Addictive-like behavioural traits in pet dogs with extreme motivation for toys

What is play? Why do many large-brained mammals engage in play throughout their lives? And what makes playing potentially addictive? Despite numerous publications on play and its possible functions, play has remained somewhat of a mystery¹, being associated with no immediate adaptive function, although it has been suggested to allow animals to practice species-typical behaviours such as hunting, mating, or fighting with a competitor in a non-serious context¹. Notwithstanding, play behaviour is ubiquitous among (at least young) mammals and some birds^2,3, and in large-brained species in particular – from humans to dogs – it persists throughout life^4,5. Still, no unified definition of play exists to date⁶, although there is some agreement that, at a proximate level, play makes us feel good⁷. Even in non-human animals, observers will often agree that playful activities look like fun⁷, with neurotransmitter systems mediating the rewarding aspects of play (opioids, cannabinoids and dopamine)^8,9 appearing to be highly conserved across mammalian taxa¹⁰.

Bateson (2014) proposed a set of characteristics that are generally accepted to define play: it is spontaneous, intrinsically rewarding and “fun”; it is separate from serious consequences; it often involves novel or exaggerated actions and role reversal; it is repetitive, but distinct from stereotypies; and it usually occurs only in healthy, stress-free animals, making it a marker of well-being⁷. However, regarding the latter point, it has been highlighted that play can also represent an attempt to cope with suboptimal conditions (e.g. in nonhuman animals, play may occur as displacement behaviour in stressful situations⁵ or may serve to reduce social tensions)¹¹. Also, in humans, playing computer games or gambling represents a way of coping with stress. Moreover, in some instances, what started as a fun activity can become compulsive and develop into a behavioural addiction^12,13,14.

A behavioural addiction can be defined as “repeated failure to resist an impulse, drive, or urge to perform an act that is rewarding to the person (at least in the short-term), despite longer-term harm to the individual or others” (ICD-11 (International Classification of Diseases 11th Revision))¹⁵. Unlike in compulsive disorders, where performance of the compulsive behaviour primarily serves to provide some relief from a negative affective state, i.e. via negative reinforcement, addictive behaviours originate because their performance generates positive affect, i.e. via positive reinforcement. However, as the addiction develops, the behaviour becomes compulsive and may even cease to be rewarding (reviewed by Freimuth et al.¹⁶. Behavioural addictions share underlying neurobiological processes¹⁷ and behavioural symptoms (such as craving, lack of self-control, tolerance, withdrawal and risk of relapse) with substance addictions^17,18. Still, they are more heterogeneous and less well-understood¹⁹.

While a wide range of behaviours have the potential to become addictive in people (e.g. exercise, sex, shopping, work, etc.)^{19,20,21,22,23,24,25}, to date, only the two disorders related to playing – gambling and internet gaming – are officially recognised as behavioural addictions in the two psychiatric manuals of psychological disorders (DSM-5 and ICD-11). The ICD-11¹⁵ included both gambling and internet gaming as behavioural addictions. In the 5th edition of DSM-5²⁶, gambling, previously classified as an impulse control disorder, was included under “substance-related and addictive disorders”^19,26, while internet gaming was listed separately as “internet gaming disorder”²⁶.

What would make behaviours related to playing so addictive? Play involves neurotransmitter systems (opioids, cannabinoids, and dopamine) that are also engaged in the rewarding aspects of food and drug rewards^8,9. Thus, video games can provide players a hedonic experience and a high degree of relaxation²⁷. Pathological gaming is an example of how seemingly normal and enjoyable behaviours can develop to disrupt regular social and environmental functioning^28,29,30.

Compared to substance addiction, there are only a few animal models of behavioural addictions. Moreover, these are restricted to controlled laboratory settings, and addictive-like behaviour has to be actively induced³¹. Rodent models have been used to investigate compulsive eating (e.g. reviewed in³¹, exercise addiction (wheel running^32,33), gambling³⁴, and responses to sexual reward³⁵. Mice selectively bred for excessive wheel-running, sometimes referred to as an addiction-prone phenotype, develop physiological withdrawal symptoms similar to those found in drug addiction after abstinence³⁶. As with excessive exercise in humans, wheel-running in rodents may become disruptive to everyday activities, leading to impaired nest-building and sheltering behaviour^37,38. The animals may continue to wheel-run despite disrupted sleep³⁹ or even in the face of injury⁴⁰, thus fulfilling the behavioural addiction criterion “persistence of the behaviour despite adverse consequences”⁴¹. This suggests that behavioural addictions are not unique to the human species.

There is, however, only one species that appears to display addictive-like behaviour spontaneously, without intentional experimental induction: the domestic dog (although inadvertent promotion of addictive-like behaviour by the caretakers cannot be ruled out). A small subset of dogs – colloquially referred to as “ball junkies” – appear to demonstrate an addictive-like desire for object play⁴².

