Exploring personality assessments of animals in human care: a case study on polar bear ages, hierarchy, and behavior

Introduction

Personality research in non-human animals (hereafter animals) is a growing field of interest for understanding and quantifying consistent individual differences in behavior or stress responses (Budaev and Brown, 2011). Specifically for animals in managed care, measuring personality can help to shape our understanding of the needs of each individual (Whitham and Washburn 2017; Kagan and Veasey, 2010; Gold and Maple, 1994). This knowledge can play a major role in improving their welfare and overall well-being (Wolfensohn et al., 2018). For instance, knowledge of an animal’s personality can reveal the social structure within groups of conspecifics, making it easier to reduce conflict and increase social wellbeing (Williams et al., 2019). Furthermore, these assessments can be used to inform compatible breeding matches, something that is of great importance for ex situ management programs (Carlstead et al., 1999; Williams et al., 2019). Quantifying the unique needs of each individual also offers the opportunity to optimize their welfare and wellbeing through such things as individually tailored choice and control or specific enrichment activities (Tetley and O’Hara, 2012). If housing or management does not meet these biological or social requirements, it can lead to boredom and stress in the individual (Blanchard et al., 2001; Burn, 2017; Meagher, 2019). This can lead to health issues and unwanted behaviors, including compulsive stereotypic behaviors like pacing or overgrooming (Wolfensohn et al., 2018).

Inter-individual differences in personality tend to remain stable and repeatable over several years (Twiss et al., 2020; Yoshida et al., 2016), but not necessarily over an individual’s lifetime. Stability in personality dimensions is often observed within a single life stage, such as adolescence or adulthood (sexual maturity), but can change over the lifetime of an individual because of major life events, the transition from one life stage to another (Bleidorn and Hopwood, 2019; Cabrera et al., 2021; Rohrer and Ferkin, 2020; Stanley et al., 2017), or due to maternal effects for species with parental care (DeRango et al., 2019). Each stage of life may require different behavioral priorities and ways of interacting with the environment and conspecifics (Groothuis and Trillmich, 2011). Studying individuals over time is essential to determine whether personality changes occur in all species as they age and to obtain a better understanding of how animal personalities are maintained (Cabrera et al., 2021). Knowledge on change points in personality offers the potential to predict changes in behavior or social structure (i.e., the hierarchy of the group) for animals in human care.

One approach to quantifying personality is the five-factor model of personality. This personality assessment has been tested and used for several animal species, including primates (Freeman and Gosling, 2010), dolphins (Tursiops truncatus, Highfill and Kuczaj, 2007), hyenas (Crocuta crocuta, Gosling, 1998), and bears (Ursus arctos arctos and Melursus ursinus inornatus, Quintavalle Pastorino et al., 2017). The five factors (hereafter referred to as traits) scored in this model are openness (also referred to as openness to experience elsewhere), conscientiousness, extraversion, neuroticism, and agreeableness (McCrae and Sutin, 2018). As defined by Quintavalle Pastorino et al. (2017) and Birgersson (2011), openness has been correlated with exploratory behavior, while conscientiousness is often linked to vigilance behavior. Behaviors such as vocalizing and locomotion have been found to be associated with the trait extraversion, while aggressive and stereotypic behavior have been better correlated to the trait neuroticism. The trait agreeableness has been generally linked with positive social behavior (i.e., affiliation; Birgersson, 2011; Quintavalle Pastorino et al., 2017).

Understanding personality and individual differences is particularly important with bear species, as they are prone to exhibit stereotypical behavior in human care (Quintavalle Pastorino et al., 2017). Wild polar bears (Ursus maritimus) spend large portions of the year as solitary animals (Nowak, 1999). However, in some regions they also are found interacting intensely (Latour, 1981). In zoological facilities, polar bears are generally housed solitarily or in breeding pairs. However, in cases where bears are housed in small social groups (Renner and Kelly, 2006), quantifying individual differences in personality can be even more important. Previous research at the Assiniboine Park Zoo (Winnipeg, Canada) used a modified five-factor personality assessment of Goldberg (1990) and Highfill and Kuczaj (2007) to create a version applicable to polar bears. This assessment demonstrated promise as a reliable way to measure individual differences in polar bear personality dimensions, but highlighted that the scores for extraversion, agreeableness, and conscientiousness increased over time (Chant, 2021; Highfill et al., 2010). Conversely, other studies indicate that certain aspects of animal personalities remain static over time (Cabrera et al., 2021). Changes in personality could also influence the group’s hierarchy, as for example, agreeableness is directly related to social behavior (Quintavalle Pastorino et al., 2017; Milewski et al., 2022). No research has yet been done on the relation between the five-factor personality assessment and hierarchy in polar bears. Moreover, the ability of the assessment to predict patterns of behavior has not yet been established, despite the critical role it could play in optimizing the welfare and wellbeing of bears in managed care.

