Behavioral Flexibility in Dogs: Adapting When Rules Change

1. What Behavioral Flexibility Is

1.1 The Definition

Flexibility is the capacity to modify behavior when contingencies change — when a strategy stops working, when new information arrives, or when the rules governing an outcome shift. It covers adapting to new rules, switching strategies, inhibiting responses that no longer pay, incorporating new information, and generalizing across contexts.

1.2 Not Simply More Change

Flexibility is the counterpart to persistence, and persistence is not a defect. Continuing a successful strategy through occasional non-reward is frequently the correct response — that is precisely what intermittent reinforcement builds (the schedule literature in detail).

Adaptive behavior depends on the balance: maintaining what works, abandoning what has genuinely stopped working. Flexibility is not "more change is better" but the capacity to detect when change is warranted and implement it efficiently.

1.3 Where Flexibility Is the Wrong Thing to Want

Persistence outperforms flexibility in a range of ordinary situations. A search dog that abandons a scent cone after two unproductive minutes finds less than one that works it. A dog that stops offering a behavior the moment reinforcement thins out has a recall that fails in exactly the conditions it exists for. Rigid routines are also how many anxious dogs manage their day, and taking that away without replacing it makes things worse rather than better.

The costly version is the inability to change when change is warranted, not the presence of stable strategies. An article recommending more flexibility without that qualification would be recommending inconsistency.

1.4 Not a Fixed Property of the Dog

Flexibility is described here as an ability, and the word invites reading it as a trait the animal has a fixed amount of. It is not. The same dog performs differently depending on the task, the reinforcer, the setting, its arousal that day and how much the outcome matters to it.

The practical consequence is that "this dog is inflexible" is a statement about a situation rather than about an animal, and that changing the situation is usually the faster route (as with temperament labels generally).

1.5 Why It Matters for Dogs

Dogs live in environments structured by human routines, in which the rules change both deliberately as training progresses and inadvertently as households shift — and adjusting to human signals is itself a domain-specific form of updating (as the social learning literature describes). And a range of presentations in behavior therapy — some reactivity, some anxiety-related behavior, compulsive-like patterns, frustration problems — can be described in part as difficulty updating behavior despite changed circumstances. That framing complements rather than replaces functional accounts, and it connects them to a construct with its own evidence base.

1.6 Why the Term Needs Care

Flexibility sounds like an unqualified good, and both ends of the range are costly. A dog that abandons a working strategy at the first difficulty is not flexible; it is inconsistent (defining a behavioral term before measuring it), and persistence with something that is working is the other half of the same ability.

What the research describes is the ability to change when the situation has changed, which requires detecting that it has. Detection is plausibly the part that fails most often and is certainly the part least visible from outside.

2. The Cognitive Components

2.1 Not One Ability

Flexibility is better understood as an emergent property of interacting processes than as a unitary capacity.

Inhibitory control — suppressing the currently prepotent response so an alternative can be expressed. Many failures of flexibility are failures of inhibition rather than failures to learn the new contingency: the animal may register that the old option no longer pays and continue responding to it anyway. This distinction is central to interpreting the evidence in §3.

Working memory — holding and manipulating information over short periods: which strategies have recently worked, what the current rule is. Extensively studied in humans; barely studied in dogs.

Attention shifting — redirecting attention from one source or stimulus dimension to another, which itself requires inhibiting an established attentional bias.

2.2 Prediction Error as the Trigger

Prediction error provides the mechanism for when flexibility gets engaged. When a previously reinforced response stops paying, or an unreinforced one starts, the discrepancy between expectation and outcome generates a teaching signal (Schultz, Dayan & Montague, 1997). Large or persistent errors signal that the current strategy is no longer appropriate — a dopaminergic teaching signal in the models this rests on (the canine dopamine evidence).

A contingency change is, on this account, simply a situation where accurate predictions become inaccurate. How fast behavior updates afterwards depends on how those errors are generated, weighted, and used — processes subject to individual variation, developmental change, and modulation by stress (the prediction-error framework in full).

2.3 Why the Components Matter Separately

Detecting that a rule has changed, inhibiting the old response and generating a new one are three operations, and a dog can be capable of one and not another. A single flexibility score merges them, which is why two dogs with the same score can need entirely different work.

