Learning Theory in Dog Training: How Dogs Really Learn

In brief

Three things define modern dog training. First, your dog never learns through consequences alone—classical conditioning, social learning and, above all, emotional appraisal are always running at the same time. Second, emotional state is not a side condition but a prerequisite: a dog who is afraid or overwhelmed learns less well—especially when it comes to taking in new information and finding flexible solutions. That holds no matter how clean your technique is. And third, there is no method that works the same for every dog—what counts as a reward is decided by your dog, not by the textbook.

Modern learning theory draws on behavioral biology, the psychology of learning and neuroscience. It has replaced older approaches built on punishment or supposed pack leadership.

👉 The two outdated counter-models: Why Drive Theories No Longer Hold Up · Why Dominance Fails as a Scientific Concept in Dogs

1. Learning is more than cue and obedience

Dogs don't learn only through direct instructions. Their behavior emerges from the interplay of several mechanisms that operate at the same time:

Operant conditioning—learning through consequences. Classical conditioning—learning by linking stimuli, for example when the clicker becomes the herald of a reward. Social learning—learning by watching other dogs or people. Emotional appraisal—learning how a situation felt.

The last point is especially important: learning never happens without emotion. Dogs are sensitive to their person's mood, and that has a considerable influence on how things go.

So training is always about communication and relationship, too.

👉 How Dogs Perceive Human Emotions · Behavior Does Not Equal Emotion

2. Positive reinforcement: the central principle

The core of modern learning theory: a desired behavior is immediately linked to a pleasant consequence—food, play, social attention or praise. The likelihood that the dog will show this behavior more often in the future goes up—under similar conditions.

That qualifier is not a formality. Reinforcement is context-bound: what was learned in the living room initially applies, as far as the dog is concerned, to the living room. That is why practicing in new places gets its own section later in this article.

The theoretical foundation was laid by Skinner (1938): behavior is shaped by its consequences. Punishment can certainly reduce a behavior—what it does not do is build a desired alternative behavior. That has to happen separately.

For context, Skinner's work was done with other species. It provides the framework, not the evidence in dogs.

2.1 What has been studied in dogs

There are now findings from dogs themselves, and they are worth a close look—not least because they differ in how robust they are.

Hiby and colleagues (2004) surveyed 364 dog owners. Rated obedience was positively associated with the number of tasks trained with reward-based methods, but not with punishment. The use of punishment was also associated with more problem behavior. As an owner survey, the study shows associations, not cause-and-effect relationships.

Rooney and Cowan (2011) went a step further and observed human–dog interactions directly instead of only asking questions. Dogs from households that relied more heavily on reward-based training performed better on a novel learning task.

Vieira de Castro and colleagues (2020) compared 92 dogs from seven dog training schools. Dogs from schools using aversive methods showed more stress-related behavior and a less favorable cortisol profile.

Taken together, this suggests that reward-based training is at least as effective for many training goals while being considerably less stressful.

The limitation belongs here too: the evidence consists mostly of observational and survey studies. Randomized comparisons, in which dogs are randomly assigned to training methods, would be ethically difficult and are correspondingly rare.

👉 Motivation in Dog Training · Reinforcement Schedules in Dogs · Aversive Training Methods in Dogs · The Fallout of Aversives

3. The four quadrants—and why only two of them are needed

First, a word of context, because the scheme is often mistaken for the whole picture: the four quadrants describe only the operant part of learning—that is, how consequences work. They say nothing about classical conditioning, social learning or emotional appraisal, all of which run alongside. The scheme is a grid for organizing one part of the picture, not modern learning theory itself.

Positive reinforcement is one of four possible cases. If that is the only one you know, you can't make sense of what is actually happening in everyday life—and above all, you'll miss the case that slips most quietly into ordinary equipment.

Two questions sort everything out: is something added or taken away? And does the behavior become more frequent or less frequent as a result?

3.1 What the terms really mean

Most of the confusion comes down to a single point: “positive” and “negative” are not value judgments here. They stand for plus and minus—adding and taking away. Negative reinforcement is not negative in the sense of bad—and it is certainly not punishment.