The analysis presented here examined the utility of a modified 5-factor personality assessment for polar bears. The first objective was to investigate the relationship between polar bear age and personality traits, with a special focus on the influence that life stage changes may have on these traits. We then aimed to evaluate the influence of polar bear personality on social hierarchy. And finally, we investigated how polar bear behavior and activity budgets, and the behavioral diversity expressed by each individual correlated with the five personality traits. Given previous work, we would expect that personality may show marked shifts as bears reached sexual maturity, that personality may have a strong effect on hierarchy, and that personality may drive the specific expression of behaviors and the diversity of those behaviors therein. Overall, we aimed to gain a greater understanding of polar bear personality and to explore the results to make recommendations on implementing polar bear personality assessments and identifying any additional knowledge gaps for bears in managed care.

Methodology

This study was reviewed and approved by the Research Ethics Review Committee of Assiniboine Park Zoo (2024-S04).

Study population

This study was conducted in Assiniboine Park Zoo (APZ; Winnipeg, Canada), where personality assessments have been completed for all present bears since 2018 (n = 9 bears). All polar bears currently or previously residing at APZ were either orphaned cubs taken in at approximately 11 months of age or were young bears involved in human-bear conflict and deemed non-releasable by wildlife management authorities in Manitoba. Suitable data on personality was therefore available from nine individuals, six of whom had been housed at the zoo for the full 7 years of the study (2018 until 2024; Table 1). Timing of life stage transitions for each bear are also noted in Table 1 and discussed later in the methods.

Table 1. Study animals.

Overview of all polar bears included in this study, including their sex, estimated year of birth, the approximate timing of sexual maturity, other notable events, and the total number of years of available personality and hierarchy data. Personality data were collected between 2018 and 2024. Data on hierarchy were collected between 2020 and 2024.

This individual was included in the behavioral observations that took place for this research.

This individual was not included in the analysis of the relation between hierarchy and personality.

Assiniboine Park Zoo is home to two polar bear habitat spaces. The Journey to Churchill exhibit space, currently houses five bears and encompasses an area of approximately 8,400 m². The main enclosure consists of three areas with access to indoor holding connecting them. Outdoor areas in this habitat feature several caves, varying topography and substrates with areas at or above the public viewing level, and access to three pools of varying sizes. The secondary enclosure is part of the Leatherdale International Polar Bear Conservation Centre (LIPBCC) at APZ and has not been accessible to the public since 2020. One bear was housed solitarily within this space within adjoining portions of this habitat, while also having free access to indoor holding. In total, the LIPBCC outdoor areas are comprised of three adjoining areas, with periodic changes to access. The outdoor habitats at LIPBCC are approximately 1,450 m² in size and feature rock formations and branches as well as filled pools during the summer months.

Personality assessments

The personality assessments were completed by animal care professionals at the APZ annually between July and October between 2018 and 2023 to provide a snapshot of the personality of each bear. An additional assessment was completed in April of 2024 to overlap the behavioral observations completed in this study (see below). This resulted in personality assessments on all nine polar bears with ages varying from 2 to 13 years old (Supplementary Materials A-1).

The personality assessment consisted of a questionnaire that was completed for each bear individually (Table 2). The questionnaire contained six statements for each of the five personality traits: openness (O), conscientiousness (C), extraversion (E), agreeableness (A), and neuroticism (N). Each of the traits were assessed with three positively formulated and three negatively formulated statements (indicated with a + or a – in Table 2). The statements were rated on a seven-point scale, where a rating of 1 on the positive statements meant that the statement is very accurate for the individual, a rating of 4 is neutral and a rating of 7 is inaccurate. For the negative statements, the scale was reversed, so a rating of 7 represented a very accurate description. All animal care staff caring for bears were asked to complete the survey each year and to make their assessments independently of other staff.

Table 2. Five-factor personality assessments for polar bears.

The five personality traits (openness –’O’, conscientiousness –’C’, extraversion –’E’, agreeableness –’A’, and neuroticism –’N’) and their corresponding statements that were used for the personality assessment of the polar bears at Assiniboine Park Zoo (modified from Highfill and Kuczaj, 2007). Each personality trait was scored based on six statements, of which three are formulated positively (indicated with a ‘+’) and three are formulated negatively (indicated with a ‘–’).

Item analysis using Cronbach’s alpha (Livingston, 2006; Tavakol and Dennick, 2011) was used to test the coherence of statements used to assess each trait. A score of 0.7 or higher was considered reliable with a high degree of coherence between statements within a personality trait, while a score lower than 0.7 was not considered to be reliable and suggests a lack of coherence for statements within a trait (McCrae et al., 2011).

To obtain an average score for each personality trait, the scores of all statements were reversed relative to the rating (i.e., a rating of 1 became a score of 7, and so on). The average score for a given trait was then calculated across all six statements. This resulted in a score of 1 to 7 for each personality trait for each personality assessment made by an animal care professional. An average score of 1 meant that a personality trait did not accurately describe the animal in question which therefore resulted in a lower overall score for that personality trait, whereas a score of 7 indicated that the trait accurately described the animal and resulted in a higher overall score for a given personality trait.

Intraclass Correlation Coefficient (ICC) was used to calculate the inter-rater reliability of the personality assessment scores over the years (2018-2024) for each bear. A score of 0.5 or higher indicated consistent responses among raters for an individual bear (Benchoufi et al., 2020; Koo and Li, 2016). The repeatability of individual trait scores were assessed using the ‘rptR’ package in R (Stoffel et al., 2017). Repeatability (R) and confidence intervals were calculated using 1000 parametric bootstrapped samples with fixed effects of individual rater ID and bear age as potential confounds. Significance of repeatabilities was assessed through likelihood ratio tests. Individual repeatabilities (R_i) were extracted following (DeRango et al., 2019).