In practice the failures look different: a dog that has not noticed the change keeps going confidently, while one that has noticed but cannot inhibit looks conflicted. That difference is visible without any equipment and tells a handler which problem they have.

3. Reversal Learning: The Main Paradigm

3.1 How It Works

The dog learns that one of two options is rewarded. Once that is reliable, the contingencies reverse without warning. Performance is measured by trials to the new criterion and by perseverative errors — continued responding to the option that no longer pays.

3.2 What the Canine Evidence Shows

Piotti et al. (2018) tested 107 medium-to-large dogs — 41 young (2.5–6.5 years) and 66 old (8–14.5 years) — on two spatial tasks designed to run in a single hour-long session. Young dogs learned both the initial discrimination and the reversal significantly faster than old dogs.

The methodological finding matters as much as the age effect: in the reversal, dogs trained on the location of stimuli learned faster than dogs trained on stimulus characteristics, and most old dogs failed to reach criterion within the 50-trial cutoff on the characteristic version. For anyone designing an assessment, location-based tasks are the more usable format.

3.3 A Correction on the Second Study

Wallis et al. (2016) is frequently cited alongside Piotti as a second reversal learning study. It is not one. Ninety-five pet Border Collies aged five months to thirteen years were tested on four touchscreen tasks: two visual discrimination tasks, inferential reasoning by exclusion, and a memory test with a six-month retention interval. There was no reversal learning task.

Its relevance to flexibility runs through a different measure. The authors discuss perseverative responding — repeating a particular response despite feedback — as an index of reduced cognitive flexibility, assessed through correction trials in the discrimination tasks. That is a genuine link, and a narrower one than a reversal study would provide. As a single-breed sample, generalizability is also limited.

3.4 What Reversal Learning Cannot Tell You

It conflates two things. Performance reflects both inhibition of the old response and acquisition of the new contingency. Most dog studies do not disentangle them, so poor performance could indicate an inhibition deficit, a learning deficit, or both — with quite different implications.

It is one operationalization among several. Whether reversal performance predicts flexibility in social behavior, novel-object problem solving, or adaptation to a changed household routine has not been tested in dogs.

The translation is unestablished. Laboratory reversal involves a single, clearly defined contingency switch in a controlled setting. Everyday flexibility involves multiple, poorly defined, gradually shifting contingencies across complex environments. The applied relevance of the laboratory findings should be understood as reasonable extrapolation, not demonstrated link.

3.5 Why the Correction on the Second Study Belongs Here

An article that quietly used a result it knew to be misreported would be easier to write and worth less. Printing the correction is what makes the rest of the citations trustworthy, and a reader has no other way to judge that.

It also illustrates how a finding acquires a life of its own once summarized, which is the recurring hazard in this collection and the reason every citation here was checked against the source.

4. How Flexibility Is Measured, and What That Costs

4.1 Two Paradigms, Two Questions

Reversal learning asks whether a dog can stop doing what worked and start doing something else. Problem-solving tasks ask what a dog does when nothing works. They are not measures of the same thing and are frequently discussed as though they were.

4.2 What Reversal Learning Actually Captures

Trials to criterion after the contingency switches, which is a number that combines several things: how strongly the first rule was learned, how quickly the change was detected, how willing the animal is to keep trying, and how motivating the reward was on that day.

A slow reversal can mean rigidity, thorough initial learning, or a dog that has stopped engaging with a repetitive task. The score does not distinguish them, which is why comparing reversal scores between studies compares procedures as much as dogs.

4.3 The Unsolvable Task Has a Specific Problem

Presenting a dog with a task that cannot be solved and measuring how long it persists is informative and ethically awkward, since the procedure is designed to produce failure, and the review literature notes considerable methodological variation between studies using it (Mendes, Resende & Savalli, 2021).

What it measures also depends on whether a person is present, because looking at the human is itself a strategy rather than a failure, and whether that counts as persistence or as giving up is an interpretive decision made by the researcher, decided differently by different groups.

4.4 Dropout Is Not Random

Dogs that stop participating are excluded, and they are plausibly the animals whose flexibility the study was trying to characterize in the first place. That selection runs through the whole literature and is rarely quantified. A study reporting how many dogs were excluded and why is more informative than one reporting only its results, and the proportion doing so improves with publication date.

