Perceptions of dairy cow–handling situations: A comparison of public and industry samples

Inappropriate cattle handling poses a reputational threat to the dairy industry. To enhance social sustainability, handling practices must resonate with societal values about animal care. However, it has yet to be determined to what extent industry and public stake-holders differ in their perception of common cattle handling situations. We administered an online survey to samples of dairy industry (IND) and public (PUB) stakeholders to examine how they perceive a variety of cow-handling scenarios ranging from positive to negative in terms of effects on animal welfare. Participants were presented with 12 brief videos depicting a range of realistic cow-handling situations and responded to measures designed to assess their attitudes and beliefs about each scenario, their perception of the emotional response of the cows depicted in each scenario, as well as their own personal emotional response. Preexisting beliefs about cow treatment on US dairy farms and demographic data, including self-reported dairy consumption, were also collected and analyzed. Before viewing the videos, 52.9% of PUB (vs. 79.0% of IND) believed cows were treated well while 27.2% (vs. 9.0% of IND) believed cows were treated badly. Within IND, believing cows were treated badly was more common among nonwhites, those with greater formal education, more liberal politics, or from urban or suburban environments. In PUB, female and younger participants were more likely to believe cows were treated badly before viewing the videos. In both samples, participants with more positive preexisting beliefs about dairy cow treatment in the US reported consuming dairy products more frequently. In both PUB and IND, scenarios which were rated more positively for attitudes or for the cows’ or respondents’ emotional experiences were also perceived as more common. Within a given cow-handling scenario, qualitative attitudes (i.e., a positive, negative, or neutral valence) did not differ between the samples. In both samples, at the participant level, overall attitudes toward cow-handling scenarios were highly correlated with both their personal emotional response to the scenario and their perception of the cows’ emotional responses. Although the participants’ overall personal emotional responses did not differ between the samples, IND rated cows as experiencing more negative emotions overall. The consensus between industry and public stakeholders around dairy cow-handling practices observed in this study could provide a common starting point for addressing other, more contentious animal welfare issues

Notably absent from this body of research have been attitudes toward routine dairy cattle handling.This is noteworthy given the tremendous effects handling has on cattle and farm worker welfare and productivity.Improper handling creates stress and fear in cows, which increases their susceptibility to disease and negatively affects milk production (Breuer et al., 2000;Hemsworth et al. 2000;2002).Stress and fear can also cause cows to behave erratically, placing both caretakers and cows at increased risk of painful injuries.Indeed, cattle-related handling accidents are one of the most significant sources of work-related injuries on US dairy farms (Douphrate et al., 2013).Poor cattle handling also presents a major public perception risk as documentary video evidence of poor handling has been the primary source of negative public attention about animal welfare on US dairy farms (Robbins et al., 2016;Rice et al., 2020).In US beef, pork, and poultry industries, this negative media attention has been shown to result in small, but significant effects on consumer demand (Tonsor and Olynk, 2011).
Currently, the development of standards within dairy animal welfare assurance programs, such as the National Milk Producer's Federation-Farmers Assuring Responsible Management (FARM) Animal Care Program (FARM, 2020) and the Validus Dairy Audit Standards (Validus, 2016), rely on the judgments of industry and academic stakeholders, with little input from public stakeholders.To enhance the social sustainability of the dairy industry, some argue animal welfare standards should incorporate input from public stakeholders to ensure they resonate with public values (Weary and von Keyserlingk 2017).Some dairy animal welfare assurance programs have made attempts to include the public perspective in their standards, such as the former Dairy Well program, which stated standards must reflect "current social norms" (Walker et al., 2017).This program presented data on public attitudes toward controversial animal welfare practices alongside much more commonly included animal research.
To address the lack of knowledge about public attitudes toward different handling practices, we set out to describe and compare public and industry perceptions toward a range of realistic dairy cow-handling scenarios.Part of the challenge in studying dairy cattle handling is how broadly the concept can be defined.Animal handling, stockmanship, and human-animal interactions are general terms which can be taken to refer to approximately all aspects of dairy cow management (e.g., hoof trimming, milking, reproduction, calving); however, in the farm animal welfare context, these terms are often interpreted more narrowly as referring to situations when caretakers are moving cattle for routine farm procedures.This more restricted view is how cow handling was conceptualized in this study.
We presented participants with a series of realistic dairy cow-handling scenarios and then assessed their attitudes, judgments of commonness, perceptions of the cows' experience, and the viewer's own emotional experience while viewing the scenarios.Consistent with the commonly discussed industry-public divide, we predicted sample differences would be found between industry experts and the public on all dependent variables (i.e., perceptions of commonness, attitudes, and emotions experienced by cattle and viewers).We also explored relationships between these outcome measures and self-reported knowledge, preexisting perceptions of general dairy cattle care, dairy consumption, and various sociodemographic variables within each sample.