Like humans, domestic dogs frequently remain playful throughout their lives¹, engaging in both social and object-related play, as well as combinations (e.g. tug-of-war⁹. Solitary object play appears to be related to predatory behaviour⁹; accordingly the development of social and object play may reflect different selective histories of dog breeds, which were selected for various purposes such as hunting, guarding, herding, and other functions^6,7,8.

Toy play is a potent reinforcer, especially in working dog training^43,44,45. For instance, detection dogs working in public settings are typically not rewarded with food due to concerns about undesirable food-seeking behaviours in the field. Still, they will work persistently for their toy rewards. It has been argued that artificial selection has exaggerated play behaviour in adult dogs, especially in working breeds or working lines, where high toy motivation is often actively selected as a predictor of performance⁴⁶. For example, in Labrador retrievers, working lines demonstrate higher playfulness than show lines, indicating a genetic basis for play motivation and potential for artificial selection⁴⁷.

Playing with toys allows dogs to express instinctive predatory sequences such as chasing, catching, possessing and “dissecting”, considered to be intrinsically rewarding to them based on their species and breed histories⁴⁸. None of this is pathological, nor is gambling or computer gaming in people. However, such highly rewarding activities have the potential to become obsessive in humans^49,50, and the same may be true for dogs.

While addictive-like behaviour towards toys in dogs has not been studied to date, the phenomenon has been described in the lay literature (where affected dogs may be referred to as ‘ball junkies’), and it has been (rarely) alluded to in the scientific literature. Lazarowski et al.⁴⁸ describe how some dogs show behavioural and physiological signs of high arousal in relation to toys, lack of self-control, and behaviours such as whining, barking, spinning, and other behavioural signs of stress when access to a toy is prevented (e.g. because the dog is restrained), suggested as an expression of their inability to manage the frustration of anticipation⁴⁶. All these signs could be interpreted as indicative of craving (and frustration when the urge cannot be fulfilled).

In humans, addictive behaviours are often associated with deficits in inhibitory control and heightened cue-reactivity and craving, which are likely key mechanisms in addiction, particularly when exposed to behaviour-specific cues^51,52,53.

In animal models of addictions, not only is an increased motivation to work for the rewarding substance notable, but the animals also continue seeking the reward even when it is signalled to be unavailable⁵⁴. Similarly, excessively toy-motivated dogs may continue to try to gain access to a toy even when the caretaker has put it away (anecdotal evidence⁵⁵. Dogs that appear obsessed with toys cannot be easily distracted from their fixation on the preferred object – demonstrating the high salience of the toy. Such dogs may even lose interest in other stimuli or social interactions as long as they have access to the toy, or sometimes even when it has been removed from reach⁵⁵ – i.e., everyday functioning may be affected. Moreover, some dogs may continue playing (e.g., running tirelessly after balls thrown for them) despite adverse consequences, such as over-exertion or even injury in the short term and damage to joints and ligaments in the longer term⁵⁶.

Thus, we suggest that ‘excessive toy motivation’ in dogs may show parallels to behavioural addictions in humans. Domestic dogs share many complex behavioural traits with us^57,and they are commonly used as model species to explore compulsive behaviours^58,59; cognitive ageing^60,61,62, ADHD^63,64,65, neuroticism⁶⁶ and autism^67,68,69.

Here, we aim to provide the first scientific evaluation of ‘excessive toy motivation’ in dogs, develop methods to assess this phenomenon, and investigate whether ‘excessive toy motivation’ in pet dogs meets the defining criteria of behavioural addictions. Due to the heterogeneity of behavioural addictions, the number and description of diagnostic criteria are inconsistent in the scientific literature, even in humans¹⁹. We decided to explore whether the most common behavioural addiction criteria can be adapted to dogs: (1) craving, (2) salience, (3) mood modification through carrying out the behaviour, (4) lack of self-control, (5) tolerance, (6) withdrawal symptoms, (7) external consequences (the addictive behaviour causes conflict with other activities, other individuals, or within the individual), and (8) relapse after abstinence from the activity (cf^28,41,70,71. Two additional criteria are used for diagnosing behavioural addictions in humans: having problems at home or work and lying to/deceiving people close to them^26,72. Since these criteria cannot be applied to animals, we focused only on the eight abovementioned criteria.

We developed a behavioural test exposing pet dogs to various situations where behavioural addiction criteria in relation to toys can be expressed. Only the first four of the criteria mentioned above can be measured in a single behavioural assessment. The remaining criteria were included in an accompanying questionnaire, in which the dogs’ owners were asked about their dogs’ everyday behaviour.

