To best understand this article in the context of behavior evaluations, see National Canine Research Council’s complete analysis here.

Article citation:

Willen, R. M., Schiml, P. A., & Hennessy, M. B. (2019). Enrichment centered on human interaction moderates fear-induced aggression and increases positive expectancy in fearful shelter dogs. Applied Animal Behaviour Science, 217, 57-62. doi:

National Canine Research Council’s Summary and Analysis:


One purpose of this study was to demonstrate that vulnerable subpopulations of dogs living in shelters could benefit from targeted enrichment treatment. This paper is included because it critically evaluates “fear-induced aggression” in shelter dogs and demonstrates that behavior elicited during an assessment can be easily modified by providing human attention prior to testing. The authors conclude that positive human interaction in the days prior to testing allows fearful dogs to pass a behavior evaluation they would have failed without the intervention. In addition, the findings raise questions regarding predictive ability and external validity of behavior evaluations in animal shelters, given the dramatic effect of the simple, short term interventions of the experiment.

Willen, Schiml, and Hennessy (2019) explain that “fear-induced aggression” is a situation-specific piece of an animal’s behavioral repertoire that occurs when the animal cannot escape the perceived threat. In other words, scared dogs in a shelter may display “aggressive” behaviors such as growling that they would not display in other contexts. This is important because such behaviors may be grounds for euthanasia in a shelter setting as was the case in the shelter whose population made up this study sample.

The authors’ previous research (Willen et al., 2017) showed that 15 minutes of positive human interaction was sufficient to temporarily reduce plasma cortisol levels for dogs in shelters. Based on this physiological finding, and in conjunction with anecdotal observation, Willen et al. (2019) hypothesized that brief human interaction during an out-of-kennel enrichment program would increase the number of dogs who would pass the SAFER (The Safety Assessment for Evaluating Rehoming) evaluation. Moreover, they hypothesized that these results would be amplified for “fear-induced aggressive” dogs.


This study was comprised of three experiments: SAFER outcome for “fearful” dogs with and without enrichment; SAFER outcome for “fearful” and “non-fearful” dogs with and without enrichment; Cognitive bias for “fearful” and “non-fearful” dogs with and without enrichment.



“Fearful-aggressive” Dogs

“Non fearful-aggressive” Dogs






Experiment 1





Experiment 2





Experiment 3*





*Same subjects were used in Experiments 2 & 3. Four “fearful-control” dogs were euthanized prior to Experiment 3, and 1 non-fearful control dog was adopted prior to Experiment 3.

“Fearful dogs” were so designated by shelter staff recommendation and observation by a Certified Applied Animal Behaviorist. Behaviors deemed “fearful” included cowering in the back of kennel, shivering, continuous panting, piloerection, and avoiding eye contact. If a dog engaged in these fearful behaviors but showed no signs of “aggression,” then they were excluded from the study. “Fearful aggression” was determined by the dogs’ behavior on approach. Approaching in a friendly manner and accepting treats, for example, were not categorized as “aggression” and thus dogs expressing these behaviors were excluded. In contrast, “fearful aggressive” behaviors included some combination of pinned ears, tucked tail, crouched body, diverted gaze, dilated eyes, and growling. On the other hand, dogs who did not express the behaviors categorized as fearful listed above, but whose warning behaviors (such as growling) were accompanied by body language interpreted as confident or “not fearful” (e.g., ears down, tail up, direct eye contact, forward movement) were excluded, along with those with a history of attempting to bite a person . In sum, dogs had to demonstrate both “fear” and “aggression” to be deemed “fearful-aggressive”.

The enrichment program (hereafter “human enrichment”) primarily consisted of sitting in a small room (3.0 X 3.7 m) with home furnishings (couch, table, lamp, wall art), toys, classical music, lavender fragrance, small treats, speaking softly to the dog, and petting or play when the dog approached the experimenter(s). Sessions were 15 minutes long and dogs received two sessions per day for approximately five days. The dogs were also given a blanket, toy, and Kong or chew treat in their kennel at the end of each testing day. Dogs in the control condition did not go to the furnished room (see Conclusion for a discussion on why this is problematic) but were given the same items in their kennel at the same time as the dogs in the human enrichment condition. SAFER was administered after the last day of human enrichment or control.

Researchers used SAFER because it is used by staff at the shelter where the research was conducted. The authors note that SAFER scores are theoretically meant to be only one piece of information collected on each dog, but, “…in practice, performing inadequately on the SAFER often can lead directly to fearful dogs being euthanized for aggression that they would be unlikely to display under other circumstances.” Of course, this is an impossible statement to prove because we cannot know the future behavior of a dog who is euthanized, but the limited research that has been done (Marder et al., 2013 and Mohan-Gibbons, Weiss, & Slater, 2012) on dogs who fail a behavior evaluation but are not euthanized does support this sentiment.

