Reward Value in Dogs: Food, Toys, Praise and Petting

1. What "Reward Value" Means

1.1 A Pleasant Thing Is Not Yet a Reinforcer

In everyday language, a reward is something the dog is assumed to like: a treat, a ball, a stroke, a cheerful "good dog". In learning theory, the relevant term is reinforcer, and it is defined differently. A reinforcer is a consequence that, when it follows a behavior, makes that behavior more likely to occur again under similar conditions. The definition is functional. It does not ask whether the consequence looks pleasant to us or whether the dog seems to enjoy it in the moment. It asks what happens to the behavior.

The difference has practical consequences. A dog can wag its tail while being praised and still not sit faster next time. A dog can be delighted by a ball at home and ignore it on a training field. In each case something the dog apparently likes fails to function as a reinforcer for that behavior, at that moment, in that place. Conversely, consequences that people would not think of as rewards, such as being allowed to sniff a lamppost, can in principle function as reinforcers too.

1.2 Value Is a Relation, Not a Property of the Item

The studies reviewed in this article point to one central idea: the value of a reward is not a property of the item. It is a property of the relation between a consequence, a particular dog, that dog's current state and the context in which the consequence is delivered. A piece of sausage is not worth a fixed amount. It is worth a lot to a hungry, food-motivated dog in a quiet room, and much less to a dog that has just eaten, is uneasy in a new place, or is watching another dog it would rather approach.

If value depends on the relation, it cannot be read off a list or assumed from the category of the reward. It has to be measured by what the behavior does, which in research means testing individual dogs. The rest of this article follows that logic: first what group-level studies found, then why individual dogs deviate, and finally how to find out what works for a particular dog.

1.3 How Researchers Measure Value

Studies on reward value in dogs use a small set of measures that do not always agree. Response rate and latency show how often and how quickly a dog performs a behavior that produces a given consequence. Concurrent choice procedures offer two alternatives at once and measure which one the dog chooses or where it spends its time. Progressive-ratio schedules raise the number of responses required for each reward until the dog stops; the last completed requirement, the break point, measures how much the dog will work for that reward. Runway tasks measure how fast a dog moves toward a reward.

Choice shows relative preference, break points show how much effort a reward will sustain, and running speed reflects incentive motivation, the pull of an expected reward. A consequence can score high on one measure and lower on another, which is one reason why studies that seem to contradict each other often measured different things (how reinforcement schedules shape responding).

2. Food vs Social Interaction

2.1 Social Interaction as an Operant Reinforcer

The most systematic comparisons of food and social interaction come from a series of experiments by Feuerbacher and Wynne. In the first, an arbitrary response, touching a target with the nose, was reinforced either with a small piece of food or with brief human social interaction (Feuerbacher & Wynne, 2012). Three populations were tested: shelter dogs, owned dogs and hand-reared wolves. Across all three populations, brief social interaction was a relatively ineffective reinforcer compared with food for most animals, producing lower response rates and longer latencies.

The important words are "brief" and "most". The study does not show that dogs do not value people. It shows that, under these conditions, a few seconds of social interaction maintained less responding than a small food item did in most individuals.

2.2 Concurrent Choice: Food or Petting

A second study offered owned dogs a direct choice between food and petting (Feuerbacher & Wynne, 2014). Each dog completed five sessions of five minutes. In the first session, both food and petting were available continuously. In the following sessions, food was made progressively less available: first on a fixed-interval schedule of 15 seconds, then of one minute, then not at all. In the last session, the original food schedule was restored. Dogs were tested in a familiar environment (a daycare) or an unfamiliar one (a laboratory room), and petting was delivered either by the owner or by a stranger.

When food was readily available, dogs preferred food to petting, as the title of the paper states. As food became scarcer, all groups spent less time at the food alternative, although there were group and individual differences in how sensitive dogs were to the change. The time dogs spent with the petting person varied with the context, with how familiar the petting person was, and with how deprived the dogs had been of social interaction.

The relative value of food and petting was therefore not fixed. It shifted with the food schedule, the place and the person.

2.3 When Both Are Withheld: Owner or Food

A later study changed one key condition (Isernia, Wynne, House & Feuerbacher, 2022). Before testing, animals were deprived of both food and social contact for at least four hours. Owned dogs then chose between their owner and food that was not connected to any person, and hand-reared wolves chose between their caregiver and food. Ten owned dogs were recruited, and six wolves were tested.

The wolves chose food first in 75% of trials. The dogs, as a group, did not choose their owner more or less often than chance. In time allocation, the median dog spent 66% of the time with the owner and 34% with food, while the median wolf spent 22% with the caregiver and 78% with food. The dogs also varied more between individuals than the wolves did.

The authors named the limits: small samples, especially of wolves, and a species comparison confounded by differences in human contact, housing and feeding (dogs were fed twice daily, wolves three times a week). Still, when food and social contact were both withheld, owned dogs did not show the clear food preference seen when food was dense and social contact readily available.