While this study is exploratory, given the lack of prior research in this area, we used convergent methodologies in an attempt to assess internal and external validity. We predicted that dogs classified as having a high tendency for addictive-like behaviour based on our continuous Addictive-like Behaviour Test score, would:

• Show higher scores for the individual behavioural addiction criteria: Salience, Craving, Mood modification, and Lack of self-control in the behaviour test,

• Show higher durations of focusing on and trying to access an unavailable toy in the behaviour test,

• Receive higher scores on the owner questionnaire designed to measure dogs’ addictive-like behaviour in everyday life,

than dogs classified as having a low tendency for addictive-like behaviour.

Play motivation test

Ethical consideration

The study was assessed and approved by the Veterinary Office of the Canton of Bern, Switzerland (Licence number BE115/17). All procedures were performed in accordance with the “Guidelines for the Treatment of Animals in Behavioral Research and Teaching” of the Association for the Study of Animal Behavior. All dog owners provided written informed consent for their participation.

Subjects

One hundred twenty-six dog-owner teams were recruited via advertisements on social media. In the first call, any play-motivated dog was welcome to participate. In a second call, we specifically sought dogs showing ‘excessive’ motivation for toy play.

Twenty-one of the 126 tested dogs were excluded from the analysis as they were (1) outside the target age range (< 1 year or > 10 years old; N = 9), (2) did not complete the test due to fatigue (N = 1), (3) did not play at all or were too fearful for pulse measurements (N = 7), or (4) due to disturbances during the test (e.g. owners bringing young children along, N = 4). The final sample (N = 105) included 56 males (34 neutered or chemically castrated, 20 intact and 2 cryptorchids) and 49 females (34 neutered and 15 intact), ranging in age from 12 months to 10 years (mean age = 5.09 years, SD = 2.6). The dogs belonged to various breeds (for demographics, see Supplementary Table 1). Eighty-two owners (72 women and 10 men) participated in the study.

Experimental set-up

Behaviour tests took place in an experimental room (Fig. 1), measuring 5.22 m x 3.36 m. A wooden partition wall divided the room into two parts so that the effective testing space was 3.60 m x 3.36 m. The room was furnished with two chairs and several shelves on the walls. One of the chairs was placed in front of the wooden partition wall (facing the entrance door), and the other was placed at a 90° angle against the wall to the left. In front of both chairs, a taped line marked a one-meter distance from the chairs. During the habituation phase, the opaque box in which a toy or food was enclosed during several subtests (hereafter, unsolvable task box) was placed next to the experimenter’s chair.

Four video cameras (IB8377-H; 4 MP, 30 fps, H.264, WDR Pro, IR, PoE, IP66, 2.8–12 mm) were placed in the room, and recordings were made using the recorder system (ND9441P NVR, 16-CH, 4HDD, H.265, HDMI/VGA, 16x PoE).

Fig. 1

Experimental room with the experimenter and owner in the starting position.

Methods

The test battery consisted of 14 subtests assessing various aspects of toy motivation in dogs. Play behaviour per se cannot be used to infer addictive-like behaviour, which is characterised primarily by reactions when the reward is unavailable; therefore, only subtests relevant to exploring behavioural addiction criteria are described in detail hereafter. The complete description of the play motivation test is available under: https://figshare.com/s/dfd6d12d922f7543b96c.

Procedure

Room habituation

After the owner and the dog had entered the test room, the dog was unleashed, and a 3-minute habituation phase commenced (Fig. 1). Meanwhile, the owner and the experimenter were seated on their allocated chairs, and the experimenter explained the test procedure. The owner signed the consent form. The owners were instructed to interact with the dog only when asked to perform one of the subtests and not to use food during testing unless absolutely necessary (such as exchanging food for a toy if the dog was unwilling to relinquish it).

Choosing the toy

After the habituation phase, the experimenter retrieved a box containing various commercial dog toys of different sizes and textures, with and without squeakers, etc., from the adjacent storage room. Only toys that might be associated with food enrichment were excluded. The owner was asked to select three toys (one ball, one tug toy and one plush toy) which they thought the dog would like the most. If the owner had brought the dog’s favourite toy from home, this toy was used in the subsequent preference test along with two other toys.

After removing the toy box from the room, the experimenter returned to the test room. The owner recalled the dog and sat down on their chair, holding the dog behind the Line marking the 1 m distance from the chair. Opposite the dog at the front of the room, the experimenter placed the three toys on the floor in a row, 40 cm apart. After the experimenter had returned to her chair, the owner released the dog, who could now explore and play with the toys for 30 s. The two people present did not interact with the dog during this time. The toy the dog spent the most time interacting with was used for subsequent testing. Forty-five dogs selected a ball, nine selected a tug toy, 39 selected a plush toy, and 12 selected a hybrid toy (plush ball: N = 3; tug with a ball: N = 6; plush tug: N = 3). On rare occasions, the dog did not show interest in any of the toys. In this case, the owner was asked to choose the type of toy the dog was usually most interested in at home. The chosen toy was used throughout the experiment, and the remaining two toys were placed on the shelf out of reach and sight of the dog. If the preferred toy was not a tug toy, the tug toy was used in subtests where the owner or experimenter played tug-of-war with the dog.