In addition to studying the effects of human interaction on overt behavior and SAFER scores, the authors investigated the effects on the dogs’ emotional states as measured via cognitive bias tasks. In the canine cognition and behavior literature, cognitive bias tasks are designed to assess whether a subject will respond to an ambiguous stimulus in a more optimistic or more pessimistic manner as demonstrated by how quickly the dog approaches a bowl not in exactly the same location as ones they’ve previously found to contain treats. A “cognitive bias” can be positive or negative and is believed to reflect an animal’s emotional state and influence their behavior. Because previous research with dogs in shelters showed that negative cognitive bias was related to undesirable behavior (specifically, separation anxiety related behaviors—dogs exhibiting this condition being assumed to be pessimistic about their owners’ return), the authors hypothesized that the human enrichment program would result in a more positive cognitive bias compared to control.

The cognitive bias testing used a common procedure in which dogs learn that a bowl in one position (P) always has bait, and a bowl at another position never has bait (N). The bowls are equidistant from the dog’s starting position. After the dog learns the distinction between bowl locations as demonstrated by a consistently shorter latency to approach the baited bowl, the experimenter then introduces three new probe locations between the existing extremes: one near the positive bowl (NP), one near the negative bowl (NN) and one directly in between the two extremes (M). The dog’s latency to approach the probe bowls is measured and a percentage of the difference between P and N is calculated (this accounts for individual differences in running speed).


In Experiment 1, the number of “fearful-aggressive” dogs who passed SAFER was more than double in the human enrichment condition compared to control: 23 of the 30 dogs who received 30 minutes of human enrichment per day passed versus only 10 of the 30 dogs who did not receive the intervention (X2 = 11.38, p = 0.001).

Experiment 2 replicated these findings: 15 of the 16 “fearful-aggressive" dogs who received human enrichment passed SAFER whereas only 2 of the 16 “fearful-aggressive” dogs in the control condition passed (X2 = 21.21, p < 0.001). Human enrichment did not appear to affect SAFER outcome for “non fearful-aggressive” dogs, perhaps due to a ceiling effect: nearly all 32 dogs passed. Specifically, 15 of the 16 dogs in the enriched condition passed, and 14 of the 16 dogs in the control condition passed.

Experiment 3 showed differences in initial performance (e.g., during training), with a Fearful X Treatment interaction (F1, 51 = 22.17, p < 0.001). For control dogs, “fearful-aggressive” dogs required more training to reach criterion compared to “non fearful-aggressive” dogs (p < 0.01), but this did not extend to enriched dogs. For “fearful-aggressive” dogs, human enrichment decreased the number of trials required to meet criterion (p < 0.01), but the opposite was true for “non fearful-aggressive” dogs (p < 0.05).

As expected, the distance of the three probe bowls from the always-positive (P) and always-negative (N) positions affected dogs’ latency to approach; the closer the probe was to P (and thus, the farther it was from N), the shorter the latency (F1.7, 98.7 = 129.50, p < 0.001). There was also a significant Fearful X Treatment Interaction on probe latencies. For fearful dogs, human enrichment reduced the latencies to reach probes at NP and M locations compared to controls. For non-fearful dogs there was no effect of human enrichment for the outer probes, but those who received human enrichment had longer latencies to reach the middle (M) probe compared to control dogs.


This study exposes a concerning reality for “fear-induced aggression” in animal shelters, particularly at shelters that rely on behavior evaluations to help determine a dog’s fate. Dogs who exhibit “fearful aggression” ("fearful aggression" was defined in this study as dogs that had some combination of pinned ears, crouched body, tucked tail, averted gaze, or growling when approached in kennel--see National Canine Research Council's literature review on "aggression") may be subjected to behavior modification, labeled as aggressive, unnecessarily excluded from adoption into families, and/or euthanized. The data presented here suggests that “fearful-aggressive” dogs in shelters can overcome their fear and pass a behavior evaluation with a relatively brief period of human enrichment.

Though the takeaway from this study is applied in nature—providing human enrichment can improve “fearful-aggressive” dogs’ performance on a behavior evaluation—there is a glaring bigger-picture issue highlighted by these results: reliance on behavior evaluations is causing fearful dogs in shelters to be euthanized. A brief period of human attention appears to allow the dogs to improve their behavior and pass a test to be eligible for adoption. As little as 30 minutes of human interaction per day for five days resulted in a significant increase in the number of “fearful-aggressive” dogs who passed the SAFER evaluation. Moreover, the data suggest that their affective state—their “outlook” or possibly their stress levels—was also changed by this human enrichment as demonstrated by performance on the cognitive bias tests.