MATERIALS AND METHODS
This study was reviewed and approved by the University of Wisconsin-Madison Educational and Social/ Behavioral Science Institutional Review Board (submission ID 2021-0393).

Subject Recruitment
Two separate samples of adults (18 years of age or older) were recruited for this study.We originally planned to collect data in a hybrid format, with a focus on the dairy industry and general public of Wisconsin; as work proceeded during the Covid-19 pandemic, we shifted to online-only data collection and expanded our dairy industry sample to the wider US.The public sample (PUB) was recruited via CloudResearch, a panel provider utilizing the Amazon Mechanical Turk worker pool to aggregate and screen participants.Mechanical Turk samples have been used extensively in the social sciences because they are more diverse and attentive (i.e., pass attention-check questions) than traditional convenience samples, which typically consist of college undergraduate students (Buhrmester et al., 2011;Paolacci and Chandler, 2014).As a legacy of our original plans for some in-person data collection, enrollment in PUB was restricted to Wisconsin residents, but was stratified according to US Census data on age, sex, educational attainment, and income levels to ensure a diverse participant pool.Anyone reporting they were currently employed in a role related to the US dairy industry was excluded from participating in the PUB sample.The amount each participant was compensated depended on their agreement with Mechanical Turk and was not disclosed to the research team.
The industry sample (IND) was recruited via research team contacts, primarily through J. Van Os' network, including emails to dairy farm owners, dairy industry vendors, nutritionists, consultants, veterinarians, animal scientists, and extension educators throughout the US; contacts were encouraged to both participate in the survey and distribute the link to the survey among their networks (i.e., "snowball technique").The IND survey was also advertised on the email listservs for the University of Wisconsin-Madison Extension dairy team, the Dairy Cattle Welfare Council, the American Association for Bovine Practitioners, and for certified Animal Care evaluators for the FARM program, as well as on the private Facebook groups for FARM evaluators and for the Dairy Girl Network Exchange.IND participants were not compensated for completing the survey.
Data collection continued until enough participants from each sample enrolled in focus group discussions, which were held on a rolling basis; qualitative outcomes from the focus groups will be reported separately.

