Identifying Behavioral Precursors To Play-Induced Aggression In Pet Dogs At Dog Parks

Principal investigator:
Lindsay R. Mehrkam, M.Sc.
Department of Psychology
PO Box 11220
Gainesville, FL 32611

Play is not just fun and games. Play is essential for a dog's welfare, but play interactions can easily lead to aggression. Accurately identifying the conditions that lead to welfare-positive play and those that predict aggressive interactions is likely to have a significant impact on applied animal behavior. Unfortunately, play has not received much empirical attention relative to other forms of dog behavior, limiting our knowledge of how and when to intervene in a play bout between two or more dogs.  Different owners – and even different trainers – may have different conceptions of appropriate and inappropriate play. Without the knowledge of behavioral precursors of aggression that reliably occur during play, the safety of both dogs and their owners may be jeopardized. The aims of this study are therefore (1) to determine the likelihood that certain play behaviors will escalate to aggression in a naturalistic setting, (2) to measure the relative likelihood of handlers responding to behavioral precursors of play-induced aggression, and finally (3) to gather data on breed-specific, size, and age-related factors in play. The results of this study will lead directly to a systematic evaluation of behaviors that will predict aggression from social play that can be utilized by owners, trainers, and behavior consultants. We are requesting funding in the amount of $1,500.00 to carry out the project in its entirety.

2. Complete proposal containing full detail on the proposed project. 4000 words maximum in a Microsoft Word or MS Word-compatible file. Please do NOT send PDFs. Proposals exceeding the word limit pages will be disqualified.
Proposals should address each of the items below under separate numbered headings.

3. Research Question

3.1 What is the research area addressed by the proposal and why is it important?

The research area addressed by our proposal is applied animal behavior, specifically, an understanding of social play and aggression prevention in domestic dogs in a naturalistic setting.  We are interested in a structural analysis of play, in terms of identifying behavioral precursors that would indicate whether play is likely to lead to aggression.

This research area is important to educate owners, trainers, and behavior consultants to differentiate between play and aggression in dogs, and to recognize a situation in which play may be escalating to aggression and require intervention. Our proposed project, if funded, would result in data that will have implications not only in applied animal behavior and animal welfare, but also dog training and behavior modification, human-animal interactions, and public health and safety.

3.2. Give a brief summary of the current state of the art in the field including references where appropriate and how the proposed project will address a specific research question.

According to the Association of Pet Dog Trainers, dog parks are an excellent resource for socializing pet dogs to other dogs and people, and provide off-leash exercise and educational advantages for both owners and the community as a whole (APDT 2012).   However, dog parks are often cautioned as potentially dangerous places for both dogs and people due to the potential of aggressive interactions between multiple dogs (APDT, 2012). Dogfights in dog parks are extremely common (Dunbar, 2012), and such incidents result in public health and safety concerns, animal welfare issues, and economic costs for the community.

The prevalence of dog fights in dog parks may be due to a failure of most owners to distinguish between play and aggression in dog social interactions, especially considering play may escalate to aggression. Tami and Gallagher (2009) found that individuals – including trainers and dog owners – often failed to differentiate between play and aggression in video recordings of domestic dogs; piloerection, biting, herding, standing over, and showing teeth in the play video were mainly interpreted as aggressive cues. Likewise, barking, bouncing, ears erected and forward, and tail-wagging were often incorrectly interpreted as playful cues. The mislabeling of these behaviors suggests that exposure to companion dogs without any theoretical knowledge of behavior is not in itself sufficient to foster the emergence of a reliable understanding of dog behavior. This appears to be especially true for aggression and play fighting, which were the behaviors most easily misunderstood (Tami & Gallagher, 2009). In addition, owners' inclination to intervene during play interactions are likely influenced by other owners (e.g., not wanting to offend or upset other owners) and false perceptions of the function of certain behaviors that appear playful or "natural" (e.g., excessive mounting, barking). Therefore, improving the ability of owners, trainers and behavioral consultants to accurately recognize behavioral precursors and furthermore know the likelihood that a particular behavior will lead to play or aggression, is important in the prevention of dogfights.