4.5 What a Practitioner Can Observe Instead

Whether the dog tries something different when the usual approach fails, how many attempts it makes, how varied those attempts are, and how quickly it gives up. Three observations, available on any walk, and closer to the question than a laboratory score. Recording them over two weeks shows movement that a single impression does not.

5. Problem Solving: The Complementary View

Where reversal learning asks how readily a dog abandons a previously correct response, problem-solving research asks what it does when no correct response has been established at all.

5.1 Means-End Tasks

Early string-pulling studies found dogs performing poorly, appearing to use a proximity rule rather than tracking the connection between string and reward. Range, Hentrup and Virányi (2011) found dogs could solve a support problem — a reward on a board versus beside it — even where proximity would have misled them. Later work (Riemer et al., 2014; Müller et al., 2014) found performance sensitive to specific perceptual and task features, with dogs sometimes attending to connectivity and sometimes falling back on proximity.

For flexibility, the interesting part is not means-end understanding as such. It is that dogs have multiple strategies available, and which one gets deployed depends on task structure and prior experience. That is exactly what flexibility as a construct describes: availability and appropriate selection among alternatives.

5.2 The Unsolvable Task

Dogs learn to open a container, then face a version that cannot be opened. The measures are how long they persist with the original strategy and whether they shift to alternatives — including, distinctively, gazing at a nearby human.

This sits precisely on the exploration–exploitation boundary: continuing to exploit what previously worked versus exploring alternatives. The tendency to shift toward human-directed gazing varies with age, experience, living situation, and prior success (reviewed in Mendes et al., 2021), so the shift is itself subject to individual variation.

One caution is essential. A dog that quickly stops trying could be making an efficient strategy shift toward seeking help — or showing a more general reduction in persistence that is hard to distinguish from giving up. The literature has generally read human-directed gazing as the former. The broader point holds regardless: reduced behavioral output is ambiguous between adaptive flexibility and shutdown (which is the central interpretive problem in the helplessness literature).

5.3 Functional Fixedness

Continuing to use an object or strategy according to its established function even when an alternative would work better. A dog that keeps working a particular route to a resource after it has been blocked, without trying available alternatives, fits this description.

It is related to perseverative error but emphasizes lock-in to a specific strategy rather than general failure to update. Dedicated studies in dogs are not well developed; the construct is useful as a framing that connects reversal learning to real-world patterns, not as an established canine literature.

5.4 What the Unsolvable Task Reveals About Method

A dog that looks at its owner rather than persisting has produced a result, and whether that counts as giving up or as an appropriate social strategy depends on the interpretation the researcher brings.

The review literature documents how much the procedures differ (Mendes et al., 2021), which limits how confidently results can be pooled. Duration before the task is declared unsolvable, whether the owner is present and what they are permitted to do all vary.

6. Stress, Arousal, and Perceived Control

6.1 Acute Stress Shifts Control

One of the more consistent findings across the stress and cognition literature — in rodents and humans — is that acute stress shifts the balance of behavioral control away from flexible, goal-directed systems toward more automatized, habitual responding. Under stress, an animal falls back on established patterns and becomes less likely to update in response to new information, even where updating would help.

Direct experimental demonstrations of this shift in dogs using flexibility paradigms under manipulated stress are limited. The broader pattern — that dogs at higher arousal are less responsive to changing contingencies and more locked into previously learned responses — is consistent with applied observation and with the cognitive bias findings below, but the mechanism is extrapolated (the arousal literature in dogs).

6.2 Affective State and Interpretation

Mendl et al. (2010) found that dogs showing separation-related behavior responded to ambiguous cues as though they predicted the less favorable outcome — a more pessimistic judgment bias than dogs without such behavior. Cognitive bias and reversal learning are not the same construct, but both are sensitive to affective state, and the finding established the link between emotional state and cognitive processing in dogs (the separation-related picture in detail).

6.3 Chronic Stress Narrows the Repertoire

Chronic stress is associated across species with behavioral rigidity: reduced exploration, a narrower repertoire, poorer updating of established patterns. This overlaps directly with what the canine chronic stress literature documents — reduced behavioral variability and reduced exploration in dogs under sustained low-control conditions (the chronic stress physiology).

The convergence matters theoretically: flexibility is not an isolated cognitive variable but is embedded in the stress-response system. Practically, it suggests that interventions aimed at building flexibility work best alongside, not instead of, reducing the underlying stress load.