Age

Individual bear ages were known across the study period and any changes in life stage are noted in Table 1. Transitional life stages are periods when there are significant changes to the physiological state of an individual (Freedman et al., 2003; Wisdom et al., 2000). These stages could include weaning, sexual maturity, breeding, or moving into different social or environmental circumstances. None of the bears in this study experienced the transition from cub to adult during the time the personality assessments were collected. Previous research has shown that polar bears can reach sexual maturity as early as 3 years of age (Ramsay and Stirling, 1986; Richardson, 2014). Wild studies suggest that female polar bears are estimated to first reproduce at around 5 years of age, whereas males reach this at around 9 years of age (Atkinson et al., 1996; Ramsay and Stirling, 1986, 1988). However, hormone analyses of captively managed bears indicate maturity occurs much earlier before successful breeding observed in the wild (Wojtusik et al., 2024). Sexual maturity in the female bears for our study was determined by observed estrus cycling by animal care professionals. All male bears included in this study were neutered when they were four years old, apart from Storm who was neutered at five years of age (Table 1). To our knowledge, there have been no studies on the effect that neutering has on polar bear development and life stages; however, at APZ, aggression was reduced but many typical breeding behaviors are still observed (i.e., pursuit/courting, mounting; APZ unpubl. data).

To investigate the effect of age on each of the five personality traits, a linear mixed model (LMM) was used (R package ‘nlme’; Pinheiro & Bates, 2025). The median scores of all individuals (n = 9) for each of the five personality traits for all years available (2018-2024) were used to test whether the age of the individuals had an effect on the personality trait. If age had a significant effect on the personality trait, a post hoc test with Bonferroni correction was used to see at which age these changes occurred and were compared to any known life stage changes noted.

Social hierarchy

Information on the social hierarchy for the polar bears at APZ prior to 2020 was excluded as group composition often changed in 2018 and 2019. As such, one individual was excluded from this analysis (n = 8). Notes from the animal care professionals on the bears’ behavior are regularly added to an internal database (ZIMS, Species360 Zoological Information Management System, https://zims.species360.org/). To gather information about the hierarchy of the group, keywords related to dominant, submissive, or aggressive behaviors were used to filter and locate the notes about negative social interactions (Supplementary Materials A-2). As the personality assessment was conducted yearly between July and October, the notes pertaining to each year’s rank determination were utilized from the 1^st of November of the preceding year to the 31^st of October of the year considered for hierarchy (e.g., the personality assessments of 2020 used hierarchy notes from 1 November 2019 to 31 October 2020). Consequently, there was no overlap between the rank provisions for the different years in which the personality assessment was conducted.

To investigate the correlations between hierarchy and the five personality traits for each year, the median score of the personality traits was calculated as above (Hawkins, 2019). Hierarchy information was translated into a matrix for each year (2020-2024). This matrix plots the individuals against each other with the winner (i.e., the dominant bear of the interaction) in the row, and the loser (i.e., the submissive bear of the interaction) in the column. Each time a social behavior related to hierarchy or dominance was named in the notes, the mentioned winner got a point in their row in the column of the corresponding loser. Given the retrospective nature of this work, it is impossible to know if all possible interactions were documented by animal care professionals. We feel confident that the corresponding information provided an accurate, randomly sampled snapshot of behavioral interactions between all bears during available work hours.

For each year, the David’s Score (David, 1988) was used to establish the position of each individual in the hierarchy (i.e., rank among conspecifics; Gammell et al., 2002). Based on these results, the most dominant bear received a rank of 1. The lowest possible rank depended on the group size for that year (i.e., it varied between 5, 7, and 8). We modeled these ordinal hierarchy ranks using a cumulative link mixed-effects model with a logit link to explore the relationship between personality traits and hierarchy. The median score of each personality trait was modeled as a fixed effect, while each individual bear and year were included as a random effect. Hierarchical rank was treated as an ordered categorical response, and model parameters were estimated using Laplace approximation as implemented in the ordinal package in R (Christensen, 2023).

Behavior

To investigate the relationship between behavior and personality, only data on the personality and the behavior of the five bears currently housed socially in the main enclosure were used (n = 5; Table 1). The individual Star was excluded from the behavioral observations as she was housed solitarily during the observation period and could not provide any information on social interactions in our ethogram.

To relate polar bear behavior with the outcomes of the 2024 personality assessment, behavioral observations were collected by two observers (AM and MP) and recorded using ZooMonitor (Lincoln Park Zoo, 2022). For a period of three weeks (April 1^st to April 19^th, 2024), behavioral observations on the polar bears were conducted Monday through Friday between 10 AM and 4 PM. Each observation day aimed to have at least twenty 16-minute sessions (ten per observer) across all bears, which resulted in 27 hours of behavioral data per week. While these observations offered a short snapshot into polar bear behavior, this was meant to be an initial exploration into the relationship between personality and behavior and will not capture seasonal differences across a full year.

