Cognitive Abilities in Dogs – Why Our Canine Companions Are Smarter Than We Think

1. Introduction

1.1 Beyond "Just Instinct"

For most of the twentieth century, serious science treated animal cognition warily, and dogs – too familiar, too domesticated – were often dismissed as reflex machines. That has decisively changed. Dogs are now a flagship model for comparative cognition, precisely because their minds were shaped alongside ours. But the correction has its own hazard: having decided dogs are smart, popular writing tends to keep inflating the claim. Getting the level right matters, because a realistic picture of what a dog understands is the foundation of fair training and reasonable expectations.

1.2 How to Read the Evidence

Four cautions run through everything below.

First, star individuals are not the average dog. The famous word-learning feats belong to a handful of exceptional Border Collies; most dogs, even bright ones, do not perform anywhere near that level.

Second, rich interpretations require lean tests to rule them out. A dog that fetches the right object may "understand a word as a name," or may simply have learned "when I hear that sound, bringing this thing pays" – distinguishing the two takes careful controls, and researchers disagree about whether they have been met.

Third, media and publication both favor the impressive. The striking positive result travels; the failed replication and the null result rarely trend.

Fourth, comparisons cut both ways. Dogs outperform chimpanzees on some social tasks and underperform them badly on some physical ones. "Smarter than we think" is true in specific domains, not as a blanket ranking.

1.3 Why "How Smart" Is the Wrong Question

Intelligence rankings of dog breeds circulate widely and rest largely on how quickly a breed learns obedience commands (Coren, 1994). That measures trainability and biddability, which are useful traits and not the same as cognition.

A breed that learns commands slowly may solve a physical problem faster, and the ranking does not capture it. The more answerable question is which abilities dogs have and under what conditions they show them (with breed explaining less than is usually assumed). Variation within a breed is also large enough that a ranking says little about any individual animal.

1.4 What This Article Is Not

It is not an argument that dogs are cleverer than assumed, nor that they are less clever. Both framings compare dogs against a human standard that the research does not use.

What follows is an account of specific capacities, each with the design that established it and the limits that come with it. Read that way, some findings turn out stronger than their reputation and others weaker, which is the ordinary result of looking at primary sources.

2. Social Cognition: Reading Humans

2.1 Understanding Human Gestures

This is the dog's genuine cognitive specialty, and the evidence is strong. In a landmark study, dogs outperformed even chimpanzees at using human communicative cues – a pointed finger, a glance, a tap – to find hidden food, and puppies with little human exposure already showed the skill, while human-raised wolves did not (Hare et al., 2002). That pattern points to something selected for during domestication rather than merely learned by each dog (how dogs understand human gestures). It is one of the firmest findings in canine cognition, though the explanation for it remains contested.

2.2 Sensitivity to Human Emotion and Attention

Dogs also track human attention and emotional state – whether a person is watching, the tone of a voice, tension in a room – and adjust accordingly. They show emotional contagion, catching something of a human's affective state. The caution here is interpretive: reading and responding to a human's emotional signals is not the same as "knowing how their human feels" in a full mentalistic sense. The behavior is real and useful; the inner life behind it is exactly the kind of thing that should not be asserted from the outside (why behavior does not equal emotion).

2.3 Why Social Cognition Is the Dog's Strength

The unifying point is that the dog's cognitive profile is lopsided toward the social. Where a dog looks brilliant, it is usually a task involving a human; where it looks limited, it is usually a task involving the physical world alone. This is not a deficiency – it is what several thousand years of co-evolution built. Keeping it in view prevents the common error of promoting a social talent into general genius.

2.4 Why the Domestication Story Is Only Half Settled

That dogs read human gestures better than expected is established. Why they do — selection during domestication, individual learning during a lifetime with humans, or both — is not, and the exchange between the two positions has run for two decades without converging (where gesture comprehension is treated in detail).

For a practitioner the answer changes nothing. For a reader trying to evaluate a claim, the difference between an established ability and a contested explanation of it is the whole point, and popular accounts present the explanation with the same confidence as the ability.