Survey Instrument and Video Clips
Survey instruments and procedures were approximately identical for both PUB and IND samples (Supplemental Document S1 contains PDF copies of both instruments: http: / / digital .library.wisc.edu/1793/ 84546).IND participants were additionally asked to specify their professional role and how often they directly interact with dairy cattle in that role.After consenting, participants answered a series of demographic questions (sex; race and ethnicity; educational attainment; household income; political views; whether they grew up on a farm; whether they spent most of their life in a predominantly urban, suburban, or rural area; vegetarianism or veganism; consumption of dairy products).Next, participants were asked to report their level of knowledge about human-cow interactions on US dairy farms ("How knowledgeable would you say you are about the interactions that occur between people and cows on dairy farms in the United States?": 1 = not at all knowledgeable; 5 = extremely knowledgeable) and their general perceptions of animal care on US dairy farms ("Generally speaking, how do you think cows are treated on dairy farms in the United States?": 1 = extremely badly; 7 = extremely well).
Participants were then presented with a series of 12 scenarios depicting a range of realistic dairy cow-handling situations.These 12 scenarios were selected from an initial pool drawn from existing publicly available videos published by both industry (e.g., worker training videos; links provided to C. Wickens by J. Van Os) and animal advocacy organizations, as well as several ad hoc videos filmed by J. Robbins on private dairy farms with the consent of farm owners and personnel appearing in videos.These candidate videos were screened and separated into 120 discrete clips by C. Wickens, based on the specific types of human-cow interactions occurring in each segment.The final 12 videos (described in Table 1) were selected, with feedback from all members of the research team and an external advisory group of dairy industry stakeholders, to reflect a range of realistic handling scenarios based on the work of Sorge et al. (2014), who surveyed dairy producers to identify challenging dairy cattle-handling situations, and on a simple classification system modeled on the research team's previous work (Hemsworth et al., 2000(Hemsworth et al., , 2002;;Breuer et al., 2000).The categories were: POS = positive, unlikely to increase fear in cows (slow, predictable movement; any physical contact is gentle, including petting, stroking, or resting hand on cow); NEG = negative, aversive, likely to increase fear in cows (fast and sudden movements, shouting, or physical contact such as slaps, pushes, and hits); NEG was subdivided into NEG1 (lighter slaps, pushes, hits) and NEG2 (forceful slaps, pushes, or hits, as well as tail-twists).
Video clips were edited (by C. Wickens) to be similar in length (14.3 ± 4.5 s, mean ± SD) and pixel size (640 × 360).All videos were in color, and all but 2 had sound available.To ensure anonymity of handlers in videos, human faces and any other identifying information (e.g., company logos) were blurred using a video editing application (Blur Video Spot).Each scenario (one per survey page) consisted of a video clip which appeared under a brief written description (29.3 ± 26.6 words, mean ± SD) of what was occurring in the video.For PUB, several video descriptions were modified with additional information to address their likely unfamiliarity with some aspects of the handling scenario (e.g., definition of a chute); descriptions were otherwise identical for both samples.
Participants were instructed to ensure the sound was enabled on their device and to watch each video clip as many times as they wanted before responding to the outcome measures that followed.The order of scenario presentation was randomized, except for the first 2 videos (i.e., petting, beating; see Table 1 for full description of all scenarios), which always appeared first, in counterbalanced order.Presenting these 2 extremes first was done to norm subjects away from using only the ends of response scales when reacting to the other video clips.
After each scenario, the following outcome measures, listed in order of their presentation to respondents, were assessed using Likert-type items/scales (Table 2): attitudes toward the scenario (3 items, 7-point scale), perceived commonness of scenario (1 item, 7-point scale), emotional experience of the cow(s) shown in the scenario (3 items, 5-point scale) and emotional experience of the study participant in response to the scenario (3 items, 5-point scale).These measures were developed specifically for this study to explore plausible key factors influencing perceptions.Although general attitudes toward the scenarios was our primary outcome of interest, we included other factors we believed might influence perceptions, such as beliefs about commonness, the psychological experience of the cow, and the viewer's own emotional response to viewing the scenario.The order of individual items comprising multi-item measures (i.e., scales) were fully random- Sometimes cows will not stand because they are sick or injured or simply because they do not want to stand.Getting the cow up is important for assessing its health status.The longer a cow stays down the less likely it is for the cow to eventually stand.Moreover, "down cows," as they are sometimes called, interfere with time sensitive daily processes similar to milking and barn cleaning and maintenance.This worker is trying to get this cow to stand up.
ized.A delay function was also included such that these outcome measures would not appear until sufficient time had elapsed to watch the video and read the text.After evaluating all scenarios, participants were invited to participate in a focus group on the topic of cow handling in the dairy industry; the focus group results will be presented in a separate publication.Finally, all participants were thanked and shown contact information for emotional distress support resources as well as links to the principal investigator's website and those for US animal care and quality assurance programs relating to cattle.The draft survey instrument was reviewed twice by a group of external advisors with dairy industry experi- How calm or agitated is the cow / are the cows in this video? 1 = very calm; 3 = neither calm nor agitated; 5 = very agitated How at-ease or distressed is the cow / are the cows in this video? 1 = very at-ease; 3 = neither at-ease nor distressed; 5 = very distressed How pleasant or unpleasant does the cow / do the cows find interacting with the person in this video? 1 = very pleasant; 3 = neither pleasant nor unpleasant; 5 = very unpleasant Viewer experience (3 items) 1,2  How calm or agitated does the interaction in this video clip make you feel? 1 = very calm; 3 = neither calm nor agitated; 5 = very agitated How at-ease or distressed does the interaction in this video clip make you feel? 1 = very at-ease; 3 = neither at-ease nor distressed; 5 = very distressed How pleasant or unpleasant was it for you to watch the interaction in this video clip? 1 = very pleasant; 3 = neither pleasant nor unpleasant; 5 = very unpleasant ence, and their feedback on the video clip selection, descriptions, and questionnaire items was incorporated before the survey was distributed to the PUB and IND sample populations.Their feedback included requests for a more even balance of videos depicting positive, neutral, and negative interactions (rather than only the latter); the replacement of ratings of specific viewer emotions (e.g., disgust) with the final cow and viewer experience questions; specific instances in which additional blurring of faces or logos was needed; suggestions on caption wording; and requests for a "back" button between video clips to revisit previous responses.Furthermore, the research team and advisors discussed whether IND participants would desire monetary compensation, with the consensus that IND participants would likely engage with the survey without compensation, to benefit research that could potentially provide insights to inform future public-industry engagement.