A scan sampling approach was used to obtain a general overview of the behaviors of multiple bears at once (Altmann, 1974; Dawkins, 2007), resulting in up to 10 intervals per session. At every interval (1 min per interval), the behaviors of all visible bears were registered by the observer following the ethogram outlined in Supplementary Materials Table A-S1 (following Quintavalle Pastorino et al., 2017; Skovlund et al., 2023). These behaviors were matched to an associated personality trait based on previous research (Birgersson, 2011; Quintavalle Pastorino et al., 2017) and the statements that were used to describe the trait in Table 3 (Highfill and Kuczaj, 2007). For example, the trait neuroticism was evaluated through a statement pertaining to aggressiveness, thereby associating the behavior ‘negative social’ with the aforementioned trait. If behaviors that do not correspond to the same personality trait took place simultaneously (e.g., walking and observing, respectively belonging to extraversion and conscientiousness), both behaviors were recorded. Moreover, vocalizing was added as an all-occurrence behavior, as it is a point behavior that can be easily overlooked when using scan sampling (Mann, 1999).

Table 3. Personality associations with behavior.

Classification of each of the behaviors into the five-factor personality traits (based on Birgersson, 2011; Highfill and Kuczaj, 2007; Quintavalle Pastorino et al., 2017; Skovlund et al., 2023). A (–) before the personality trait indicates that the behavior matches the negatively formulated statements of that personality trait, and a (+) indicates that the behavior matches the positively formulated statements of that personality trait. A full ethogram which defines these behaviors can be found in Supplementary Materials A.

Visitor areas were used for observations to minimize the effect of observer presence on the bears’ behavior (Soriano et al., 2013). Prior to the start of each session, the location of all bears was established, and two observation points were chosen based on the highest number of visible individuals. If an observation point was not used regularly, this observation point was prioritized in later sessions to minimize bias in behaviors observed for specific locations (e.g., the presence of a pool in that area). If within two minutes after starting the session all bears disappeared from view, and no bear was observed within the next three minutes, the session was terminated, and another observation point was chosen from which a new session could be started.

The interobserver reliability (Cohen’s kappa; k) for behavioral observations was calculated based on 3 hours of test observations to verify that the collected behavioral data used were reliable across observers. The resulting k was 0.899, indicating a near perfect agreement between the two observers (Bajpai et al., 2015; Fleiss et al., 2003).

To investigate the correlation between behavior and personality, the frequency of each behavior associated with an individual personality trait (i.e., O+, O-, C+, C-, etc.; Table 3) was calculated per bear per day. For example, both walking and swimming belong to the code E+, so the frequencies of the behaviors walking and swimming were added. Frequencies of behaviors belonging to multiple personality codes were added to all corresponding codes (e.g., social play belongs to A+ and E+, so the total scores of both A+ and E+ contained the frequencies of social playing). Next, the total scores of each code were divided by the sum of all the frequencies of all behaviors that were seen that day to generate an activity budget of each personality code. The mean daily activity budgets were calculated separately for both the negatively formulated statements and the positively formulated statements (Tables 2, 3). The full ethogram of all behavioral categories can be found within the Supplementary Material Tables A–S1. If the personality assessments are documenting consistent individual differences in behavior, we would expect positive correlations between the personality trait score and the time spent in the associated behavioral states.

Based on previous personality research (Birgersson, 2011), a Kruskal-Wallis test was used for each individual to see if there was a significant difference between the ten behavioral and personality codes. If a significant difference was found, a Dunn’s post hoc test was used to test which codes differed significantly. To investigate the strength of the correlation between personality and behavior, Spearman’s correlation was used based on the recommendation of Quintavalle Pastorino et al. (2017). Results of this analysis are explored in the Supplementary Materials.

We also explored the relationship between long-term trends in personality and overall behavioral diversity as this can be a useful way to track overall animal wellbeing of mammals in managed setting (Hall et al., 2021). To do this, we used all recorded behavioral activity budgets for each individual to calculate an overall behavioral Shannon Diversity index statistic (H), excluding behaviors related to rest and stereopathic behaviors, like pacing, as outlined in Hall et al. (2021). We used this Shannon Diversity index to explore patterns associated with median ratings of each personality dimension across all years measured (Shuert et al., 2020; Twiss et al., 2020). As we are limited in sample sizes, our discussion is focused on the relative trend across individuals.

Results

Over the years, a total of 15 animal care professionals filled in the personality assessments for the bears (Animal care professionals n₂₀₁₈ = 8; n₂₀₁₉ = 7; n₂₀₂₀ = 7; n₂₀₂₁ = 6; n₂₀₂₂ = 3; n₂₀₂₃ = 5; n₂₀₂₄ = 5). Across all nine bears, this resulted in a total of 311 personality assessments completed across all years, with a range of two to eight assessments per bear per year.

Item analyses were conducted for statements associated with each trait. The average Cronbach’s Alpha exceeded the recommended threshold of 0.7 in 59% of cases. The statements within the traits conscientiousness and extraversion were not highly correlated, as extraversion did not have a Cronbach’s alpha above 0.7 in any of the seven years and conscientiousness only in two of the seven years (2023 and 2024).

