Beyond WEIRD Humans and STRANGE Dogs: Using Big-Team Science to Improve Generalizability and Reproducibility in Comparative Psychology

Generalizability Across Locations

Dogs inhabit every continent and almost every island on Earth (Wandeler et al., 1993), making them an accessible species to study worldwide. However, is a dog in New York City the same as one in Vienna, Kyoto, Kolkata, or Buenos Aires? Across the world, dogs vary substantially, making findings in one location less generalizable to others (Fig. 1).

OPEN IN VIEWER

Fig. 1.

Differences between dogs in the Global North and Global South.

One of the most obvious and potentially strongest sources of variation in dogs is the hundreds of distinct breeds. Breeds differ dramatically not only in morphology (a 75-fold difference between smallest and largest breeds) but also in behavior. While some breeds excel at detecting scents, others show higher levels of trainability, impulsivity, or problem behaviors (Pongrácz & Dobos, 2025). Variation in breeds alone probably makes dogs the most behaviorally diverse species on the planet. And breeds are not evenly distributed across the globe, leading to variation in breed composition and hybridization in different locations. Interestingly, countries may vary in the frequency of hybrids and mixed breeds, with potentially more in the Global South and the United States than in Europe.

Alongside pronounced differences between dog breeds, human guardians exhibit clear cultural differences within and between countries. The dog-human bond is a critical component of companion dogs’ behavior, with guardians shaping their lives from nutrition to socialization to training. There are massive differences between and within cultures in how dogs are perceived, valued, and treated (Serpell, 2004). Some herding and guarding dogs live outside with the livestock that they tend, whereas some urban “fur babies” live in luxury, with every need met. The intricate layering of dog and guardian diversity results in a level of variation not observed in other species.

This extreme variation observed between rural/urban dogs and across countries and cultures is rarely accounted for in dog behavior and cognition studies. Dozens of canine science research sites exist in North America, South America, Europe, Asia, and Oceania. Would we expect behavior in one site to generalize to another? Research teams in different regions likely recruit different subpopulations of guardians. For instance, the ManyDogs 1 study recruited guardians from 20 different research sites in North America, South America, and Europe (ManyDogs Project et al., 2023b). Yet the guardian populations differed across sites in terms of their age distribution (Fig. 2a). Although sites in New England had fairly uniform age distributions, New York, Budapest, and Messina skewed toward younger guardians, whereas Arizona skewed toward older guardians. Guardians’ age likely affects their dog’s behavior (e.g., older, retired guardians may spend more time with their dogs) and even which breeds they selected (e.g., older guardians may choose smaller, less energetic breeds). Thus, different subpopulations of guardians can result in different subpopulations of dogs being tested across research sites.

OPEN IN VIEWER

Fig. 2.

Guardian age distribution and study effects across research sites. The vertical bar graphs on the left show (a) the guardian age distribution for the 17 research sites involved in the ManyDogs 1 study that tested pet dogs (the three sites that tested working dogs without guardians are not shown here). The bars in these graphs represent the relative frequency of different age groups tested at each site; North American, European, and South American sites are shaded in blue, green, and pink, respectively. The dots and bars in the horizontal bar graph on the right show (b) mean differences and standard deviations, respectively, between conditions across research sites. In general, the variation within a site (standard deviation) was larger than the variance between sites (difference between the smallest and largest mean difference).

These research-site differences are important to account for because some dog-behavior studies have failed to replicate (Stevens et al., 2022). Although it is possible that unreplicated effects result from idiosyncratic designs or analysis, truly absent effects, or random error, another possibility is that effects found in one sample of dogs are indicators of socioecological differences that do not generalize to all dogs. That is, research sites may be testing different subpopulations and getting different results, which could be due to breed base rates in the population, different recruitment strategies of the researchers, or broader environmental and cultural factors such as demographic differences in the human populations. However, it remains unknown whether these population differences across sites translate to differences in behavior. A big-team science group studying infants found that variation within research sites exceeded variation between sites (ManyBabies Consortium, 2020), which seems to also be the case for ManyDogs 1 (Fig. 2b), despite the site differences in demographics.