6.4 Anxiety and Reactivity

Anxious and reactive responses are almost by definition behaviorally fixed: a consistent, often escalating response to a trigger, with limited sensitivity to information supporting an alternative. Whether that reflects a general reduction in flexibility, a domain-specific failure concentrated on threat-related stimuli, or simply a heavily reinforced pattern that has never been directly challenged is unresolved — and a full account would need to include the negative-reinforcement history that typically maintains avoidance (the reactivity picture; the anxiety neurobiology).

6.5 Perceived Control as a Background Condition

A theme running through the above deserves separate statement. If an animal's history has established that its behavior generally produces effects — that testing alternatives is worth the effort — that expectation may itself support the willingness to explore and update that defines flexibility.

Conversely, a history in which behavior and outcome were decoupled may reduce exploration not through acute arousal or chronic stress physiology specifically, but because the basic expectation that behavioral variation is productive has weakened. On this reading, perceived control operates as a background condition for flexibility rather than a moment-to-moment modulator — which would mean that agency-focused interventions support flexibility through a route additional to stress reduction (Maier & Seligman, 2016).

This is theoretically motivated and untested in dogs.

6.6 Emotional Regulation Is Not Cognitive Flexibility

The two are routinely merged, and separating them changes what gets worked on. Cognitive flexibility is about updating a rule: detecting that the contingency changed, inhibiting the old response, generating a new one. Emotional regulation is about the intensity and duration of an affective response and how quickly an animal returns to baseline after one.

A dog can be strong at one and weak at the other. One that solves novel problems readily and needs twenty minutes to settle after a doorbell has good flexibility and poor regulation. One that recovers within seconds but keeps trying the same blocked route has the reverse.

They interact rather than being the same thing: a dog at high arousal will underperform on flexibility measures whatever its flexibility is, which is why the two are hard to tell apart from outside and why regulation is usually the thing to address first.

6.7 Why This Chapter Carries the Practical Weight

Of everything here, the finding that stress and arousal narrow the behavioral repertoire is the one that changes what a practitioner does. It relocates the target from the dog's cognition to the dog's circumstances, which is a cheaper target and one a household can actually move.

A dog that seems inflexible in a demanding week may be perfectly flexible in a quieter one, and testing that is easier than training anything. It requires a quieter week rather than a technique.

7. Lifespan and Individual Differences

7.1 Puppies

High neural plasticity and strong exploratory tendency, in a period where the behavioral repertoire is still being constructed and the cost of trying alternatives is low because few responses are yet established (the sensitive period in detail).

7.2 Adults

A relative balance between accumulated experience — efficient established responses for familiar situations — and continued capacity to adjust when circumstances genuinely change.

7.3 Older Dogs

The most robust dog-specific finding. Piotti et al. (2018) documented significantly slower reversal learning in older dogs, and Wallis et al. (2016) found age-related decline in discrimination learning with perseverative responding as a flexibility-relevant index. This connects to the broader picture of canine cognitive aging (including cognitive dysfunction syndrome).

7.4 Sources of Individual Variation

Emotional state is the best supported, through both the acute and chronic mechanisms in §6 and the canine cognitive bias evidence.

Genetics, early experience, environmental complexity, and training history are all plausible contributors, each supported by findings in adjacent literatures rather than by direct study of flexibility in dogs. Training approaches that specify a single fixed solution with no scope for variation plausibly provide fewer opportunities to develop flexibility-relevant skills than approaches incorporating variation — a hypothesis, not a demonstrated finding.

7.5 What the Aging Findings Do and Do Not Support

That cognitive flexibility declines with age in dogs is established (Wallis et al., 2016; Piotti et al., 2018). That a given older dog's difficulty is age rather than pain, hearing loss or reduced sleep is not established by those studies.

Age is a population finding and a poor first explanation for an individual animal. The medical questions are cheaper to ask and more often productive.

8. Which Findings Come From Which Species

8.1 An Unusually Canine Article

Seven of the nine sources measured dogs, which is a better balance than most cognition topics achieve and is worth stating because the framework chapters read like extrapolation. The findings are canine; the language used to organize them is not.