Statistical Analysis
All statistical analyses were conducted using R software (version 3.6.2).Descriptive statistics were summarized using the psych package.To ensure adequate cell size and to ease interpretation, several sociodemographic variables were collapsed before inferential analysis (Table 3); all variables were treated as categorical except age, which was continuous.Internal reliability within each multi-item outcome measure (scale) was assessed separately for each population, with a threshold of 0.70 indicating acceptable reliability.When this threshold was met, PUB and IND responses were combined and Cronbach's α scores were calculated.Cronbach α scores for the multi-item scales assessing attitudes, perceptions of the cows' experience, and the viewer's own personal experience were 0.96, 0.90, and 0.94, respectively, and thus the items within the scales were collapsed.
To investigate factors influencing general perceptions of the cow-handling scenarios, composite attitude, cow, and viewer experience scores were created by averaging responses to all scenarios for each participant across both samples.Relationships between demographic variables and outcome measures were analyzed separately for the PUB and IND samples.All variables and their distributions were visually inspected to determine the most appropriate statistical test.Between-sample (PUB vs. IND) comparisons of multi-item or composite variables were conducted using t-tests.Wilcoxon-Mann Whitney tests were used to test for associations between singleitem ordinal outcome measures and 2-level, categorical sociodemographic variables.Kruskal-Wallis tests were applied when the independent variables had >2 levels.
To compare PUB and IND perceptions of the 12 scenarios, Wilcoxon-Mann Whitney tests were conducted for each outcome measure, with Benjamini and Hochberg false discovery rate P-value correction (n = 52).In addition, the ratings for each video were re-coded qualitatively into 3 categories.For each participant's average attitude, ratings of 1.0 to 3.9 = negative (inappropriate, inhumane, unacceptable), 4.0 = neutral, and 4.1 to 7.0 = positive (appropriate, humane, acceptable).For commonness, ratings 1 to 3 = uncommon, 4 = neither, and 5 to 7 = common.For average cow and viewer experiences, 1 to 2.9 = negative (agitated, distressed, and unpleasant), 3.0 = neutral, and 3.1 to 5.0 = positive (calm, at-ease, and pleasant); these values were reversed from the scales presented to respondents for ease of comparison to their attitude and commonness ratings.
Spearman correlations, separately by PUB versus IND, were computed to assess: (1) within-subject associations between composite attitude scores versus perceived cow experience or viewer experience; (2) scenario-level associations between perceived commonness and attitudes, viewer experience, and perceived cow experience.Significant effects were defined as P < 0.05 and tendencies as P ≤ 0.10.

Demographics of Samples
After exclusion of incomplete responses, the final PUB sample consisted of 136 people and the IND sample consisted of 201 people.A total of 14 people  4).For PUB, preexisting perceptions about the care of dairy cattle in the US were more negative among females (Wilcoxon rank sum test [W] = 2,753, P = 0.04) and younger participants (rs = 0.27, P < 0.01).For IND, believing cows on US dairy farms were treated "badly" was more common among: nonwhite (vs.white) participants (W = 2,483.5,P = 0.05); those holding a bachelor's degree or greater (W = 5,965, P = 0.01); those with more liberal politics (χ 2 = 32.3,P < 0.01), and those who spent the majority of their life living in an urban or suburban environment (χ 2 = 7.89, P = 0.01).Within IND, there was a tendency for existing perceptions of animal care to be more positive among those who reported handling dairy cows more frequently (χ 2 = 8.079, P = 0.09, df = 4).More frequent dairy consumption was associated with more positive preexisting perceptions of dairy cattle treatment in both PUB (χ 2 = 19.38,P < 0.01, df = 6) and IND (χ 2 = 13.33,P = 0.02, df = 5).Selfreported knowledge was not associated with preexisting perceptions in PUB (χ 2 = 10.5, P = 0.11, df = 6) or IND (χ 2 = 8.73, P = 0.12, df = 5).No other variables were associated with preexisting beliefs about dairy cattle treatment (P ≥ 0.11).

