Controllability and Predictability in Dogs: Two Variables, Often Confused

1. Four Terms That Need to Stay Separate

1.1 Predictability

Predictability means that information in the environment reduces uncertainty about what is likely to happen next. The information may be temporal - an event tends to occur at a particular time - or signalled, where a cue precedes it. Predictability does not imply that the animal can alter the event.

1.2 Controllability

Controllability means that the animal's behavior reliably changes an outcome: an action can start, stop, postpone, shorten, avoid or otherwise alter what happens. In welfare science, control is therefore about a response-outcome contingency, not simply about whether the animal was offered two objects or two routes.

1.3 Choice

Choice and control overlap, but they are not synonyms. A choice provides meaningful control only when the available actions reliably lead to different outcomes that matter to the animal. Choosing between two options that both end in the same compulsory procedure may provide preference information without giving control over the procedure itself (Englund & Cronin, 2023).

1.4 Agency

Agency is the broader concept: the animal can act and have those actions make a difference in its environment. Choice and control can contribute to agency, but neither term should be used as a vague synonym for "being nice to the dog." Keeping the constructs separate matters because different studies manipulate different combinations of them.

1.5 Why Predictability and Control Get Merged

In ordinary life they often travel together. A dog that knows a routine may also know what behavior makes it continue or stop. But a dog can know exactly what is about to happen and still be unable to affect it. Conversely, a dog may be able to stop an interaction without knowing in advance exactly when it will begin.

The distinction is therefore not academic. A handling protocol can be perfectly predictable and still be uncontrollable, while a free-choice interaction may increase control but simultaneously change movement, distance and exposure. Those differences determine what a study can legitimately claim.

1.6 Detecting a Contingency Is Part of the Problem

Control only exists functionally if the dog can detect the relationship between its behavior and the outcome. Predictive cues likewise need to be sufficiently reliable and salient to be learned. There is no universal number of seconds that makes a cue useful, and intermittent outcomes should not automatically be described as "gambling." The relevant question is whether the dog can learn the contingency under the arrangement being used.

1.7 Why the Pair Is Worth a Whole Article

Few concepts in applied behavior work are invoked as often and examined as rarely. "Give the dog control" and "dogs need predictability" appear in course descriptions, welfare guidelines and marketing copy, usually as self-evident goods and usually without anyone saying which of the two is being supplied.

The distinction is not academic. A handling protocol can be perfectly predictable and leave the dog no influence at all, and the resulting behavior — a dog that holds still throughout — is easily mistaken for a successful outcome.

2. Where the Concept Came From

2.1 The Original Dog Experiments

The learned-helplessness literature began with experiments in which dogs exposed to inescapable electric shock later showed impaired escape or avoidance when escape became possible (Overmier & Seligman, 1967; Seligman & Maier, 1967). The work established that previous experience with aversive events could alter later responding.

The procedures involved severe aversive stimulation and would not meet modern welfare standards. They are relevant historically because the entire discussion of uncontrollability grew from them, not because they provide a model to reproduce.

2.1a Why the Origin Has to Be Named

The founding experiments restrained dogs and delivered shock they could not escape. There is no way of citing this work that makes that acceptable, and no honest way to discuss control in animals without it: the concept exists because those experiments were done (what the learned-helplessness work showed).

Two things follow. The findings are treated here as established knowledge that does not need producing again, and the modern successors deliberately avoid the design — which is also why the canine evidence base has grown so slowly since.

2.2 Why Yoked Designs Matter

A yoked design gives two subjects closely matched physical outcomes while arranging different response-outcome contingencies. One subject can produce or terminate an outcome; the yoked subject receives a matched outcome that is not contingent on its own behavior.

This is a powerful way to study control, but the interpretation still requires care. The physical outcome may be equated while feedback, response-outcome contingency and subjective predictability differ. A yoked design therefore isolates an experimentally defined contingency better than ordinary observation does, but it does not automatically reveal the animal's internal representation of that contingency.