2.5 The Comparison Group Matters

Claims that dogs outperform chimpanzees or wolves on social tasks depend on which animals were tested, how they were raised and how much testing experience they had. Research wolves live in a small number of institutions and are tested repeatedly across many studies; pet dogs recruited from the public are typically naive.

Whichever way such a comparison comes out, testing experience differs systematically between the groups (with individual variation doing more work than group means suggest). That does not make the comparisons worthless; it means the difference they measure is not purely a species difference.

3. Communication and Word Learning

3.1 What Most Dogs Actually Do

Many dogs respond to a substantial number of words, distinguish commands partly by tone, and can learn new signals through structured training. But a great deal of this is a word functioning as a cue to act ("sit," "come") rather than as a name that refers to a thing. That distinction is central to the science, and blurring it is the root of most word-count inflation.

3.2 The Gifted Word Learners

A few exceptional dogs go further. A Border Collie named Rico knew labels for more than 200 objects and could infer a new object's name by exclusion – picking the unfamiliar item when he heard an unfamiliar word – and still retrieve it weeks later, a pattern researchers likened to the "fast mapping" seen in children (Kaminski et al., 2004). Another Border Collie, Chaser, was trained over three years to the names of 1,022 objects and showed she could treat words as referring to categories (Pilley & Reid, 2011). These are real, rigorously documented feats – and they belong to rare "gifted word learner" individuals, not to dogs in general.

3.3 A Careful Claim About "Understanding Language"

Two qualifications keep this honest. First, even for the gifted dogs, whether they grasp words as true symbolic referents or as sophisticated fetch-cues has been debated by language researchers, and the leaner reading has not been fully excluded. Second, the widely repeated figure that "the average dog understands about 165 words" traces to a popular book (Coren, 1994), not to a controlled study, and should be treated as a rough illustration rather than a measured fact. The defensible statement is that dogs can learn to associate many sounds with actions and objects, that a rare few show something close to referential understanding, and that this rests on general learning and memory rather than a human-style language faculty. Structured demonstration methods can even teach dogs to copy actions on cue, a form of social learning that doubles as a window onto how they process what they observe.

3.4 What "Fast Mapping" Means and Does Not Mean

The term describes inferring the referent of a new word by exclusion — among known objects and one unknown one, the new word must refer to the unfamiliar item. That is a real inferential step and it is narrower than word learning in the human sense.

It does not establish that the dog has a concept attached to the label, that it could use the word itself, or that it understands the word outside the retrieval task (because transfer across contexts is not automatic). Whether a dog that fetches the right toy on hearing its name knows anything about the toy when the toy is absent has not been tested.

3.5 Why the Gifted Dogs Are Unrepresentative

The dogs in these studies are exceptional by selection: they came to attention because they already showed the ability. Reporting their performance as a fact about dogs is the same error as describing human memory from the performance of memory champions.

The interesting question the single-subject work raises is what the exceptional dogs have that others do not, and that has not been answered. Candidate explanations run from breed and individual temperament to the amount and structure of teaching, and no design has separated them. Recruiting such dogs through public appeals, as later work has done, is a step toward it and inherits the same selection problem.

4. Memory

4.1 Everyday Memory

Dogs plainly remember people, places, routines, and – given their sensory world – smells, and they apply those memories in new situations. Memory is also consolidated during rest, so what a dog learns is reorganized and stabilized during sleep, which is part of why a skill often looks better the next day. The gradual erosion of these systems in old age is its own well-characterized condition (cognitive dysfunction syndrome).

4.2 Episodic-like Memory

More striking is evidence that dogs can recall events they had no reason to memorize. Using a "Do as I Do" method, researchers had dogs watch a human action and then – unexpectedly, without any prior signal that memory would be tested – asked them to reproduce it; the dogs could, after both short and longer delays, though their performance faded with time (Fugazza et al., 2016). Because the encoding was incidental, this is taken as evidence of episodic-like memory. The hedge is deliberate and important: the "-like" marks that we cannot verify the conscious, autobiographical "mental time travel" that defines human episodic memory. What is shown is the behavioral signature of recalling an incidentally encoded event – impressive, and carefully stated.