Perceptions of Specific Handling Scenarios
Mean PUB and IND attitude ratings differed for 11/12 scenarios; commonness scores for 10/12; cow experience scores for 11/12, and viewer experience scores for 10/12 scenarios (P < 0.05; Table 5).Examination of standard deviations showed PUB responses were more variable across all scenarios compared with IND.However, when the means for each outcome variable were sorted in rank order (Figure 1), PUB and IND ranked most of the 12 scenarios similarly for perceived commonness, attitudes, and cow and viewer experience.The scenarios which were perceived as the most common were also rated most positively for attitudes, perceived cow experiences, and viewer experience by both PUB (r = 0.91 to 0.93) and IND (r = 0.75 to 0.86).Furthermore, despite the quantitative differences between PUB and IND ratings, qualitative categorical recoding of attitude, cow, and participant emotional experience measures according to their valence (negative, neutral, or positive) showed the valence of responses did not differ between PUB and IND for many of the scenarios (Figure 2A, B, C).Specifically, for the scenarios we a priori predicted would be perceived as the most positive (POS; i.e., petting, easy-walk1, 2, and 3) or most negative (NEG2; i.e., hot-shot-downer, beating), the vast majority of both PUB and IND respondents had attitudes and cow and viewer experience ratings that were positive versus negative, respectively.The qualitative responses were more disparate for some of the intermediate scenarios; for example, for both aggromove 1 and 2, most IND participants rated the videos negatively for attitudes and cow and viewer experience, whereas most PUB respondents had positive ratings.