Positive reinforcement. Something is added, and the behavior becomes more frequent. The dog sits, and a treat appears.

Negative reinforcement. Something is taken away, and the behavior becomes more frequent. The pressure on the leash stops as soon as the dog gives in. For this to work, something unpleasant has to have been there beforehand—otherwise there would be nothing to take away. That is exactly what makes this case so tricky.

Positive punishment. Something is added, and the behavior becomes less frequent. A jerk on the leash, yelling, a startling stimulus.

Negative punishment. Something is taken away, and the behavior becomes less frequent. The dog jumps up, and the attention disappears; play is interrupted when it gets too rough.

3.2 The definition is about the effect, not the intention

This point is regularly overlooked, even though it is the actual content of the terms. Whether something was a reinforcer is not decided by what was intended, but by what happens to the behavior afterward.

An everyday example: the dog barks, the person scolds, the barking increases. Punishment was intended. Reinforcement is what happened—the attention was the consequence that mattered. Conversely, well-meant petting can have no effect at all on an insecure dog, because in that situation it is worth nothing to him.

This leads to a rule that saves a lot of time in training: if a behavior isn't developing in the direction you want, the consequence wasn't what you thought it was. The dog isn't making a mistake—your assessment was wrong.

3.3 Why negative reinforcement is the least conspicuous case

Positive punishment is argued about because it is visible. Negative reinforcement, by contrast, often goes unnoticed—and it does so in equipment and techniques that are considered harmless.

Any tool that applies pressure and releases it when the dog shows the desired behavior works on this pattern. For it to work, the state before must have been unpleasant. The dog reliably learns something in the process—but what he learns is how to make something unpleasant stop, not what pays off.

Such techniques can therefore produce rapid behavior change. Whether the behavior persists once the pressure is no longer part of the situation depends on the learning history, the context and what reinforces it afterward.

And there is a side effect that doesn't show up in the quadrant scheme: whatever feels unpleasant also gets linked through classical conditioning—to the place, the situation and, often enough, the person. That association can't be dosed.

3.4 What this means in practice

Modern training essentially works with two of the four fields: positive reinforcement for what you want to build, and—used sparingly and cleanly—negative punishment when you want a behavior to stop for which a reward would itself be the reinforcement.

The decisive difference between the two kinds of punishment is not how well they work but what they produce on the side. If you add something unpleasant, you create an emotional association you can't control. If you briefly take away something pleasant, that can also cause frustration, and situations, cues or people can become emotionally charged this way too—but the risk is typically lower and the emotional intensity smaller.

And in every case the same principle applies as above: suppression tells the dog what to stop doing. For him to know what he can do instead, an alternative behavior has to be built and reinforced.

4. Markers and cues: where classical and operant learning meet

A marker signal such as a click or a marker word gets its meaning through classical conditioning—the principle described by Pavlov (1927): it is paired with a reward until it announces one. A cue such as “sit” works differently. It becomes the signal that this particular behavior will now be reinforced—that is operant stimulus control. In practice, both processes run side by side, but they are not the same thing.

In training, we use this pairing to connect a clicker, a word or a hand movement with an upcoming reward. A cleanly built marker signal is a precise tool: it pinpoints the exact moment of the correct behavior, while the reward is allowed to come a moment later.

The order matters: first establish the behavior reliably, then put the cue in front of it. If you say the word while the dog is still guessing, you link it to confusion.

And a second rule follows from the same principle: a new cue should be built on its own. If you give it at the same time as one that is already established, the dog may not learn it at all—after all, the familiar cue already fully predicts the reward.

What a marker signal is not, however, is a superior tool in itself. Chiandetti and colleagues trained 51 dogs to open a lunch box—one group with a clicker and food, one with the spoken word “bravo” and food, and one with food alone. All three groups learned the task and transferred it to a new device to the same extent (Chiandetti, Avella, Fongaro & Cerri, 2016).

So the value of a marker lies not in the sound but in the precision it gives the human: it bridges the gap between the right moment and the reward. If you achieve that precision with a word, you don't need a device. If you don't achieve it at all, a clicker won't help either.

