Pongrácz & Dobos (2024) — Biological Reviews of the Cambridge Philosophical Society
Behavioural Differences and Similarities Between Dog Breeds: Proposing an Ecologically Valid Approach for Canine Behavioural Research
Published: August 26, 2026
A large proportion of the research on dog breed behaviour has a methodological problem that limits how useful its conclusions actually are — and a critical review of 97 publications has now identified and categorised exactly what that problem is. The findings matter not just for researchers but for anyone who uses breed behaviour research to make decisions about dogs: breeders, owners, trainers, and veterinary professionals. 🐾
Researchers Péter Pongrácz and Petra Dobos surveyed recent publications involving dog breed behaviour comparisons and applied a SWOT analysis framework — Strengths, Weaknesses, Opportunities, Threats — to three primary methods of breed selection used by researchers. The goal was not simply to catalogue which studies were well-designed but to identify when particular methodological approaches are structurally incapable of producing the conclusions they are drawn toward — regardless of statistical rigour — due to a fundamental mismatch between the breeds chosen and the research question being asked.
Three Breed Selection Methods and What Each Can and Cannot Establish
The first method identified across the 97 publications is convenience sampling — selecting whatever breeds are available and accessible for the study rather than choosing breeds that are specifically relevant to the hypothesis. Convenience sampling has obvious practical advantages: it is faster, cheaper, and easier to recruit than selecting specific breeds for hypothesis-driven reasons. Its structural weakness is that the breeds included may have no particular logical connection to the question being investigated, making the results difficult to interpret even when they are statistically significant.
A study that finds breed differences in a convenience sample has demonstrated that these particular breeds differ on this measure — but without a theoretical basis for why these breeds were compared, the result is difficult to connect to any broader claim about breed function, ancestry, or selection history. If the breeds chosen happen to vary on the outcome measure for reasons unrelated to the research question, significant differences can emerge that appear to support conclusions they do not actually establish.
The second method is hypothesis-driven ancestry-based sampling — selecting breeds based on their shared or divergent evolutionary history to test hypotheses about how domestication and selective breeding have shaped behaviour. This approach is theoretically grounded and produces conclusions that are biologically meaningful when the ancestry choices are well matched to the hypothesis. Its weakness lies in the accuracy of the ancestry classification itself: breed registries and public ancestry data are imperfect, and breeds whose current behavioural profiles have diverged from their ancestral classification through recent breeding history can produce misleading results when classified by ancestry alone.
The third method is hypothesis-driven functional sampling — selecting breeds based on their current functional role or purpose rather than their ancestry. This is arguably the most ecologically valid approach for questions about how selection for specific tasks has shaped behaviour, because it selects breeds based on what they are currently selected to do rather than what their ancestors were selected to do. Its limitation is that functional classification is also imperfect — breed registries do not always reflect actual working function, and many breeds occupy multiple or ambiguous functional categories.
When Method and Research Question Are Mismatched ⚠️
The review’s most practically significant contribution is identifying the specific conditions under which a particular sampling method is structurally unable to produce the conclusions being drawn — a mismatch between breed choice and research goal that no amount of statistical sophistication can correct.
A study investigating whether herding breed selection produced superior human social cognition that uses convenience sampling — with no principled reason for which herding breeds and which non-herding breeds were included — cannot establish that conclusion even if significant differences are found. The differences may reflect any number of confounded breed characteristics that have nothing to do with herding selection specifically. The methodology cannot isolate the variable the research question requires to be isolated.
Conversely, a study using ancestry-based sampling to investigate current functional behaviour may compare breeds whose ancestry predicts one functional profile but whose recent breeding history has produced a completely different one. The result accurately reflects the ancestry comparison but does not answer questions about current behavioural function.
The connection to findings covered earlier in this series is direct. The Morrill et al. Science study established that breed explains only 9 percent of individual behavioural variation. The GWAS research identified that behavioural genomic regions are not unusually differentiated between breeds, while morphological regions are. Both of those findings are easier to interpret knowing that much of the breed behaviour research preceding them used convenience sampling designs whose conclusions may have been built on mismatched methodology. The limited explanatory power of breed for individual behaviour may partly reflect the genuine biological reality, and partly reflect the poor ecological validity of studies that have shaped the field’s assumptions.
What Ecologically Valid Breed Research Looks Like 🐕
The review’s positive contribution alongside its critical analysis is the articulation of what ecologically valid breed behaviour research should look like — what the researchers describe as biologically relevant research questions matched with breed selection methods that can actually address those questions, leading to conclusions that are valid within the ecological context of the breeds studied.
For owners, the practical implication is a form of methodological literacy: not every study that finds breed differences in behaviour is telling you something reliable about breeds in general. The breeds compared, the reason they were chosen, and the alignment between the research question and the breed selection method all determine whether the finding can be generalised beyond the specific sample studied. A finding from a convenience sample tells you what these breeds do in this context. It does not necessarily tell you what breeds-that-were-selected-for-function do in contexts that activate that function.
At Zoeta Dogsoul, accurate understanding of the dog in front of you requires accurate understanding of what the research about their breed actually establishes. NeuroBond is built on seeing the real individual — and the real individual exists in a breed context whose research literature is now being held to a higher methodological standard. That standard benefits owners as much as it benefits scientists. 🐾
Source: Pongrácz, P., & Dobos, P. (2024). Behavioural differences and similarities between dog breeds: proposing an ecologically valid approach for canine behavioural research. Biological Reviews of the Cambridge Philosophical Society. Published August 5, 2024.