Only recently has play behavior between dogs been systematically researched. Most of this research has been in naturalistic settings. Bekoff (1995) attempted to identify social communicative play signals - such as the play bow - in social play between dogs. Horowitz (2009) reported that interspecific attention between dogs in a dog park greatly influenced the dogs' likelihood to engage in play. Bauer (2010) studied dogs with established dominance relationships and found that role reversal and self-handicapping were most strongly correlated with the maintenance of play bouts. In contrast, there are no scientific studies that have identified what behaviors during play may be precursors to aggression. While knowing what play behaviors are involved in cooperation during play is worthwhile, this contributes to an understanding of only one possible outcome of play  - more play. It is also of much practical interest as to what terminates play due to the onset of aggression.  In other words, what is needed is the objective identification of inappropriate and appropriate play.

The likelihood that certain behaviors will lead to aggression during play can be assessed by first recording what behaviors were observed during play bouts; then measuring the frequency of aggression immediately following these bouts; and subsequently calculating the conditional probabilities that these behaviors lead to aggression. Conditional probability analysis has been used successfully to identify behavioral precursors to aggression (Borrerro & Borrerro 2008) and other forms of severe problem behavior (Herscovitch et al. 2009) in human applied settings. In addition, conditional probabilities have been used to assess canid behavior, for example, predatory behavior in domestic dogs  (Coppinger et al., 2010) and social communicative play signals (Bekoff, 1995).

If a comprehensive set of behavioral precursors to play-induced aggression can be identified, this would greatly empower dog owners to predict aggressive interactions resulting from play. These findings would be useful not only to dog owners, trainers, and behavioral consultants, but to any member of a community that has dog parks or other off-leash settings for dogs. Conditional probability analysis is a statistically powerful method for detecting a valid, legitimate contingency between behavior (in this case, discrete play behavior) and its consequences (aggression or non-aggression), and is superior to purely observational or correlational methods alone (Watson, 1997). In this way, conditional probabilities are a relatively simple statistical procedure that can produce a sophisticated, objective, and informative analysis of play and aggression while still permitting study in a naturalistic, rather than a relatively artificial experimental, setting.

Given the applied implications and the methods available, the data obtained from this project  will be able to 1) identify behavioral precursors to play-induced aggression 2) assess the likelihood that owners respond to these precursors and 3) determine whether these precursors correlate with other relevant variables, including the dogs' size, gender, neuter status, and breed. We will use conditional probability analyses to address Aims 1 and 2 and descriptive and statistical analyses to address Aim 3.

Summary of Methodology. Research assistants will be recruited from undergraduate students at the University of Florida. Research assistants will be trained to use an ethogram that includes a comprehensive list of behaviors that may potentially be behavioral precursors to play and aggression (collected from empirical studies of aggressive behaviors and approximately 100 hours of preliminary observations) and their definitions (compiled from peer-reviewed scientific papers on play, as well as extensive pilot observations). Ten research assistants will be responsible for conducting focal observations of dogs in five dog parks in Gainesville, FL, daily and video recording these observations.  Five to 10 research assistants will be appointed to code every occurrence of a target behaviors and precursors from video. Reliability for live coding and video coding will be assessed throughout the study. 

After recording the frequencies of behaviors observed before, during, and after all social play occurrences, conditional probability analyses will be conducted (Borrero & Borrero, 2008; Herscovitch et al., 2009). To calculate the probability of aggression given a particular precursor behavior (behavior during social play bouts), the number of occurrences of the precursor response that were followed by an occurrence of the target response within 10 seconds (s) (Borrerro & Borrerro, 2008)  would be divided by the total number of occurrences of that precursor response. To calculate the probability of precursor behavior given target behavior, the number of occurrences of the target behavior that occurred within 10 s following a potential precursor would be divided by the total number of occurrences of the target behavior. To calculate the unconditional probability of the target behavior, the number of 10-s intervals that contained the target response would be divided by the total number of 10-s intervals in the assessment. To calculate the unconditional probability of the precursor response, the number of 10-s intervals that contained the precursor response would be divided by the total number of 10-s intervals in the assessment.

In addition, these conditional probability methods will be used to determine the likelihood that owners respond to these precursors. This will also allow us to identify what behaviors owners are most likely to attend to during social play, and if these behaviors are precursors to aggression.