Description of the subtests and their relevance for addiction criteria coding

A description of the subtests and, when applicable, their relevance for addiction criteria coding is given in Table 1.

Behavioural coding

Videos were coded using Solomon Coder (Solomon Coder beta 19.08.02, Copyright 2006–2019 by Andràs Péter).

For most subtests, the starting point for coding was when the experimenter and the owner were sitting on their chairs, and the dog was behind the Line, which marked a 1 m distance from the owner’s chair (Fig. 1).

Qualitative and quantitative coding was performed by coders who were not involved in the experiments.

Three different coding approaches were employed:

1. a.

   Scoring of individual variables that may be indicative of addictive behaviour each minute, which were later summed up as Addictive-like Behaviour Test score (Table 2); coder: KS.

2. b.

   Coding of presence/absence of the four addiction criteria in each minute of each subtest; coder: KS.

3. c.

   Quantitative coding.

• Scoring and point sampling of behaviours during the subtests “Social play” and “Dog alone” (Table 4); coder: FL.

• Coding of absolute durations of behaviours in subtests where the toy was rendered inaccessible (unsolvable task box and toy on a shelf, see Table 4); coders: DZ and AH.

A second coder (AM) performed reliability coding of addictive-like behaviours and behavioural addiction criteria and point sampling for 15 dogs. Reliability between the two coders who coded the durations was also analysed for 15 dogs. Reliability was good or excellent for all included variables (ICC, absolute agreement, single measures, two-way mixed-effects model, computed in IBM SPSS Statistics Version 23 (IBM Corporation and its Licensors 1989, 2015) (see Supplementary Table 3 for full results).

Sub-criteria to generate an Addictive-like behaviour test score (AB-T score)

To quantify dogs’ propensity for addictive-like behaviour as objectively as possible, we introduced the Addictive-like Behaviour Test score (AB-T score). Applicable sub-criteria were rated for each minute of the test, and for analysis, each sub-criterion was assigned a score between 0 and 2 points, as detailed in Table 2. The points from all the subtests, including the cool-down period, were added to yield the AB-T score. The maximum possible value of the AB-T score was 120 points. A cut-off point for addictive-like behaviour was defined by a data range split divided into two halves. Dogs scoring equal to or above the mid-point (44.2 points) are referred to as dogs showing a high tendency for addictive-like behaviour or high-AB dogs.

Dogs scoring less than 44.2 points are referred to as low-AB dogs (dogs with a low tendency for addictive-like behaviour). The sub-criteria included in the Addictive-like Behaviour Test score (as detailed in Table 2) were selected as they were assumed to be relatively independent of the level of obedience and training. Dogs might have been trained to drop a toy on a cue and to exert impulse control and refrain from jumping towards the toy in the experimenter’s hand; therefore, these variables were not included in the AB-T score. However, behaviours such as staring at the toy or pacing are believed to be less subject to training and were included.

Presence/absence of behavioural addiction criteria

Separately from the AB-T variables, in each subtest, the addiction criteria Salience, Craving, Mood modification, and Lack of self-control were rated each minute as present or absent based on the occurrence of pre-defined behaviours. For instance, Salience was inferred from searching for a toy although there was an »attractive alternative« (food, owner inviting the dog to play). Craving was based on the dog focusing mainly on the toy (> 50%) and medium to high arousal directed at the inaccessible toy (inferred from behaviours such as panting, restlessness, and high muscle tension). We further coded behaviours that are usually characteristic of high arousal in some dogs, e.g., pacing, jumping towards the toy, and vocalising (see Table 3). If at least one of the pre-defined behaviours was expressed, the respective addiction criterion was coded as present. Tolerance, Withdrawal symptoms and Risk of relapse after abstinence could not be tested in the setting of the play motivation test since they develop over time.

For each subtest, a summary score for each of the four addiction criteria was computed by summing up the points for each minute of the subtest.

Quantitative coding

Selected behaviours during the subtests “Social play” and “Dog alone” were coded by point sampling at 3-second intervals and then extrapolated to proportions of time (Table 4). Based on a subsample of dogs coded using both point sampling and absolute durations of behaviour, we determined sufficient agreement between the two measurement methods, justifying the use of point sampling. During subtests where a reward was inaccessible in the unsolvable task box, the absolute duration of interacting with the box (with low or high effort) was coded (Table 4). For detailed definitions of quantitatively coded variables, see Supplementary Table 2.