👉 Clicker Training for Dogs · Prediction Error in Dogs · Extinction in Dog Behavior

5. Extinction: why ignoring rarely works the way people think

A behavior that no longer has a consequence becomes less frequent. This process is called extinction, and it is the reason behind the most common piece of advice there is: just ignore it.

In principle the advice is right, and in practice it regularly fails because of three things.

5.1 First it may get worse

When the usual consequence fails to appear, the behavior may at first become more frequent, louder or more intense. This so-called extinction burst can occur, but it doesn't have to. If it does, it is not a setback but a common course—the dog first tries the proven strategy harder before giving it up.

In practice, this means that if it happens, most people give in exactly when it is most unpleasant. And with that, they haven't just failed to extinguish anything—they have rewarded the more intense version. That is the most common way to turn quiet demanding into loud demanding.

5.2 Extinguished behavior comes back

Extinction doesn't remove anything; it puts something on hold. After a break, the behavior often reappears on its own, and in the original setting it is especially easy to call up again.

If you expect this, you won't be caught off guard and won't slip back into old patterns. If you take it for failure, you start over from scratch.

5.3 Extinction alone leaves a gap

The most important objection, though, is a different one. A dog who used to bark to get something had a reason for doing so. If the behavior disappears without him knowing another way, the need remains.

That is why, in modern training, extinction never stands alone but always goes hand in hand with building an alternative behavior that achieves the same thing—just differently. A dog only gives up a strategy when he has a better one.

6. Reinforcement schedules: from every time to sometimes

At the beginning, every correct behavior is rewarded. That is usually the fastest way to build an association—but behavior learned this way tends to fade more quickly once the reward stops.

If the behavior is later rewarded only occasionally, it can become more resistant to extinction. The reason is obvious: if you are used to things not always working out, you don't stop the first time they don't.

Two points are crucial in practice. First, thinning out has to happen gradually. If you jump abruptly from every time to every fifth time, you risk the behavior breaking down instead of becoming more stable. Second, it usually helps if the reward becomes less predictable, not just less frequent. With a fixed pattern, the dog can learn the pattern along with the behavior—and then work mainly toward the repetitions where something pays off.

Every household knows the uncomfortable flip side: unwanted behavior often runs on exactly this schedule. If you usually stay firm when your dog begs at the table but occasionally hand something over after all, you haven't merely given in now and then—you have put the behavior on an intermittent schedule, and intermittent reinforcement can make behavior particularly resistant to extinction.

In practice, one sequence has proven useful: first build reliably with continuous reward, then thin out, then make it harder. If you make it harder and thin out at the same time, the behavior often falls apart—and you wonder why it doesn't hold up outdoors.

7. Generalization: why home doesn't count for outdoors

This is the point announced in section 2, and in everyday life it is the most common source of frustration. A dog who sits reliably in the living room doesn't do it in the park. That is not disobedience, nor is it proof that he “really can do it and just doesn't want to.”

The reason is that learning is bound to the context in which it took place. For the dog, everything that was present at the time belongs to the learned behavior: the room, the floor surface, the person's posture, the time of day, the bag with the treats. What we think of as a cue is, for him, a whole bundle.

7.1 Generalizing means varying deliberately

A behavior doesn't become independent of location because you practice it long enough. It does so because you practice it under changing conditions and the dog learns which part of the bundle actually counts.

In practice, this means changing only one variable at a time: first the same room facing a different direction, then a different room, then the yard, then a quiet street. If you increase location, distraction and distance all at once, you don't get progress—you get a breakdown.

And when you change location, it also means lowering the criteria and rewarding more often again. For the dog, a new place is a new task, not a repetition of the old one.

7.2 Stimulus control: when a cue is really a cue

The opposite direction to generalization is just as important. A cue is only under stimulus control when four things are true at the same time: the dog shows the behavior in response to the cue, he doesn't show it without the cue, he doesn't show it in response to a different cue, and he doesn't show a different behavior in response to this cue.