The main training period for research assistants will begin at the time the funds are received and last approximately four weeks. Data collection will begin after Daylight Savings Time starts on March 10, 2013 and end on November 2, 2013. This will be to control for the time of day of observations (conducted between 4:30 p.m. and 8:00 p.m.). Conditional probabilities will be calculated by the principal investigator using the data obtained from video coding. To ensure the safety of the research assistants and the integrity of the data collected, assistants will be instructed not to bring their own dogs to the dog park during observation times, and not to intervene in any social behavior between two or more dogs. A notice will be posted at dog parks informing owners that they will be videotaped during their interactions with their dogs, and that they should inform the research assistant if they did not want themselves or their dog(s) videotaped. The notice will also ask that owners interact with their dogs as they normally would.

3.3. What is the potential contribution to the field of the project if successful?

This project has the potential to contribute immensely to both the theoretical and applied aspects of understanding social play in dogs.

Theoretically, play has received little empirical attention. To further complicate matters, the proximate functions (i.e., immediate benefits) of engaging in social play are not known. The results of this study would contribute to a more comprehensive and scientific, rather than anecdotal, understanding of the structure of play in domestic dogs, as well as the functions of play and aggression.

These theoretical gaps have also restricted our applied knowledge of play in pet dogs. The results of our study would help to identify not only what behaviors are important to the cooperation or maintenance of play, but also behaviors during play which may lead to an aggressive interaction. This has important implications for the field of applied animal behavior as a whole. Because conditional probabilities are quantitative, numerical values, owners, dog trainers, and behavior consultants can easily compare the relative likelihood of each behavior leading to aggression. For example, raised hackles during play may only lead to aggression 40% of the time it is observed, whereas excessive mounting during play may lead to aggression 85% of the time it is observed. Thus, a trainer or behavior consultant may recommend that raised hackles warrant owners' attending closely to their dogs, but not necessarily intervention, whereas excessive mounting, because of its higher conditional probability of leading to aggression during play, may warrant immediate intervention by the owner. In this way, conditional probabilities are a relatively simple statistical procedure that can produce a sophisticated and informative analysis of play and aggression while still permitting study in a naturalistic setting. This gives dog owners, trainers and consultants more direct, objective measure than observations alone.

By determining each behavior's respective likelihood of leading to aggression, the topography of the behavior (i.e., how the behavior appears) will be linked to its function (i.e., why it occurs). This has critical implications for animal welfare because behaviors with the same topography do not always share the same function. In addition, the outcomes of our project could greatly empower dog owners and trainers, by providing empirical data on what behaviors will most likely lead to aggression during play. This would make dog parks safer places to socialize dogs, given that owners could be made more knowledgeable about how to distinguish between appropriate and inappropriate play behaviors, as well as between play and aggression. Finally, although our project will be conducted in various dog parks in Gainesville, Florida, these findings would have generality for other off-leash settings, and would likely be useful for any setting in which dogs are permitted to play, such as households, dog daycare centers, and boarding facilities.

4. Expected outcomes

4.1.  What tangible assets, if any, will be created or produced as a result of the proposed project?

The results of this project will be reliable, non-anecdotal information on the probability that certain behaviors during play may lead to aggression. Trainers and behavior consultants would be able to provide a comprehensive list of validated behavioral precursors that are operationally and objectively defined to dog owners. This means that dog owners will be able to more easily recognize appropriate and inappropriate play behaviors, and predict the occurrence of aggression escalating from social play.

4.2. How will the results of this project be disseminated?

The results of this study will be disseminated via presentation at professional conferences (e.g. Association of Pet Dog Trainers, Animal Behavior Management Alliance, Animal Behavior Society) and prepared in a manuscript for publication in a peer-reviewed journal of appropriate scope (e.g., Applied Animal Behaviour Science, Journal of Veterinary Behavior: Clinical Research and Applications). In addition, we would be very willing to submit a piece to The APDT Chronicle of the Dog magazine describing our aims, methods, findings, and applied implications for readers - or to be interviewed about the study by writers from the Chronicle or other media targeting the dog-training and -owning public. If funds are awarded, APDT will be acknowledged in the dissemination of all data obtained for this project.

