Our study was approved by the management of the Centre Feron-Vrau at the Lille Catholic University (France), a group of five nursing homes including Saint-Antoine de Padoue where our observations took place. The authorization is based on a methodology similar to that of a previous research project conducted in 2021 at the Université de Strasbourg (UNISTRA/CER/2021-10). Residents and caregivers observed gave their verbal agreement to participate in the study. As for the cats, they have food and water ad libitum, reserved sleeping areas, enrichments (i.e., cat trees, carton boxes) in the unit, and access to veterinary care.

Established in 2017 in Lille (France), the Saint-Antoine de Padoue nursing home (private institution) decided to welcome cats after the admission of a resident who was forced to part with her cat. The resident’s physical and emotional discomfort led health professionals to support the inclusion of her cat in the unit. This decision subsequently allowed other cats to be welcomed into the institution’s closed units. The institution is home to 319 residents in 14 living units: eight conventional units (CU), four Alzheimer’s disease living units (ADU, two of which have one cat), one reinforced accommodation unit (RAU; welcomes residents with severe behavioral disorders—e.g., hallucinations, aggression, disinhibition—such as elderly people with bipolar disorder, Parkinson’s disease, or schizophrenia), and one living unit for disabled older adults (DISU, which has two cats). The data used in this study come from data collected in February 2023 in the DISU (Vesque-Annear et al., 2024).

A total of nine caregivers (two men and seven women) and 12 residents (four men, including the new resident, and eight women) were observed in the DISU. Information on the residents collected from the unit’s referent educator is summarized in Table 1. In this study, two European cats (2-year-old castrated males) were observed. These two cats, residents of the nursing home, were named Gizmo (ginger cat) and Felix (black and white cat).

The DISU is a restricted-access unit (with a badge) with 12 bedrooms whose doors are open and accessible to cats during the day. The unit has a living room with a kitchen and an outdoor terrace that is also accessible to the cats in the unit (Figure 1). Some enrichments were available: two cat trees and two carton boxes (50 cm × 50 cm reused and customized boxes with a blanket inside). Carton boxes are placed high up on a cupboard along the bay window facing the outdoor terrace.

The entire unit (cats, residents, and caregivers) was observed by one observer (HVA) for a total of 45 h over 30 days, divided into 7 weeks of observation between 13 March and 28 April 2023 (Vesque-Annear et al., 2024). A new resident arrived on 28 March 2023, 2 weeks after the start of the observations. The arrival of this resident was indicated by event (T) in Figure 2. To study the effect of this new arrival, the data were divided into three periods (T − 11 days: before the arrival of the resident, T + 10 days: first period after the arrival of the resident, and T > 10 days: second period after the arrival of the resident). Since the arrival of the new resident took place on the 12th day of observation, the observation days from d12 to d30 were divided into two periods (T + 10 days and T > 10 days) in order to obtain three periods that were relatively equivalent in terms of number of observation days. We voluntarily chose to keep the full 30 days of observation rather than arbitrarily shortening the dataset to create three strictly equal periods of 9 days each. Retaining all days of observation avoids introducing bias that could result from arbitrarily removing days or selecting them based on subjective criteria. Randomly excluding certain days could distort the dataset by unintentionally omitting relevant behavioral variations that naturally occur over time. In addition, reducing the number of observation days to fit equal time periods could lead to inconsistent results depending on which days are excluded, requiring multiple analyses to verify the impact of different selection choices. By retaining the full dataset, we ensured a more robust and reliable analysis, reducing the risk of selection bias and preserving the integrity of the temporal progression of behaviors observed in the unit. Temporal subdivision of the dataset into shorter intervals (e.g., 3−day periods) was considered, as several behavioral changes appeared to occur shortly after the arrival of the new resident. However, such subdivision would have produced intervals with very low numbers of events, particularly for infrequent behaviors and dyadic interactions. Preliminary tests showed that models based on these shorter windows lacked statistical power and often produced unstable or non−convergent estimates. For this reason, we adopted 10−day periods as a compromise that preserved analytical reliability while still allowing us to examine temporal changes in behavior.

Behavioral observations of the cats (including general activity and areas frequented) were recorded all along the three periods (T − 11 days, T + 10 days, and T > 10 days) following the all-occurrence sampling method (Altmann, 1974; for details, see Vesque-Annear et al., 2024).

The general activity of each cat is based on seven behavioral categories (resting, locomotion, play, avoidance, feeding, exploration, and maintenance; Stanton et al., 2015). Each category included behaviors coded as state behavior (expressed as a percentage of time spent performing the behavior) or event behavior (expressed as a number of occurrences). To clarify the behaviors described in this article, Supplementary Material 1.A, inspired by the ethogram of Vesque-Annear et al. (2024), specifies the coding of behaviors as state or event behaviors. In addition, two resting behaviors have been renamed for better explanation. The behavior “lying awake” was renamed “lying awake, relaxed”, and the behavior “lying with all four paws on the ground” was renamed “lying on his belly, in alert” in the present study.