The second condition is the one most often overlooked. A dog who keeps sitting on his own to beg for food hasn't learned “sit” particularly well—he has learned that sitting pays off in general. That becomes obvious as soon as you want something else from him.

A useful test for practice follows from this: wait for a moment without saying anything. What the dog offers then is the behavior he considers the safest bet—and not necessarily the one you were about to ask for.

👉 Generalization in Dogs

8. Learning by observation

Dogs learn not only from their own experience but also by watching others.

The prerequisite is looking to the human. In a comparison with hand-raised wolves, Miklósi and colleagues (2003) showed that dogs faced with an unsolvable task turn to the human and seek eye contact, whereas wolves hardly do so. This human-directed attention is the precondition for social learning between humans and dogs to work at all.

Demonstrators make a difference. Pongrácz and colleagues (2001) had dogs solve a detour task around a fence. Animals that had first watched a person walk the route solved the task considerably faster.

And they don't imitate blindly but selectively. Range and colleagues (2007) had dogs watch a female demonstrator dog—so here the model was a conspecific, not a human—operate a lever with her paw instead of with the preferred muzzle.

When the demonstrator was carrying a ball in her mouth, using the paw had an understandable reason: only about one in five observer dogs adopted the unusual action. When there was no recognizable reason, however, more than 80 percent did.

This finding is more remarkable than it may seem at first glance: the dogs apparently took the circumstances of the observed action into account instead of simply copying it.

In practice, this means that observational learning can be used deliberately, for example in group classes or with a calm demonstrator dog. But it doesn't replace practicing yourself—your dog has to perform the behavior himself and be reinforced for it.

👉 Social Learning in Dogs · Overimitation in Dogs · Cognitive Abilities in Dogs

9. The role of emotions

Emotions are not a sideshow of learning but a prerequisite for it.

Under high arousal, attention narrows, and behavior becomes more rigid and less flexible. That is not unwillingness; it is physiology.

The precise wording matters here: a dog under stress doesn't learn nothing—he learns differently. Well-practiced, ingrained routines often still work. What gets harder is taking in new information and finding flexible solutions. That is exactly why training in agitated situations fails at the very thing you want to achieve there.

One clarification that is often missing is important here: the goal is not for the dog to be as calm as possible. Learning works best in a moderate range of arousal—too little arousal impairs attention just as too much impairs processing. And the favorable range shifts with the task: the more demanding it is, the lower it lies.

That is why modern approaches are emotion-based: they first make sure the dog stays below his stress threshold, and then they train.

👉 Emotion-Based Dog Training · Arousal Regulation in Dogs · Chronic Stress in Dogs · Leash Reactivity in Dogs

10. What neuroscience contributes

Learning processes are linked to the brain's reward system. Two messenger substances come up again and again in this context—with very different quality of evidence.

10.1 Dopamine and prediction error

Dopamine is not a happiness hormone; it is mainly involved in motivation, expectation and learning.

The key is the principle of prediction error: Schultz and colleagues (1997) showed that dopaminergic signals reflect not so much the reward itself as the deviation from what was expected. An unexpectedly good reward produces a strong signal; a fully expected one hardly produces any.

That explains why training that is always the same loses its effect and why surprises work.

For context: this work was not done in dogs but in primates. The basic mechanisms of the reward system are considered well established across mammalian species, but direct measurements in dogs are not available.

10.2 Oxytocin—and why caution is warranted here

Oxytocin can be released during friendly contact between humans and dogs. The widely cited study on this described a mutual increase during prolonged eye contact (Nagasawa et al., 2015).

However, this study does not hold up under methodological scrutiny. The crucial comparison arm with wolves was too small, and the owners were unevenly distributed by sex between the study arms—in a methodological commentary on the original paper that reanalyzed its data, the effect of owner sex was larger than the effect of species (Kekecs et al., 2016). This is not a separate measurement but a reexamination of the same data. Several follow-up studies were unable to confirm the mutual increase.

For training, this changes little: friendly contact is still right. What doesn't hold up is the idea that you can release bonding hormones through particular actions and thereby create a bond.