2.3 Predictability Needed Its Own Manipulation

This distinction became especially important because controllability and predictability had often been confounded in earlier research. Dess et al. (1983) addressed that directly in dogs by manipulating the two variables independently. That experiment remains the clearest canine demonstration that the variables can produce different physiological effects.

3. The 2016 Neurobiological Reinterpretation

3.1 What Maier and Seligman Changed

Maier and Seligman (2016) revisited learned helplessness using decades of neuroscience. In the rodent model they reviewed, prolonged aversive stimulation activates stress-related systems including the dorsal raphe nucleus. The newer account does not require an animal to learn "nothing I do matters" in order for passivity and anxiety-like effects to emerge. Instead, detected control recruits medial prefrontal circuitry that can inhibit parts of that response.

3.2 It Is a Reinterpretation, Not a Reversal of the Behavioral Finding

The behavioral phenomenon was not erased. What changed was the proposed mechanism. The older account emphasized learning uncontrollability; the revised rodent account emphasizes learning control and its protective neural consequences.

Calling this a complete reversal can therefore be misleading. It is more accurate to call it a neurobiological reinterpretation of why uncontrollable aversive experience has the effects that it does.

3.2a Why the Distinction Reaches Ordinary Training

If detecting control is what is learned, then any arrangement in which a dog's behavior reliably produces an outcome contributes something beyond the specific behavior being taught. That is a claim about training in general rather than about a particular method, and it is the strongest theoretical argument available for contingency-based work (how contingencies are arranged in practice).

It also sets a quality standard for that work. A contingency the dog cannot detect, or one that holds only sometimes, does not deliver what the account describes (how expectation and outcome interact).

3.3 The Species Boundary Matters

The neural circuitry discussed by Maier and Seligman was worked out primarily in rodents. Comparable circuitry has not been demonstrated experimentally in dogs. It is reasonable to treat the rodent model as a mechanistic hypothesis relevant to mammals, but not as a direct canine finding.

3.4 What This Does Not Let Us Diagnose

A motionless dog at a clinic cannot be diagnosed as "learned helpless" from stillness alone. Stillness may reflect voluntary cooperation, freezing, behavioral inhibition, fatigue, restraint, learned immobility or another state. The important lesson is symmetrical: stillness does not prove acceptance, but it also does not prove helplessness (behavior is not the same thing as emotion).

4. The Best Direct Dog Evidence: Dess et al. 1983

4.1 Control and Predictability Were Manipulated Separately

Dess, Linwick, Patterson, Overmier and Levine (1983) exposed 18 adult male dogs to an aversive stress-induction procedure in which controllability and predictability were independently varied. Plasma cortisol was measured during that phase. Later, the dogs received identical shocks in a novel context so that previous experience could be related to a new cortisol response.

4.2 Immediate Cortisol: Control Mattered, Predictability Did Not

During stress induction, uncontrollable shocks produced significantly greater cortisol elevations than controllable shocks. Predictability did not produce a significant immediate cortisol effect in that phase (Dess et al., 1983).

4.3 Later Cortisol: Predictability Mattered, Prior Control Did Not

The pattern changed at the later test. Previous exposure to unpredictable shock increased cortisol responding to novel test shocks, whereas prior controllability did not significantly alter that later cortisol response (Dess et al., 1983).

4.4 Why This Study Matters

The result is not that control is always more important than predictability or vice versa. It is almost the opposite: the variables affected different parts of the stress response in this experiment. That is the strongest reason to stop treating them as interchangeable welfare slogans.

4.5 What It Does Not Tell Us

The study used severe aversive stimulation, a small and highly specific sample, and cortisol as its central outcome. It does not tell us how much everyday choice a pet dog needs, whether a start-button behavior protects against later veterinary stress, or whether the same pattern appears with appetitive events.