8.2 What Was Measured in Dogs

Means-end and string-pulling tasks (Range, Hentrup & Virányi, 2011; Riemer et al., 2014; Müller et al., 2014), the unsolvable task across studies (Mendes et al., 2021), aging effects on cognition (Wallis et al., 2016; Piotti et al., 2018) and judgment bias in relation to behavior (Mendl et al., 2010).

8.3 What Was Established Elsewhere

The prediction-error account of when a rule change is detected comes from primate recording (Schultz, Dayan & Montague, 1997), and the control-detection material is a synthesis from rodent work (Maier & Seligman, 2016).

Two borrowed sources carrying both mechanism chapters, which is the usual pattern and unusually easy to see here. The canine sources carry the findings and the borrowed ones carry the explanation, which is the arrangement a reader should look for rather than the reverse.

8.4 The Laboratory Concentration

Most of the canine work comes from a small number of connected groups using related paradigms. Agreement between such studies reflects shared methodology as much as independent confirmation, and replication by unconnected laboratories is largely absent.

That is normal for a young field and it constrains how far a single finding should travel. It is also improving as the number of groups working on canine cognition grows.

8.5 What Survives

That flexibility is measurable, varies between individuals, declines with age and is reduced by stress and arousal are canine findings. The account of why they hold is borrowed and would not change the practical recommendations if revised.

9. Applied Implications

9.1 Make Room for Exploration

Approaches in which the dog generates, tests, and adjusts its own strategies — shaping being the clearest example, where behavioral variability is a resource rather than an obstacle — plausibly support flexibility. This follows from the theoretical framework rather than from evidence that such training improves measured flexibility, which has not been established.

9.2 Avoid Total Rigidity

Highly rigid routines produce reliable performance inside that specific context and may not support adapting when context changes. The balance mirrors §1.2: consistency is necessary for efficient learning, but routines that eliminate all variability may cost adaptive capacity.

9.3 Train Across Varied Contexts

Varying locations, handler positions, distraction levels, and minor cue variations is the established route to generalization. It can also be read as practice in a core flexibility component: recognizing that a response remains appropriate despite surface variation, and adjusting where it does not.

9.4 Manage Arousal

Given §6.1, keeping arousal in a workable range during training is directly relevant to whether the dog can acquire and adjust behavior at all. A dog at high arousal shows reduced sensitivity to the contingency being trained and falls back on established patterns — which may not be the one you are working on.

9.5 Play as Low-Cost Practice

Play has three features that make it a plausible flexibility context. It is highly variable — the same sequence rarely repeats, and partners continuously introduce change that must be responded to. It is low-stakes — the cost of an ineffective response is minimal, no lost reward, just a shift in the interaction, which makes exploration unusually cheap. And it involves rapid role switching — chasing and being chased, offering and withholding, escalating and settling — each requiring inhibition of an ongoing response and a shift to another.

Together these support the hypothesis that play exercises flexibility through repeated low-cost practice. This fits the broader ethological account of play as practice, and direct canine evidence linking play to measured flexibility does not exist (the play development literature).

9.6 In Behavior Therapy

Reactivity, some anxiety presentations, compulsive-like behavior, and frustration problems can each be described as involving a fixed pattern persisting despite changed circumstances or available alternatives. On that framing, part of behavior modification is building flexibility specifically: providing experiences where alternative responses are available, reinforced, and successful.

The practical emphasis follows directly. Programs that focus only on suppressing an undesired response, without building and reinforcing alternatives, may address the surface behavior while leaving the dog's repertoire of accessible options unchanged (which is also the argument for differential reinforcement).

10. What to Do With a Rigid Dog

10.1 Check What Rigidity Is Costing

A dog with strong routines that copes well is not a problem to be solved. Rigidity becomes a target where it prevents adaptation the dog actually needs — a changed household, a new route, a reduced ability to do what it used to. Routines that serve a dog well are not a deficit in need of correction, and predictability is a resource for many animals rather than a limitation.

10.2 Rule Out the Body and the Load

Reduced variability of behavior is one of the earlier visible effects of chronic stress and of discomfort, and both are cheaper to address than any cognitive training. Neither requires the dog to learn anything for the situation to improve.

Where a dog has become more rigid over weeks, that ordering applies before anything else is attempted. A gradual narrowing of what a dog is willing to do is one of the more useful early signals available to a household.