Representativeness in WEIRD Contexts and STRANGE Samples

Similar to other fields, canine science has inherent sampling biases that influence which guardians contribute to research (e.g., Elston, 2021; Kaźmierczak et al., 2023). This seems partially caused by location. Canine research is often conducted at university campuses in the Global North, which have proximity biases for the nearby student population and accessibility barriers that can be impossible to overcome without a personal vehicle or pet-friendly public transit. Recruitment strategies also play a role, with some researchers leveraging social media platforms popular with younger generations, such as Instagram or TikTok, whereas others connect with older generations through dog interest groups on Facebook or via referrals within local training schools or breed clubs. Even within these communities, sociodemographics and individual beliefs about animal minds may drive self-selection for research participation. Guardians who are more interested in dog behavior, value novel enrichment opportunities, share a closer bond with their dog, or have invested more in training while also having the disposable time and resources will be disproportionately represented in research samples. Additionally, certain samples may favor some breeds because demographics and cultural trends can shape preferences for certain physical features or behavioral traits in companion dogs (e.g., Özcan et al., 2017). On the other hand, whereas well-socialized, confident dogs with highly invested guardians may be overrepresented, guardians with socially anxious dogs that find transportation stressful will be underrepresented. This suggests that researchers should collect and report standard demographic data about the dogs and owners that they study, as well as information about recruitment strategies. These data can then be compared across locations to account for potential site differences as well as to the local population demographics to assess local sample representativeness.

Because of these factors, only a narrow range of companion dogs and guardians are included in research at a given site. These constraints are similar to biases observed in human psychological research, in which findings are often based on readily available student samples rather than representative cross-sections of the population (Hanel & Vione, 2016). Therefore, the current state of knowledge in canine science does not explain dogs as a species and limits our understanding of human-animal interactions. These WEIRD biases in guardian participation are further compounded by their companion dogs fitting several of the categories described in the STRANGE framework (Webster & Rutz, 2020). Particularly relevant categories in the context of research on human-animal interactions include social background, rearing history, genetic makeup, and experience. Most dogs included in today’s canine research could be described as STRANGE in these categories, and this limited representativeness poses another threat to reproducibility.

Research subjects are frequently from recognizable single breeds or intentional hybrids originating in Europe or North America (e.g., Border Collie or Goldendoodle). These dogs have experienced selective breeding and genetic bottlenecks (Dutrow et al., 2022; Marsden et al., 2016) causing significant, breed-level differences in behavior and personality (MacLean et al., 2019). Further, companion animals’ interactions with their environment are a reflection of their human guardian’s preferences, socioeconomic status, and lifestyle. In contrast to this STRANGE subpopulation, the majority of domestic dogs in the world are free-living, freely breeding individuals whose patterns of behavior are determined by their need to access food, shelter, and mating opportunities (e.g., Majumder et al., 2014). Although free-living dogs make up approximately 75% of the world’s domestic dog population (Hughes & Macdonald, 2013), they are rarely studied outside of behavioral ecology and conservation contexts. As a result, we know little about normative dog behavior and cognition in the general population.

The final dimension limiting representativeness in canine science comes from the researchers themselves. Often from WEIRD cultures or trained in Western scientific traditions, canine scientists may find themselves perpetuating STRANGE-ness in their study populations because of their educational background or current research environment. Further, a researcher’s access to institutional resources as well as their personal connection to dogs may shape their choice of study population and methods, such as working in shelters, visiting guardian homes, or recruiting for in-lab participation. At a higher level, cultural values and perspectives also drive research priorities, define incentive structures, and influence methodological innovation, which do not represent the breadth of epistemologies on animal behavior and human-animal interactions. For example, a country’s emphasis on security may generate funds to study detection dogs, which focuses efforts on odor detection in a population of working dogs. To broaden the field, corresponding shifts need to happen in the funding landscape and general scientific community to create space for globally relevant and representative research.

Big-Team Science and Other Solutions

Big-team science has the potential to overcome many challenges of generalizability and representativeness within canine science. Big-team science is large-scale collaborative research conducted across multiple independent research sites. In comparative psychology, big-team science initiatives such as ManyDogs, ManyBirds, and ManyPrimates actively work to reduce sampling biases, increase transparency, and integrate diverse research perspectives to advance their fields (Lambert et al., 2022; Many Primates et al., 2019). In the context of canine science, leveraging the power of larger samples and increasing the diversity of sampled populations and researcher identities can address many of the issues identified above, in addition to the systemic issues, with justice, equity, diversity, and inclusion in research.