5. Predictability on Its Own

5.1 There Is No Single "Predictability Effect"

Bassett and Buchanan-Smith (2007) reviewed predictability across captive animals and distinguished temporal predictability from signalled predictability. Outcomes were not uniform. Effects depended on the type of event, the type of cue, the species and what other forms of control were available.

5.1a Temporal and Signaled Predictability Are Different Arrangements

Knowing that something happens at a fixed time and being told each time that it is about to happen are not the same provision. Strictly fixed timing gives an animal a precise moment to become aroused about, and any deviation is itself an event.

For households the practical version is familiar: reliable sequences with flexible clock times are usually easier to live with than a fixed schedule the day cannot always honor (how dogs organize their day).

5.2 Safety Signals Are a Useful Model, Not a Universal Rule

A cue that reliably announces an aversive event can make non-cue periods informative: the absence of the cue may indicate relative safety. But the literature does not justify the blanket claim that an imperfect cue is always worse than no cue at all. Reliability matters because it changes information value, not because there is a universal cutoff at which predictability becomes harmful.

5.3 Predictable Pleasant Events Can Also Produce Anticipation

Predictability of appetitive events can support anticipatory behavior. Whether that anticipation is experienced positively, becomes frustrating, or is behaviorally neutral depends on the arrangement. Fixed schedules and variable schedules can therefore have different effects without one being universally superior.

5.4 Predictable Aversive Events Are Still Aversive

A cue for an unpleasant procedure can itself acquire aversive value through conditioning. That does not mean unpredictability would be better. The practical implication is to distinguish the informational value of a cue from the valence of what it predicts (how predictive fear learning works).

5.4a When Predictability Backfires

A reliable signal for something unpleasant remains a signal for something unpleasant. Where the announced event is aversive and entirely uncontrollable, the cue itself becomes a source of distress, which is the ordinary picture in dogs that react to the sight of the clippers or the route to the clinic.

The answer is not to make the event unpredictable, which removes the safe periods and makes matters worse, but to change what the cue predicts or to give the dog something to do about it (how predictive cues are worked with).

5.5 Household Routines Should Not Be Oversold

Consistent sequences can make daily life easier to interpret, but the evidence does not establish a universal rule that flexible clock times are better than fixed times or that a particular delay between cue and event is optimal. Those are management choices that should be assessed by the dog's behavior rather than presented as direct products of the predictability literature.

6. Modern Canine Evidence: What Has Actually Been Tested

6.1 Housing and Restriction

Beerda and colleagues documented behavioral, hormonal and immunological changes in dogs exposed to sustained social and spatial restriction (Beerda et al., 1999a, 1999b). These studies are relevant to welfare, but they do not isolate controllability. Housing, social contact, space and environmental stimulation changed together.

They therefore support the broader conclusion that restrictive environments can have measurable costs, not the narrower claim that those costs were specifically caused by lack of control.

6.1a What Restriction Studies Can and Cannot Attribute

Social and spatial restriction changes many things at once: space, company, stimulation and the ability to withdraw. The physiological and behavioral changes documented in those dogs are therefore consistent with an effect of uncontrollability without isolating it (what chronic stress does).

The framing is still useful for kennels and shelters, because it reorders priorities: choice of resting place, the option to withdraw and some influence over social contact are not extras added after the basics but part of them.

6.2 Training-Method Comparisons

Vieira de Castro et al. (2020) reported poorer welfare-related measures in dogs trained in schools using more aversive-based methods compared with reward-based schools. Such comparisons matter, but they simultaneously alter reinforcement history, aversive exposure, predictability, handler behavior and response-outcome relationships. They cannot identify control as the causal variable.

6.3 McGowan 2014: Producing a Reward Versus Receiving It

McGowan et al. (2014) used 12 beagles in a yoked problem-solving design. Dogs in the experimental condition learned operant responses that produced access to food, human contact or dog contact. In the yoked condition, the same kinds of rewards were delivered after delays matched to a partner rather than being contingent on the dog's own action.