Age and personality

The personality assessment was consistent among raters over time as the raters did not demonstrate large fluctuations in how they scored each bear year-over-year (ICC for all individuals >0.8). When controlling for the age of the bear and the personality assessment rater, repeatability of personality traits was significant, but variable across individual bears (Openness R = 0.347, CI = 0.111 – 0.536, p < 0.001; Conscientiousness R = 0.319, CI = 0.099 – 0.529, p < 0.001; Extraversion R = 0.441, CI = 0.166 – 0.637, p < 0.001; Agreeableness R = 0.346, CI = 0.117 – 0.536, p < 0.001; Neuroticism R = 0.470, CI = 0.184 – 0.676, p < 0.001). Individual repeatabilities were also varied across individual traits and are presented in the Supplementary Materials (Supplementary Figures B–S3).

The age of the bear had a small, but significant effect on the personality traits openness (β_age = 0.08 ± 0.03, t = 2.47, p = 0.01), conscientiousness (β_age = 0.13 ± 0.02, t = 5.09, p < 0.001), and agreeableness (β_age = 0.08 ± 0.04, t = 2.01, p = 0.049; Figure 1). The lowest scores for all three traits were measured at age 2 (with a mean median of 4.49, 3.95, and 3.32 respectively; Supplementary Table B–S1). No significant effect of age was found for the personality traits extraversion (β_age = 0.047 ± 0.03, t = 1.44, p = 0.15), and neuroticism (β_age = -0.06 ± 0.0, t = -1.404, p = 0.16; Figure 1). However, all traits showed significant differences between at least two ages (these results can be found in Supplementary Materials B-1), though there was little consistency in the pattern of these differences across bears (Supplementary Figures B–S2). We were unable to evaluate the additional random effect of year on personality traits nested within each individual due to a failure of model convergence. A significant contribution to variance was explained by the random effect of each bear for all models, including openness (54.9%), conscientiousness (67.1%), extraversion (63.1%), agreeableness (55.3%), and neuroticism (57.9%), indicating individual variation exceeded any mean effect across individuals.

Changes in personality scores varied across individuals as they aged. While most bears showed some significant fluctuations in their personality scores over time (Supplementary Material Tables B–S2; Figures B–S2), Kaska was the only individual whose personality scores did not differ significantly as she aged. Baffin and Willow were the only individuals that showed fluctuations in personality scores that coincided with a change in life stage. Baffin’s conscientiousness and agreeableness score showed a strong increase between age 2 and 3, and a strong decrease was observed in his extraversion and neuroticism score during this same period. Willow, on the other hand, showed a significant difference in her conscientiousness and openness score between age 2 and 3. Other significant differences that were found could not be explained by changes in life stages. Full results of individual personality scores are provided in Supplementary Materials B-1.

Social hierarchy and personality

For each year, a minimum of 55 hierarchy-related notes were available to determine the rank (Total notes sampled n₂₀₂₀ = 55; n₂₀₂₁ = 70; n₂₀₂₂ = 56; n₂₀₂₃ = 109; n₂₀₂₄ = 64). The hierarchical structure of the group of polar bears exhibited general stability throughout the years (Table 4). Most individuals experienced minimal shifts, typically only ascending or descending by one position within the hierarchical rank. Notably, one individual, Nanuq, demonstrated a deviation from this stable trend, ascending by three positions between 2022 and 2023.

Table 4. Polar bear social hierarchy.

Hierarchy structure of the group of polar bears at Assiniboine Park Zoo, based on ZIMS animal care notes from 2020 until 2024. The individual that is listed first holds the highest rank. The individual, Nanuq, is highlighted in bold as the only bear to change relative hierarchy position over time.

Across all five predictors, none of the personality dimensions significantly predicted dominance rank (all p > 0.12). However, effect directions were consistent with biologically meaningful trends. Openness showed the largest positive effect (β = 1.46, SE = 1.03, p = 0.155), suggesting that individuals scoring higher on this trait tended to occupy higher dominance ranks, although the effect was not statistically significant. The effect size was substantive (odds ratio ≈ 4.3), but uncertainty was large. Conscientiousness had a negative but non−significant association with dominance rank (β = –1.15, SE = 0.99, p = 0.242), indicating that more conscientious individuals may be less dominant, though evidence for this effect was weak. Extraversion also showed a negative, non−significant trend (β = –0.94, SE = 0.96, p = 0.326). Agreeableness exhibited the weakest effect in the model (β = –0.55, SE = 1.27, p = 0.663). Neuroticism produced the strongest negative trend (β = –1.56, SE = 1.02, p = 0.128). Although non−significant, the effect size was sizeable (odds ratio ≈ 0.21), suggesting that bears that were scored higher in the neuroticism trait may be less likely to attain higher rank.

Random effects explained a substantial portion of the variance in dominance rank. Bear identity showed very large variability (SD = 5.06, variance = 25.64), indicating that stable individual differences strongly structured dominance outcomes, beyond what was explained by any single personality trait assessed here. In contrast, year−to−year variation was relatively small (SD = 0.57, variance = 0.33), suggesting that annual environmental or social conditions had minimal influence compared to intrinsic differences among individuals.

Behavior and personality

To investigate the correlation between behavior and personality, the bears were observed for approximately 27 hours per week for a three-week period (total 86 hours). Kaska’s behavior was observed for 106 sessions, Nanuq’s for 144, Storm’s for 114, Willow’s for 108, and York’s for 135.