A questionnaire was developed (available in English and German) (for inter- and intra-rater reliability based on over 1500 dogs recruited via an online survey; see Supplementary Table 6). The questions relating to behavioural addiction criteria relevant to the present manuscript are shown in Table 5. They were rated on a 5-point Likert scale indicating the extent of agreement with the statement (1 – strongly disagree; 2 – partly disagree; 3 – neither agree nor disagree; 4 – partly agree; 5 – strongly agree). The dog owners were asked to complete this questionnaire during the cool-down period of the behavioural test.

SPSS Statistics Version 23 (IBM Corporation and its Licensors 1989, 2015) was used to compute a Categorical Principal Component Analysis and Mann-Whitney U tests. R version 4.1.0 (The R Foundation for Statistical Computing, 2021) was used to create boxplots and to calculate linear models.

Assessment of differences in summary scores of individual behavioural addiction criteria between high-AB dogs and low-AB dogs

We calculated summary scores for the four addiction criteria for each subtest by summing up the points for each minute. Using Mann-Whitney U tests, we tested whether there was a difference in the addiction criteria Craving, Lack of self-control, Mood modification and Salience between dogs classified as high-AB dogs (AB-T score ≥ 44.2 points) and low-AB dogs (AB-T score < 44.2 points). Note that although components of Salience and craving were used to calculate the AB-T score, these are not identical to the 1 − 0 variables of Salience and craving here. While the addiction criteria were coded as 1/0 for each minute, the variables included in the AB-T score were more detailed, and individual elements potentially indicative of addictive-like behaviour were differentiated. Mood modification and Lack of self-control were not used in the designation of the AB-T score. See Sect. 1.6 and Table 3 for more details.

Assessment of differences in durations of toy-directed behaviours between high-AB dogs and low-AB dogs

We performed Mann-Whitney U tests to assess whether high-AB and low-AB dogs differed in quantitatively coded variables such as time engaging with the toy during different subtests, attempting to attain an unavailable toy, etc. (see Supplementary Tables 4 and 5).

Associations between questionnaire and behaviour test results and calculation of an Addictive-like Behaviour Questionnaire score (AB-Q score)

Linear models were used to assess associations between the addictive-like behaviour score and the 19 questionnaire questions targeting addictive-like behaviour. Model requirements were checked by visually assessing normality and homoscedasticity of the residuals. If applicable, the dependent variable was transformed.

Additionally, Mann-Whitney U tests were used to test whether the 19 questionnaire scores differed between high-AB and low-AB dogs. This was the case for fifteen questions; therefore, these were summed up to generate an Addictive-like Behaviour Questionnaire score (AB-Q score). Cohen’s R was used as a measure of effect size.

Both intra-rater reliability (available for 274 dogs, including dogs from the online survey) and inter-rater reliability (available for 24 dogs) of the AB-Q score were very good (see Supplementary Table 6).

Due to the exploratory nature of this study, no correction for multiple testing was performed (as recommended by⁷³.

Addictive-like behaviour test score (AB-T score)

The mid-point of the data range of the AB-T score was 44.2 (range 6.6–95). Therefore, dogs scoring 44.2 or higher were classified as showing a high tendency for addictive-like behaviour (high-AB dogs). This was the case for thirty-three of the 105 highly play-motivated dogs tested, with a mean score of 59.7 points and a median of 58.6. The mean AB-T score for low-AB dogs (< 44.2 points) was 23.1, and the median was 22.8. For descriptive statistics, see Supplementary Table 7.

Assessment of differences between high-AB dogs and low-AB dogs in summary scores of individual behavioural addiction criteria

Mann-Whitney U tests indicated that high-AB dogs scored significantly higher than low-AB dogs on craving (U = 217, p < 0.0001), salience (U = 208, p < 0.0001), and lack of self-control (U = 756.5, p = 0.002), but not mood modification (U = 1022, p = 0.157), in the behaviour test (see Supplementary Table 4, Figs. 2a-d). For descriptive statistics, see Supplementary Table 8.

Fig. 2

Behavioural addiction criteria and toy-directed behaviors: Multiple sub-figures (a–g) illustrate differences between high-AB and low-AB dogs across various metrics. (a) Craving summary score (b) Salience summary score (c) Lack of self-control summary score (d) Mood modification summary (e) Interaction with the box while the toy is enclosed – total interaction time (s) (f) Toy on the shelf– duration of looking at the toy (s).

Quantitatively coded variables

High-AB dogs interacted significantly longer with the box than low-AB dogs in the ‘toy in the box’ subtest (U = 675.5, p < 0.0001). They also spent more time looking at the toy on the shelf during the ‘toy on shelf’ subtest (U = 414.5, p < 0.0001) and the ‘social play without toys’ subtest (U = 942.5, p = 0.021), while focusing less on the owner in the latter (U = 819.5, p = 0.011) compared to low-AB dogs (Supplementary Table 5, Fig. 2e and f). However, time spent interacting with the toy while the owner and experimenter were out of the room did not differ significantly between high-AB and low-AB dogs (U = 994, p = 0.135; Supplementary Table 5). For descriptive statistics, see Supplementary Table 9.