Dogs in the achievement condition showed more activity and tail wagging, whereas control conditions included frustration-like behaviors such as chewing at the apparatus. This is important positive-valence evidence that contingent achievement differs behaviorally from receiving a similar reward without producing it.

6.3a Why the Eureka Design Matters Beyond Its Size

Almost all of this literature rests on aversive events, which limits both what can be studied today and what can be concluded about ordinary life. A design that equates the reward and varies only whether the dog produced it is therefore disproportionately valuable, and it supplies the template for the studies that are missing (how affective state is measured independently).

6.4 McGowan Does Not Isolate "Agency Only"

The control condition was not simply identical experience minus an abstract sense of agency. For the control dog, reward timing was not contingent on its behavior and was therefore less predictable from its own actions. The authors themselves discussed frustration arising from that unpredictability. The study is best described as a test of contingent achievement versus yoked, noncontingent reward, not as a pure manipulation of agency with every other psychological variable held constant.

6.5 Cooperative Care: Wess et al. 2022

Wess et al. (2022) conducted a blinded controlled pilot study of cooperative-care training around veterinary examination. Forty-seven dogs entered and 40 completed the study. The training group received repeated cooperative-care sessions plus home training between two standardized veterinary visits.

The results were mixed rather than a simple success story. Clear beneficial HR/HRV effects were not demonstrated, transfer from training to the blinded veterinary examination was limited, and compliance did not uniformly improve. The study is valuable precisely because it shows that cooperative-care training as a package cannot be assumed to produce every expected physiological or procedural benefit.

6.6 Adaptive and Collaborative Veterinary Handling: 2023-2024

Squair and colleagues randomized 28 dogs to routine veterinary care or an intervention package that combined modified handling, low-stress procedures and collaborative-care elements over four visits. Physiologically, the intervention group showed a greater reduction in cortisol from the first to the fourth visit and an improvement in a composite stress index, although heart rate and several other individual measures did not differ between groups (Squair et al., 2023).

Behaviorally, the intervention group later showed lower clinical stress scores during several parts of the examination than controls (Squair et al., 2024). This is relevant evidence that changing the overall veterinary interaction can improve outcomes. It does not isolate the specific causal contribution of choice, control, predictability or any one opt-out behavior because the intervention deliberately changed multiple components together.

6.7 Free-Choice Versus Forced Touch: Sarrafchi et al. 2025

Sarrafchi et al. (2025) compared forced and free-choice touch in 18 certified therapy dogs interacting with 44 people. During forced touch, dogs were leashed and continuously touched. During the free-choice condition, dogs could move freely and people touched them only when they came within reach.

Ear-back behavior occurred more frequently during forced touch. During free-choice trials, dogs showed more avoidance and sniffing - behaviors that were possible because movement was available - yet they still spent 77.9% of the time within reach of participants.

The study provides direct modern evidence relevant to choice and voluntary contact. It is not a pure controllability experiment because leash restriction, movement freedom and amount of touch changed together. The authors themselves identify leash condition as a confound.

6.8 A Coping Option During Training: Lukavská et al. 2026

Lukavská et al. (2026) tested operant training with or without a previously reinforced "Default Variant of Behaviour" - a simple, low-effort behavior available to the dog as an alternative during training. Dogs with access to this option showed reductions in observed stress-related behaviors compared with training without it. Cardiovascular measures did not show the same significant pattern.

This study is particularly useful because it is modern, non-aversive and directly concerned with a control-like coping option. It also warns against reducing welfare to one measurement channel: behavioral and cardiovascular stress indicators did not move in lockstep.

6.9 What the Modern Studies Add

It is no longer accurate to say that ethical modern canine research has not manipulated anything relevant to choice or control. It has. But each study changes more than one construct or tests a package rather than an isolated mechanism. The evidence therefore supports a more precise statement: choice-, collaboration- and control-related arrangements can alter canine behavior and some stress-related measures, while the independent contribution of a specific controllability mechanism remains incompletely isolated.