4.3 Why the Incidental Encoding Design Matters

The strength of the episodic-like memory result is in the method: the dogs were not trained to remember the action, and the test came unexpectedly (Fugazza, Pogány & Miklósi, 2016). A memory that was never rehearsed is much harder to explain as a trained response.

The term episodic-like is doing careful work. Human episodic memory involves conscious recollection of a past experience as one's own, and no animal design can test that (where the same inference problem appears). Researchers in this area chose the hedged term deliberately, and popular accounts drop it.

4.4 What Is Unknown About Canine Memory

How long such memories persist, whether they are updated on recall, and whether dogs can access them deliberately are all open. Owner reports of dogs recognizing people after years are common and are not evidence, since recognition and recollection are different capacities. Recognizing a familiar person requires a stored representation; recalling the occasion on which the dog last met them requires something considerably more.

The systematic work on canine memory duration is thin relative to how confidently the topic is discussed. Claims that dogs remember for a specific number of minutes circulate widely and have no traceable source in the research.

5. Problem-Solving and Causal Reasoning

5.1 Flexibility and Persistence

Dogs do solve problems – opening latches, working around obstacles, retrieving food from awkward places – and they show real behavioral flexibility, trying alternative strategies and adjusting after failure. This much of the popular claim holds.

5.2 The Limits of Physical Causal Reasoning

But the further claim – that dogs demonstrate a "genuine understanding of cause and effect" – is where the popular account overreaches. In controlled tests of physical causal reasoning (how objects connect, support, and act on one another), dogs often perform poorly and are readily misled by surface cues; their grasp of physical causality is limited and genuinely debated among researchers (what canine causal reasoning does and doesn't show). Persistence and flexibility are not the same as understanding why a mechanism works, and the evidence for the latter is thin.

5.3 The Social Shortcut

There is a revealing twist. Faced with an unsolvable task, dogs characteristically stop and look back at a human – recruiting a social partner – where hand-raised wolves keep working the problem alone (a much-cited comparison from Miklósi and colleagues, 2003). This captures the whole shape of canine cognition: the dog's answer to a hard problem is often another creature, not another lever. It is a brilliant strategy for an animal that lives with humans – and further reason to see the dog's intelligence as social first. The neural machinery for the deliberate, self-controlled side of problem-solving is itself an active area of study (the prefrontal cortex and canine self-control), as is the question of whether dogs monitor their own knowledge (metacognition in dogs).

5.4 The Social Shortcut Is a Finding, Not a Failure

That dogs turn to a person when a physical problem becomes difficult is often reported as a deficit. It is a strategy, and in the environment dogs actually live in it is usually the correct one.

Comparing it against a wolf's persistence measures which strategy each species defaults to, not which animal is more capable (where causal reasoning is examined at length). A wolf that keeps working at a sealed container and a dog that looks at its person are both doing something sensible given the life each leads.

5.5 What Would Distinguish the Two Accounts

Whether dogs fail physical tasks because they lack the reasoning or because they prefer the social route is testable: remove the person and see what happens, and compare performance when a human is present against when none is available.

Studies doing this exist and their results are mixed, which is itself informative — a pure deficit account would predict failure either way. Mixed results across similar designs usually mean the effect depends on something the designs differ in, which in this case is most likely the dog's history with people and with problem-solving.

6. Quantities and "Fairness"

6.1 Numerical Cognition

Dogs show rudimentary numerical ability. They reliably choose the larger of two food quantities, and in a preferential-looking study they looked longer at an impossible arithmetic outcome ("1+1=3") than at an expected one, suggesting they anticipated the result and noticed the violation (West & Young, 2002). This is real but modest: it is relative-quantity sensitivity and expectation, not counting or arithmetic in any human sense, and the authors themselves called it rudimentary.