2.4 Individual Food Motivation

Food itself does not have the same value for every dog. A study grouped 34 dogs by eating style: fast eaters (n = 15), slow eaters (n = 13) and dogs that left food in the bowl (n = 6) (Okamoto, Ohtani, Uchiyama & Ohta, 2009). The fast eaters, judged to be the most food motivated, responded best to food rewards. Dogs that tended to leave food showed little interest in food and responded poorly regardless of the type of reward. Slow eaters responded similarly across reward types.

Eating style was observed, not assigned, so the grouping is correlational, and the subgroups are small. But it fits the larger picture: "food works better than social interaction" describes most dogs in particular tests, and for a dog with low food motivation it may not hold.

3. Petting vs Praise

3.1 Dogs Preferred Petting to Vocal Praise

If food and social interaction differ, so do different forms of social interaction. Feuerbacher and Wynne (2015) compared petting and vocal praise directly. As the title of the paper states, dogs preferred petting to vocal praise, both in concurrent choice, where both were available at the same time from different people, and in single-alternative procedures, where only one type of interaction was available in a session.

These were choice procedures, not training tasks. The study shows that petting was the preferred social consequence in this format, not how well either would maintain a trained behavior.

3.2 The Same Petting, Different People

Who delivers a social consequence matters. In a further concurrent choice study, two people provided identical petting (Feuerbacher & Wynne, 2017). Owned dogs preferred their owners in an unfamiliar environment but spent more time with the stranger in a familiar environment. Shelter dogs and owned dogs tested with two strangers could, as the title of the paper states, quickly form strong preferences for one stranger over the other, with preference magnitudes comparable to owned dogs' preference for their owner in the unfamiliar setting.

The physical consequence was the same; only the person and the place changed, and that alone shifted the value of petting. Social contact has also been studied from a hormonal angle (what is known about oxytocin and social contact), but the behavioral data here show something simpler: social contact is not a single reward with a single value.

3.3 Praise Added to Food

Many trainers deliver food together with praise, assuming the combination is worth more than food alone. A study tested this (Feuerbacher, Stone & Friedel, 2022). Dogs chose between food delivered piece by piece with praise and the same food delivered all at once without praise, and between large and small amounts; progressive-ratio break points were then measured for each option.

Large food portions produced break points nearly twice those of small portions. The delivery method had only minor and inconsistent effects, and praise had no measurable impact. Preference tests generally predicted which option would sustain more work. In this study, adding praise to food delivery did not add measurable value.

3.4 Why Vocal Praise Is Often Relatively Low-Value

Taken together, these results show that vocal praise was often relatively low-value in the choice and reinforcement procedures tested. Dogs preferred petting to vocal praise in choice procedures (Feuerbacher & Wynne, 2015), and when it was added to food it did not raise the work dogs were willing to do (Feuerbacher et al., 2022). Brief social interaction in general maintained less responding than food for most dogs (Feuerbacher & Wynne, 2012). How effective verbal praise is as a learned reinforcer can still depend on the individual dog, its learning history and the situation.

From a learning perspective, this is not surprising. Words have no built-in value for a dog. "Good dog" can become a conditioned reinforcer if it reliably predicts something the dog values, but if it is used constantly and without consistent follow-up, it predicts little. Some dogs may respond more to a cheerful tone than others, but the data do not support the assumption that praise alone is enough to maintain trained behavior in most dogs.

This does not mean that social attention is irrelevant. Functional analysis methods, which test candidate consequences experimentally, have shown that attention from a person can function as a reinforcer for individual dogs (Feuerbacher & Wynne, 2016). Attention can also maintain behavior owners do not want, such as jumping up (how attention can maintain unwanted behavior). Social consequences have value, but it differs by form, person and situation. In direct comparison studies, vocal praise performed less strongly than petting or food.

4. Food vs Toys and Play

4.1 Food and Toys in a Work Task

Toys are often recommended as an alternative to food. A study with ten pet dogs compared the two directly (Lazaro, Winter, Fernand, Cox & Dorey, 2023). Each dog first completed preference assessments with six foods and six leisure items, such as a ball, a tug toy and a squeaky toy. The most preferred food and leisure item were then tested as consequences for a simple response, in alternating food, leisure and extinction phases, followed by progressive-ratio sessions.

Nine of the ten dogs responded at higher rates for food than for the leisure item. Seven of the ten showed an individually significant difference in favor of food. In the alternating design, all dogs completed the maximum of 15 trials per session when working for food, but only two did so when working for their preferred leisure item.

The authors noted two limits: food and toys were never offered at the same time, so the study measures work, not choice, and dogs with a toy-based training history were not included. For most dogs in this task, the best food sustained more work than the best toy, but there were exceptions.

4.2 Toy Preferences Can Be Measured

Toy preferences can be measured. In a shelter study, play with specific toys during a brief preference assessment predicted play with those toys in naturalistic settings (Protopopova, Brandifino & Wynne, 2016). The validation sample was small (n = 20) and consisted of shelter dogs, and the study does not show how well the toys work as reinforcers in training.