7. What This Means for Cooperative Care

7.1 Cooperative Care Is a Package

Cooperative care usually combines desensitization, reinforcement, predictable sequences, trained stationing, start cues, pauses, opt-outs, changed handling and altered session length. When the package works, the scientific question is not whether the dog benefited, but which components were responsible for how much of the effect.

7.2 The Opt-Out Question Is Still Open

The central applied claim has not yet been cleanly isolated: the same procedure, with the same handling and reinforcement, compared under a reliably honored interruption signal versus no interruption option. Existing veterinary studies support collaborative approaches as packages, but they do not answer that narrower causal question.

7.2a Predictability Is Cheap, Control Is Not

One asymmetry explains much of practice. Making something predictable costs only consistency: the same cue, the same order, the same ending. Making something controllable requires giving up the ability to proceed whenever it suits the handler.

That is why predictability is everywhere in applied advice while control is largely rhetorical. A household can deliver the first by deciding to; the second means accepting that a session sometimes ends earlier than planned (what cooperative care research shows).

7.3 Control Has to Be Functional

A "stop" behavior provides control only if the dog's response reliably changes what happens. If a withdrawal or stop response is sometimes ignored, the arrangement is not reliably controllable from the dog's perspective. That statement follows from the definition of control; it does not require claiming that a single ignored signal produces helplessness.

7.3a Where Control Should Not Be Handed Over

Some situations cannot be decided by the dog's preference: medication that has to be given, a wound that has to be treated, a road that has to be crossed. Pretending otherwise produces a worse outcome than being honest about it.

What remains available in those cases is everything around the event — the announcement, the bounded duration, the recovery afterwards, and the many other situations in the same week where the dog's behavior can be allowed to matter (what handling without agency costs).

7.4 Stillness Is Not the Outcome Measure

A successful procedure should not be defined solely as one during which the dog remains motionless. Relevant outcomes include voluntary approach, ability to disengage, latency to re-engage, body language, completion of the procedure, recovery afterwards and, where feasible, physiological measures. No single marker establishes emotional state.

8. Practical Use Without Overselling the Evidence

8.1 Build Reliable Response-Outcome Contingencies Where They Are Safe

Low-cost opportunities for influence include choosing whether to approach social contact, selecting among safe walking directions, leaving a resting place, or ending optional handling. These arrangements are plausible welfare improvements and are consistent with contemporary agency frameworks. They should not be marketed as proven immunization against future stress.

8.1a Agency Is Not the Same as Fewer Rules

Control here means a contingency between behavior and outcome, not an absence of structure. A dog left to do as it likes in a situation it cannot influence has been given freedom without agency, which is the arrangement most likely to produce the passivity the concept warns about.

What the dog needs is a response that works, not fewer boundaries. The two are regularly conflated in discussions of choice in training (why the mechanism changes the plan).

8.2 Make Unavoidable Events Legible

Medication, emergency treatment and some veterinary procedures cannot always be made optional. Predictive cues, clear beginnings and endings, gradual preparation and reduced restraint can still improve how interpretable the event is. A predictable event is not automatically pleasant, but predictability can be one useful component of the plan.

8.3 Nail Care

For nail care, distinguish the variables explicitly. A repeated station and consistent sequence provide predictability. A dog that can move out of the station and thereby pause the procedure may also have control. Whether that control produces an independent long-term benefit beyond the rest of the training package remains an open experimental question.

8.4 Veterinary Handling

The modern veterinary studies justify using collaborative, low-stress handling approaches more strongly than older versions of this article could. They do not justify claiming that one start-button behavior is the proven active ingredient. Measure the whole dog and the whole procedure rather than the presence of a fashionable cue.