6.2 Inequity Aversion – Fairness or Something Simpler?

The claim most in need of correction is the "moral" one. Dogs asked to give a paw will work happily when unrewarded – until they see a partner being rewarded for the same action, at which point they stop cooperating (Range et al., 2009). This is a genuine and replicated effect, and it is easy to narrate as "dogs have a sense of fairness." The evidence does not support that rich reading (inequity aversion: fairness or something else?). Crucially, dogs react only to the disadvantage of getting nothing while another is paid; unlike some primates, they do not object to receiving a worse-quality reward, and the response can be explained by simpler mechanisms – social frustration, disappointed expectation – without invoking a moral concept of fairness. What dogs demonstrably have is sensitivity to a rewarded partner; whether that amounts to anything like a sense of justice is unproven and, on current evidence, doubtful.

6.3 What the Inequity Result Requires

The finding is that the dogs tested stopped working when a partner was rewarded for the same action and they were not (Range et al., 2009). Calling that fairness requires the dog to be comparing its treatment against another's, which is one explanation among several.

Simpler accounts are available: the presence of a rewarded partner changes what the situation predicts, and frustration at a reduced reward rate produces the same behavior without any social comparison (which is what operationalizing a behavioral claim is for). Distinguishing them requires conditions in which the two accounts predict different outcomes, and the canine work on this is still working toward such designs.

6.4 Numerical Cognition Is a Narrow Claim

The counting work demonstrates sensitivity to quantity differences, which many species share and which does not involve counting in any sense a person would recognize (West & Young, 2002).

It is a real perceptual capacity and it is routinely reported as though dogs could count. The distinction matters because approximate quantity discrimination is widespread across the animal kingdom, including in species nobody credits with arithmetic, and the canine result is unremarkable set beside them.

7. What Replication Has Shown

7.1 Why This Chapter Belongs Here

Canine cognition has produced a series of striking findings over twenty-five years, and several of them have been retested at larger scale since. The retests have not always agreed, and an article surveying what dogs can do owes the reader that part of the record (which is what operationalizing a behavioral claim is for).

7.2 The Multi-Lab Pointing Study

Point-following is the most cited result in this field. It was tested across 20 research sites with data from 455 dogs, comparing an ostensive condition — eye contact and the dog's name before the point — against a non-ostensive control (ManyDogs Project et al., 2023).

Both conditions were significantly above chance and there was no significant difference between them, consistently across sites. Dogs also followed contralateral, momentary pointing at lower rates than earlier research had described (ManyDogs Project et al., 2023).

7.3 What That Changes

The dogs in that sample followed points. What the study removes is the evidence that they do so because the gesture is framed as communication, which is the interpretation the popular version rests on (where gesture comprehension is treated in detail). A null difference is not proof that the conditions are equivalent to the dog, and it does mean the strong reading has lost the support it was resting on.

It also shows that not every pointing style works equally well, which matters for anyone quoting a single success rate as though it described the ability. A figure from a study using close, sustained, same-side pointing is not a figure about point-following generally.

7.4 The Solidity Finding

An earlier study had reported that dogs spontaneously use the solidity principle — that one solid object cannot pass through another. A follow-up presented 112 dogs with seven different versions of the task and reached a conclusion in stark contrast to that claim (Müller et al., 2014).

One hundred and twelve dogs across seven variants is an unusually serious attempt at a question in this field, and the correction has travelled far less widely than the original (where causal reasoning is examined at length). That asymmetry is structural rather than anyone's fault: a positive finding about a familiar species is interesting to write about, and a null result is not.

7.5 What a Failed Replication Establishes

It establishes that the earlier evidence does not support the claim. It does not establish that dogs lack the competence, which is a weaker and more useful statement.

The honest position after such a result is that the question is open. That distinction is routinely lost in both directions, and losing it in the sceptical direction is no more accurate than losing it in the credulous one. A field that overcorrects is not more rigorous than one that overstates; it is wrong in the other direction.