4.3 Interest in a Toy Fades

Toy value also declines quickly with repeated exposure. Sixteen kennel-housed Labrador Retrievers were given one toy in successive short play periods until their interest waned (Pullen, Merrill & Bradshaw, 2012). A toy differing in color, odor or both restored play, regardless of the type of contrast, the interval between presentations or the delay. The authors concluded that dogs habituate to the toy as a whole. The study involved one breed in kennels, but the principle is relevant: a toy that is always available and always the same tends to lose value.

4.4 Play With a Person

In training, the reward is more often a game with a person than a static toy. In a study of 14 Golden Retrievers, dogs played tug-of-war sessions in which they either won or lost the toy (Rooney & Bradshaw, 2002). Measures of "confidence", the trait closest to what is often called dominance, were not affected by winning or losing. Obedience was higher after play, regardless of who won, and the most playful dogs showed significantly more attention-seeking after games they had won.

This small single-breed study does not measure reward value directly, but it suggests that play with a person has short-term effects on how dogs respond to that person, and winning did not change the confidence measure. As a reward, play combines object, movement and social contact (what play does in the developing dog).

In practice: Play rewards often raise arousal more than food does, which can help with energetic behaviors such as a fast recall but can make calm behaviors, such as a settled down-stay, harder to reinforce. This is an observation from training practice, not a result of the studies above (how arousal affects learning and self-control).

5. Not All Food Is Equal

5.1 Quality or Quantity

"Food" is not a single category. A runway study compared quality and quantity separately (Riemer, Ellis, Thompson & Burman, 2018). In the quantity comparison, dogs ran for either one or five pieces of dry food. In the quality comparison, they ran for sausage or dry food. Dogs ran significantly faster for the higher quality food, while the amount of dry food had no significant effect on running speed. The authors concluded that higher quality rewards appear to produce greater incentive motivation in dogs than a larger quantity of a lower value reward.

5.2 Amount and Work Requirement

At first sight, this conflicts with the study in which large portions produced break points nearly twice those of small portions (Feuerbacher et al., 2022). But the studies measured different things. Running speed reflects how strongly an expected reward pulls the dog; a break point reflects how much total work a reward will sustain as the requirement rises. Quality may matter more for the immediate pull, amount for how much work a reward can support. Neither finding says that one piece of sausage is always better than five pieces of kibble, or the reverse.

5.3 Can Dogs Tell the Difference?

Amount can only affect value if dogs notice it. In a study of four dogs, each dog chose between a constant amount of four pieces and alternatives of one to seven pieces (McGuire, Bizo, McBride & Kocek, 2018). The dogs reliably chose the larger amount, and discriminating one from four pieces was easier than discriminating four from seven. The authors pointed out the limits: four dogs, few trials, possible satiation within sessions, and the possibility that the pattern reflects incentive value rather than pure discrimination.

5.4 Variety or the Best Food Every Time

A common recommendation is to vary rewards to keep dogs interested. A study with 16 pet dogs tested this (Bremhorst, Bütler, Würbel & Riemer, 2018). After a preference test with sausage, cheese and a liver treat, each dog chose repeatedly between an option that always delivered its most preferred food and one that delivered the three foods in a semi-random sequence, in six blocks of ten trials in one session.

At the group level, there was no preference for either option. Individually, six dogs preferred the varied option, six preferred the constant option, and four showed no preference. The preference for the varied option increased across the session: two dogs preferred it in the first 30 trials, five in the last 30. The authors noted that all three foods were of high value, which may have slowed habituation to the constant option, and that the study used a single session.

The result neither supports nor refutes variety as a general rule. Dogs differ, and the appeal of variety may grow as a session goes on.

In practice: High-value food rewards are often energy-dense, and a long training session can add considerably to the daily intake, which matters for dogs prone to weight gain (what is known about overweight in dogs). This is a practical consideration, not a finding of the reward studies.

6. State and Context

6.1 Deprivation and Satiation

The value of any reward depends on how much of it the dog has recently had. This is why both food and social contact were withheld in the owner-or-food study (Isernia et al., 2022), and the time dogs spent with a petting person varied with their relative deprivation from social interaction (Feuerbacher & Wynne, 2014). Satiation can also occur within a session, as the amount-discrimination study noted (McGuire et al., 2018).

6.2 The Breakfast Effect

It is often assumed that a hungry dog works best for food. A study on search performance complicates this (Miller & Bender, 2012). In a counterbalanced within-subjects design, dogs were tested either after fasting for 12 hours or after a breakfast of half their daily energy requirement, eaten 30 or 90 minutes before testing. Before the search task, each dog completed a ten-minute sit-stay. Thirteen dogs were analyzed. In the group tested 30 minutes after eating, dogs searched more accurately when fed than when fasted (73.3% vs 64.2% correct). In the group tested 90 minutes after eating, there was no difference (65.2% vs 64.8%).

The study is small, the preceding sit-stay prevents separating the effect of a recent meal from that of an inhibition task, and reinforcer value was not measured directly. But it is a counterweight to the idea that more deprivation always means better performance.

6.3 Who Delivers It and Where

Context changes value for social rewards as much as for food. Petting by the owner was preferred to petting by a stranger in an unfamiliar room but not in a familiar one (Feuerbacher & Wynne, 2017), and the time dogs spent with a petting person depended on the place and the person (Feuerbacher & Wynne, 2014). One possible interpretation, not tested directly, is that the owner provides security in an unfamiliar setting.