It is important to reiterate that some dogs were deemed “aggressive” but not fearful and were excluded from the study as the researchers were specifically interested in “fear-aggression.” However, the findings could have been more robust if the researchers did include dogs from this group, as well as “fearful” but not “aggressive” dogs; it would be useful to know whether all dogs—regardless of the combination of fear and “aggression”—would benefit from the brief human enrichment procedure. These data could also be useful in clarifying the underlying cause of the “aggressive” behavior. For example, if the dogs deemed “aggressive,” but not also deemed fearful, were affected in the same way as the “fearful-aggressive” dogs, then perhaps all of the dogs were initially acting out of fear, as some behavior experts suggest. We believe this was a missed opportunity to gain information on canine behavior in shelter settings.

Moreover, there were four “fearful-aggressive” dogs in Experiment 2 who were euthanized without being tested on SAFER because shelter staff deemed them too dangerous to test. This is interesting for two reasons: all four of these “fearful-aggressive” dogs were in the control condition (did not receive human enrichment), and no dogs in the human enrichment condition exhibited such extreme behavior that staff were unwilling to test them. Second, for the dogs who exhibited this extreme behavior, SAFER was not used; the shelter staff believed they already had the information they needed to make a decision. This is important as it relates to the ongoing debate on whether behavior evaluations are necessary for use in shelters—see No better than flipping a coin: Reconsidering canine behavior evaluations in animal shelters (plus analysis) and What is the evidence for reliability and validity of behavior evaluations for shelter dogs? A prequel to no better than flipping a coin (plus analysis). (Please note that the authors of both papers are affiliated with National Canine Research Council.)

Another interesting note revealed in this paper is that the shelter uses a modified version of SAFER: the subscale involving a toy is not included. Though this has no impact on the results of this study, it is an important piece of information for researchers and animal welfare stakeholders, because with respect to research on behavior evaluations explicitly, modifications of scales prevent comparisons, and any claims of validity or reliability of the standardized scale cannot be extended to modified versions. By including this statement in their study, the authors have provided one more piece of the puzzle because it is often suggested that—at least anecdotally—shelters tend to use their own modified versions of behavior evaluations, but this is difficult to document. 

One weakness in this study is the confounds of music, odor, home-like environment, and kennel reprieve for dogs in the human enrichment versus control conditions. If the researchers wanted to specifically address human interaction, then odor and music should have been removed, and control dogs should have been placed in the furnished room for the same number of trials and amount of time as the enriched dogs. With the experimental procedures used, one cannot say for certain whether the improved behavior was due to human attention, olfactory exposure, classical music, or kennel reprieve or perhaps a cumulative effect of multiple stimuli (the lavender scent introduced is perhaps the least likely to represent such a confound since the research on which the authors based this choice is itself very weak (lacking appropriate controls), providing little evidence to support its effect on behavior). However, the most important conclusion—that “fear-induced aggression” can be mitigated, and dogs more likely to pass a behavior evaluation following a very brief and easy treatment—remains the same, further highlighting the potential harm in using provocative behavior evaluations to make decisions regarding dogs living in shelters.

Abstract and Link to Full Text of the Original Article:

Additional References:

Willen, R. M., Mutwill, A., MacDonald, L. J., Schiml, P. A., & Hennessy, M. B. (2017). Factors determining the effects of human interaction on the cortisol levels of shelter dogs. Applied Animal Behaviour Science186, 41-48.

Patronek, G. J., & Bradley, J. (2016). No better than flipping a coin: Reconsidering canine behavior evaluations in animal shelters. Journal of Veterinary Behavior15, 66-77.

Patronek, G. J., Bradley, J., & Arps, E. (2019). What is the evidence for reliability and validity of behavior evaluations for shelter dogs? A prequel to “No better than flipping a coin.” Journal of Veterinary Behavior, 31, 43-58.

Marder, A. R., Shabelansky, A., Patronek, G. J., Dowling-Guyer, S., D’Arpino, S. S. (2013). Food-related aggression in shelter dogs: a comparison of behavior identified by a behavior evaluation in the shelter and owner reports after adoption. Applied Animal Behaviour Science 148(1-2), 150–156. doi:

Mohan-Gibbons, H., Weiss, E., & Slater, M. (2012). Preliminary investigation of food guarding behavior in shelter dogs in the United States. Animals, 2, 331-346. doi:10.3390/ani2030331

Page last updated May 28, 2020