The social proximities between unit members (caregivers, residents, and cat(s)) were recorded every 10 min following instantaneous scan sampling (Altmann, 1974). The position of each individual (i.e., cats, residents, caregivers) and their social proximity to others were hand-noted by HVA on the unit plan (similar to Figure 1). Physical proximities were noted visually and indicated on the same plan (Supplementary Material 1.B). A total of 110 scans per individual at T − 11 days, 100 scans per individual at T + 10 days, and 90 scans per individual at T > 10 days were recorded. Subsequently, social proximities were grouped into three categories: “in contact”, “nearby”, and “talking to/looking at each other” in order to visualize social networks (Supplementary Material 1.C).

Analysis and statistical processing were carried out using R software (version 4.4.2).

General activity and areas frequented by the cats: For each observation period, the proportion of time spent on general activity by both cats (i.e., resting, playing, avoidance, locomotion, feeding, exploration, and maintenance) and on visiting each area (i.e., living room, bedroom, corridor, outside) was calculated. These proportions were also calculated and detailed for each day of observation (Supplementary Material 2). To obtain these percentages, only state behaviors were considered. Resting behaviors were considered separately to illustrate the differences between lying asleep and awake resting behaviors. Apart from a few sequences of play, no affiliative behavior between the two cats was observed in this study. The time (in seconds) that Felix and Gizmo spent in each behavior category, resting behavior, and areas visited was recorded daily over the three periods of observation.

Generalized linear models (GLMs) were employed to assess the influence of the period (T − 11 days, T + 10 days, and T > 10 days) on the time spent by Felix and Gizmo performing each behavioral category and resting behaviors and on the time spent by the two cats in each unit’s areas. Days of observation during which no behavior was recorded were excluded from the analysis. These exclusions reflected when the cats were outside or in a bedroom with a resident and/or caregiver. For Felix, 8 days of observation were removed: 3 days at T − 11 days (14 March, 2:30–4:00 p.m.; 16 and 21 March, 10:15–11:45 a.m.), 3 days at T + 10 days (30 March, 10:15–11:45 a.m.; 31 March and 13 April, 2:30–4:00 p.m.), and 2 days at T > 10 days (14 and 28 April, 12:00–1:30 p.m.). On seven of these days, Felix was absent because he was outside the unit, which was not the case for Gizmo. Additionally, for the GLM on resting behaviors, 1 day of observation was excluded for Gizmo at T + 10 days (28 March, 12:00–1:30 p.m.: the day the new resident was admitted to the unit), as no resting behaviors were observed.

GLM tests adapted for data following a negative binomial distribution were conducted using the MASS package (Venables and Ripley, 2002). These tests evaluated the influence of our three periods of observation on the general activity (behavioral categories) of both cats and the areas frequented by Gizmo. To account for overdispersion and excess zeros in the data, zero-inflated negative binomial (ZINB) models were applied using the zeroinfl() function from the pscl package (Jackman, 2024; Zeileis et al., 2008). The ZINB models were specifically used to analyze the influence of the period (T − 11 days, T + 10 days, and T > 10 days) on the resting behaviors of both cats and on the areas frequented by Felix. Post-hoc pairwise comparisons were performed on adjusted means using Tukey correction for multiple testing (emmeans package; Lenth, 2024). These comparisons accounted for the interaction of the periods of observation with behaviors or visited areas. For each period, the mean time spent in each behavior and area frequented was indicated as follows: mean ± standard error of the mean [min; max]. A significance level of p ≤ 0.05 was used for all analyses. A probability between 0.05 < p < 0.09 was qualified as a tendency. Although the APA discourages qualifying results with p-values between 0.05 and 0.10 as “trends” or “tendencies”, this practice remains common and scientifically justified in several disciplines, notably in behavioral ecology, biology, and ethology (Nakagawa, 2004; Dushoff et al., 2019; Amrhein et al., 2019). Many researchers argue that an arbitrary threshold at p = 0.05 should not constitute a strict dividing line between significance and non-significance, and that it is meaningful to report near-threshold results as weak signals—provided they are interpreted cautiously and not ignored. As Amrhein et al. (2019) emphasize, p-values are better understood as continuous measures of compatibility with the null hypothesis, not as binary verdicts. In this context, reporting tendencies helps to contextualize results and foster nuanced discussion, especially in exploratory studies or those with small sample sizes, as is the case here with only two individuals. Furthermore, Dushoff et al. (2019) point out that rigid thresholds can lead to the loss of potentially informative patterns, particularly in studies with complex or naturalistic designs. Finally, Nakagawa (2004) highlights that the term “trend” is widely used in behavioral ecology, provided it is clearly defined in the methods and interpreted prudently, avoiding overstatement.

