195–203. https://doi.org/10.1016/j.anbehav.2010.10.001

Mackie, L., & Huber, L. (2023). Socially priming dogs in an overimitation task. Frontiers in Psychology, 14, 1063132. https://doi.org/10.3389/fpsyg.2023.1063132

Mackie, L., & Huber, L. (2024). Dogs with prior experience of a task still overimitate their caregiver. Scientific Reports, 14, 20806. https://doi.org/10.1038/s41598-024-70700-3

Mackie, L., Trehorel, J., & Huber, L. (2025). Watched or not: Overimitation in dogs under different attentional states. Learning & Behavior, 53, 171–182. https://doi.org/10.3758/s13420-024-00635-2

Range, F., Virányi, Z., & Huber, L. (2007). Selective imitation in domestic dogs. Current Biology, 17(10), 868–872. https://doi.org/10.1016/j.cub.2007.04.026