👉 Dopamine in Dogs · Oxytocin in Dogs · The Neurology of Dog Behavior

11. No two dogs learn the same way

Dogs differ in temperament and sensitivity, in their motivation, in learning speed, in past experiences and in genetic predispositions.

The last point deserves a clarification: behavioral traits are indeed heritable—but breed is a weak proxy for them and explains only a small part of the differences between individual dogs.

What matters is tailoring training to the individual human–dog team, not to a breed description.

👉 Breed and Behavior · Temperament, Personality, and Coping Styles in Dogs

12. What this looks like in practice

Clicker training. An acoustic signal marks the exact moment of the desired behavior. The clicker is classically conditioned and then used operantly—a good example of how both forms of learning mesh.

Recall training. One of the most important cues in everyday life, and one that can be reliably established with positive reinforcement and patient building.

Dummy training. Retrieve training uses existing motivation, without force and with lots of experiences of success.

Dogility. Obstacle training promotes coordination, self-confidence and the bond—playfully and without pressure.

Social walk. Structured group walks help dogs practice relaxed social behavior.

👉 Recall Training · Retrieve Training with Dummies · Brain Games and Scent Work

13. The challenges lie with the human

Time and repetition. Learning takes regular, short sessions. And here there is a counterintuitive finding from a study of 44 laboratory beagles learning one obedience task: with the same number of sessions, dogs trained once or twice a week showed better acquisition than dogs trained daily, and one session at a time did better than three sessions in a row (Demant, Ladewig, Balsby & Dabelsteen, 2011).

Two clarifications belong here. The effect concerns acquisition, not retention: the schedule had no effect on long-term memory. And what the animals had learned stuck remarkably well—after four weeks without any further training, they could still do the task.

For practice, this is good news: having little time doesn't necessarily put you at a disadvantage. What the study does not show is that training twice a week is generally better than daily training—it tested one specific task under laboratory conditions, not family dogs with everyday goals. What can be taken from it: spacing sessions out did not slow learning down there.

Timing. The reward acts on whatever came immediately before it—not on what was intended. The later it comes, the more likely it is to reinforce something else: getting up, looking over, stepping forward. There is no fixed cutoff in seconds—the shorter the delay, the easier it is for the dog to connect behavior and consequence, although depending on the task, available cues and learning history, dogs can also learn with delayed consequences. That is exactly what marker signals are for: they bridge the gap between the right moment and the reward.

Even more important than sheer speed is reliability. A behavior that is sometimes followed by something and sometimes not, even though it was performed the same way, is harder for the dog to figure out than one with a somewhat later but consistent consequence.

Consistency. Cues have to be clear and uniform. Changing words for the same exercise makes learning unnecessarily hard. And a cue should never announce both something good and something unpleasant—otherwise it loses its value.

Know-how. The better you understand how your dog learns, the more purposefully you can intervene.

Self-regulation. Your own state affects learning progress.

Realistic expectations. Setbacks in new places are the norm, not a failure—section 7 explains why.

👉 Emotional Contagion in Dogs · Controllability and Predictability in Dogs · Why Dog Training Doesn't Work

14. What the research does not show

• The evidence rests mostly on observation and surveys. Randomized comparisons of different training methods are ethically difficult and hardly exist.
• The classical foundations do not come from dog training. Skinner worked with rats and pigeons; Pavlov did work with dogs, but in laboratory experiments on the salivary reflex, not in training. They provide the framework; the dog-specific training findings come from other studies.
• Prediction error has not been measured in dogs. The findings come from recordings in primates; in dogs, only compatible imaging data are available.
• The oxytocin study does not support its interpretation. Anyone who cites it should cite the methodological reexamination along with it.
• The quadrants are a classification scheme, not a finding. They describe how consequences work and, on their own, say nothing about what will serve as a reinforcer for a particular dog. That can only be observed, not deduced.
• The findings on training frequency concern simple tasks. What was studied was the learning of manageable exercises under controlled conditions. Whether the same applies to behavior therapy for fear or aggression has not been shown.
• “Positive reinforcement always works” would be too strong. What is established is that, for many training goals, it is at least as effective and less stressful.