4.3. How can companion dog trainers and behavior consultants benefit from the results?

Owners often seek advice from companion dog trainers and behavior consultants about how to safely monitor their dogs' play in an off-leash setting. Providing trainers and behavior consultants with information about specific, observable behavioral precursors that have been demonstrated statistically to predict aggression from play would allow them to make recommendations based on scientific findings to their clients. As previously stated, trainers and behavior consultants would also able to provide a comprehensive list of behavioral precursors that are operationally defined. This would allow all dog owners – whether novice or experienced – to be able to easily recognize these behavioral precursors. In this way, these findings could empower dog owners to predict an onset of aggression and potentially prevent dogfights and attacks in their community.

Another important benefit of this study for dog trainers and behavioral consultants is that, once the relevant behavioral precursors and other variables have been identified, trainers can give more informed advice to their consumers, and target the relevant behaviors for modification. This could potentially lead to many future directions for trainers to develop assessment and treatment plans for targeting specific behavioral precursors and regulating play. This ultimately improves trainers' control over a behavior that is so poorly understood both structurally and functionally.

Finally, trainers and consultants will gain information on other factors influencing play structure, including breed-specific play styles, as well as the influences of owners on their dogs' behavior in the shaping, maintenance, and generalization of appropriate and cooperative social play in dogs. Clients who are seeking help with dogs that play too rough or inappropriately can be educated by trainers and consultants as to what behaviors require behavior modification and which play behaviors are appropriate or permissible and contribute to healthy social play between pet dogs. Ultimately, the results of this project will allow trainers and consultants to contribute in an even broader sense to public safety and making dog parks and other naturalistic play settings manageable places to apply their skills.

5. Schedule.
5.1. When is the project to be started and completed? What milestones will be used to measure progress of the project and when will they be completed?

We anticipate the proposed project to take approximately one year to complete. If funded, the project will begin in February 2013 and end in January 2014. Specific milestones are as follows:

February 2013: Begin training of 15-20 research assistants for data collection, video coding and data analysis. Research assistants will be recruited from undergraduate psychology classes and current volunteer databases. Purchase of necessary equipment.

March 2013: Complete training of research assistants. Data collection period begins. Video coding begins (conducted simultaneously with data collection).

November 2013: Complete data collection period. Finalize video coding.

December 2013: Complete video coding. Conduct data analysis using conditional probabilities, descriptive, and inferential statistics. Begin writing up results for publication.

January 2014: Complete writing up of manuscript and submit to peer-reviewed journal of appropriate scope. Review process may be expected to take 2-6 months.

6. Use of Funds: Provide a budget ($US) describing how the award will be used, including purchases of supplies, salaries and other costs. The budget should be presented as a table with the total budget request indicated. Indirect costs or overhead will not be awarded.

Item                                                                 Quantity (Itemized Cost)                           Total Cost

Kodak Playsport zx5 HD                                     4     ($70.00 each)*                                  $280.00
Waterproof Pocket Video Camera
SanDisk 16GB SDHC Memory Cards                4     ($17.95 each)**                                   $71.80
Extreme Class

WD My Passport 1 TB External                           1    ($79.99 each)*                                   $79.99
Hard Drive Storage USB 3.0 Blue                                                                 

One-month stipend for doctoral student/
applied animal behaviorist to train RAs                                                                             $1,068.21
                                                                                                            Total Expenses:    $1500.00


7. Evaluation: How will the results of this project be evaluated?

The results of this project will be evaluated with respect to our three original aims: 1) To determine the likelihood that certain play behaviors will escalate to aggression in a naturalistic setting, (2) to measure the relative likelihood of handlers responding to behavioral precursors of play-induced aggression, and finally (3) to gather data on breed-specific, size, and age-related factors in play.

First, the results of conditional probability analyses will allow us determine which behaviors with the highest conditional are most likely to lead to aggression. This will support our hypothesis that such precursors do exist, and can be used to effectively manage social play between dogs in an off-leash setting.