8.5 Touch and Social Interaction

The 2025 therapy-dog study supports allowing voluntary approach and avoidance during touch where the setting permits it. But because free-choice and forced-touch conditions differed in leash status and movement freedom, the study should not be translated into a universal rule that every touch interaction must follow the same setup.

8.6 Alone Time

Departure cues can be predictable without being helpful if they already predict distress. For dogs with separation-related problems, announcing departure is not automatically therapeutic; the meaning of the cue depends on learning history. Predictability should therefore be assessed rather than assumed (what is known about separation-related behavior).

8.6a Why Alone Time Is the Hardest Case

Departure is the situation in which a dog has the least of both: it cannot predict when the person returns and cannot influence whether they go. What can be arranged is the announcement, a bounded and consistent return ritual and free choice of where to settle (what is known about separation-related behavior).

How far that helps has not been tested directly. It follows from the two variables rather than from a canine study, which is the appropriate way to present it to a household (what dogs register about duration).

8.7 Leash Encounters

A leash encounter is not by definition unpredictable and uncontrollable. Some dogs receive early information, can choose distance through the handler's movement and have rehearsed responses that alter the interaction. Others do not. The useful question is case-specific: what can the dog anticipate, and which of its responses reliably change the situation?

8.8 What to Record

For a clinical case, record both constructs separately. What predicted the event? How reliable was that cue? What actions were available to the dog? Which actions actually changed the outcome? Then record behavioral response, recovery and whether the pattern changes across repeated representative situations.

A fixed one-week logging period is not a validated diagnostic rule. The goal is enough representative observation to identify recurring contingencies.

8.8a What This Looks Like in a Week

Concretely, a household working on this changes a handful of small things. The dog decides the direction at two junctions on a walk. Handling sessions end when the dog moves away, not when the handler is finished. The departure gets a consistent announcement and a consistent return. Feeding and the main rest period keep their order even when the clock moves.

None of it is dramatic and none of it requires equipment. The point is not the individual measures but that the dog's behavior starts making a difference to several things a day rather than to none.

8.9 What Progress Looks Like

No single measure is sufficient. Useful changes can include easier voluntary approach, lower behavioral stress scores, faster recovery, less need to disengage, improved procedural completion and stable physiological measures. Fewer opt-outs may reflect greater comfort - or reduced signaling - so the number should never be interpreted on its own.

8.10 A Note on Language

"Giving the dog a choice" is the phrase in circulation and it is looser than the concept requires. A choice between two options a handler has arranged is not the same as a contingency in which the dog's behavior changes what happens, and the second is what this literature is about (Englund & Cronin, 2023).

Using the precise term costs nothing and prevents the most common misapplication: a setup that offers alternatives while the procedure proceeds regardless (why the term decides what is being claimed).

9. What We Can Say With Confidence

Predictability and controllability are distinct constructs. They can be manipulated separately and can affect different outcomes.

Dess et al. 1983 remains the clearest direct dog experiment separating them. Uncontrollable shock increased immediate cortisol more than controllable shock; prior unpredictability, rather than prior uncontrollability, increased cortisol responding to later novel shock.

The 2016 mechanism is primarily rodent evidence. It is important mechanistic work, not direct proof of the same neural process in dogs.

Modern dog research is no longer absent. Problem-solving, veterinary handling, free-choice touch and operant training now provide direct canine data relevant to choice, collaboration and control-like arrangements.

Those modern studies do not isolate a single "control molecule." Movement freedom, contingency, handling style, reinforcement, predictability and social context often change together.

Cooperative care has empirical support as an overall approach, but the independent effect of a reliably respected opt-out remains unresolved.

Stillness alone is not consent and is not a diagnosis of helplessness.

10. Research Gaps and Critical Appraisal

The direct canine control-by-predictability experiment is still old and aversive. Dess et al. (1983) remains unusually informative because it independently manipulated both variables, but its procedure limits direct translation to companion-dog care.