15. Conclusion

Modern learning theory is not a rigid set of rules but an attitude. It means seeing the dog as an individual capable of learning and guided by emotion, who grows not through pressure but through understanding and motivation.

The evidence is fairly clear on this: dogs learn best when they feel safe, when they understand what is meant, and when they succeed.

And the sentence that sums up the whole thing: suppression tells the dog what to stop doing. Reinforcement tells him what he can do instead.

👉 Force-Free Dog Training · Building Trust with Your Dog · Dog Trainer or Behavior Consultant?

Key takeaways on modern learning theory

Learning never runs on consequences alone. Operant and classical conditioning, social learning and emotional appraisal all operate at the same time. If you only think about reward and punishment, you miss half of it.

For many goals, reward-based training is at least as effective and less stressful. This is shown by surveys, direct observation and cortisol profiles (Hiby et al., 2004; Rooney & Cowan, 2011; Vieira de Castro et al., 2020)—mostly observational data, not randomized comparisons.

“Positive” and “negative” are arithmetic signs, not value judgments. They stand for adding and taking away. Negative reinforcement requires that something unpleasant was there beforehand—which is why it is the least conspicuous of the four cases.

The dog decides what is a reinforcer. The definition is about the effect, not the intention. If a behavior doesn't develop as intended, the consequence wasn't what you thought it was.

Under extinction, behavior may initially increase before it declines—and extinguished behavior can come back. Without an alternative behavior built in its place, a gap remains, because the need is still there.

Occasional reward can make behavior durable. That applies to wanted behavior as much as unwanted behavior—intermittent reinforcement can make behavior particularly resistant to extinction, including begging that is occasionally rewarded.

The clicker is not the active ingredient. Clicker, spoken word and food alone led to the same learning success in 51 dogs (Chiandetti et al., 2016). What counts is the precision of the marking, not the device.

Prediction error explains why routine loses its effect. Dopaminergic signals reflect the deviation from expectation, not the reward itself (Schultz et al., 1997)—measured in primates, not in dogs.

Less training can be more. In laboratory beagles learning one task, training once or twice a week and one session at a time led to better acquisition than daily training or three sessions in a row (Demant et al., 2011)—a finding for that setting, not a general rule.

Stress doesn't prevent learning; it changes it. Ingrained routines keep working; what suffers is taking in new things and finding flexible solutions.

References

Chiandetti, C., Avella, S., Fongaro, E., & Cerri, F. (2016). Can clicker training facilitate conditioning in dogs? Applied Animal Behaviour Science, 184, 109–116. https://doi.org/10.1016/j.applanim.2016.08.006

Demant, H., Ladewig, J., Balsby, T. J. S., & Dabelsteen, T. (2011). The effect of frequency and duration of training sessions on acquisition and long-term memory in dogs. Applied Animal Behaviour Science, 133(3–4), 228–234. https://doi.org/10.1016/j.applanim.2011.05.010

Hiby, E. F., Rooney, N. J., & Bradshaw, J. W. S. (2004). Dog training methods: Their use, effectiveness and interaction with behaviour and welfare. Animal Welfare, 13(1), 63–69. No DOI has been assigned to this work.

Kekecs, Z., Szollosi, A., Palfi, B., Szaszi, B., Kovacs, K. J., Dienes, Z., & Aczel, B. (2016). Commentary: Oxytocin-gaze positive loop and the coevolution of human–dog bonds. Frontiers in Neuroscience, 10, 155. https://doi.org/10.3389/fnins.2016.00155

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. https://doi.org/10.1016/S0960-9822(03)00263-X

Nagasawa, M., Mitsui, S., En, S., Ohtani, N., Ohta, M., Sakuma, Y., Onaka, T., Mogi, K., & Kikusui, T. (2015). Oxytocin-gaze positive loop and the coevolution of human-dog bonds. Science, 348(6232), 333–336. https://doi.org/10.1126/science.1261022

Pavlov, I. P. (1927). Conditioned reflexes: An investigation of the physiological activity of the cerebral cortex. Oxford University Press. No DOI has been assigned to this book.