Big-team science provides a solution to low generalizability across research sites. Instead of sites conducting studies independently and consistently reinforcing their own results, big-team science allows data to be combined and applicable effects to be collectively investigated. In addition to testing for generalizability, large-scale collaboration allows features across sites that might account for differences to be tested and observed, which also facilitates metadata collection on recruitment and testing strategies and thus a broader assessment of general phenomena and smaller-scale investigations into what factors may moderate observed effects in a given population.

Large-scale collaborations can address fundamental limitations in participant representativeness. Initiatives such as ManyDogs encourage researchers to collectively develop best practices for recruitment, integrate new communication modalities and outreach tools, and innovate methods specifically designed to improve sample diversity for both guardians and dogs across a range of cultures and research contexts. For example, to complement in-lab studies, researchers could also test a subset of dogs in the home, or via online participation. These approaches reduce transportation costs and remove some accessibility barriers while increasing the possibility for anxiety-prone or fearful dogs to participate. Online methods may be particularly effective at increasing the representativeness of dogs and guardians; however, they necessarily limit the type of research that can be done. Notably, big-team science cannot solve all representativeness issues because guardian self-selection biases may persist, with participation skewed toward highly motivated individuals with overtrained dogs.

The international nature of big-team science facilitates the co-creation of globally relevant research questions, mitigates STRANGE factors in certain populations, and promotes researcher diversity and inclusion. Large-scale collaborative research builds up partnerships spanning the Global South and North, simultaneously improving scientific rigor and researcher equity and inclusion. In the special case of domestic dogs, we propose that multisite collaborative studies may be the key for understanding the species as a whole. Multisite studies open the door to comparing differences in the ethology of free-living dogs from different populations, diverse cultures, and habitats. Including dogs from different regions, such as sled dogs in Northern Canada, street dogs in Mexico or Italy, or free-ranging dogs in India or Morocco will advance canine science by directly addressing some of the STRANGE problems with companion dog populations. Big-team science creates unparalleled opportunities for addressing biases in canine science (e.g., Winder et al., 2025), increasing research reproducibility and representativeness.

Although big-team science offers exciting possibilities, it comes with a number of challenges (Forscher et al., 2023). Research funding is scarce for large, distributed initiatives, and research groups are generally expected to use their existing resources (i.e., space, supplies, and personnel) in multilab collaborations. Between the Global South and Global North there are marked resource inequities that impose additional barriers to inclusion. Big-team science projects must also move slowly and often do not align with traditional incentive structures in academia, making it particularly risky for early-career researchers or those facing high publication demands and low job security to invest time without quicker outcomes to advance in their careers.

Final Considerations

Canine science has severe limitations and biases, both in the researchers carrying out the work and in the populations being studied. In particular, generalizability is hampered by limited access to diverse samples, and the representativeness of the research participants is influenced by socioecological factors and biological traits. Current findings are difficult to generalize even within the studied populations (primarily dogs in the Global North) because of breed differences and cultural variation in guardianship and training norms. Further, the limited inclusion of guardians from diverse backgrounds and socioeconomic classes and narrow focus on companion animals prevents a deep understanding of domestic dogs as a species.

Big-team-science initiatives offer the potential for expanding collaborations and overcoming systematic issues with researcher and population diversity and inclusion. Although there is still much that can be improved, big-team science is starting to change the norms around research collaborations and open the door for underrepresented research populations and researchers to have a voice.

Recommended Reading

Arden, R., Bensky, M. K., & Adams, M. J. (2016). (See References). A review of canine cognition that calls for additional work that can identify and quantify sources of variability such as breed, training background, and other individual differences.

Bhadra, A., & Sarkar, R. (2023). A dog’s life in the human jungle. In J. R. Stevens (Ed.), Canine cognition and the human bond (pp. 63–90). Springer. A review of free-ranging dog social ecology and behaviour that highlights areas for future research.

Coles, N. A., Hamlin, J. K., Sullivan, L. L., Parker, T. H., & Altschul, D. (2022). Build up big-team science. Nature, 601(7894), 505–507. An opinion piece that highlights the ways that big team science can benefit the broader science community and ways that it can be better supported.

Udell, M. A. R., & Brubaker, L. (2016). Are dogs social generalists? Canine social cognition, attachment, and the dog-human bond. Current Directions in Psychological Science, 25(5), 327–333. A review of literature on the origins of canine social cognition and dog-human interaction that highlights gaps for future cross-cultural work to address.