Journal Article: Are Shelter Dogs More Aggressive? and “Perfectly Safe” vs. “Vicious”

Title:  Human directed aggression in domestic dogs (Canis familiaris): Occurrence in different contexts and risk factors

Authors: Rachel A. Casey, Bethany Loftus, Christine Bolster, Gemma J. Richards, Emily J. Blackwell

Published: December 11, 2013 in Applied Animal Behaviour Science

General overview:

These researchers distributed 14,566 surveys and received 3897 completed, legible responses (26.6% response rate). The surveys assessed owner demographics, basic info about the owner’s youngest dog (origin, age, breed, neuter status) as well as training methods (grouped into positive reinforcement/negative punishment or positive punishment/negative reinforcement) and instances of human directed aggression or avoidance of humans in multiple contexts. The researchers used their survey results to create three multivariate models for aggression directed towards family members, towards unfamiliar people entering the home and towards unfamiliar people outside of the home.

[Brief explanation of multivariate models: In this case, this means that the authors took all of the questions they asked about in the surveys and entered them into a statistical model as variables, such as “breed type”, “age”, “neuter status”, etc. These variables were assessed for their ability to predict an outcome variable, which were the aggressive responses in various contexts from the surveys for this research. Using exclusion criteria, variables are eliminated if they didn’t predict the outcome variable very well. The final model contains only those variables that, in combination, predict the outcome variable “well”, according to various statistical standards. For more information, see this NIH article on multivariate analysis.]

The model with the outcome variable of aggression directed towards family members included owner age, dog age and neuter status, training method, attendance to training classes (except puppy classes), breed type and origin of dog. The model with aggression directed at unfamiliar humans entering the household included owner gender, owner age, dog age and neuter status, attendance to puppy classes and breed type. Finally, the model with aggression directed at unfamiliar humans outside the household included dog age and neuter status, puppy classes, ring craft classes, training category and breed type.

My comments:

Let’s start with my complaints about this article. I felt “aggression”, as assessed through the surveys, was again poorly defined. The authors briefly acknowledged that there may be discrepancies between owner interpretation of dog behavior but seemed to feel that differences might be gender based. I’m not sure how substantiated this theory is, but this issue could be circumvented by clearly defining “aggressive” behaviors in a certain way. Obviously, recall (a respondent’s imperfect memory)  and reporting (a respondent’s voluntary suppression of information) biases will still exist but these are issues inherent with surveys.

Although the authors clearly define their study population as a convenience sample, I didn’t think this was appropriately included in the interpretation of the results. Half of the respondents had received this survey at a dog show or dog-related event, which is very unrepresentative of the general dog-owning population in my personal experience. I’m not sure if wide-scale dog owner demographics are available but I feel like the inclusion of so many people who were motivated to attend a dog show makes the generalizability of this study to the general population of dogs and owners is questionable.

My final critique is the interpretation of owner gender, owner age, dog gender/neuter status and dog breed variables in the final multivariate models. The authors did a nice job of comparing their results with previous studies but for these variables in particular, the discussion got confusing. It basically came down to the fact that previous studies have both agreed and disagreed with the results from this research for these particular factors. So do these variables contribute to a dog’s risk of human direct aggression, and why is that? The authors just sort of shrugged in answer. I’m all for being open and honest about your findings, but…really? That’s about half of the final variables in all three models! If there really isn’t anything you can conclude from these variables except that more research needs to be conducted into each of these variables, why were they included to begin with?

Finally, the interesting bits of the study. First, there was the origin of the dog. Compared to dogs obtained from breeders, dogs from rescue groups had a 2.6 times increased risk of aggression towards humans and dogs obtained from an “other” source (pet stores, internet sites, etc.) had a 1.8 times increased risk. Are shelter dogs at greater risk of human directed aggression? There are two potential explanations to this apparent finding:

1. Yes, shelter dogs have an increased risk of human directed aggression because A) human-aggressive dogs are more likely to surrendered to a shelter and/or B) being in a shelter could somehow make a dog more aggressive towards humans.

2. No, shelter dogs are not actually more likely to exhibit human directed aggression but it appeared in this study as the result of owners of dogs from rescue centers being more likely to report human directed aggression because A) they may feel less “responsible” for the dog’s behavior since (in most cases) they aren’t the first owner (i.e. less reporting bias than owners who obtained their dogs from a breeder) and/or B) they may be more attuned to their dog’s behavior because of its unknown background (i.e. less recall bias than owners who obtained their dogs from a breeder).