In practice: A reward that works at home often loses value outdoors, where smells, other dogs and moving objects compete with it, so many trainers use a higher-value reward in distracting places. This is a rule from practice, not a directly tested finding.

7. When Value Changes: Contrast, Omission, Frustration and Partial Reward

7.1 Incentive Contrast: Three Studies, Three Answers

In many species, a sudden drop in reward quality leads to responding below the level of animals that always received the lower reward, a phenomenon called successive negative contrast. Whether this applies to dogs is less clear than it is often presented.

The first study in dogs reinforced gazing at an experimenter with either dried beef liver or dog pellets (Bentosela, Jakovcevic, Elgier, Mustaca & Papini, 2009). Liver produced longer gazing than pellets. When dogs that had received liver were switched to pellets, they rejected food more often, their gazing dropped below the level of dogs that had always received pellets, and they showed more withdrawal and lying down. This is the classic contrast effect.

A later study did not find it (Riemer, Ellis, Ryan, Thompson & Burman, 2016). With 24 owned pet dogs and 21 shelter dogs, neither population showed a successive negative contrast effect. A within-subjects analysis even showed a reverse effect, and the authors concluded that the paradigm may not be robust enough to replicate easily across contexts and populations.

A third study tested 36 dogs from three populations, 12 laboratory dogs, 12 shelter dogs and 12 owned dogs, in a puzzle task, first with a high-value food, then with a low-value food, then again with the high-value food (Vieira de Castro, Baria-Unwalla, Cabrita & Olsson, 2025). Shelter dogs and three of the twelve owned dogs showed reduced engagement or longer completion times after the downshift. Laboratory dogs and the other nine owned dogs did not.

Read together, the evidence on incentive contrast in dogs is inconsistent: one positive result, one failed replication with a reverse effect, and one result that depended on population and individual. Some dogs, in some situations, respond to a drop in reward quality with reduced effort, but this is not a reliable rule. The underlying idea, that responses depend on the difference between expected and received outcomes, is central to learning theory (how prediction errors drive learning). A related question is how dogs respond when another dog receives a better reward than they do (what is known about inequity aversion in dogs).

7.2 Omission, Extinction and Frustration

A stronger form of devaluation is omission: the expected reward does not come at all. Studies with the same gaze response showed that reinforcer omission and extinction changed this learned communicative behavior (Bentosela, Barrera, Jakovcevic, Elgier & Mustaca, 2008). When gazing that had previously produced food no longer did so, dogs moved away, oriented sideways, lay down, walked around, sniffed and vocalized more, behaviors the authors described as frustration responses (Jakovcevic, Elgier, Mustaca & Bentosela, 2013).

Frustration is a normal part of extinction (what happens when a behavior stops being reinforced) and has its own neurobiology (what is known about frustration in dogs). In a reward-omission task, frustration responses increased when food was withheld, and the presence of the owner reduced them (Dzik, Jousset, Vietti & Bentosela, 2025). This fits a broader picture of social buffering, in which a familiar person can soften stress responses (how a familiar person can buffer stress).

7.3 Partial Reward During Training

A popular idea is that rewarding only some correct responses makes dogs work harder. For new behaviors, one study tested this (Cimarelli, Schoesswender, Vitiello, Huber & Virányi, 2021). Naive dogs were clicker trained, with food following either every click or 60% of clicks. Partial rewarding did not speed up learning, and partially rewarded dogs showed a more pessimistic-like response in a later cognitive bias test (what cognitive bias tests can and cannot show).

Reducing reward delivery during acquisition brought no learning benefit and was associated with a more negative affective state. A signal that only sometimes predicts a reward is a less reliable predictor, and each missing reward is a small omission.

8. Contingency and Neural Measures

8.1 Earning vs Receiving

Reward value may depend not only on the reward but also on how it is obtained. A study of 12 beagles used a yoked design (McGowan, Rehn, Norling & Keeling, 2014). Experimental dogs learned operant responses that gave them access to food, human contact or contact with another dog. Each experimental dog had a yoked partner that received the same reward without having to perform the response, after a matched delay. The dogs that produced the reward through their own behavior responded differently from those that simply received it, a difference the authors interpreted as positive emotional arousal linked to achievement.

The study is small and limited to one breed, and control and predictability were not fully separated. But it suggests that what a reward is worth may depend partly on whether the dog's behavior made it happen (why control and predictability matter).

8.2 Caudate Responses to Praise and Food Cues

Neuroimaging offers a different way to ask about reward value. In a study with 15 awake, unrestrained dogs, objects were first paired with consequences: one object predicted praise from the handler, another predicted a piece of hot dog, and a third predicted nothing (Cook, Prichard, Spivak & Berns, 2016). Activation in the ventral caudate, a brain region involved in reward processing, was then compared between the cues. Thirteen of the fifteen dogs showed equal or greater activation to the praise cue than to the food cue.