DISCUSSION
This study described and compared public and industry perceptions toward a range of specific, realistic Video descriptions found in Table 1.
3 Cow and viewer experiences: 1 = very agitated, distressed, and unpleasant; 5 = very calm, at-ease, and pleasant.Both experience scales in this table are reversed from those presented to respondents, for ease of comparison to attitude and commonness responses.dairy cattle-handling scenarios.Previous research suggested producers generally report a positive view of the welfare of livestock, whereas the public or consumers hold more negative views (Te Velde et al., 2002;Vanhonacker et al., 2008).In the present study, however, attitudes toward cow-handling practices were generally similar between the US dairy industry and Wisconsin general public, despite differences in knowledge of industry practices and in sociodemographic factors.Despite having identical mean ages, the industry sample was considerably more male, white, educated, conservative, and had substantially higher incomes than the pub-lic sample.In addition, the racial and ethnic makeup of the nonwhite participants within the industry and public samples differed.Although the public sample was broadly representative of the US general population, we cannot comment on the representativeness of industry sample because demographic information for this population is not available.That said, the industry sample contained a diverse range of roles including farm owners, workers, consultants, and scientists, so it seems likely we gathered a diverse range of perspectives across these professional roles.Although our relatively small public sample was fairly representative in terms of viewer's experience (averages of calmness, ease, pleasantness), and (D) commonness.Respondents' ratings were converted to 3 categories, as follows, with the percentage of responses in each category shown.For attitude: 1.0 to 3.9 = negative (inappropriate, inhumane, unacceptable), 4.0 = neutral, 4.1 to 7.0 = positive (appropriate, humane, acceptable).For cow and viewer experiences: 1 to 2.9 = negative (agitated, distressed, and unpleasant), 3.0 = neutral, 3.1 to 5.0 = positive (calm, at-ease, and pleasant); these values are reversed from the scales presented to respondents, for ease of comparison to attitude and commonness ratings.For commonness: 1 to 3 = uncommon, 4 = neither, 5 to 7 = common).Video descriptions are in Table 1.demographics, all participants were from Wisconsin, a major dairy producing state, which may not reflect the US population more generally in terms of attitudes and perceptions of dairy farming.Furthermore, the scales used in this study were ad hoc and were only assessed for face validity.Future research should attempt to use measures subjected to more comprehensive validation.
Not surprisingly, participant beliefs about the care and welfare of cattle on US dairy farms before viewing the scenarios were substantially more positive within the dairy industry than the public.Despite these sample differences, more than half of the public respondents believed cattle were treated well.Within the dairy industry, preexisting attitudes toward the status of animal care in the US were less favorable among more educated, liberal, female, and nonwhite participants from urban/suburban environments.Similarly, composite attitude scores in response to the handling scenarios were less positive for nonwhite or vegetarian participants and tended to be more negative for those from nonrural environments.For the public sample, female and younger participants held more negative preexisting beliefs about how well dairy animals are cared for.Composite attitude scores in response to the scenarios were also more negative for female and younger participants, as well as for nonwhite participants, vegetarians, those living in nonrural environments, and those with lower incomes.
These demographic associations with pre-existing attitudes and attitudes toward the cow-handling scenarios are relatively consistent with previous research.For example, females tend to be more concerned about animal welfare (Herzog, 2007;Apostol et al., 2013), and more recent research suggests this effect is mediated by well-established sex differences in empathy (Graça et al. 2018).Conversely, racial or ethnic differences in attitudes toward dairy cattle welfare have not received much scholarly attention.For both the public and dairy industry samples, we found composite attitude scores were more positive for white (vs.nonwhite) participants.Although earlier research found more interest and concern and affection for animals among whites (Kellert, 1980), more recent work is equivocal (Richardson et al., 2020).Political differences in concern for animal welfare have been noted elsewhere (Lusk, 2012).Moral psychology research has shown liberals tend to rank prevention of harm and welfare as the primary criterion for ethical decision making whereas conservatives tend to consider these one among several other values worthy of serious moral consideration (Graham et al., 2009), which could explain why animal welfare concerns often diverge along political lines.
Although public and dairy industry attitudes toward specific handling scenarios consistently differed quanti-tatively, the average valence of these attitudes (positive, neutral, or negative) were largely the same, indicating overall, differences were in degree, rather than kind.Furthermore, the relative rankings of the scenarios, from most to least positive in terms of attitudes and both cow and viewer experience, were similar between the 2 samples.The quantitative differences observed may be at least partially attributable to higher variability of the public responses.It is worth noting that although qualitative differences in the valence of attitudes (e.g., liking vs. disliking; positive vs. negative) frequently predict conflict and diametrically opposed behavioral outcomes (Hepler, 2015), this is not the case with quantitative differences (i.e., degree of liking vs. degree of disliking).Within our study, in the specific instances in which most industry versus public respondents differed in qualitative valence, the latter actually held more positive views of those handling practices (i.e, aggro-move1 and 2).Thus, our results suggest the public and industry are likely to agree on the appropriateness of the scenarios they evaluated, with the public in some instances appearing to give dairy-cow handlers the benefit of the doubt, whereas industry respondents disapproved.This implies that, overall, attending to valence in public-industry dialog about cow-handling practices may offer common ground in building important consensus.
In both samples, attitudes toward the cow-handling scenarios, both overall and by scenario, were highly correlated with perceived emotional experience of the cows and the viewer.A large volume of research attests to the central role emotions play in attitude formation (Petty et al., 2003), with research showing strong (vs.weak) emotions lead to more stable attitudes over time and better predict attitude-relevant behaviors (Rocklage and Luttrell, 2021).Furthermore, the extent to which different animals are believed to experience emotions is a reliable predictor of attitudes about their moral standing and morality of their treatment (Knight et al., 2004).Because our measures of both the cows' and viewers' emotional experience were ad hoc and aimed at assessing general emotional valence (i.e., how positive or negative the scenario made them feel while watching), we cannot comment on which specific emotions may have been elicited.Previous research by Nabi (1998) has shown that disgust predicts attitudes toward animal testing.In designing the survey, we initially included an array of discrete emotions (e.g., disgust) for participants to rate their experience but decided against this with feedback from our external advisors, as it would have significantly increased respondent burden, which can harm data quality.Further work should establish which specific, discrete emotions may be activated when viewing such scenarios.