Pongrácz, P., Miklósi, Á., Kubinyi, E., Gurobi, K., Topál, J., & Csányi, V. (2001). Social learning in dogs: The effect of a human demonstrator on the performance of dogs in a detour task. Animal Behaviour, 62(6), 1109–1117. https://doi.org/10.1006/anbe.2001.1866

Range, F., Viranyi, 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

Rooney, N. J., & Cowan, S. (2011). Training methods and owner–dog interactions: Links with dog behaviour and learning ability. Applied Animal Behaviour Science, 132(3–4), 169–177. https://doi.org/10.1016/j.applanim.2011.03.007

Schultz, W., Dayan, P., & Montague, P. R. (1997). A neural substrate of prediction and reward. Science, 275(5306), 1593–1599. https://doi.org/10.1126/science.275.5306.1593

Skinner, B. F. (1938). The behavior of organisms: An experimental analysis. Appleton-Century-Crofts. No DOI has been assigned to this book.

Vieira de Castro, A. C., Fuchs, D., Morello, G. M., Pastur, S., de Sousa, L., & Olsson, I. A. S. (2020). Does training method matter? Evidence for the negative impact of aversive-based methods on companion dog welfare. PLOS ONE, 15(12), e0225023. https://doi.org/10.1371/journal.pone.0225023

Notes on the sources

The statements on training frequency were checked against the publication (Demant et al., 2011): training once or twice a week led to better acquisition than daily training, and one session at a time better than three in a row. The study involved 44 laboratory beagles, one obedience task and 18 sessions per dog; whether the result transfers to other tasks, family dogs or other schedules was not tested. The effect concerned the acquisition phase; the schedule had no effect on long-term memory. The animals still mastered the task after four weeks without further training.

Kekecs et al. (2016) is not a study of its own but a methodological commentary that reanalyzes the data of the original paper. It is labeled accordingly here and is not presented as a primary finding.

On Chiandetti et al. (2016): 51 dogs, three conditions—clicker plus food, spoken word plus food, food alone. The task was opening a lunch box, followed by a test on a new device. All three groups learned and transferred to the same extent.

The presentation of the four quadrants and of the sections on extinction and reinforcement schedules follows the established textbook knowledge of learning psychology. It is not based on individual dog-specific studies and is therefore given without citations; the linked individual articles cover the respective evidence in detail.

The statements on Range et al. (2007) were checked against the publication. It should be stressed that the demonstration was performed by a conspecific, not a human: the study demonstrates selective imitation between dogs. The proportions were about 21 percent adopting the unusual action when a reason for it was recognizable, compared with over 83 percent when it was not.

The oxytocin passage requires a correction compared with common accounts. The study by Nagasawa et al. (2015) is often cited as dog-specific evidence for a bonding feedback loop. The methodological commentary by Kekecs et al. (2016), however, shows from the original data that the wolf comparison arm was too small for the conclusions drawn and that the unequal sex distribution of the owners between the study arms produces an effect larger than the species effect. Several follow-up studies were unable to confirm the mutual increase. The claim that “the oxytocin findings are established as dog-specific” is therefore not supported.

Hiby et al. (2004) is an owner survey and shows associations, not causes. Rooney and Cowan (2011) use direct behavioral observation and are more robust in that respect, but remain a cross-sectional study. Vieira de Castro et al. (2020) compares existing training groups without random assignment. For the overall conclusion, therefore: the direction is consistent across several independent approaches; causality, methodologically, is not established.

The figures on training frequency and the number of sessions in a row come from a study of laboratory beagles and are listed individually in the article on the neurology of dog behavior.

Skinner (1938) is a foundational work based on experiments with rats; his later work used pigeons. Pavlov (1927) did work with dogs, but in laboratory experiments on the salivary reflex, not in training. Both provide the theoretical framework; the transfer to dog training rests on the cross-species validity of the principles of learning.

Frequently asked questions about modern learning theory in dog training