Assessing personality traits in dogs : conceptual and methodological issues
Year: 2017
Type: dissertation
Abstract: Animal personality is defined by consistent between-individual differences in behaviour through time or across contexts. Behaviour is further organised into broader behavioural dimensions referred to as personality traits (e.g. fearfulness, aggressive- ness or boldness). While animal personality is a relatively new field, researchers have been interested in quantifying and predicting stable behavioural traits or dimensions in domestic dogs (Canis lupus familiaris) for over fifty years. Nonetheless, deciding which personality traits are most relevant or which traits behaviours reflect remains a difficult task for animal (as well as human) personality researchers. Largely, this is because personality is something we infer from behavioural data rather than directly observe, which depends on the conceptual and methodological approach taken. For dogs in particular, the predictive validity of personality assessments has been inconsis- tent, such as in behavioural assessments of shelter dogs. Moreover, there have been a diverse number of traits and behavioural dimensions proposed, with little consensus across studies on which traits are most relevant for describing dog behaviour.\nThis thesis evaluated conceptual and methodological issues of assessing personality and personality traits in dogs. In particular, the papers addressed key aspects of the statistical analysis of behavioural data on dogs for making inferences about personality and personality traits, drawing upon perspectives across both ethology and human psychology. The papers demonstrate three broad results.\n First, research to understand which personality traits underlie dog behaviour would benefit from moving from largely exploratory-based to hypothesis-driven frameworks. Personality traits in dogs are usually inferred by using exploratory latent variable sta- tistical models, such as principal components analysis, and studies have applied a mix- ture of latent variable models that have differing underlying assumptions. Confirma- tory, reflective latent variable models provide a more powerful framework for testing competing hypotheses about the latent structure of behavioural data in dogs and for verifying the robustness of the derived personality traits. Using data on inter-context aggressive behaviour towards people and dogs in shelter dogs, we found two, corre- lated latent variables: aggressiveness towards people and dogs, respectively. However, these posited traits failed to account for all of the co-variation between aggressive be- haviour across contexts, violating the assumption of local independence. Moreover, in- teractions between aggression contexts and the sex and age of the dogs demonstrated a violation of measurement invariance. That is, sex and age differences in aggressive behaviour could not be simply explained by differences in latent aggressiveness traits. The robustness and reproducibility of other personality traits in dogs could be verified by applying similar approaches to multivariate data.\nSecondly, dogs do not only differ in how they behave on average (i.e. personality), but in the amount they change their behaviour across time (behavioural plasticity) and the amount of day-to-day fluctuation around their average behaviour (predictability). By applying the framework of behavioural reaction norms, popular within behavioural and evolutionary ecology, we studied these different components of variation in dogs’ reactions to meeting unfamiliar people over time at a shelter. Accounting for indi- vidual differences in intra-individual behaviour (i.e. plasticity and predictability) in addition to personality improved the predictive accuracy of our results compared to focusing on personality variation only. The results also highlighted the importance of gathering repeated measurements on individuals when estimating behavioural varia- tion. Specifically, behavioural predictions at the individual level were highly uncertain compared to those at the group-level (aggregating data across dogs), since the amount of data available on each dog individually was often small. Together, these results em- phasised the benefits of longitudinal assessments of dog behaviour in shelters, and the importance of systematic modelling of both inter-individual (i.e. personality) and intra-individual variation in dog behaviour.\nThirdly, predominant approaches to conceptualising of animal personality traits are faced with a number of challenges. Inspired by recent work in human psychology, we elucidated how animal personality, and integrated behavioural phenotypes in general, can be re-conceptualised using a network perspective. The network perspective repre- sents the behavioural repertoire of individuals as a system of causally connected, au- tonomous behaviours. Behavioural dimensions or traits are, thus, viewed as emergent patterns of causally related clusters of behaviours, rather than separate underlying variables. We demonstrated the application of network analysis to survey data col- lected on behavioural and motivational characteristics of police patrol and detection dogs. Our analyses emphasised a number of close, functional relationships between variables consistent with previous research on dog personality, as well as unique in- sights from novel network statistics into the organisation of police dog behaviour. We highlighted the merits of this perspective for furthering work on the organisation of behavioural phenotypes and animal personality, and situating this research within work on a diverse range of complex systems across science.\nIn summary, this thesis has drawn upon advancements across ethology and human psychology to present novel directions for understanding personality in dogs. The work will be of benefit to researchers determining which personality traits explain individual differences in dog behaviour and those aiming to predict future dog be- haviour. Lastly, the results should stimulate a greater awareness of the conceptual issues involved in making inferences about personality in dogs and other animals.
Source: NVA (Nasjonalt vitenarkiv)
Author: Conor Goold
Institution: Norwegian University of Life Sciences
Language: English
Subfield: Animal Science and Zoology
Domain: Life Sciences
Keywords: dog personality, animal personality, personality traits, shelter dogs, aggression, fearfulness, boldness, latent variable models, confirmatory factor analysis, measurement invariance, local independence, interindividual variability, sex differences, age differences, predictive validity, human-dog interaction
Open Access status: green
OpenAlex ID: W2808226341