7.6 How to Read the Rest of This Article

Nothing above undermines the findings described in the other chapters. It sets the standard for reading them: a striking result from a small sample is a hypothesis, and the studies that carry the most weight here are the ones with the largest and most varied samples.

Where this article reports a figure, the sample behind it is worth noticing. That is not a counsel of scepticism about the field. It is how a reader tells a demonstration that one dog can do something from a demonstration that dogs generally do, and the two are constantly reported in the same register.

8. What Kind of Study Produced Each Finding

8.1 Three Kinds, Three Weights

The evidence in this article comes from designs that differ enormously in what they can support, and the difference is not visible in how the findings are usually quoted.

8.2 Single-Subject Studies

The word-learning results come from individual animals studied intensively (Kaminski, Call & Fischer, 2004; Pilley & Reid, 2011). A single dog demonstrating a capacity establishes that the capacity is possible in the species and says nothing about how common it is.

Both dogs were also trained extensively by owners with an unusual investment in the task, which is part of the finding rather than a confound to be set aside (with individual variation doing more work than group means suggest). What was demonstrated is what a dog can reach with that much structured teaching, not what a dog arrives with.

8.3 Small-Group Laboratory Studies

Most of the classic results — the gesture work, the inequity paradigm, the counting study — involve tens of animals in one laboratory. That is enough to detect a substantial effect and not enough to establish how robust it is.

The replication chapter above shows what happens when several of these are retested at scale, which is the reason to hold them lightly rather than to dismiss them (a problem early testing runs into as well). Some will survive retesting and some will not, and at present nobody knows which are which.

8.4 Large Multi-Site Studies

Only one result in this article rests on a multi-site design with hundreds of animals (ManyDogs Project et al., 2023). Its findings should carry correspondingly more weight, including the parts that are inconvenient. Multi-site designs also control for something a single laboratory cannot: whether an effect depends on one team's procedure, apparatus or population.

8.5 What Is Missing Entirely

No study in this field has followed the same dogs across several cognitive tasks and reported how performance on one relates to performance on another. Whether canine cognition has a general factor, as human intelligence testing describes, or consists of separable abilities, is therefore unknown.

That is a large gap for an article titled around how smart dogs are, and it is worth stating plainly: the question has not been asked in a form that could answer it (which is what operationalizing a behavioral claim is for). Test batteries covering several tasks in the same animals exist in the assistance-dog literature, and they were built to predict working success rather than to describe the structure of canine cognition.

9. Research Gaps and Methodological Challenges

The field's limits are unusually relevant here, because the public picture runs well ahead of the data.

Star-subject studies. Several of the most famous findings rest on one exceptional individual (Rico, Chaser). Single-subject work is valuable but cannot tell us what dogs in general can do.

Rich versus lean interpretation. For almost every headline ability – "understands words," "grasps cause and effect," "sense of fairness" – a leaner explanation exists, and the debate over which is correct is often unsettled. Responsible reading defaults to the leaner account until the richer one is earned.

Publication and media bias. Positive, surprising results are published and amplified; failures to replicate and null results are not. The visible literature therefore skews optimistic.

Sampling and breed. Much work uses convenience samples and a narrow range of breeds (Border Collies are heavily overrepresented in word-learning research), complicating generalization (breed is a weak predictor of behavior).

Measuring cognition at all. Turning "intelligence" into a testable, repeatable measure is genuinely hard, and different tasks can rank the same dogs differently (operationalizing dog behavior).

Several landmark findings have been retested at scale with different results. The ostensive-communication interpretation of point-following was not supported across 20 sites (ManyDogs Project et al., 2023), and a basic physical-cognition claim did not survive a larger follow-up (Müller et al., 2014).

The word-learning evidence rests on individual animals. Single-subject designs establish that a capacity is possible and cannot establish how common it is (Kaminski et al., 2004; Pilley & Reid, 2011).

No study has measured performance across tasks in the same dogs. Whether canine cognition has a general factor or consists of separable abilities is untested.