Although no significant differences were found between the personality and corresponding behavioral codes, the expected trend was found for three of the five traits (Figure 2). Positive correlations were found between scores for both positive and negative trait associations for three of the personality trait categories (openness, agreeableness, and neuroticism) and the time spent in associated behavior categories. For all individuals with a higher score on the code O+ than on the code O- in the personality assessment meant that the individual spent more time on behaviors belonging to the code O +. The same was found for agreeableness and neuroticism. For the traits conscientiousness and extraversion, a reversed pattern was seen. For example, when an individual scored lower on C+ than on C- in the personality assessment, the opposite was seen in the behavior, meaning that an individual exhibited more behaviors that belonged to C+ than to C-.

Behavioral differences within individuals were variable. Kaska and Storm did not show any significant differences between the ten personality codes. All other remaining bears showed multiple significant differences in both their personality and behavioral codes. The significance between the personality codes and the behavioral codes was not always the same. For example, Willow scored significantly higher on the behavioral code A+ than A-, but there was no significant difference between A+ and A- in the personality codes. However, for some individuals and codes it did align. For example, Nanuq scored significantly higher on the code N- than on code A-, for both personality and behavior. Moreover, Willow scored significantly higher on O+ and C+, than on O-, A- and N+ for both the personality and the behavioral codes. Detailed results for individual bears can be found in Supplementary Materials B-3.

Overall, we found that behavioral diversity (Shannon-H) was similar among all bears (increasing behavioral diversity: Storm = 1.385, Kaska = 1.422, York = 1.445, Willow = 1.508, Nanuq = 1.529). Shannon-H was negatively associated with neuroticism and extraversion, while Shannon-H was positively associated with openness, conscientiousness, and agreeableness (Figure 3).

Discussion

Assessing the personality of animals in managed care offers a powerful tool to monitoring and improving an individual’s welfare and overall wellbeing (Chadwick, 2014; Finkemeier et al., 2018; Fureix et al., 2010; Rose et al., 2017). This is especially true for large carnivores, where complex physical and mental enrichment are necessary for optimal care and health (Clubb and Mason, 2007; Hoy et al., 2010; Skibiel et al., 2007). Here, we assessed if and how a five-factor personality assessment for bears in managed care was able to capture individual differences in personality, their stability over time, and how they might relate to aspects of hierarchy and observed behavior. The results showed that the personality scores accurately described individual differences in personality between bears. We also noted some differences between certain ages, depending on the individual and trait, but these did not consistently indicate shared developmental periods as they aged. Personality was not significantly indicative of position or movement within the hierarchy. However, while no statistically significant correlation was found between personality and associated behaviors, trends indicated that the measured personality differences corresponded to expected behaviors and demonstrated inter-individual differences in activity budgets and behavioral diversity.

The five-factor personality assessment was able to accurately describe inter-individual differences among the polar bears measured here. When examining the effect of age, we found that the majority of variance within these models was attributed to individual bears. While this meant that we were unable to find a consistent pattern in personality as the bears aged, it does suggest that the personality assessment is capturing individual differences among the polar bears included in our study. Notably, we saw differences among individuals in the variability and consistency of personality dimensions over time (see Supplementary Materials). Some individuals, like Kaska and Storm, showed consistency in most, if not all, dimensions in the personality assessment (Supplementary Tables B–S2). Other individuals, like Willow and Star, displayed greater variability in personality dimensions as they aged. This could indicate that this assessment is capturing other aspects of personality than originally anticipated, such as stress-coping styles or stress reaction norms (Koolhaas et al., 1999), where individuals may be found to be either consistent in their expression of behavior and physiology (e.g., proactive stress-coping) or more variable in order to match the physical environment or density-dependent effects (e.g., reactive stress-coping) as has been found in other species (Coppens et al., 2010; Mazza et al., 2019; Monestier et al., 2015; Shuert et al., 2020; Twiss et al., 2020). Monitoring resting heart rate variability has become a promising approach to determining personality dimensions along the pro-reactive continuum of stress-coping styles (Shuert et al., 2020; Twiss et al., 2021). This approach leverages differences among individuals in hypothalamic-pituitary-adrenal (HPA) axis pathways to quantitatively assess personality differences and stress-reactivity, the physiological mechanisms that underpin behavioral differences among individuals (e.g. Koolhaas et al., 1999). While this approach may overcome the issues in evaluating personality using solely qualitative assessments, it comes at a greater cost of requiring further animal manipulation to attach and remove monitors.

Differences in consistency or variability in behavior are likely why little evidence was found to suggest that changes in personality traits consistently aligned with critical life periods, such as reaching sexual maturity, within our small population of bears. Even though research suggests that life stages can cause noticeable changes within the measured personality of an individual (Bleidorn and Hopwood, 2019; Cabrera et al., 2021; Rohrer and Ferkin, 2020; Stanley et al., 2017), only one of the nine bears (Baffin) showed changes in personality that could possibly be explained by a transition between life stages. For three of the five traits, conscientiousness, agreeableness, and neuroticism, he showed a significant difference in his scores between ages 2 and 3 which we attributed to the transition between adolescence and maturity, after which these traits stabilized for several years. To add complexity, all males in this study were neutered around 4 years of age which may have impacted these results (Gosling, 2008; Stamps and Groothuis, 2009). In addition to this, we also possess no information as to the personality of their dams as all bears were brought to the facility from the wild between 11 months and 2 years of age (Owen, 2021). Early life experiences, such as rearing environments (Robinson et al., 2017), or maternal personality traits (DeRango et al., 2019), have been shown to impact personality development in other species and it is unclear to what extent this may have impacted the bears in our study. Further research is needed to assess the impact of development on personality in polar bears, which may be complemented by other markers, such as testosterone (Wojtusik et al., 2024).