10.3 Start With Contingencies That Cannot Fail

A dog that has stopped trying different things needs experiences where trying produces an outcome, and the content matters less than the reliability. Ambitious problems are counterproductive, because a failed attempt is another instance of the thing being worked against, and setting the bar embarrassingly low is the correct call even though it feels wrong to most handlers.

10.4 Vary One Thing at a Time

Changing the location, the handler and the task simultaneously produces a session in which nothing is learned about any of them, and a dog that fails such a session has been told nothing useful. Small, deliberate variation builds what large changes do not. One new element per session is a workable rule of thumb and feels slower than it is.

Play is the low-cost version of this and is the recommendation the article makes for exactly that reason. Nothing in a game has to be got right, which is what makes variation cheap there and expensive elsewhere.

10.5 What the Article Does Not Claim

That flexibility can be trained as a general capacity is not established in dogs. What the evidence supports is that specific flexibility improves with practice at specific things, and that the conditions around a dog affect how much of it is available on a given day.

The distinction matters because exercises marketed as building cognitive flexibility are selling the first while demonstrating at most the second. Practice at a puzzle produces a dog that is better at that puzzle, which is worth having and is a considerably narrower claim than the one being sold.

10.6 Why Play Earns Its Place in the Recommendations

It supplies variation, uncertainty and a low cost of failure at the same time, which is difficult to arrange deliberately and happens on its own in play.

That is an argument from structure rather than from a canine study, and the article says so where it makes it. Recommendations of that kind are defensible as long as their basis is stated.

10.7 The Version Worth Remembering

Flexibility is several abilities rather than one. It is measurable and the measures disagree. It declines with age and narrows under stress. And the fastest route to more of it in a given dog usually runs through its circumstances rather than its cognition, which is the least marketable finding in the article.

11. Summary at a Glance

The construct — Capacity to change strategy when contingencies change. Emergent from inhibitory control, working memory, attention shifting, and prediction-error-driven updating. Counterpart to persistence, not opposite of it.

The main paradigm — Reversal learning, measured by trials to new criterion and perseverative errors. Conflates inhibition and new learning; represents one operationalization among several.

What canine research establishes — Age-related decline in reversal learning (Piotti et al., 2018, n = 107) and in discrimination learning with perseverative responding (Wallis et al., 2016, n = 95 Border Collies); affective state affects interpretation of ambiguity (Mendl et al., 2010); dogs deploy multiple strategies on physical problems depending on task structure.

Stress effects — Acute stress shifts control toward automatized responding; chronic stress narrows the repertoire and reduces exploration. Both are well established across species and extrapolated to dogs.

Applied orientation — Make room for exploration, avoid total rigidity, train across contexts, manage arousal, build alternatives rather than only suppressing responses.

12. Research Gaps and Critical Appraisal

The paradigm base is narrow. Almost all direct canine evidence comes from reversal learning in spatial or simple visual formats. Whether it predicts flexibility in social behavior, novel problem solving, or household adaptation is untested.

Inhibition and learning are confounded. Reversal performance mixes two processes with different implications, and dog studies rarely separate them.

Wallis et al. (2016) is not a reversal study. It is regularly cited as one. It tested two discrimination tasks, reasoning by exclusion, and memory; its flexibility relevance runs through perseverative responding, not reversal.

Stress effects on flexibility are untested in dogs. The relationship is well established in rodents and humans and consistent with the canine cognitive bias literature, but direct experimental manipulation of stress with measured flexibility outcomes in dogs is lacking — arguably the most consequential gap given the applied stakes.

The translation gap is unclosed. Laboratory reversal and everyday flexibility differ in structure, timescale, and complexity, and no study has connected them.

Testing is cross-sectional and single-session. Given evidence that performance varies across occasions within individuals, one-time testing may not characterize a dog's typical functioning (a measurement reliability problem).

Research group concentration. The canine reversal literature is concentrated among a small number of predominantly European groups using broadly similar paradigms, which limits how far specific quantitative findings can be assumed to generalize.

Individual difference sources are inferred. Genetics, early experience, environmental complexity, and training history are plausible contributors supported by adjacent literatures rather than by adequately powered direct study.

The two paradigms measure different things. Reversal learning and problem-solving tasks are discussed together and answer separate questions, which limits how far results can be combined into a single account of canine flexibility.

Dropout is a selection problem. Dogs that stop participating are excluded, and they are plausibly the animals the studies most needed to characterize.

