"Science Friday!": Research Paper Review

Anyone up for a little dog science? Dive on in with me!

Research Paper Review:

Echterling-Savage, K., DiGennaro Reed, F.D., Miller, L.K., & Savage, S. (2015). Effects of Caregiver-Implemented Aggression Reduction Procedure on Problem Behavior of Dogs. Journal of Applied Animal Welfare Science, 18, 181-197.


Many studies in applied animal behavior rely on correlational designs or survey research, rather than experimental manipulations involving direct observation and systematic measurement of pet behavior. Even fewer are those studies that experimentally manipulate owners’ implementation of behavioral interventions while simultaneously observing and evaluating pets’ responses to those interventions. So this 2015 paper by Echterling-Savage et al. is quite refreshing, as it does all of these things.

In this paper, published in The Journal of Applied Animal Welfare Science, researchers open by discussing the common recommendation given by pet behavior professionals to pair a pleasant stimulus (typically food) with a stimulus that elicits aggression (e.g., another dog) as a way to reduce aggression through classical conditioning. In theory, the process works by changing the dog’s adverse physiological response to another dog (heightened arousal, tension, and overt responses including barking, growling, and lunging) to one of relaxation, low arousal, and food-focused anticipatory behavior. Although this procedure may be utilized commonly in behavior therapy with dogs, empirical support for its efficacy is lacking.

The authors set out to test the ability of owners to carry out an aggression reduction procedure with their own dogs, and to measure its effect on the dogs’ behavior. Four dog-owner dyads were utilized. All dogs had a history of aggression toward other dogs in the form of barking and lunging on leash.

There were two phases to the experiment, each involving baseline, intervention, and follow-up data collection. Dogs were exposed to a multicomponent intervention in a time-lagged fashion across subjects (a concurrent multiple baseline design).

Before beginning the experiment proper, a baseline measurement of dog aggression on leash was taken at a local park. Owners were instructed to “do what they normally do” and dog aggression (defined as any vocalization and/or lunging directed toward a passing stimulus dog) was measured using a partial interval recording system. Researchers also measured precursor behavior, which they defined as those responses directed toward a passing stimulus dog that immediately preceded aggression, the topography of which varied across dogs (e.g., prolonged stare, ears pointed toward stimulus dog).

Preintervention training then was provided to all dyads. This involved pairing “Yes” with an edible (squeeze cheese, which allowed for fast delivery and had been determined to function as a reinforcer for all dogs in previous testing). Dogs were fitted with a Gentle Leader and provided with reinforcement inside the house for wearing the head collar for short periods. They later wore the Gentle Leader throughout all experimental sessions. Owners were also instructed to deliver daily meal content contingent upon compliance with simple commands (e.g., sit) throughout the day and this protocol continued through all phases of the experiment.  

During the intervention phase, researchers taught owners how to implement an aggression reduction procedure that required the owner to deliver the bridge word and edible within 1 second of the dog engaging in precursor behavior (before aggression) toward a passing stimulus dog. If the dog continued to engage in precursor behavior without aggression, the owner was instructed to continue to provide the bridge word and edible every second. Owners’ implementation of the procedure and their dogs’ behavior was recorded over multiple 60-min sessions. Trials were dependent upon the passing of another dog because these observations took place in a naturalistic setting. Subjects were exposed to between 3 and 10 passing dogs/trials per session.

The second phase of the experiment involved generalization programming. Here, researchers required owners to carry out the aggression reduction procedure in more challenging circumstances that included walking their dogs (rather than remaining stationary as in the intervention phase), talking to another person while remaining stationary, and going in and out of doors or cars. Such situations simulate real-life scenarios in which dog-dog aggression is likely to occur and were staged at a dog training facility.

Baseline levels of aggression were high and variable across stimulus dogs for most subjects. Aggression was reduced substantially when intervention was introduced, and was at 0% by the end of intervention. It remained at zero or near-zero levels during follow-up and generalization programming for all dogs. These differences were statistically significant. Levels of precursor behavior remained high from baseline to intervention. This latter finding is not surprising, given that precursor behavior (looking at the other dog) was required for reinforcement (bridge + edible) to be delivered.

At the conclusion of the study, researchers obtained measures of social validity (i.e., the acceptability of goals, procedures, and training effects) as assessed by surveys of caregivers, dog trainers, rescue group personnel, veterinary care professionals, and researchers. Social validity scores were high. Interobserver agreement, fidelity of caregiver implementation of the protocol, and fidelity of experimenter delivery of caregiver training were also high.

The authors discuss limitations of their findings, including their definition of precursor behavior as any behavior oriented toward the other dog. Clearly, this can range in intensity from casual glances to hard stare with stiff, chest-forward stance. Individual stimulus dogs likely produced different levels of intensity in the precursor behavior as well as the aggressive behavior of dogs, and researchers did not tease apart the independent effects of various stimulus dogs. It is also possible that aggression was reduced so dramatically early on because food was delivered at such a high rate that food consumption interfered with the opportunity for aggression. Yet the authors mention that in generalization programming, the rate of food delivery was reduced over time and aggression remained at zero or near zero levels. Finally, a component analysis was not conducted to tease apart other elements of the intervention that were present throughout, including the dogs’ wearing of the Gentle Leader and compliance-dependent food delivery in daily life at home.


As stated above, this study is unique in applied animal behavior research in its use of experimental manipulation (as opposed to mere correlational analysis), its direct observation of both owner and dog behavior, its operational definitions of independent and dependent variables, and its systematic measurement of a specific behavioral intervention used to address canine aggression in a naturalistic setting.

It also draws attention to the essential but often neglected element of generalization programming in behavior therapy for dogs. Pet behavior professionals often provide instruction and demonstration of strategies for specific circumstances in which problem behavior occurs. Yet much more instruction and coaching is typically required to produce behavior change that is sustainable and reliable across settings, triggers, and handlers. For the behavioral gains of the dog AND the owner to be persistently maintained across time, both must practice implementation of behavior therapy and training strategies in multiple relevant contexts and conditions. To incorporate direct observation during intervention, follow-up, and generalization programming (rather than, for example, relying on owner report) is a distinct and important advantage of this study.

Echterling-Savage and her colleagues demonstrated that caregivers can successfully implement this particular procedure for reducing canine aggression on leash exhibited toward other dogs (fidelity of caregiver implementation was 82.9%). It also demonstrates that this particular procedure reduces aggression rapidly and is sustainable across many trials, across time, across trigger dogs, and across contexts. A food pairing procedure of one sort or another is commonly recommended in behavior therapy for dogs and this research provides empirical support for its ongoing use.