Unexpectedly, the industry sample had lower composite cow experience scores, suggesting they viewed cows as experiencing more negative affect than the public sample did.Previously, it has been argued that economically motivated animal use similar to dairy farming (as opposed to noneconomic human-animal interactions based on companionship) leads to diminished perceptions of the capacity for animals to experience negative affect (i.e., suffer), perhaps to mitigate uncomfortable psychological dissonance (Serpell, 2004); our results contradict the predictions based on this theory.Instead, our finding is consistent with previous research showing pig farmers did not ascribe their animals with diminished capacity to suffer (Peden et al., 2020).
The overall difference between the industry and public samples in our study appeared to be primarily driven by responses to the aggro-move 1 and 2 videos (which 68% and 79% of industry respondents rated as negative for cow experience, vs.only 24% and 22% of the public), and to a lesser extent, the water-spray and twist-slap-yell videos (rated negative by over half of industry respondents, but by less than a third of public respondents).We speculate that this could reflect awareness among many industry respondents of factors that cows find aversive, such as yelling (Pajor et al., 2000) or water spray on their faces (e.g., Chen et al., 2016).Interestingly, in specific instances, more than half of industry respondents perceived certain handling practices as generating negative experiences for the cows (i.e., slap-face, twist-slap-yell, slap-downer), yet found those interactions to be acceptable, appropriate, or humane.We interpret these descriptive findings to suggest that many industry respondents find certain practices that negatively affect cow welfare to be acceptable when used in specific contexts.This speculation could be further interrogated in future research, including in probes within focus group discussions.Taken together, these findings run counter to the idea that exposure to the practical realities within animal-use industries necessarily leads to a diminishment of perceiving animals' emotional experiences.Our findings may have also been driven by the public sample's lack of familiarity with cattle, leading to greater uncertainty about their mental experiences.The relationship between experience, attitudes, and perceptions of animal experience is clearly an area that deserves greater research attention.
We selected the 12 video clips to reflect a range of practices based on previous research and industry recommendations for encouraging or discouraging specific handling practices.In the interest of transparency and repeatability of our results, we will make the video clips publicly accessible after the completion of a subsequent objective of this grant project, in which we plan to use the same video clips.The 4 clips our research team rated as positive or neutral using our simple classification system (i.e., easy-walk 1, 2, 3; petting) were likewise the top 3 or 4 most positively ranked handling scenarios (including a tie) by both the public and industry samples.In addition, 2 of the 4 scenarios our team rated as the most aversive to cattle (beating, hot-shotdowner) were similarly ranked most poorly by both public and industry participants; however, the other 2 (twist-slap-yell, slap-downer) were ranked higher than some of the scenarios we had coded as only moderately aversive to cattle.This discrepancy between participant ratings and our internal classification system reflects the broad categories included in the latter, as well as our focus on the valence of the interaction (i.e., was the interaction likely perceived by the cow as positive, neutral, or negative) as inferred from previous research directly measuring cows' responses (Breuer et al., 2000;Hemsworth et al., 2000).Although perceived cow experience was correlated with both viewer experience and attitudes, this discrepancy underscores the need for the present study explicitly evaluating both public and industry perceptions.
Within both the public and industry samples, scenarios which were rated more negatively were perceived to be less common.However, although the beating scenario was ranked as the least common overall by both public and industry participants, 40% of the public sample believed this scenario was common, as did 25% of the industry sample.This scenario, obtained from an undercover investigation conducted on a US dairy farm and publicized by an animal advocacy group, depicted workers beating cattle with sticks or metal rods.Such abusive behavior would likely violate state animal cruelty statutes throughout the US and would certainly constitute an immediate failure if observed during any farm animal welfare audit for industry quality assurance programs.The actual frequency of such behaviors has not been documented in the literature and is almost never observed during second-party FARM Animal Care program evaluations or third-party audits, while workers know they are being observed by assessors.
The high proportion of public participants who perceive beating as common could be influenced by news reports of undercover exposé videos filmed by animal advocacy groups (Robbins et al., 2016;Rice et al., 2020).Because of the public sample's lack of familiarity with the scenarios, we provided brief text explaining the context of the scenarios.It is possible that alternative framings could have resulted in different response patterns.In future research, it would be interesting to examine how different levels of context provision affects public perceptions.Emphasizing or deemphasizing different tradeoffs inherent in cattle-handling scenarios could provide a more nuanced understanding Robbins et al.: PERCEPTIONS OF COW HANDLING of attitudes.Interestingly, more public participants, compared with industry participants, rated both the beating and hot-shot-downer videos as neutral or even positive for attitudes and cow and viewer experience, which could reflect the aforementioned uncertainty about cows' mental experiences, including their ability to feel negative emotions such as pain or fear.Concerningly, however, a quarter of dairy industry participants, many of whom have greater insight into what actually happens on US farms, also rated abuse as common.This finding highlights a potential opportunity to improve cattle-handling practices and animal welfare on US dairy farms.
In both the public and industry samples, participants who held more positive perceptions of general dairy cow treatment in the US tended to consume more dairy products.This is an important finding as it suggests there is a relationship between dairy consumption and perceptions about the care and welfare of dairy cows.It is possible those who believe cows are treated well are more willing to consume dairy products.Or conversely, those who consume more dairy products justify their consumption with an internally consistent rationale that the cows are likely treated well.Future research could more explicitly explore causal relationships or other factors explaining this association.However, this finding could also be an artifact of our survey design, which asked participants about their dairy consumption before asking about their perceptions of dairy cow welfare.Self-perception theory argues people may infer their attitudes based on their own attitude-related behaviors (Bem, 1972).Thus, if participants lacked well-developed initial opinions, they may have inferred their general perceptions of dairy animal care based on their response to our dairy consumption question.This methodological pitfall could be addressed in future work by counterbalancing the order of the survey questions.