In addition, we will evaluate play bouts in which an owner intervened by calculating the proportion of play bouts in which an owner intervened and comparing this with the respective conditional probability for each behavior. This will allow us to evaluate the likelihood that owners respond to behaviors that reliably predict aggression resulting from play.          
Finally, we will evaluate the relative frequencies of each potential precursor for specific breed or breed-types of dogs involved in play bouts. This will provide information on whether there are specific demographic factors in dogs that correspond to the precursors of play-induced aggression, as well as its prevalence. We will also use our data to evaluate the prevalence of total play-induced aggressive interactions in the five local dog parks we sampled our data from. This will be evaluated by calculating the total number of play bouts we recorded during the study, and the proportion of those bouts that resulted in aggression.

8. Qualifications of Principal Investigator: Include a brief description of any relevant prior research, teaching, publication or other professional experience.

I am a Ph.D. candidate in Psychology/Behavior Analysis at the University of Florida. I currently hold a Bachelor of Arts in Animal Behavior from Franklin & Marshall College and a Master's of Science in Psychology/Behavior Analysis from the University of Florida. I have nearly 10 years of research experience in the field of applied animal behavior science, especially with respect to canids. Prior to attending graduate school, I conducted ethological research on captive wolves and wolf-dog hybrids and held a year-long internship assisting with applied animal behavior research at Disney's Animal Kingdom under the direction of senior scientists. I currently conduct research in the fields of both basic and applied canine behavior under the supervision of Dr. Clive D.L. Wynne in the Canine Cognition and Behavior Lab. My dissertation research currently investigates proximate mechanisms of social play in domestic dogs, the role of human interaction on canine social behavior, and the evaluation of play as a welfare indicator. I carry out several projects on behavioral welfare in captive canid populations in Florida as wel, including assessing the behavioral welfare of shelter dogs, human interaction as environmental enrichment (Mehrkam, Wynne & Verdi, in review, Journal of Applied Animal Welfare Science), and behavior modification for fearful and aggressive wolves and wolf-dog hybrids in private sanctuaries. I also conduct behavioral research at the AZA-accredited Santa Fe College Teaching Zoo, where I lead a series of projects with Dr. Nicole Dorey evaluating the use of preference assessments in enrichment evaluations for a wide range of species (Mehrkam & Dorey, in review, Zoo Biology). In addition to research, I have teaching assistant experience at both the undergraduate and graduate levels, with the majority of my experience being in psychology and animal behavior courses. I have also served as an adjunct instructor and lead several teaching seminars at the Santa Fe College Teaching Zoo and online seminars through E-training for Dogs. I also serve as an ad hoc reviewer for the journal Behavioural Processes.

Association of Pet Dog Trainers.  2012. "Dog Park Pros and Cons". <>. Accessed 23 December 2012.
Bauer, E.B. 2007. Cooperation and competition during dyadic play in domestic dogs, Canis familiaris. Animal Behaviour 73: 489–499

Bekoff, M. 1995. Play signals as punctuation: the structure of social play in canids. Behaviour 132(5–6):419–429.

Borrero C.S.W, Borrero J.C. 2008. Descriptive and experimental analyses of potential precursors to problem behavior. Journal of Applied Behavior Analysis. 41:83–96.

Dunbar, I. "Ask an Expert- Dog Park Dog Fight." In Modern Dog Magazine. <> Accessed 23 December 2012.
Herscovitz, B., Roscoe, E.M., Libby, M.E., Bourret, J.C., Ahearn, W.H. 2009. A procedure for identifying precursors to problem behavior. Journal of Applied Behavior Analysis. 42(3): 697-702.

Horowitz, A.C. 2009. Domestic dogs (Canis familiaris) use visual attention cues when play signaling. Journal of Veterinary Behavior: Clinical Applications and Research, 4(2) 53-54.

Komuro et al. 2009. Behaviors of domestic dogs and their keepers in an off-leash dog park. Japanese Journal of Animal Psychology. 59(1): 47-55.

Tami, G. & Gallagher, A. 2009. Description of the behaviour of the domestic dog (Canis familiaris) by experienced and inexperienced people. 120(3-4): 159-169.

Ward, C., Bauer, E.B., Smuts. B.B.. 2008. Partner preferences and asymmetries in social play among domestic dog, Canis lupus familiaris, littermates. Animal Behaviour 76:1187–1199.

Watson, J.A. 1997. Contingency and its two indices within conditional probability analysis. The Behavior Analyst 20(2): 129-140.