In a second experiment with 13 dogs, praise was withheld on some trials. The difference between praise and food responses in the first experiment predicted how the dogs' brains responded to the praise cue in the second. In a third experiment, dogs chose in a Y-maze between their owner and a bowl of food over 20 trials, and the individual neural preference predicted their choices.

Two points are essential. First, the "praise" alternative in the Y-maze included the owner's presence, so it was not pure verbal praise. Second, the result is an individual brain-behavior link. The study does not show that most dogs prefer praise to food. The authors also named limits: 15 dogs is a small sample, a prior association between owners and food could not be excluded, and attachment, training history and breed were not examined (how dopamine and reward processing shape learning).

8.3 Individual Differences in the Brain

Earlier imaging work had already shown how variable reward responses are. In a study of 13 dogs, eight of them showed a positive differential caudate response to a hand signal that predicted food (Berns, Brooks & Spivak, 2013). Neural data tell the same story as the behavioral studies: averages exist, but individual dogs differ.

9. Measuring Preference

9.1 Paired-Stimulus Assessments

The most widely used method in behavior analysis is the paired-stimulus preference assessment: two items are presented at once, the dog chooses one, and over many pairings a ranking emerges. In a study with eight dogs, a paired-stimulus assessment of foods was followed by concurrent and progressive-ratio tests of reinforcer effectiveness (Vicars, Miguel & Sobie, 2014). The paired-stimulus assessment predicted reinforcer effectiveness in all subjects.

The study is small, but preference tests also generally predicted efficacy in the study on food delivery (Feuerbacher et al., 2022).

9.2 Owners Can Run the Test

Preference assessments do not have to be run by researchers. In one study, owners conducted paired-stimulus food preference assessments with their own dogs at home (Waite & Kodak, 2023). Fifteen dogs were enrolled, and eleven dogs from nine owners completed the study with video recordings. The owners carried out the procedure with an average integrity of 99.8%, and high- and low-preferred foods were then tested as reinforcers on progressive-ratio or fixed-ratio schedules.

The most important finding for practice was that the preference hierarchies identified by the assessment were often different from what the owners had guessed. The sample was small, but the finding supports the central argument of this article: value has to be tested, not assumed.

9.3 Preference and Approach Speed

A third study linked preference to a different measure. Eight dogs first completed a paired-stimulus assessment of six raw foods, then ran a five-meter runway for their most preferred food, their least preferred food and two staple foods (Cameron, de Garnham, Jensen & Bizo, 2019). Dogs approached their most preferred food significantly faster than their least preferred food and both staple foods, while the two staple foods did not differ. The authors noted possible order effects, and data from one dog were partly omitted because of a procedural error.

9.4 The Owner as Part of the Test

A preference test is only informative if the dog chooses freely. A study of 54 pet dogs showed how easily that can be undermined (Prato-Previde, Marshall-Pescini & Valsecchi, 2008). When choosing independently, dogs picked the larger of two food portions significantly more often. When their owners showed a preference for the smaller portion, dogs chose the larger portion significantly less often, and the owner's influence was strongest when the portions were equal.

In a preference test at home, a dog may follow the owner's cues rather than its own preference, especially when the options are similar.

9.5 Preference Is Not the Same as Efficacy

Preference and efficacy are related but not identical. A dog may choose a toy over a treat in a free choice and still work less for the toy, and the toy-versus-food study measured work, not choice (Lazaro et al., 2023). In most studies the two measures pointed in the same direction (Vicars et al., 2014; Feuerbacher et al., 2022), but they answer different questions.

10. What Follows for Practice

10.1 Measure, Do Not Assume

The value of a reward for a particular dog cannot be read off a list. Food outperformed brief social interaction and toys for most dogs in controlled tests (Feuerbacher & Wynne, 2012; Lazaro et al., 2023), but not for all, and when both food and social contact were withheld, owned dogs showed no clear food preference (Isernia et al., 2022). Owners' guesses about their dog's favorite food were often wrong (Waite & Kodak, 2023). The reliable way to know what works is to watch what the behavior does.

Training with rewards depends on consequences that actually function as reinforcers for that dog, not on consequences that are meant as rewards.

10.2 Testing Preferences at Home

In practice: A simple paired choice can be run at home, following the logic of the paired-stimulus assessments described above. Two small pieces of different foods are presented at the same time, equally spaced, and the dog's choice is noted. Over several rounds, with positions swapped, each food is paired with each other food, and the food chosen most often is the best candidate for a high-value reward. To reduce owner influence, the person presenting the food should look at the dog rather than at either piece and stay quiet (Prato-Previde et al., 2008). This procedure is an adaptation for everyday use, not one of the tested research protocols.

10.3 Adding a Work Requirement

In practice: A choice shows what a dog prefers, not how much it will do for it. A second step is to ask for a simple, well-known behavior, such as a hand touch or a sit, and to raise the requirement gradually with each candidate reward, for example by asking for two, then three repetitions before the reward comes. A reward that keeps the dog working as the requirement rises is a stronger reinforcer than one the dog stops working for early. This follows the logic of the progressive-ratio tests in the studies above (Vicars et al., 2014; Lazaro et al., 2023), but it is a simplified version from practice and should be kept short to avoid frustration.