Our expectations for assessing hierarchy were gathered from other species, such as dogs (Canis familiaris, Vékony et al., 2022), chimpanzees (Pan troglodytes, Anestis, 2005), Barbary macaques (Macaca sylvanus, Konečná et al., 2012), and domestic chickens (Gallus gallus domesticus, Favati et al., 2014), where personality differences were found to predict differences in rankings among individuals. Other studies present mixed results to these outcomes, as no significant difference was found in personality relative to hierarchy in female baboons (Papio hamadryas ursinus; Seyfarth et al., 2012) and various bird species (Sturnus vulgaris, Boogert et al., 2006; Poecile atricapillus, Devost et al., 2016; Estrilda astrild, Funghi et al., 2015). We found no significant correlation between personality and hierarchy in our small sample size of bears. Only one bear, Nanuq, went through a major change in rank throughout the duration of the study, which appeared to be unrelated to housing or social causes. Another bear (Kaska) was remarkably consistent in personality traits and remained stable in her ranking among individuals. In other contexts, these results may become clearer and present an opportunity for further study. For example, it was demonstrated that among other factors, contact seeking and social interactions influenced the social rank of males in bottlenose dolphins (Tursiops truncatus), whereas for females, factors such as playing with objects and avoiding humans impacted their social rank (Frick, 2016). Research on the relationship between hierarchy and personality has focused primarily on social species. Polar bears are primarily solitary (Nowak, 1999) during most of their life history. Given the variability of personality dimensions noted across individuals as they aged, it is possible that we are not capturing behavioral or physiological differences which may predict hierarchy. As a result of the mixed results found for predicting hierarchy and a lack of consistency across studies for the ways in which personality assessment data is collected and scored, we suggest that the use of other methods that capture aspects of personality will better elucidate inter-individual differences in personality.

Relating personality to behavior has largely seen mixed results across species. For example, a study on gorillas (Gorilla gorilla) did find a significant correlation between personality traits and behavior (Kuhar et al., 2006). Consistent individual differences in personality were also found to dictate gross differences in activity budgets for wild grey seals (Halichoerus grypus), but only with respect to rest and vigilance; nursing activity was consistent across all personality types (Shuert et al., 2020). More broadly, strong correlations between aspects of personality and behaviors or whole activity budgets seem to largely context dependent, where behaviors associated with energy gain (Careau et al., 2010), reproductive success (Twiss et al., 2012), environmental context (Monestier et al., 2015), or predation risk (Huntingford et al., 2010) often see the greatest correlations associated with aspects of personality. It is likely that the specific social housing conditions relaxes selection for gross differences in activity budgets, meaning that we could not capture statistically significant differences for behavior and personality associations. Our study was also limited to daylight hours during the spring of 2024, which only represents a portion of their daily activity budget and may have missed nuance related to these conditions as well as with and without the presence of the public. Despite these challenges, we noted trends for three of the five traits (openness, agreeableness and neuroticism). For all individuals, the positive codes of openness (O+) and agreeableness (A+) were higher than the negative codes of these traits (O- and A-), in both behavior and personality. More specifically, an individual that scored higher on behaviors belonging to the codes O+ than O-, also scored higher on O+ than O- in the personality assessment. Agreeableness and conscientiousness displayed a positive trend with behavioral diversity, while those with a higher neuroticism score tended to score the lowest behavioral diversity index (Figure 3). This largely suggests that overall behavioral diversity may better capture inter-individual differences relative to personality dimensions than correlations to single behaviors alone.

We also noted that behavioral codes did not appear to align with the personality traits of conscientiousness and extraversion. For all five individuals, time spent in behavioral codes C- and E- were higher than C+ and E+ and were associated with the opposite personality code across all individuals. The negative codes of both extraversion and conscientiousness consisted of, among other behaviors, resting and sleeping. The polar bears exhibited these behaviors frequently, which is consistent with the known tendency of polar bears to rest a lot during the warmer months (COSEWIC, 2008). However, the negative statements that were used on the personality assessment to score extraversion and conscientiousness were focused on how unexcitable and lazy an animal is. An animal that rests a lot is not necessarily unexcitable or lazy, as this behavior depends on factors such as the species being evaluated, and environmental influences (Cirelli and Tononi, 2008; Keene and Duboue, 2018). This may call into question whether resting behavior correlates with the negative statements of these two traits for polar bears and may only be appropriate in certain contexts. For example, polar bears use ambush as a hunting style which requires long periods of motionless waiting (Davis and Pagano, 2021). However, studies evaluating behavior of both captive and wild polar bears found that resting behaviors occupied a large portion of a daily activity budget, regardless of environmental contexts (Pagano et al., 2017). We observed similar activity budgets in our study here. Furthermore, other negative statements that are used to score conscientiousness focused on inconsistency and undependability. While we did not necessarily find consistent correlations across individuals, we argue that these assessments may still be used by animal care professionals to provide more insights into personality than can be achieved by solely ad hoc behavioral observations (Vazire et al., 2007).