So which is it? I have no idea – I suspect it’s a combination – and I wish the authors’ had delved into this finding a little more. Or maybe someone could pay me to do that!

Next, the authors found that dogs that were aggressive in one context were not likely to show aggression in another context. This is really interesting finding because it supports what all those behaviorists have been harping about (that dogs exhibit aggressive behavior in response to certain perceived threats in certain contexts) and goes against the popular theory that some dogs are “vicious” while other dogs are “perfectly safe.” These researchers furthermore added that their finding that pit bull-type breeds (Staffordshire Bull terriers, other bull breeds and mastiff breeds) that are frequently deemed vicious by popular culture and breed-specific legislation did not have an increased risk of human directed aggression compared to the baseline group (cross breeds).

Attendance to training classes (except puppy classes) was associated with an increased risk of human directed aggression in this study. The authors concluded that this could be because A) these classes somehow increased dogs’ aggressive behavior or that B) owners with dogs who exhibited human directed aggression were more likely to take them to training classes. Additionally, I think there could be an increased awareness of dog behavior in owners who have been instructed by a qualified dog trainer.

In a somewhat related vein, this study also found that certain training methods were associated with an increased risk of human directed aggression. Dogs of owners who reported using any kind of positive punishment or negative reinforcement had a 2.2 increased risk of aggression direct at unfamiliar people outside the household and a 2.9 increased risk of family directed aggression. The relationship of factors is unknown: A) do owners with aggressive dogs “resort to” punitive training methods more often or B) do dogs trained with punitive measures develop aggressive behaviors more often?


A small note of my own personal opinion: it would be useful to know if shelter dogs are actually more likely to demonstrate human directed aggression. Why?

  1. Human directed aggression from dogs puts the humans around them at risk of being bitten, which could result in psychological trauma, serious injury or even death.
  2. Aggressive dogs aren’t happy dogs. Behaviorists tend to think that dogs exhibiting aggressive behavior are fearful and anxious.
  3. If aggressive behaviors are being exhibited in a limited number of contexts, they could possibly be resolved with fairly minimal behavior modification* or management. Small(ish) effort = happier dog + safer humans. That seems like a good deal to me!
  4. Rescue groups try their darnedest to send their adopters home with as much information as they need to succeed with their new family addition. If dogs from shelters really are more likely to demonstrate human directed aggression, adopters ought to be prepared for it!
  5. If dogs from rescue centers really aren’t more likely to exhibit human directed aggression and it’s really that owners who got their dogs from breeders are unaware of or unwilling to admit to their dog’s aggressive behavior, this also needs to be addressed. Reading a dog’s warning signals can keep dogs happier and humans safer!

*Owners of dogs exhibiting any kind of aggressive, territorial or otherwise concerning behaviors should seek the help of a trained veterinary behaviorist or certified animal behaviorist before attempting any kind of behavioral modification!


Journal Article: Pet Owners Lack Basic Knowledge of Pets; Young, Intact Pets at Risk for Shelter Surrender

Title: Characteristics of Shelter-Relinquished Animals and Their Owners Compared With Animals and Their Owners in U.S. Pet-Owning HouseholdsTitle:

Authors: John C. New, Jr., M.D. Salman, Mike King, Janet M. Scarlett, Philip H. Kass, Jennifer M. Hutchison

Published: 2000 in the Journal of Applied Animal Welfare Science

General overview: The authors wanted to better understand the risk factors for relinquishing a cat or a dog to an animal shelter. Using interview data from a sample of people who relinquished dogs and cats in 12 shelters in four regions and a sample of U.S. households with companion animals, the investigators compared animal characteristics and human characteristics between the relinquished and owned animals and their owners. The authors found that relinquished animals were more likely to be intact (not spayed for neutered), younger, mixed breed and owned for a shorter duration of time. People who relinquished animals tended to be men under the age of 35. This study was sponsored by the National Council on Pet Population Study and Policy.

My comments: 

Interview data issues

There are some pretty basic issues with interview data, which fall into the categories reliability, fairness and validity.

