This survey was conducted with members of the general public, undergraduates from North Carolina State University, and veterinary students, faculty, and staff at eight colleges of veterinary medicine in the United States (North Carolina State University, Auburn University, University of Georgia, Louisiana State University, Oregon State University, Tufts University, and Virginia Tech University). Data from one school (North Carolina State University) were collected from undergraduates as well as veterinary students, staff, and faculty; data were not collected from undergraduates at any other institution, but were collected for veterinary students, staff, and faculty. All responses were collected anonymously and written informed consent was obtained from all participants. All surveys were reviewed by North Carolina State University’s Institutional Review Board (Protocol #22285) and were performed in accordance with relevant guidelines and regulations; all participants were required to be over 18 years of age and reside in the United States. General public members were recruited using Amazon’s Mechanical Turk (mTurk) between October 25 and November 4, 2020; respondents were paid $0.75 for their participation. mTurk is a crowdsourcing website where workers are employed to complete on-line tasks including surveys. mTurk respondents had an identical survey to the other populations, with the exception of an additional attention check question embedded in the questionnaire; respondents who failed this attention check were excluded from analysis. This population of respondents has been employed in many online surveys, and attention checks have proven to be useful for ensuring the authenticity of responses^33,34.

Undergraduate respondents were recruited from NC State University using listservs specific to students majoring in Animal Science, Zoology, and Biology. These were selected in order to recruit respondents with a high level of interest in and knowledge of animals, and are the majors with the highest proportion of pre-veterinary students. Veterinary students, faculty, and staff (including interns and residents) were recruited using listservs administered by each participating college of veterinary medicine. While invitations were sent to all veterinary colleges, responses were received from eight; participating colleges included: NC State University, Auburn University, University of Georgia, Iowa State University, Louisiana State University, Oregon State University, Tufts University, and Virginia Tech University; surveys remained open for an average of 3.5 weeks at each participating university. While responses were collected anonymously, interested student participants were invited to provide their email address so that they could be entered into a raffle to win one of 39 Amazon gift cards ($20 each).

The survey was adapted from Gruen et al. and developed using Qualtrics® survey software. Pilot testing was performed to ensure readability and feedback on question structure and disambiguation were incorporated into the final survey. The survey was structured in seven blocks. The first block included information about the survey and respondents gave informed consent before moving on to the second block. The second block included directions and an example for the picture and scale. The third block presented standardized pictures of ten different breeds of dogs (Siberian husky, Labrador retriever, border collie, Boston terrier, German shepherd, golden retriever, Jack Russel terrier, maltese, pitbull-type, and chihuahua) and six mixed breed dogs and asked respondents to rate their pain sensitivity on a scale from “Not at all sensitive” to “Most sensitive imaginable” (Fig. 1). Selection of these breeds and results from this block are discussed in the companion article (Caddiell, et al.). The fourth block presented the same ten purebred and six mixed-breed dogs but respondents were asked to rate the dogs on a scale from 0 = Not at all likely to 10 = Very likely for the variables: Trust this dog with young children; Adopt this dog into your house; Trust this dog with a cat or small animal; Trust this dog in a crowd of people; Take this dog to a park.

In order to evaluate the effect of breed makeup on these ratings, two forms of the survey were created (A and B). In form A, respondents saw a pie chart representing the breed composition (results of an Embark® DNA panel) for the first three mixed-breed dogs (11–13) while seeing only pictures of the second three mixed-breed dogs (14–16); form B showed the breed composition for the second three mixed-breed dogs (14–16) and pictures for the first three (11–13) (Fig. 2).

The fifth survey block asked respondents to rate how warm or cool they felt toward a list of breeds/groups that included the ten breeds included in the survey and four groups (small/toy dogs, medium/large dogs, purebred dogs, mixed-breed dogs). Respondents indicated their responses on this “Feelings thermometer” by moving a slider along a scale from 0 to 100 and were instructed that, “Ratings between 0 and 49 mean that you do not feel particularly warm or favorable toward the group. A rating of 50 means you feel neutral, neither warm nor cold toward the group. Ratings between 51 and 100 mean that you feel warm or favorable toward the group.” Finally, the sixth block asked respondents to complete demographic questions and the seventh had space for respondents to provide any general feedback. Demographic questions were different for the different respondent populations, with mTurk respondents asked to provide their age, race/ethnicity, gender, region of residence in the United States, highest level of education, and annual household income; academic populations were asked to provide their year in school, major (undergraduates only), whether they were planning to pursue a veterinary degree (undergraduates only), if they had previous experience working at a veterinary clinic (undergraduates and veterinary students), specialty (if applicable), and degrees obtained. Faculty/staff with veterinary degrees were also asked how long it had been since they graduated and where (regionally) they had obtained their veterinary degrees.

Statistical analysis

Data analysis included descriptive and inferential statistics. Descriptive statistics were calculated for all demographic questions and examined by participant population. Participant populations included general public, undergraduates, 1st and 2nd-year veterinary students (1st/2nd), 3rd and 4th-year veterinary students (3rd/4th), and veterinary faculty and staff. The decision to group 1st and 2nd-year veterinary students and 3rd and 4th-year veterinary students was made in order to best classify veterinary students based on their course work. Across veterinary schools, the 1st and 2nd-year curriculum is focused largely on didactic material while the 3rd and 4th years engage students in applied or clinical learning and may represent a change in views that is important to capture.

Linear mixed effects regression models (R software, R Core Team) were used to assess survey form and its interaction with population and breed on likelihood of adoption and trust ratings (from 0 – 10) accounting for a random effect for subject and university. An ANOVA using Satterthwaite’s approximation to calculate degrees of freedom was used to evaluate all of the adoption/trust ratings together to assess the effect of population, breed, and their interaction. This method was then used to further evaluate the effect of feelings thermometer scores, participant population, and their interaction for each of the likelihood of adoption and trust ratings. Likelihood ratio tests comparing nested linear mixed effects models were used to compare models. Linear mixed models were used to assess likelihood of adoption and trust ratings as predicted by the interaction between the feelings thermometer scores and participant population; in these models, subject and university were accounted for as random effects. When making pairwise comparisons among populations or breeds, linear contrasts were used. Tests of individual contrasts used z-tests, while global tests used Wald tests.