10.4 What Weak Praise Means for Training

If vocal praise has little reinforcing value for a particular dog in a particular situation, a training plan that relies on it almost exclusively will often deliver less than expected. A dog that does not respond reliably may simply not be receiving a consequence worth the effort. When food was dense, it was preferred to petting (Feuerbacher & Wynne, 2014), and adding praise to food did not raise its efficacy (Feuerbacher et al., 2022).

Praise is not useless: it can carry information if it reliably predicts a valued consequence, and dogs preferred petting to vocal praise in choice procedures (Feuerbacher & Wynne, 2015). The implication is to pair praise with a consequence the dog actually works for, or to use petting or play where the dog clearly values them, and to check whether the behavior is maintained.

10.5 Matching the Reward to the Situation

Different rewards may suit different purposes. In difficult environments or with difficult tasks, a high-value reward is more likely to compete with distractions. In procedures such as counterconditioning, where a reward is paired with a trigger the dog is afraid of, the reward has to be strong enough to matter to the dog in that state (how counterconditioning uses rewards). Play may suit energetic behaviors, quieter food delivery may suit calm ones, and petting may be welcome in one context and an intrusion in another. The choice of reward is not a fixed decision but one revisited as circumstances change.

11. Summary at a Glance

Reward value is a property of the relation between a consequence, a dog, its current state and the context, not of the item, and it is defined by what the consequence does to behavior.

Brief social interaction was a weaker reinforcer than food for most dogs (Feuerbacher & Wynne, 2012), and dogs preferred food to petting when food was readily available (Feuerbacher & Wynne, 2014). When both food and social contact were withheld, owned dogs showed no clear preference between owner and food (Isernia et al., 2022). Dogs preferred petting to vocal praise (Feuerbacher & Wynne, 2015), and adding praise to food delivery had no measurable effect (Feuerbacher et al., 2022). The value of the same petting depended on who delivered it and where (Feuerbacher & Wynne, 2017). For most pet dogs in one study, the best food sustained more work than the best toy (Lazaro et al., 2023), and interest in a toy faded quickly with repeated exposure (Pullen et al., 2012).

Food quality raised running speed where quantity did not (Riemer et al., 2018), while larger portions raised break points (Feuerbacher et al., 2022). Variety and constancy were preferred by different dogs (Bremhorst et al., 2018). Evidence on incentive contrast is inconsistent (Bentosela et al., 2009; Riemer et al., 2016; Vieira de Castro et al., 2025), while withholding an expected reward produced frustration responses (Jakovcevic et al., 2013). Partial rewarding during acquisition did not speed learning (Cimarelli et al., 2021). Paired-choice tests predicted reinforcer efficacy (Vicars et al., 2014), and owners' guesses often did not match tested preferences (Waite & Kodak, 2023).

12. Research Gaps and Critical Appraisal

12.1 Small Samples

Almost every study in this field is small. The toy-versus-food study tested ten dogs (Lazaro et al., 2023), the paired-stimulus validation eight (Vicars et al., 2014), the runway preference study eight (Cameron et al., 2019), the amount-discrimination study four (McGuire et al., 2018), and the imaging study fifteen (Cook et al., 2016). Even the larger studies, such as the variety study with 16 dogs (Bremhorst et al., 2018) or the contrast studies with 36 to 45 dogs (Riemer et al., 2016; Vieira de Castro et al., 2025), are small by the standards of most behavioral research. Single-subject designs show effects in individuals but cannot estimate how common a pattern is.

12.2 Short Tests and Artificial Tasks

Most studies measured value over minutes or a few sessions, often with arbitrary responses such as nose touches or gazing at an experimenter. Whether the same rankings hold over weeks of training, with complex behaviors and real distractions, is largely untested. The petting-versus-praise study used choice procedures, not training performance (Feuerbacher & Wynne, 2015), and the breakfast study measured search accuracy, not reinforcer value (Miller & Bender, 2012).

12.3 Populations

Several studies included shelter dogs, laboratory dogs or kennel-housed dogs (Feuerbacher & Wynne, 2012; Pullen et al., 2012; Protopopova et al., 2016; Vieira de Castro et al., 2025). These study populations have particular histories, housing and access to social contact, and responses differed by population in the contrast study (Vieira de Castro et al., 2025). Findings from these groups cannot be transferred directly to family dogs.

12.4 Training History

Reward value is partly learned. A dog trained for years with tug toys may value play differently from a dog trained with food, and the toy-versus-food study explicitly excluded dogs with a toy-based training history (Lazaro et al., 2023). The imaging study could not rule out that owners had previously been associated with food (Cook et al., 2016). Studies that systematically compare dogs with different reward histories are largely missing.

12.5 Inconsistent Results

The research on incentive contrast is the clearest example of inconsistent findings: one study found a contrast effect, one found none and even a reverse effect, and one found effects in some populations and individuals but not in others (Bentosela et al., 2009; Riemer et al., 2016; Vieira de Castro et al., 2025).