  1. Reliability in interview data focuses on whether an interviewer will score similar observations the same (intra-interviewer reliability) or different interviewers will score similar observations the same (inter-interviewer reliability). This study avoids much of this issue because they use a standardized questionnaire without, as far as I can tell, open-ended questions.
  2. Fairness issues with interview data have to deal with the representativeness of the subset of the general population (also known as the sample population) that was offered the questionnaire and the makeup of people who did fill out the questionnaire. I’m not sure this study has done quite a good job at addressing this one.
    • The shelters used as study sites for this research were located in California, Colorado, Tennessee, Kentucky, New Jersey and New York. There is no explanation for the selection of this shelters, leaving me to assume that they were convenient shelters to sample based on the researchers’ locations. (The first author is from the University of Tennessee, the second is from Colorado State University, etc.) This is probably not fair as a standalone issue but especially because of the regional differences that may exist throughout the United States in animal sheltering trends, which are not addressed in this paper.
      • The authors contend that the use of shelters across the U.S. aids the paper’s generalizability in the conclusions. However, I think the only real way to achieve a generalizable conclusion would have been if the shelters were equally representative of major geographic areas in the U.S. (both urban and rural, as well as north/south/east/west).
  3. The last major concern with interview data is validity, which I also this study fails to adequately address. The households that were sampled to achieve data on owned animals to compare to relinquished animals contained households that had relinquished animals to a shelter within the past year, which the authors acknowledge may not represent the general population of animal-owning households and may impact the interpretation of the findings.
    • In plain English, this means that the comparison population overlapped with the study population. So, when interpretation of the findings, I guess it would just be best not to compare these two populations at all because they are not clearly distinct populations.

Interesting Findings

Based on the methods section, I’m not comfortable with the comparison between the “owned” pet households and the relinquished pet individuals so I’m just going to highlight some of the findings the authors presented without the comparison.

Knowledge deficiency

I think one of the most interesting findings of this piece is the knowledge deficit displayed by survey respondents. Many people felt that female animals were better off having one litter before being spayed (WHY?), had fundamental misunderstandings of normal animal behavior, such as play behavior, and did not know “appropriate methods” of training. (The authors do not detail what they mean by “appropriate methods”.)

This highlights a real animal welfare issue and a substantial area for interventions. It’s an animal welfare issue because one of the five freedoms expounded by animal welfare advocates is the freedom to express normal behavior. If owners don’t understand what normal behavior is, an animal may not be permitted to express that behavior out of the owner’s preference. This finding emphasizes the need for increased humane education for pet owners to advance pets’ quality of life and possibly reduce the number of unwanted pets.

Young, intact animals are at risk for relinquishment

The authors found that young, intact animals were overrepresented in the relinquished animals population. The term “overrepresent” in the context of a survey means that a certain selection of a population appeared more frequently that its actual distribution in the general population, so I’m not sure how the authors determined this – you’d have to know the age/neuter status of the all pets in the U.S., which a quick Google search tells me are not known quantities. What can certainly be stated is that there were a lot of unaltered and young animals in the relinquished animals population. I worked at an animal shelter previously and this aligns with my personal experience as well. I wonder if this may relate to the knowledge deficit of normal animal behavior since young, intact animals are generally more active and untrained/disobedient than older animals.

I wish the authors had included a question about maturity level in pets because one area of knowledge deficiency that I identified during my employment at a shelter was the age at which animals should be considered as “adults”. Dogs and cats really aren’t socially/intellectually mature until around age two. I saw a number of one-year-old dogs dropped off at the shelter possibly because their families did not understand that they were dealing with an adolescent dog that would soon grow out of the unruly, moody temperament she currently exhibited.

Many pet owners could benefit from pet training and management education

The authors try to make a case for length of ownership being correlated with owner attachment, but since this directly involves comparison with the “owned” pet households, I can’t really commit myself to this finding. The authors note that behavior factors (notably house-soiling and biting) may play a role in the relinquishment of animals but they did find that behavior problems existed in the “owned” animal population as well. I’m not really sure what conclusions you can draw from this when you consider the lack of distinction between these two populations but I do agree that the existence of these behaviors indicates improper training or management practices and an area of needed improvement for both the pet’s quality of life and the owner’s.


So, in all, I think this research identified several areas of animal ownership that could be focused on for animal welfare and public health interventions as well as areas for future research. I don’t agree with the analysis the authors made between the study population and the comparison population but some interesting findings are still elucidated by the publication.