Neural transfer remains untested. The prefrontal-dorsal-raphe account central to the modern learned-helplessness model has not been demonstrated in dogs.

The opt-out effect has not been isolated. No study has yet compared an otherwise identical handling procedure with versus without a reliably honored interruption response while measuring both behavioral and physiological outcomes.

Long-term transfer is still largely open. It remains unclear whether repeated benign experiences of control in daily life protect a dog in a later unrelated stressful context.

Dose is unknown. There is no established amount, frequency or distribution of control that produces a clinically meaningful welfare effect.

Individual differences need more work. The 2026 training study adds evidence of individual variability and illustrates that behavioral and cardiovascular responses can diverge, but moderators such as history, temperament and clinical status remain poorly mapped.

Choice and control still need cleaner separation. Free-choice interaction studies often change leash status, movement, distance and exposure at the same time. Future designs should preserve the benefits of ethical testing while isolating the response-outcome contingency more precisely.

Frustration is the neglected side. An interrupted contingency — a response that used to work and now does not — produces its own state, and the transition from control to no control has not been examined in dogs (how frustration is described).

Generalization is assumed. Whether agency built in ordinary situations transfers to a clinic or a grooming table is the practically decisive question and has not been tested in this species (why transfer cannot be assumed).

10.1 The Most Informative Next Experiments

Three studies would answer unusually practical questions. First, compare an otherwise identical low-intensity handling procedure with and without a reliably honored opt-out. Second, give dogs a period of benign contingency training and test whether that history changes response to a later standardized, mildly challenging but ethical event. Third, extend yoked appetitive designs so that the reward is equated while the dog's ability to produce it is varied across multiple contexts and measured with both behavior and physiology.

Modern work has already moved the field beyond "there are no ethical dog data." The next step is not more slogans about choice. It is cleaner experimental separation of the variables practitioners already claim to use.

11. Conclusion

Predictability and controllability are related but different. Predictability reduces uncertainty about what is coming; controllability means the dog's behavior can reliably alter what happens. Dog research has demonstrated that the distinction matters: in the 1983 experiment the two variables influenced cortisol in different ways. The influential 2016 reinterpretation of learned helplessness adds a mechanistic model in which learned control has protective effects, but that neural account comes primarily from rodents and should not be written as a canine fact.

The modern dog literature is now broader than a single old shock experiment. Contingent problem-solving differs behaviorally from yoked reward delivery; collaborative veterinary-care packages can reduce some behavioral and physiological stress measures; free-choice touch changes how therapy dogs behave; and a control-like coping option during operant training can reduce observable stress behavior. At the same time, these studies do not yet prove that one isolated opt-out or a general quantity called "agency" is the causal ingredient behind every benefit.

The practical conclusion is deliberately modest. Make important events as interpretable as possible. Where it is safe, let the dog's behavior genuinely affect optional interactions. Measure more than stillness. And when calling something "control," check that the dog's response actually changes the outcome.

Key Insights (Takeaways)

• Predictability, controllability, choice and agency overlap, but they are not interchangeable terms.

• Dess et al. (1983) showed directly in dogs that controllability and predictability can affect cortisol differently.

• The modern learned-helplessness mechanism comes mainly from rodent neuroscience; its neural details have not been demonstrated in dogs.

• McGowan et al. (2014) showed that producing a reward differs behaviorally from receiving a yoked, noncontingent reward, but control and predictability are not perfectly separated in that design.

• Cooperative and collaborative veterinary approaches now have controlled canine evidence as packages, while the independent effect of a respected opt-out remains unisolated.

• Free-choice touch and a control-like coping option in operant training provide modern canine evidence relevant to agency and welfare, but each design has important confounds.

• Stillness proves neither consent nor learned helplessness. Use multiple behavioral and, where possible, physiological outcomes.