12.6 Measures That Disagree

Choice, running speed and break points do not always agree. Quality affected running speed while quantity did not (Riemer et al., 2018), whereas amount affected break points (Feuerbacher et al., 2022). This shows that "reward value" is not one quantity, and studies using several measures in the same dogs are needed.

13. Conclusion

The research on reward value in dogs does not produce a fixed ranking of rewards. Food tended to outperform brief social interaction and toys in controlled tests, but for most dogs in particular procedures, not for all dogs in all situations. Dogs preferred petting to praise, and praise added nothing measurable when combined with food. The value of every reward shifts with the dog's food motivation, recent experience, the person delivering it and the place. Contrast effects are inconsistent, while withholding an expected reward produced frustration responses.

The most useful lesson is methodological. Because reward value is a relation rather than a property of the item, it has to be measured in the individual dog. Paired choices and simple work tests, adapted from the research, offer a way to do this at home. A reward is not what the trainer intends it to be. It is what the dog's behavior shows it to be.

Key Insights (Takeaways)

• Reward value is a property of the relation between a consequence, a dog, its state and the context, not of the item itself, and it is defined by what the consequence does to behavior.

• Brief social interaction was a weaker reinforcer than food for most dogs, and food was preferred to petting when food was readily available (Feuerbacher & Wynne, 2012, 2014).

• When food and social contact were both withheld, owned dogs as a group showed no clear preference between owner and food (Isernia et al., 2022).

• Dogs preferred petting to vocal praise, and adding praise to food delivery had no measurable effect on how much dogs worked (Feuerbacher & Wynne, 2015; Feuerbacher et al., 2022).

• For most pet dogs in one study, the best food sustained more work than the best toy, but individual exceptions existed (Lazaro et al., 2023).

• Food quality, amount and variety affect value in different ways, and dogs differ in whether they prefer variety or their favorite food every time (Riemer et al., 2018; Bremhorst et al., 2018).

• Evidence on incentive contrast in dogs is inconsistent, while withholding an expected reward produced frustration responses (Riemer et al., 2016; Jakovcevic et al., 2013).

• Owners often guessed wrong about their dog's preferred food; short paired-choice tests predicted which food dogs would work for (Waite & Kodak, 2023; Vicars et al., 2014).

References

Bentosela, M., Barrera, G., Jakovcevic, A., Elgier, A. M., & Mustaca, A. E. (2008). Effect of reinforcement, reinforcer omission and extinction on a communicative response in domestic dogs (Canis familiaris). Behavioural Processes, 78(3), 464–469. https://doi.org/10.1016/j.beproc.2008.03.004

Bentosela, M., Jakovcevic, A., Elgier, A. M., Mustaca, A. E., & Papini, M. R. (2009). Incentive contrast in domestic dogs (Canis familiaris). Journal of Comparative Psychology, 123(2), 125–130. https://doi.org/10.1037/a0013340

Berns, G. S., Brooks, A., & Spivak, M. (2013). Replicability and heterogeneity of awake unrestrained canine fMRI responses. PLoS ONE, 8(12), e81698. https://doi.org/10.1371/journal.pone.0081698

Bremhorst, A., Bütler, S., Würbel, H., & Riemer, S. (2018). Incentive motivation in pet dogs – preference for constant vs varied food rewards. Scientific Reports, 8, 9756. https://doi.org/10.1038/s41598-018-28079-5

Cameron, K. E., de Garnham, J., Jensen, K., & Bizo, L. A. (2019). Food preference predicts speed of approach on a runway task by dogs. Pet Behaviour Science, 8, 1–10. https://doi.org/10.21071/pbs.v0i8.11179

Cimarelli, G., Schoesswender, J., Vitiello, R., Huber, L., & Virányi, Z. (2021). Partial rewarding during clicker training does not improve naïve dogs' learning speed and induces a pessimistic-like affective state. Animal Cognition, 24(1), 107–119. https://doi.org/10.1007/s10071-020-01425-9

Cook, P. F., Prichard, A., Spivak, M., & Berns, G. S. (2016). Awake canine fMRI predicts dogs' preference for praise vs food. Social Cognitive and Affective Neuroscience, 11(12), 1853–1862. https://doi.org/10.1093/scan/nsw102

Dzik, M. V., Jousset, M., Vietti, J., & Bentosela, M. (2025). Effects of the presence of a human on frustration in domestic dogs: Is there a social buffering phenomenon? Applied Animal Behaviour Science, 283, 106503. https://doi.org/10.1016/j.applanim.2025.106503

Feuerbacher, E. N., Stone, C., & Friedel, J. E. (2022). Give the dog a big bone: Magnitude but not delivery method of food impacts preference and reinforcer efficacy in dogs. Behavior Analysis: Research and Practice, 22(1), 31–49. https://doi.org/10.1037/bar0000237

Feuerbacher, E. N., & Wynne, C. D. L. (2012). Relative efficacy of human social interaction and food as reinforcers for domestic dogs and hand-reared wolves. Journal of the Experimental Analysis of Behavior, 98(1), 105–129. https://doi.org/10.1901/jeab.2012.98-105