Addictive-like behaviour questionnaire score (AB-Q score)

Linear models demonstrated significant associations between the AB-T score and 18 out of 19 individual questions (Table 6). However, according to Mann-Whitney U tests, only fifteen questions differed significantly between dogs classified as showing a high tendency for addictive-like behaviour in the behaviour test (AB-T score ≥ 44.2) and those that did not. These fifteen questions (Cohen’s R > 0.2 – see Table 6) were summed up into the Addictive-like Behaviour Questionnaire score (AB-Q score) (see Table 6).

This study represents the beginning of the exploration of addictive-like behaviour in domestic dogs. Convergent behavioural measures support the existence of an addictive-like behavioural phenotype in 33 of the 105 tested highly play-motivated dogs. Note that we specifically sought dogs exhibiting extreme behaviour; thus, this proportion is not a reflection of the general population. Perhaps not surprising, working breeds – many of which are known to have been artificially selected for high toy or predatory motivation^74,75,76 – were overrepresented in the sample.

As predicted, dogs classified as high-AB dogs based on the detailed AB-T score (Addictive-like Behaviour Test score) scored significantly higher than low-AB dogs on the individual criteria craving, salience, and lack of self-control in the behaviour test. Contrary to the prediction, mood modification (when given access to a toy) did not differ between high and low-AB dogs. In retrospect, however, this lack of difference between the two groups strengthens our argument that we were measuring a phenotype beyond mere enjoyment of play. Still, despite the significant differences between high- and low-AB dogs in the other investigated addiction criteria, Salience, Craving and Loss of Self Control, there was generally high variation between individuals.

In line with the predictions, high-AB dogs showed higher durations of focusing on and trying to access an inaccessible toy than low-AB dogs, often prioritising attempting to access the toy over eating or interacting with the owner. Thus, there was general agreement between the three alternative methods of coding the data (detailed behaviour score, addiction criteria, and quantitative coding), indicating internal consistency.

The external validity of the behaviour test was demonstrated by significant associations of the AB-T score with 18 out of 19 questions from the addictive-like behaviour questionnaire filled in by the dogs’ owners, intended to measure addictive-like behaviour in everyday life. Nonetheless, although significant, the effect sizes were relatively low, indicating that no single question would have predictive value for assessing a tendency for addictive-like behaviour in dogs.

In studies using animal models of substance addiction, one way to differentiate an addiction from drug use that occurs due to lack of choice is to present the subject with a choice between the addictive substance and other highly desirable stimuli. If an individual continues to take the drug at the expense of these other options (such as consumption of a food reward), this points to the possibility of addictive-like behaviour^77,78. Consistent with this, high-AB dogs showed a loss of interest in other relevant stimuli, focusing on the inaccessible toy and foregoing the opportunity to consume food or to engage with their owner. The latter is also reminiscent of behavioural addictions in humans, leading to a decline in social interactions⁷⁹.

The intense toy-seeking and loss of interest in other stimuli, despite the availability of food or social interaction – considered as indicators for salience and persistence – might resemble “hyperfocus,” a trait associated with ADHD and autism in humans^80,81. However, unlike typical hyperfocus, which often emerges in the absence of competing stimuli, dogs in our study were presented with alternative salient rewards (e.g., the toy was placed on a shelf while the owner actively invited the dog to engage in social play; in another subtest, food was available in a puzzle toy while the preferred toy was inaccessible in a closed container), and they still showed a preference for the inaccessible toy. Like dogs with ADHD, dogs in the current study with high AB-T scores in general exhibited high impulsivity (labelled as “loss of self-control”), and some individuals displayed heightened activity (which could be interpreted as the hyperactivity component of ADHD^64,65 in particular during the cool-down period. Thus, further research is needed to explore commonalities and differences between addictive-like behaviour and ADHD-like behaviour in dogs. While dogs with a high tendency for addictive-like behaviour might exhibit many characteristics of dogs with ADHD, the converse is not necessarily true – dogs might show ADHD-like behaviour without displaying any hyperfixation on toys.