Corrections travel less far than original findings. Popular accounts of canine cognition were largely written from the first wave of results and have not been updated as those results were revisited.

10. Practical Implications

10.1 What Cognition Means for Training

Understanding how dogs actually process information sharpens training. Because much "word learning" is cue-to-action learning, signals should be consistent and clearly tied to consequences; because dopamine-driven reward learning depends on precise timing, the marker must land on the right behavior (dopamine and canine learning). Working with the dog's real cognitive strengths – its social attention, its associative learning – beats demanding human-style comprehension it does not have.

10.2 Communication and Clarity

Most training breakdowns are communication failures, not stupidity. A dog that "won't listen" is often a dog receiving inconsistent or ambiguous signals. Clarity, consistency, and reasonable expectations reduce the frustration on both ends of the leash.

10.3 Assistance and Therapy Work

The dog's social-cognitive talent is exactly what makes assistance and therapy roles possible: reading human cues, attending to human state, and cooperating are the very abilities the research identifies as canine strengths (the bond that underpins this work). Training tailored to those strengths, rather than to inflated notions of canine reasoning, is what actually works.

10.4 Against Both Underestimating and Overselling

The practical sweet spot is a narrow one. Underestimating dogs leads to crude, coercion-based handling; overselling them ("he knows exactly what he did," "she understands every word") leads to unfair expectations and misread behavior. Respecting a dog means seeing its mind accurately – neither a reflex machine nor a furry person.

10.5 What Cognition Research Does Not Give a Handler

Almost nothing in this article yields a training technique. The findings describe capacities under laboratory conditions; training works with an individual dog in a specific environment, and the gap between the two is large.

What the research does supply is a set of corrections to common assumptions — that failure to follow a signal is disobedience, that a dog which solves one problem should solve another, that a breed ranking predicts what an animal can learn (because transfer across contexts is not automatic). Each of those assumptions produces a specific handling error, and removing it is worth more than any technique.

10.6 The Two Errors Are Not Equally Common

Underestimating dogs and overselling them are both possible, and in current popular coverage the second is far more frequent. Claims about dogs understanding language, feeling fairness or reasoning about physics all run ahead of what has been shown — and each of them has a narrower version that is supported.

Correcting them is not a diminishment of the species. It is the difference between an account a reader can rely on and one that collapses when checked. Dogs are remarkable for what they demonstrably do — reading people, adapting to households, learning across a lifetime — and none of that needs inflating.

11. Summary at a Glance

Point-following is real and its interpretation is contested — Across 455 dogs at 20 sites, performance was above chance with no difference between ostensive and non-ostensive conditions (ManyDogs Project et al., 2023).

Not every pointing style works equally — Contralateral, momentary pointing produced lower rates than earlier research had described (ManyDogs Project et al., 2023).

Word learning rests on individual dogs — The landmark results come from single animals studied intensively (Kaminski et al., 2004; Pilley & Reid, 2011), which establishes possibility rather than prevalence.

A basic physical-cognition claim did not replicate — A study of 112 dogs across seven task versions contradicted the earlier finding that dogs spontaneously use the solidity principle (Müller et al., 2014).

Episodic-like memory has been demonstrated — Dogs recalled others' actions after incidental encoding (Fugazza, Pogány & Miklósi, 2016), using a design that avoids training the memory being tested.

Inequity aversion is a contested interpretation — Dogs tested in these paradigms stop working when a partner is rewarded and they are not (Range et al., 2009); whether that reflects fairness or something simpler is disputed.

Social cognition is the strength — Where dogs outperform expectations it is generally on tasks involving human beings, and where they underperform it is generally on physical problems.

No study has measured across tasks — Whether canine cognition has a general factor or consists of separable abilities has not been tested.

12. Conclusion

Dogs are genuinely, measurably intelligent – above all in the social domain, where they read human communication better than any other species and recruit us as partners in solving their problems. They learn associations readily, a rare few learn words at a startling scale, they remember incidentally encoded events, they sense quantity, and they respond to a rewarded partner. Every one of those statements is supported by real research. What the science does not support is the inflated version: that the average dog understands language, reasons about physical cause and effect, or possesses a moral sense of fairness. The honest picture is more modest and more interesting than the myth – a mind exquisitely tuned to living with humans. Recognizing that mind for what it is, rather than for what we wish it were, is the real foundation of a respectful relationship built on understanding rather than on either dominance or fantasy.

Key Insights (Takeaways)

• The dog's real cognitive strength is social: dogs read human gestures and attention better than chimpanzees do, a skill present in young puppies but not in human-raised wolves (Hare et al., 2002). Where dogs look brilliant, a human is usually involved; where they look limited, the task is usually purely physical.

• Word-learning feats belong to rare, exceptional individuals – Rico's 200+ labels (Kaminski et al., 2004) and Chaser's 1,022 (Pilley & Reid, 2011) – not to dogs in general. Whether even they grasp words as true names or as elaborate fetch-cues is debated, and the popular "average dog knows ~165 words" figure comes from a trade book (Coren, 1994), not a study.

• Dogs show episodic-like memory: they can reproduce an incidentally observed action when unexpectedly asked, with performance fading over time (Fugazza et al., 2016). The "-like" is deliberate – the behavioral signature is real, the conscious autobiographical experience is not something we can verify.

• The popular claims about "cause-and-effect understanding" and a "sense of fairness" overreach. Dogs are flexible, persistent problem-solvers but perform poorly on physical causal-reasoning tasks and tend to recruit a human instead (Miklósi et al., 2003). Their inequity response is real (Range et al., 2009) but asymmetric and better explained by social frustration than by morality; numerical ability is rudimentary (West & Young, 2002).

• The practical goal is to neither underestimate nor oversell. Train to the dog's genuine strengths – social attention and associative learning – with clear, consistent, well-timed signals, and hold expectations that match a mind built for living with humans rather than for human-style reasoning.

References

Coren, S. (1994). The Intelligence of Dogs: A Guide to the Thoughts, Emotions, and Inner Lives of Our Canine Companions. Bantam Books. 

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

Hare, B., Brown, M., Williamson, C., & Tomasello, M. (2002). The domestication of social cognition in dogs. Science, 298(5598), 1634–1636. https://doi.org/10.1126/science.1072702

Kaminski, J., Call, J., & Fischer, J. (2004). Word learning in a domestic dog: Evidence for "fast mapping." Science, 304(5677), 1682–1683. https://doi.org/10.1126/science.1097859

ManyDogs Project, Espinosa, J., Stevens, J. R., et al. (2023). ManyDogs 1: A multi-lab replication study of dogs' pointing comprehension. Animal Behavior and Cognition, 10(3), 232–286. https://doi.org/10.26451/abc.10.03.03.2023

Miklósi, Á., Kubinyi, E., Topál, J., Gácsi, M., Virányi, Z., & Csányi, V. (2003). A simple reason for a big difference: Wolves do not look back at humans, but dogs do. Current Biology, 13(9), 763–766. 

Müller, C. A., Riemer, S., Range, F., & Huber, L. (2014). Dogs' use of the solidity principle: Revisited. Animal Cognition, 17(3), 821–825. https://doi.org/10.1007/s10071-013-0709-9

Pilley, J. W., & Reid, A. K. (2011). Border collie comprehends object names as verbal referents. Behavioural Processes, 86(2), 184–195. https://doi.org/10.1016/j.beproc.2010.11.007

Range, F., Horn, L., Virányi, Z., & Huber, L. (2009). The absence of reward induces inequity aversion in dogs. Proceedings of the National Academy of Sciences, 106(1), 340–345. https://doi.org/10.1073/pnas.0810957105

West, R. E., & Young, R. J. (2002). Do domestic dogs show any evidence of being able to count? Animal Cognition, 5(3), 183–186. https://doi.org/10.1007/s10071-002-0140-0