Feuerbacher, E. N., & Wynne, C. D. L. (2014). Most domestic dogs (Canis lupus familiaris) prefer food to petting: Population, context, and schedule effects in concurrent choice. Journal of the Experimental Analysis of Behavior, 101(3), 385–405. https://doi.org/10.1002/jeab.81

Feuerbacher, E. N., & Wynne, C. D. L. (2015). Shut up and pet me! Domestic dogs (Canis lupus familiaris) prefer petting to vocal praise in concurrent and single-alternative choice procedures. Behavioural Processes, 110, 47–59. https://doi.org/10.1016/j.beproc.2014.08.019

Feuerbacher, E. N., & Wynne, C. D. L. (2016). Application of functional analysis methods to assess human–dog interactions. Journal of Applied Behavior Analysis, 49(4), 970–974. https://doi.org/10.1002/jaba.318

Feuerbacher, E. N., & Wynne, C. D. L. (2017). Dogs don't always prefer their owners and can quickly form strong preferences for certain strangers over others. Journal of the Experimental Analysis of Behavior, 108(3), 305–317. https://doi.org/10.1002/jeab.280

Isernia, L., Wynne, C. D. L., House, L., & Feuerbacher, E. N. (2022). Dogs and wolves differ in their response allocation to their owner/caregiver or food in a concurrent choice procedure. PeerJ, 10, e12834. https://doi.org/10.7717/peerj.12834

Jakovcevic, A., Elgier, A. M., Mustaca, A. E., & Bentosela, M. (2013). Frustration behaviors in domestic dogs. Journal of Applied Animal Welfare Science, 16(1), 19–34. https://doi.org/10.1080/10888705.2013.740974

Lazaro, X. A., Winter, J. M., Fernand, J. K., Cox, D. J., & Dorey, N. R. (2023). Efficacy of edible and leisure reinforcers with domestic dogs. Animals, 13(19), 3073. https://doi.org/10.3390/ani13193073

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

McGuire, K., Bizo, L. A., McBride, E. A., & Kocek, T. B. (2018). Discrimination of food amounts by the domestic dog (Canis familiaris). International Journal of Comparative Psychology, 31, Article 04.03. https://doi.org/10.46867/ijcp.2018.31.04.03

Miller, H. C., & Bender, C. (2012). The breakfast effect: Dogs (Canis familiaris) search more accurately when they are less hungry. Behavioural Processes, 91, 313–317. https://doi.org/10.1016/j.beproc.2012.09.012

Okamoto, Y., Ohtani, N., Uchiyama, H., & Ohta, M. (2009). The feeding behavior of dogs correlates with their responses to commands. Journal of Veterinary Medical Science, 71(12), 1617–1621. https://doi.org/10.1292/jvms.001617

Prato-Previde, E., Marshall-Pescini, S., & Valsecchi, P. (2008). Is your choice my choice? The owners' effect on pet dogs' (Canis lupus familiaris) performance in a food choice task. Animal Cognition, 11(1), 167–174. https://doi.org/10.1007/s10071-007-0102-7

Protopopova, A., Brandifino, M., & Wynne, C. D. L. (2016). Preference assessments and structured potential adopter-dog interactions increase adoptions. Applied Animal Behaviour Science, 176, 87–95. https://doi.org/10.1016/j.applanim.2015.12.003

Pullen, A. J., Merrill, R. J. N., & Bradshaw, J. W. S. (2012). Habituation and dishabituation during object play in kennel-housed dogs. Animal Cognition, 15(6), 1143–1150. https://doi.org/10.1007/s10071-012-0538-2

Riemer, S., Ellis, S. L. H., Ryan, S., Thompson, H., & Burman, O. H. P. (2016). A reappraisal of successive negative contrast in two populations of domestic dogs. Animal Cognition, 19(3), 471–481. https://doi.org/10.1007/s10071-015-0947-0

Riemer, S., Ellis, S. L. H., Thompson, H., & Burman, O. H. P. (2018). Reinforcer effectiveness in dogs – The influence of quantity and quality. Applied Animal Behaviour Science, 206, 87–93. https://doi.org/10.1016/j.applanim.2018.05.016

Rooney, N. J., & Bradshaw, J. W. S. (2002). An experimental study of the effects of play upon the dog–human relationship. Applied Animal Behaviour Science, 75(2), 161–176. https://doi.org/10.1016/S0168-1591(01)00192-7

Vicars, S. M., Miguel, C. F., & Sobie, J. L. (2014). Assessing preference and reinforcer effectiveness in dogs. Behavioural Processes, 103, 75–83. https://doi.org/10.1016/j.beproc.2013.11.006

Vieira de Castro, A. C., Baria-Unwalla, P., Cabrita, A. R., & Olsson, I. A. (2025). Successive negative contrast: An appropriate approach to measure affective state in dogs? Animal Welfare, 34, e34. https://doi.org/10.1017/awf.2025.29

Waite, M. R., & Kodak, T. M. (2023). Owner-implemented paired-stimulus food preference assessments for companion dogs. Journal of the Experimental Analysis of Behavior, 120(1), 62–77. https://doi.org/10.1002/jeab.846