Another characteristic of addicted individuals is that they are willing to pursue their addiction even if it has adverse consequences⁸². In the current study, “adversity” was elicited by the owner and the experimenter leaving the room in order to assess the effect of social isolation on the behavioural addiction criteria. Isolation in an unfamiliar place is well-established as a stressful experience for dogs^83,84,85,86. However, this subtest was not a good measure of addictive-like behaviour: Time spent interacting with the toy while the dog was alone did not differ significantly between high-AB and low-AB dogs. For welfare reasons, we decided against exposing the dogs to more severe stressors; however, it cannot be ruled out that this subtest was not “aversive” enough. The dog was left alone for only 30 seconds, and the subtest took place in the middle of the test when the dogs were already habituated to the test room. It is also possible that individual differences in subjects’ separation distress, independent of play motivation, affected the results. Additionally, there was no clear contingency between interacting with the toy and the ‘adverse’ outcome (owner leaving). Future studies could potentially enhance the design by providing the dog with an explicit choice, such as by placing the toy in a separate room, away from the owner and the experimenter. This could help determine whether the dog is willing to risk being alone in an unusual or new environment when it normally prefers the safety of being near its owner. Such a design would better reflect the conflict between competing motivations (social security vs. reward seeking) and could offer a more valid test of the criterion of persistence under adversity.

Still, the importance of continued efforts to engage in the behaviour despite adverse consequences was demonstrated in the questionnaire, where one of the highest associations with the AB-T score was found with the question, “My dog will continue to play with a ball/toy despite adverse consequences”. This suggests that some dogs may fulfil the criterion of continuing the addictive-like behaviour despite adverse consequences in real life, even if this could not be demonstrated in the behaviour test.

A critical factor in addiction is the propensity to attribute incentive salience to classically conditioned cues predicting rewards^87,88. In humans, cues associated with addictive behaviours, such as specific locations or objects, can induce craving and drug administration^88,89. In dogs, a toy such as a ball could represent such as a conditioned cue. It may achieve its value, for example, by the experience of chasing and catching. For many domestic dogs, balls or other toys possess incentive salience, according to the three criteria by Robinson and Berridge⁴⁹: they (1) “elicit approach” (i.e. they become “wanted” and act as “motivational magnets”); (2) “they can energise ongoing actions by eliciting cue-triggered wanting”; (3) “they can act as reinforcers in their own right, reinforcing the acquisition of a new instrumental response (measurable by conditioned reinforcement)” (cf⁴⁹, p. 3139].).

The perceived value of the toy was demonstrated in our study by many dogs having difficulty relinquishing the toy. It can be speculated that balls become ‘motivational magnets’ by being associated with species-typical predatory behaviour (cf⁴⁸. The high salience of the toy was especially apparent in subtests where dogs were foregoing available alternatives such as freely available food or social play with the owner, at the expense of trying to regain their inaccessible toy.

In both rodents and humans^90,91, individuals with a higher tendency to attribute incentive salience to classically conditioned cues predicting rewards (sign trackers) are more vulnerable to addiction than goal trackers, who focus primarily on the (location of the) reward itself^88,92, see the meta-analysis by⁹³. Tendency to sign-track vs. goal-track is associated with the risk of addiction and is also related to variations in the dopaminergic system and stress physiology⁸⁸.

While it was not explicitly measured in the current study, in dogs, a tendency to sign track might be advantageous in a training context – i.e., maintaining motivation would be easier in dogs that are not only sensitive to rewards but also attribute value to the cues predicting these rewards, even if not always followed by a primary reinforcer. Sensitivity to reward – and propensity to attribute incentive salience to reward-predictive cues – would thus be highly relevant traits in relation to trainability and might be selected for especially in working dog breeds.

Several publications state the importance of ‘obsessive’ play motivation for working dog success^{42,94,95,96,97,98,99}. Dogs with extreme toy motivation are believed to show better focus, reduced distractibility and greater trainability^97,99. However, if such motivation becomes addictive-like, it needs to be questioned whether the well-being of these dogs may be compromised. If dogs prioritise toy interactions over other essential aspects of their daily lives this may have maladaptive effects, as is the case in humans with behavioural addictions^100,101. Certainly, adverse health consequences may arise from repetitive ball chasing, like straining ligaments and joints⁵⁶. Moreover, welfare would be compromised when dogs experience a high level of frustration when access to their reward is prevented (cf¹⁰².

Anecdotally, when play motivation becomes excessive, irritability, high arousal levels, and frustration may negatively affect dogs’ trainability and work¹⁰³. Indeed, as stated by Mathews⁹⁶, the high ‘drive’ of search dogs often makes them unsuitable as family pets, which is also supported by owner reports that pet dogs with extreme motivation for toys are often problematic to control¹⁰².

Thus, it needs to be questioned when play becomes maladaptive. Do high-AB dogs still ‘like’ to play, or have they progressed to primarily ‘wanting’ and the compulsive need to continue¹⁰⁴?. Similar to the escalating engagement seen in human behavioural addictions¹⁰⁵, some dogs would repeatedly spin, jump, focus or bark towards the unavailable toy on the shelf for the duration of the subtest. Two dogs even managed to destroy the box enclosing their favourite toy. These behaviours might be likened to the repetitive actions observed in individuals with behavioural addictions⁴¹. Nonetheless, such behaviours may also occur in other behavioural phenotypes such as canine compulsive disorder or autism spectrum-like behaviours¹⁰⁶. Further research is needed to elucidate commonalities and differences between such phenotypes in dogs.