Within and across individuals, we noted several interesting patterns and consistencies for assessed personality traits which provide useful landmarks for optimizing animal wellbeing. Across all years, we found that most individuals had a consistent score for openness and conscientiousness, generally only deviating in median score by +/- 0.5. This has been found in other bear species in previous work (Huber, 2010; Morić and Huber, 1989; Stirling et al., 2021), indicating that this may be a consistent trait for the species and may not be as useful for addressing individual needs. Other traits were found to be much more variable, such as neuroticism where the average score ranged from 1.88 to 6.22. Individual differences were noticeable in this trait with 57% of the explained variance captured across individual bears; some bears (such as Siku) were noted as being more aggressive and temperamental, while others (such as York and Nanuq) were found to be calmer and more easy-going (APC, unpubl. data). We also found that neuroticism was negatively correlated with behavioral diversity scores. These results suggest that differences within the neuroticism trait could be used to optimize management for the individual, and to predict how an individual will respond to new situations. For example, rhesus macaques (Macaca mulatta) that scored high on neuroticism reacted strongly to novel stimuli and were therefore more prone to anxiety and fear in new situations (Suomi, 1997). Moreover, the trait has been characterized by emotions such as anxiety, fear, irritability, and aggression in humans (Barlow et al., 2014; Costa and McCrae, 1992; Delval et al., 2024; Rothbart et al., 2001). Stress can be reduced in individuals that score high on the trait neuroticism by making changes in management gradually and maintaining established routines (Bassett and Buchanan-Smith, 2007). For other traits, such as openness, it is known that individuals with a higher score tended to exhibit more intelligence and curiosity than lower scoring individuals (Goldberg, 1990). Higher scoring individuals for openness could therefore benefit from providing sufficient stimulation from challenges or cognitive enrichment (e.g., Krebs and Watters, 2017; Powell and Svoke, 2008) to prevent the development of negative stereotypic behaviors (Shyne, 2006). Ultimately, we argue that these personality scores can be used to optimize management by addressing every individual’s needs, and subsequently improve animal wellbeing (Williams et al., 2019; Wolfensohn et al., 2018).

While we argue that this personality assessment provides a useful tool for assessing individual needs, we have provided some recommendations for its future use. Item analysis showed that the Cronbach’s alpha was low (<0.7 for all or multiple years) for two of the five traits, extraversion and conscientiousness, suggesting that there is weak correlation between the statements used to test these traits. To obtain an alpha higher than 0.7 for extraversion, we recommend four statements of this trait should be rewritten. The only statements that correlated with each other were 1 and 4 (respectively, assertive and timid), whereas the negative statements are supposed to be a counterpart to the positive ones. Item analysis showed that for polar bears the rating of how playful and active an individual is does not correlate with how excitable and vocal they are. Adult male and non-maternal female polar bears vocalize rarely, and vocalizations are usually related to courting or aggressive interactions (Derocher, 2012; Wemmer et al., 1976). In polar bears, therefore, the statement on how vocal an individual is may be a better fit for the trait of neuroticism than for extraversion. To accurately score extraversion, the statements will need to be reviewed and adjusted based on the specific ethology of polar bears. Extraversion is characterized by activity level, seeking stimulation, and the intensity of interpersonal interaction (Highfill and Kuczaj, 2007; McCrae and Costa, 2006). Whether an individual is assertive or timid does not correlate with this trait, since highly assertive individuals are not automatically more active or playful than timid individuals. The statements of this trait, especially the negative ones, might be better focused on how playful and active an individual is. Instead of assertive and timid, statements could focus on whether an individual initiates social play behavior. In addition, the highest possible alpha that could be achieved after removing statements for the trait conscientiousness was still below 0.7. This suggests that all six of the statements do not correlate enough for polar bears. However, both traits had a high repeatability, allowing the results of this study to still be considered reliable (Bruton et al., 2000; Senthil Kumar and Shahraz, 2023).

Apart from the traits conscientiousness and extraversion, this study showed a trend between behavior and personality was present and shows promise for evaluating personality in human managed settings. After some adjustments for extraversion and conscientiousness, we argue that this assessment is a useful way to quantify polar bear personality and better tailor care to individual needs of polar bears. A larger sample size of behavioral observations and individuals is needed to further explore correlation between behavior and personality in polar bears. Behavioral diversity may better capture the desire to tailor personality assessments to the needs of bears in human care, where it is widely accepted that environmental enrichment and animal wellbeing aim for a higher behavioral diversity index overall. Overall, this study found that dimensions of personality can be measured in polar bears using a simplified assessment. This also suggests that when using personality assessments across species, we recommend a pilot study to verify and optimize those results for each species.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fetho.2026.1867861/full#supplementary-material

Supplementary material Supplementary material at the source

Footnotes

Edited by: Yann Henaut, ECOSUR Conservation and Biodiversity, Mexico

Reviewed by: Scott Nunes, University of San Francisco, United States

Karen Lucchini, Federal University of Pernambuco, Brazil