References

Bassett, L., & Buchanan-Smith, H. M. (2007). Effects of predictability on the welfare of captive animals. Applied Animal Behaviour Science, 102(3-4), 223-245. https://doi.org/10.1016/j.applanim.2006.05.029

Beerda, B., Schilder, M. B. H., Bernadina, W., van Hooff, J. A. R. A. M., de Vries, H. W., & Mol, J. A. (1999b). Chronic stress in dogs subjected to social and spatial restriction. II. Hormonal and immunological responses. Physiology & Behavior, 66(2), 243-254. https://doi.org/10.1016/S0031-9384(98)00290-X

Beerda, B., Schilder, M. B. H., van Hooff, J. A. R. A. M., de Vries, H. W., & Mol, J. A. (1999a). Chronic stress in dogs subjected to social and spatial restriction. I. Behavioral responses. Physiology & Behavior, 66(2), 233-242. https://doi.org/10.1016/S0031-9384(98)00289-3

Dess, N. K., Linwick, D., Patterson, J., Overmier, J. B., & Levine, S. (1983). Immediate and proactive effects of controllability and predictability on plasma cortisol responses to shocks in dogs. Behavioral Neuroscience, 97(6), 1005-1016. https://doi.org/10.1037/0735-7044.97.6.1005

Englund, M. D., & Cronin, K. A. (2023). Choice, control, and animal welfare: definitions and essential inquiries to advance animal welfare science. Frontiers in Veterinary Science, 10, 1250251. https://doi.org/10.3389/fvets.2023.1250251

Lukavská, M., Rudolfová, V., Bušina, T., Eretová, P., & Nekovářová, T. (2026). Behavioural adaptation and stress regulation in operant training: The role of the Default Variant of Behaviour. Applied Animal Behaviour Science, 299, 106971. https://doi.org/10.1016/j.applanim.2026.106971

Maier, S. F., & Seligman, M. E. P. (2016). Learned helplessness at fifty: Insights from neuroscience. Psychological Review, 123(4), 349-367. https://doi.org/10.1037/rev0000033

McGowan, R. T. S., Rehn, T., Norling, Y., & Keeling, L. J. (2014). Positive affect and learning: Exploring the "Eureka Effect" in dogs. Animal Cognition, 17(3), 577-587. https://doi.org/10.1007/s10071-013-0688-x

Overmier, J. B., & Seligman, M. E. P. (1967). Effects of inescapable shock upon subsequent escape and avoidance responding. Journal of Comparative and Physiological Psychology, 63(1), 28-33. https://doi.org/10.1037/h0024166

Sarrafchi, A., de Zwaan, N., Tucker, M., & Merkies, K. (2025). Can I touch you? A pilot study comparing consensual and non-consensual human-dog touch interactions. Applied Animal Behaviour Science, 285, 106560. https://doi.org/10.1016/j.applanim.2025.106560

Seligman, M. E. P., & Maier, S. F. (1967). Failure to escape traumatic shock. Journal of Experimental Psychology, 74(1), 1-9. https://doi.org/10.1037/h0024514

Squair, C., Proudfoot, K., Montelpare, W., & Overall, K. L. (2023). Effects of changing veterinary handling techniques on canine behaviour and physiology. Part 1: Physiological measurements. Animals, 13(7), 1253. https://doi.org/10.3390/ani13071253

Squair, C., Proudfoot, K., Montelpare, W., Doucette, T., & Overall, K. L. (2024). Effects of changing veterinary handling techniques on canine behaviour and physiology. Part 2: Behavioural measurements. Animal Welfare, 33, e43. https://doi.org/10.1017/awf.2024.37

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

Wess, L., Böhm, A., Schützinger, M., Riemer, S., Yee, J. R., Affenzeller, N., & Arhant, C. (2022). Effect of cooperative care training on physiological parameters and compliance in dogs undergoing a veterinary examination - A pilot study. Applied Animal Behaviour Science, 250, 105615. https://doi.org/10.1016/j.applanim.2022.105615