Behavioural addictions in humans often involve emotional dependency on specific activities¹⁰⁷. Whether dogs similarly seek comfort, stimulation, or stress relief through persistent engagement with the toy could not be determined in the context of the behaviour test. In the questionnaire, “Is attached to the favourite toy” was the only question not significantly associated with the AB-T score. Thus, further research is required to determine whether dogs develop an emotional dependency on their toys (as described anecdotally).

To better understand the origin and possible functional underpinnings of excessive toy-directed behaviour in dogs, future research should examine whether similar patterns of excessive object play occur in non-domesticated canids. While data are limited, recent studies have shown that both hand-reared and wild wolf pups engage in object play¹⁰⁸. For instance, wolf pups have been observed developing a preference for toys and spending increasing amounts of time with them over time¹⁰⁸. Hand-reared wolf pups will even retrieve objects to humans¹⁰⁹. In the wild, wolves have also been seen interacting with human-made objects¹¹⁰. These findings suggest that object play is not unique to dogs but rather could represent a broader trait shared by canids. Comparative studies are needed to assess how common and functionally relevant such behaviours are in wolves, which would help clarify the biological basis of the addictive-like behaviours observed in some dogs.

Being the first of its kind, this study has its limitations. As no gold standard exists, the study is exploratory, and our categorisation of dogs into high and low addictive-like behaviour groups, determined by a data range split, was somewhat arbitrary. Nonetheless, the assignment of high- and low-AB categories corresponded well to the first author’s personal assessment of addictive-like tendencies in the participant dogs.

In interpreting the questionnaire results, it is important to acknowledge the potential biases associated with using owner-reported questionnaires. Owners may unintentionally project their perceptions or expectations onto their dogs’ behaviours, potentially leading to discrepancies between reported and observed behaviours in behavioural test. This is particularly relevant in cases where owners have multiple dogs, as they are likely to compare their pets to one another, influencing their assessment, such as by underestimating or overestimating certain behaviours. For instance, an owner with a highly active dog may rate their less active dog as overly calm. Integrating owner reports and objective testing allows for a more comprehensive and accurate canine behaviour evaluation.

To conclude, there appear to be parallels between excessive toy motivation in dogs and behavioural addictions in humans. Interestingly, also in humans, the first officially recognised behavioural addictions (gambling and internet gaming) originate in play^28,29,30,111. Generally, play is an activity that induces a pleasurable emotional state⁶. In humans, much evidence suggests that video games can affect people’s lives positively. They make players feel better about themselves, help raise their self-esteem and assist people in dealing with everyday stress¹¹¹. Some people are excessive gamers, but only a minority would be classified as addicts^111,112,113. Similarly, many dogs may greatly enjoy toy play without developing harmful compulsions (cf. in humans^28,29,30,111).

Despite the observed parallels between high-AB dogs and humans affected by behavioural addictions, we refrain from conclusively characterising high-AB dogs as exhibiting addictive behaviour, given the absence of established benchmarks or standardised criteria. It is important to be cautious when pathologising behaviour, especially given that even in humans, addictive behaviours are still difficult to define and measure¹¹⁴. To further understand possible parallels in the processes underlying behavioural addictions in humans and excessive toy motivation in dogs, subsequent research endeavours should seek to correlate individual differences in addictive-like behaviour in dogs with characteristics associated with addictive behaviours in humans, such as high impulsivity, impaired reversal learning, heightened perseveration, and delayed extinction of previously rewarded responses^115,116.

References

Authors and Affiliations

1. Division of Animal Welfare, Vetsuisse Faculty, Veterinary Public Health Institute, University of Bern, Bern, 3012, Switzerland

   Alja Mazzini, Katja Senn & Stefanie Riemer

2. Division of Behavioural Ecology, Institute of Ecology and Evolution, University of Bern, Hinterkappelen, 3032, Switzerland

   Federico Monteleone

3. Messerli Research Institute, Department of Interdisciplinary Life Sciences, Vetmeduni Vienna, Vienna, 1210, Austria

   Stefanie Riemer

Contributions

S.R. and A.M. contributed to the conception and design of the research. A.M. drafted the original manuscript, and S.R. provided revisions. A.M. carried out the experiments. K.S., F.M. and A.M. coded the videos. A.M. and S.R. interpreted the data. All authors have read and approved the final version of the manuscript.

Corresponding author

Correspondence to Stefanie Riemer.

Competing interests

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions