Perception biases for problematic behaviors in dogs due to owners’ attributes

Abstract

The abnormal or undesirable behaviors of owned dogs are not always considered problematic; it depends on the perception bias of their owners. To demonstrate the perception bias in dog owners’ attributes, 133 dog owners in Aomori (rural) and Tokyo (urban) were surveyed through questionnaires distributed via seven animal hospitals regarding the frequency of potentially problematic behaviors and their perceived difficulty with them. The interaction effects of the lived location (urban, rural), age (20s–50s, 60s or later), and sex (male, female) of the owners were evaluated through a hierarchical multiple regression model. The analyses of 115 responses demonstrated that the tendency of perception regarding the five major behaviors under consideration varied with these attributes. Our results indicated that owners living in Aomori undervalued destruction behaviors of their dogs both when family members were and were not at home, while they overvalued jumping on people. Senior owners tended to undervalue nuisance barking when family members were at home along with uncontrollable hyperactivity. Male owners also undervalued destructive behavior when family members were not home. The study concludes that perception bias due to dog owners’ attributes should be taken into account in epidemiological surveys and during medical interviews by veterinarians or other behavioral specialists. Further exhaustive investigation and exploration of the cultural background of these perception differences should be conducted.

In recent years, many dog owners have consulted veterinarians or other professionals that specialize in animal behavior about problematic behaviors [34]. Some dogs show abnormal behaviors that deviate from the normal range for canines, while others may display normal behaviors that their owners consider unacceptable based on their lifestyle [2]. These abnormal or generally undesirable behaviors can be termed potentially problematic behaviors. Furthermore, such behaviors are designated as problematic behaviors when they are recognized to cause difficulty for the owners and must be addressed. Therefore, even though owners are aware of their dogs’ behaviors that could be evaluated as problematic by others, they might not consult veterinarians. For instance, while many owners noticed that their dogs displayed behavioral symptoms associated with dementia, they considered such a change unproblematic [1, 18]. Furthermore, Pirrone et al. [22] demonstrated that many dog owners (65%) perceived their dogs to exhibit undesirable behaviors that were not problematic.

Severe problematic behaviors could eventually become a matter of death [4], although the Japanese tend to be reluctant to euthanize [13, 28]. As highlighted in the behavior management guidelines [9], regular check-ups from puppyhood and client education can prevent the development of problematic behaviors and eventual euthanasia [9, 14, 16, 21]. However, during consultation, abnormal or generally undesirable behaviors can be overlooked because they have not been perceived as problematic by the owners. Therefore, it is important to know what kind of owners are more likely to overlook these behaviors. Once such predisposition has been identified, a proactive approach of asking and educating owners about these behaviors could be expected to both control these behaviors and improve treatment outcomes.

Although various epidemiological studies have examined the frequency of abnormal or undesirable behaviors and the presence of behavioral problems, research focusing on the factors behind perceiving the abnormal or undesirable behaviors as problem is limited. In fact, Pirrone et al. [22] demonstrated that the perception of being problematic varied according to the type of behavior and the various attributes of the dog owners or their dogs. The caregiver burden for animal sickness, including problematic behavior, was associated with the owners’ age but not their sex or income [26]. The degree of perception of problematic behavior in human infancy to adolescence life-stage is subject to variation based on the evaluator’s role in society [10, 19]. These findings suggest the hypothesis that dog owners’ attributes quantitatively influenced the degree of perceiving abnormal or undesirable behavior as problematic.

To test the hypothesis, we investigated the relationship between the frequency of potentially problematic behaviors and perceived difficulty using a linear regression model that included the attributes as interaction terms. The study particularly focused on location, age, and sex as representative attributes that allow for categorization into dichotomies. Unlike the body frame of the owner and the dog, these could be examined even with relatively small sample size.

MATERIALS AND METHODS

A questionnaire survey was performed from April 29 to May 28 in 2017 and 2018. In 2017, veterinarians or staff members in five primary animal hospitals in Aomori (population density: 132.5 person/km²; population of 20s to 50s: 573,000 persons; 60s and over: 505,000 persons in 2017 [27]), a rural area, distributed questionnaires among dog owners. In 2018, the same questionnaires were distributed to two hospitals in the wards area of Tokyo (population density: 6,300.0 person/km²; population of 20s to 50s: 7,843,000 persons; 60s and over: 3,875,000 persons in 2018 [27]), an urban city. Each facility received 50 sheets. The survey period ended when all questionnaires had been completed or when one month had elapsed.

No live dog was involved in the study; therefore, it was not subject to the institutional animal ethics review. The study was carried out in accordance with the Ethical Guidelines for Medical and Health Research Involving Human Subjects and the Declaration of Helsinki. The participating dog owners only answered the questionnaires when they consented to the study outline. These were compiled by hospital staffs, and only the questionnaires with no personal information were given to the authors for analysis. Then, the collected data were treated with caution.

Questionnaire

The front page of the questionnaire (English translation is shown in Supplementary File 1) stated that the present survey was conducted anonymously; whether or not dog owners chose to participate would not affect the treatment provided to their dogs. Then, it was deemed that consent was obtained when the following questions were answered. The first part of the questionnaire surveyed owner characteristics: age (in the 10s, 20s, 30s, 40s, 50s, 60s, 70s, 80s, or older), sex (male or female), number of housemates, and number of dogs owned in the past. Next, information was gathered about the participants’ dogs: present age (year and month; can be approximate), age at the time the dog was adopted (year and month; can be approximate), origin (pet shop, breeder, found a stray dog, or other [with free description]), sex (male, female), whether neutered (if so, year and month; can be approximate), breed, main space occupied (indoor or outdoor), and number of dogs or cats living together.

The next page of the questionnaire listed questions about potential problematic behaviors. With references to several sources [17, 22, 31], the following 16 common behaviors were listed as items: A) biting people; B) growling or barking at people; C) biting dogs or other animals; D) growling or barking at dogs or other animals; E) destruction of furniture, building materials, or toys when family members are home; F) nuisance barking when family members are home; G) inappropriate elimination when family members are home; H) destruction of furniture, building materials, or toys when family members are not home; I) nuisance barking when family members are not home; J) inappropriate elimination when family members are not home; K) fear response to unfamiliar people; L) fear response to noise, including thunder; M) pica, including coprophagy; N) opportunistic escape; O) jumping on people; and P) uncontrollable hyperactivity. Using a 4-point Likert scale, the participants were asked to score the frequency (1 =never, 2 =less, 3 =sometimes, 4 =usually) and difficulty (1 =never, 2 =seldom, 3 =somewhat, 4 =very) of dealing with each behavior.

Analysis

R 4.0.2 (The R Foundation, Vienna, Austria) was used for the analysis, with a significance level of P<0.05. Age categories were classified into junior (20s to 50s) and senior (60s or older). The reason for this was that most occupations in Japan had a mandatory retirement age of 60 years, and people’s lifestyles and positions would differ substantially before and after this age. Dog’s breeds were grouped according to Tonoike et al. [30]. Independences of demographic characteristics were evaluated by Fisher’s exact test.

The frequency and difficulty of dealing with the behaviors were deemed as continuous values, and their means were compared with respect to location, age, and sex using the Mann–Whitney U test with Bonferroni corrections. The behaviors were then analyzed using a hierarchical multiple regression model to explore the interaction effect of the owner’s location, sex, and age. The response variables in all models were perceived difficulty of each behavior. The explanatory variables in the first step models (Model 1) were the frequency of each behavior, location (Tokyo=0, Aomori=1), age (junior=0, senior=1), and sex (female=0, male=1). To avoid multicollinearity, the frequency measured using the 4-point Likert scale underwent mean centering.

The second-step models (Model 2) included the addition of two-way interaction terms between the frequency and other variables: frequency × location (Aomori), frequency × age (senior), and frequency ×sex (male). Mean centering was applied to the frequency as aforementioned. Prior to performing these analyses, multiple imputations with predictive mean matching for missing values of behavior variables were performed for a more reliable estimation under the presumption that the probability of missing did not depend on the actual values of its missed data (missing at random; MAR). The mice 3.12.0 package [5] created 100 imputed datasets comprising the frequency and difficulty of the behaviors, location, age, sex, and the interaction terms. Pooled statistics calculated for imputed datasets were combined using Rubin’s rule. The variance inflation factor (VIF) of each explanatory variable over 10 was considered as multicollinearity, and such a model was rejected.

To determine whether the addition of the interaction terms changed the coefficient of determination (R²), Models 1 and 2 regarding the same behavior were compared with the D1 multivariate Wald test using the testModesl function of the mitml 0.3–7 package [8]. When the R² was statistically increased in Model 2 from Model 1, a simple slope analysis was conducted for the significant interaction effects to explore the nature of the interaction. Additionally, the main effects of the behavior frequency were calculated with respect to the location (Tokyo and Aomori), age (junior and senior), or sex (female and male). The same calculations were performed for the complete case dataset that deleted all cases including any missing values in the behavior variables to evaluate the appropriateness of multiple imputation under the missing at random presumption.

RESULTS

Of 133 participants, 121 owners answered the questionnaire. The analyses included 115 valid responses, which answered with 0 to 2 missing values. The missing percentage for each variable related to the behaviors was 0% to 10.7%. In particular, 14 participants had missing values for question I). A possible reason could be that there were limited ways for dog owners to notice nuisance barking in their absence; they would not have known about them without neighborhood contacts. Therefore, the question item was excluded and remaining 15 behaviors were analyzed.

Demographics of participants

The subsequent demographic characteristics are summarized in Table 1. A larger proportion of senior owners (≥60s) were male (45.8%), unlike junior (<60s) ones (20.9%, P=0.019). In contrast, no difference was detected in the proportion of senior owners in each region (10.0% in Tokyo, 23.2% in Aomori, P=0.239). The participants had experience of the ownership of 2.7 dogs on average (standard deviation [SD]=2.1) besides the subject dog. The age of the dogs ranged from 0.3 to 18.3 (mean ± SD: 7.9 ± 4.6 years), and they were adopted between 0 to 9.0 (0.8 ± 1.5) years of age. The age category of dogs (<6 or ≥6 years old) was not biased by location (Table 1, P=0.614), age (54.9% of junior, 66.7% of senior owners, P=0.471) and sex of their owners (60.0% of male, 56.4% of female owners, P=1.000). The origin of the dogs was clearly different between locations (Table 1, P=0.022). The majority of dogs in Tokyo came from pet shops, followed by breeders. However, dogs in Aomori mostly came from other places of origin, which meant that most of them were given to the participants by acquaintances. There were no differences in the sex of the dogs by location (male dogs were 65.0% in Tokyo, 56.7% in Aomori, P=0.619), age (62.8% of junior, 41.7% of senior owners, P=0.100), or sex of their owners (67.9% of male, 54.9% of female owners, P=0.272). The sterilization procedure was performed on the dogs when they were approximately 2 to 3 years old. The seven invalid responses of living area of the dogs in Aomori were expected to be for dogs that lived both inside and outside the house. The subject dogs were kept with another 0.5 dogs/cats on average. Dog owners in Tokyo had a maximum of two animals, while those in Aomori had a maximum of seven.

Table 1. Demographics of the participants.

		Location		P
		Tokyo (n=20)	Aomori (n=95)
Age of owner
	20s	2 (10%)	1 (1%)	0.293
	30s	4 (20%)	18 (19%)
	40s	8 (40%)	30 (32%)
	50s	4 (20%)	24 (25%)
	60s	2 (10%)	17 (18%)
	70s	0 (0%)	5 (5%)
Gender
	Female	17 (85%)	68 (72%)	0.272
	Male	3 (15%)	27 (28%)
Housemate
	0	1 (5%)	1 (1%)	0.818
	1	5 (25%)	27 (28%)
	2	6 (30%)	33 (35%)
	3	3 (15%)	16 (17%)
	>3	3 (15%)	16 (17%)
	NA	2 (10%)	2 (2%)
Age of dog
	<6 year	11 (55%)	55 (58%)	0.614
	≥6 year	9 (45%)	33 (35%)
	NA	0 (0%)	7 (7%)
Origin of the dog
	Breeder	7 (35%)	15 (16%)	0.022
	Petshop	10 (50%)	27 (28%)
	Custody	1 (5%)	8 (8%)
	Other	2 (10%)	35 (37%)
	NA	0 (0%)	10 (11%)
Breed groups
	Ancient and spitz breeds (Shiba Inu: 9)	1 (5%)	8 (8%)	0.769
	Toy dogs (Shih-Tzu: 1, Papillon: 1, Pug: 2, Pomeranian: 2, Chihuahua: 8)	4 (20%)	10 (11%)
	Spaniels, scent hounds, and poodles (Beagle: 2, Bolognese: 1, Dachshund: 16, Maltese: 1, Poodle: 20)	10 (50%)	30 (32%)
	Working dogs (German Shepherd Dog: 1, Miniature Schnauzer: 2)	0 (0%)	3 (3%)
	Small Terriers (Yorkshire Terrier: 1)	0 (0%)	1 (1%)
	Sight hounds and herding dogs (Italian Greyhound: 1, Shetland Sheepdog: 1, Welsh Corgi: 2)	0 (0%)	4 (4%)
	Retriever (Labrador Retriever: 3, Golden Retriever: 3)	0 (0%)	6 (6%)
	Mastiff-like breeds (French Bulldog: 1)	0 (0%)	1 (1%)
	Mongrel (Small dogs originated from Chihuahuas, Dachshunds, etc.: 12, unspecified due to lack of information: 21)	5 (25%)	28 (29%)
	NA	0 (0%)	4 (4%)
Sex of the dog
	Intact female	0 (0%)	7 (7%)	0.186
	Intact male	3 (15%)	26 (27%)
	Spayed female	7 (35%)	32 (34%)
	Castrated male	10 (50%)	25 (26%)
	NA	0 (0%)	5 (5%)
Area the dog live
	Indoor	19 (95%)	80 (84%)	1.000
	Outdoor	1 (5%)	8 (8%)
	NA	0 (0%)	7 (7%)

Independences of location and each factor were evaluated by Fisher’s exact test. The bold numbers show that P<0.05. NA: Not available.

Frequency and difficulty of potential problematic behaviors

The responses regarding the 15 behaviors with respect to the location, age, and sex are shown in Fig. 1 and Supplementary File 2. Nuisance barking clearly troubled dog owners’ families in Tokyo when they were home, while the opportunistic escape of dogs was statistically significant only for that frequency. Younger owners in the 20s to 50s age group scored worse for 7 of the 15 dog behaviors. Growling or barking at other animals, nuisance barking, destructive behaviors, inappropriate elimination both when family members were home, and uncontrollable hyperactivity were all frequent and distressed younger owners. Finally, female owners reported more frequent dogs biting people, although that difficulty was not significant. On the contrary, there was no difference in reported frequency of growling or barking at people, although females expressed higher difficulty.

Fig. 1.

Violin plots combining with boxplots of frequency and perceived difficulty of potential problematic behaviors of dogs. Scores within each attribute are compared using Mann–Whitney U test with Bonferroni corrections (*: adjusted P<0.05, **: P<0.01). The behaviors having any significant difference are only shown.

Interaction effect of the owner’s attributes

Hierarchical multiple regression analyses presented five significant models with interaction terms (Table 2); all the results of the analyses, including insignificant models, are shown in the Supplementary File 3. The simple slope analyses proved the significance of all the partial regression coefficients of the frequency to perceived difficulty, clearly mediated by the interaction term (P<0.01). Figure 2 shows the differences in the slopes based on the owner’s attributes. These results were generally similar for the complete case dataset and the imputed dataset, although there were slight variations in the coefficients and P values.

Table 2. Estimated partial regression coefficient (standard error) in significant interaction models for imputed and complete case dataset.

Fig. 2.

Significant main effects of behavior frequency on perceived difficulty, mediated by dog owner’s attributes. The graph on the left side are reference graphs (Tokyo, junior, female). B means estimated partial regression coefficient (standard error). F: female, M: male, Jr: junior, Sr: senior, AO: Aomori, TY: Tokyo.

DISCUSSION

Analyses of interaction terms between the frequency of behaviors and attributes indicated that the perceived difficulty even for the same degree of frequency was different across the attributes. The lived location, age, and sex of the owners were suggested to modulate the perception of the difficulty of problematic behaviors. However, the opposite deduction that they reported more because in trouble could be possible. It should be considered in relation to the cognitive capacity and cultural context where each behavior and subsequent problem is interpreted. Furthermore, dog owners’ personality and psychological status might also influence their perception [7, 20, 26].

Senior owners tended to disregard growling at people and barking at family members. The hearing acuity of the elderly is particularly impaired when the sounds are in a higher frequency range, from 700 Hz and above, which is a major vocal range of dogs that causes disturbance [23, 33]. This can make it hard to recognize such vocalizations as a problem, even if heard as a sound. Contrary to Pongrácz et al. [23], who suggested that males perceived high-pitch barking as more annoying, the male owners in the present study generally responded less or mildly to dogs’ growling sounds (Fig. 1). Females may respond more emotionally than males in the actual context [24, 29], thus showed different responses in the experimental environment. This tendency might partially account for female owners’ perception of destructive behavior as more problematic. However, as there is little evidence on gender differences in perceptions of animal behavior, future discussions should focus on differences in perceptions according to sex, while being careful to not discriminate.

Uncontrolled hyperactivity was not a concern for elderly owners, despite the decline in their physical function. It is possible that the elderly considered vigorous temperament to be natural and not something worth their concern. As for the low frequency of uncontrolled hyperactivity of their dogs among senior owners could be because many of them were retired and had trained their dogs. However, as Lord et al. [15] reported more problematic behaviors in dogs with unemployed owners, it would be necessary to elaborate including the background.

Male and Aomori were significant factors to decrease attention for destruction behavior. The regression model also showed a significant negative interaction in seniors in complete-case analysis. However, the analysis for imputed cases was not significant, suggesting that the missing values might occurred with regards to aging rather than at random. Furthermore, most of senior owners in the complete-case analysis scored one for both frequency and difficulty for destructive behavior while away home, then these behaviors can be elusive for the elderly people to evaluate. Although the difference in perception by gender may be explained similarly to the above, it should be carefully interpreted because males dominated among the elderly in the present study. While awareness of destructive behavior was low in Aomori, the opposite result was obtained with jumping on people being disregarded in Tokyo. Such difference by location requires further investigation. Although there was no significant breed difference between the two locations as a whole, it was impossible to conclude precisely because mongrel dogs were common in Aomori. The possibility of perception bias due to dogs’ attributes rather than owners’ attributes also cannot be dismissed, such as the fact that large dogs tend to annoy their owners more than small dogs. Body frame and weight of dogs should be considered in conjunction with in the future.

As such, a limitation of this study is the potential influence of dogs’ attributes. A problem’s severity can vary depending on sex and the size and breed of the dogs, even for the same degree of behavior. Sampling was not controlled for such attributes in this study, allowing for bias. And, since these factors were not included in the present analyses, a more extensive study would be required. This study also has a sensitivity limitation. In univariate analyses (Fig. 1), inconsistent findings significant only for frequency or difficulty were observed for biting people, growling or barking at people, and opportunistic escape. However, both values showed similar trends, so inadequate test sensitivity or small sample size may have obscured the differences. Also, in the multiple regression analyses (Table 2), some effects may have been overlooked because of lower analytic power, including interaction terms [3]. Although we only examined the interaction effect for the three factors, further studies on various interaction effects could be meaningful. However, very large sample sizes would be required to incorporate variables involving multiple options (breeds, sex taking into account with or without neutralization) and classification with ambiguous criteria (different breeds have different life stages and obesity assessments). As noted in some other studies [6, 32], it would also be reasonable on that occasion to adopt P<0.10 as the significant level or to increase the sample size.

A straightforward interpretation of the results suggests that dog owners who live in rural areas and are elderly or male presumably have less perceived difficulty with the same level of potential problematic behaviors. Such owners could overlook the possibility that potentially treatable behavior can escalate and cause major problems. Therefore, veterinarians or other behavioral professionals should proactively inquire about these potential problematic behaviors. Proper client education should also be provided because negligent perception of the problem is likely to interfere with treatment willingness and compliance. This is especially important for the Japanese, who do not typically take initiative in reporting the problem, but instead rely on physicians to make decisions [11, 25]. It can be preferable to ask other family members about such issues; they may perceive behaviors differently even within the same household. There appears to be a preference for family members to be involved in the decision-making process for medical treatments [12]. However, these findings and considerations may only be unique to Japan, and are not necessarily common across cultures. Hence, further evaluation involving individual cultures could develop this research issue.

CONFLICT OF INTEREST

The authors declare no conflict of interest.