CONCLUSIONS
Determining appropriate handling practices is not simple and is typically based on the judgments of industry and academic stakeholders.Incorporating data on the public's perspectives could help enhance the social sustainability of the dairy industry.This study explored industry and public perceptions of 12 realistic dairy cow-handling scenarios.Results suggest there was general agreement between the 2 samples regarding good versus bad cow-handling practices.The approach taken here provides an example of how public perspectives can be assessed.This information could later be integrated into decision making processes to better ensure animal welfare assurance programs reflect the public perspectives they were originally developed to address.
Robbins et al.: PERCEPTIONS OF COW HANDLING Table 5. Ratings (mean ± SEM) of attitudes, cow and viewer experience, and commonness for each video 1 for both public (PUB; n = 136) and industry (IND; n = 201

Figure 1 .
Figure 1.Bump charts showing relative ratings of attitudes, commonness, and cow and viewer experience for all videos in both (A) public (n = 136) and (B) dairy industry samples (n = 201).Ranking of 1 = most positive attitude, cow, and viewer experience, and perceived as most common; 12 = most negative attitude, cow, and viewer experience, and perceived as most uncommon.Multiple videos occupying the same ranking indicates a tie.When a line crosses another line, it indicates a change in rank.For example, for the public sample, the scenario Easywalk1 (red) received the most positive attitude and cow experience scores (tie), was perceived as the most common, and received the second-most positive viewer experience score.
Robbins et al.: PERCEPTIONS OF COW HANDLING

Table 1 .
Descriptions of dairy cattle-handling scenarios shown to industry and public participants

Table 1 (
Robbins et al.: PERCEPTIONS OF COW HANDLING Continued).Descriptions of dairy cattle-handling scenarios shown to industry and public participants 1Subjective classifications (POS = positive or neutral; NEG = negative, subdivided into NEG1 and NEG2 based on greater aversiveness in the latter) used to select the scenarios presented to ensure variety.Videos collected by the research team were filmed on commercial dairy farms with permission.Industry training videos and animal advocacy group exposés were publicly accessible online.All human faces and company logos were blurred in the video clips. 2 Text in brackets provided to public sample only.

Table 2 .
Outcome measures used to assess responses to 12 dairy cattle-handling scenarios.Sections of questions beneath each video appeared in the order presented below 1

Table 3 .
Robbins et al.: PERCEPTIONS OF COW HANDLING Demographic variable coding used in inferential statistical tests

Table 4 .
Robbins et al.: PERCEPTIONS OF COW HANDLING Demographics for both Wisconsin general public (n = 136) and US dairy industry (n = 201) samples 1 1 Some categories do not sum to 100% due to rounding.2Responses of slightly, moderately, or extremely badly were combined into badly; slightly, moderately, or extremely well were combined into well.