The canine work is concentrated. Most of it comes from a small number of connected groups using related paradigms, so agreement reflects shared methodology as much as confirmation.

The mechanism chapters are borrowed. Prediction error comes from primate recording (Schultz et al., 1997) and the control account from rodent work (Maier & Seligman, 2016), and neither has been tested in dogs.

13. Conclusion

Behavioral flexibility connects several things that are usually discussed separately: prediction error and reinforcement learning, the effects of acute and chronic stress, cognitive aging, and the agency-focused approaches that follow from the helplessness literature. It is not a specialized laboratory capacity but something embedded in the systems governing learning, stress, and motivation — which is why it declines under stress, declines with age, and shows up in behavior therapy as a fixed response that will not update. The dog-specific evidence is narrower than that framework deserves. One paradigm dominates, it confounds two processes, one of the two studies most often cited for it does not actually use it, and the connection to everyday flexibility has never been tested. The practical orientations that follow are nonetheless coherent and low-risk: leave room for the dog to try things, do not make everything invariant, train across contexts, keep arousal workable, and build alternatives rather than only removing responses. Held as a working framework with its limits visible, that is a reasonable basis for practice — which is the most that can honestly be claimed for it at present.

Key Insights (Takeaways)

• Flexibility is not "more change is better." It is the capacity to detect when change is warranted, and its counterpart — persistence through occasional non-reward — is frequently the correct response. Adaptive behavior depends on the balance.

• Failures of flexibility are often failures of inhibition. The dog may register that the old option no longer pays and continue responding to it anyway, which is why reversal learning performance conflates two processes that canine studies rarely separate.

• Age-related decline is the most robust canine finding. Younger dogs learned both discrimination and reversal significantly faster than older dogs (Piotti et al., 2018, n = 107), and location-based tasks proved more usable than characteristic-based ones for older animals — a point worth knowing for assessment design.

• Wallis et al. (2016) is widely miscited. It is a discrimination, reasoning, and memory study in 95 Border Collies with no reversal learning task; its bearing on flexibility runs through perseverative responding in the discrimination tasks.

• Stress narrows the repertoire, and reduced output is ambiguous. Acute stress shifts control toward automatized responding and chronic stress reduces exploration — but a dog that stops trying may be shifting strategy efficiently or may have given up, and telling those apart is the same problem the helplessness literature runs into.

References

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

Mendes, J. W. W., Resende, B., & Savalli, C. (2021). A review of the unsolvable task in dog communication and cognition: Comparing different methodologies. Animal Cognition, 24(5), 907–922. https://doi.org/10.1007/s10071-021-01501-8

Mendl, M., Brooks, J., Basse, C., Burman, O., Paul, E., Blackwell, E., & Casey, R. (2010). Dogs showing separation-related behaviour exhibit a "pessimistic" cognitive bias. Current Biology, 20(19), R839–R840. https://doi.org/10.1016/j.cub.2010.08.030

Müller, C. A., Riemer, S., Virányi, Z., Huber, L., & Range, F. (2014). Dogs learn to solve the support problem based on perceptual cues. Animal Cognition, 17(5), 1071–1080. https://doi.org/10.1007/s10071-014-0739-y

Piotti, P., Szabó, D., Bognár, Z., Egerer, A., Hulsbosch, P., Carson, R. S., & Kubinyi, E. (2018). Effect of age on discrimination learning, reversal learning, and cognitive bias in family dogs. Learning & Behavior, 46(4), 537–553. https://doi.org/10.3758/s13420-018-0357-7

Range, F., Hentrup, M., & Virányi, Z. (2011). Dogs are able to solve a means-end task. Animal Cognition, 14(4), 575–583. https://doi.org/10.1007/s10071-011-0394-5

Riemer, S., Müller, C., Range, F., & Huber, L. (2014). Dogs (Canis familiaris) can learn to attend to connectivity in string-pulling tasks. Journal of Comparative Psychology, 128(1), 31–39. https://doi.org/10.1037/a0033202

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

Wallis, L. J., Virányi, Z., Müller, C. A., Serisier, S., Huber, L., & Range, F. (2016). Aging effects on discrimination learning, logical reasoning and memory in pet dogs. Age, 38(1), 6. https://doi.org/10.1007/s11357-015-9866-x