Social network analysis: For the visualization and statistical analyses of the social networks, an initial analysis was carried out on the 22 individuals present, including 11 residents (but excluding the new resident to better look at his impact within the network), 9 caregivers, and the 2 cats. Among the nine caregivers, three were present for less than 7.7% of the total observation time (≤3 h 30), and the results with these three caregivers contributed to an overestimation of the correlations obtained. Therefore, in order to remain consistent with the observation context, the visualization and statistical analyses were performed without these three caregivers. Then, our analysis included 19 individuals in the analyses, with 6 caregivers, 11 residents, and 2 cats.

Social network matrices were produced for the “in contact”, “nearby”, and “talking to/looking at each other” networks (Vesque-Annear et al., 2024) for each of our three periods of observation (T − 11 days, T + 10 days, and T > 10 days). The matrices included the number of social interactions recorded for each individual (caregiver, resident, cat on the x-axis) with the other individuals in the unit (caregiver, resident, cat on the y-axis). Correlation tests of the social network matrices at T − 11, T + 10, and T > 10 days were conducted using Socprog software (Whitehead, 2009) to assess changes in the social network structure (N = 19) across these three periods. Dietz R-tests (Dietz, 1983), which are rank-based and similar to Spearman’s correlation, were used to evaluate the effect of the period (T − 11 days, T + 10 days, and T > 10 days) on the structure of the network. Specifically, these tests assess whether individuals maintain a consistent number of interactions across periods while also examining the overall stability of the social network structure over time. This method is particularly suitable for comparing matrices, as it accounts for rank-order relationships between elements and is robust to non-parametric data (Sosa et al., 2021).

Visualization of the social networks “in contact”, “nearby”, and “talking to/looking at each other” was carried out thanks to matrices for each period T − 11 days (11 days), T + 10 days (10 days), and T > 10 days (9 days) using Gephi software (Bastian et al., 2009; version 0.10.1). On the social network visualization, each node represents an individual; the larger the node, the more the individual interacts with other individuals. The thickness of the link between two individuals reflects the amount of interaction. Finally, an individual at the center of the network interacts strongly with other individuals who themselves interact strongly with each other (Wasserman and Faust, 1994). For the “talking to/looking at each other” social networks, only interactions emitted by residents and caregivers to the cat were recorded. Gazes from the cat to a resident or caregiver were not recorded. The probability of observing these behaviors during scan sampling was relatively limited, considering that mutual gazes between humans and cats are rare: cats producing occasionally glances (less than 1 s) and gazes (a few seconds; Grandgeorge et al., 2020). The results obtained for the social networks “talking to/looking at each other” were indicated in Supplementary Material 3.

The weighted degree, a descriptive measure of the network defined as the total number of social interactions per individual, weighted by the frequency of interactions with each individual in the unit (Sosa et al., 2021), was recorded during each period of observation. This measure was calculated for all social networks analyzed, including “in contact”, “nearby”, and “talking to/looking at each other”. Spearman’s rank correlations were performed on the weighted degree to assess whether individuals within the unit (residents, caregivers, and cats) maintained consistent interaction patterns with the same partners across the three periods (T − 11, T + 10, and T > 10 days). A low correlation between two periods indicates a shift in the social network, suggesting that individuals change their interaction patterns, thereby reflecting a reorganization of social partners.

A total of 1,466 behaviors emitted by Gizmo were recorded during the 45 h of observation. Gizmo performed 380 behaviors (a mean of 35 behaviors/day) before the arrival of the new resident (T − 11 days), 405 behaviors (41 behaviors/day) during the first period following his arrival (T + 10 days), and 681 behaviors (76 behaviors/day) during the second period (T > 10 days). Details of Gizmo’s general activity and general daily activity over these three periods are shown in Figure 3 and Supplementary Material 2.A, respectively.

No significant interaction was measured between the time spent in exploration, in locomotion, in feeding, in playing, or in maintenance behavior and the period of observation (GLM tests: p > 0.05 and Tukey post-hoc tests: p > 0.05). Likewise, no significant interaction was found between the time spent in resting behavior and the period of observation (GLM tests: p > 0.05 and Tukey post-hoc tests: p > 0.05). However, a detailed analysis of resting behaviors revealed differences between our three periods of observation. For example, the time spent “lying awake, relaxed” was higher at T + 10 days (702 ± 308 s [0; 2,853]) compared to T − 11 days (204 ± 87 s [0; 724]) according to the GLM (p = 0.038; z = 2.07), although this difference was not confirmed by the post-hoc tests (Tukey: p = 0.221) (mean ± standard error of the mean [min; max]). Moreover, Gizmo spent 1.57 times less time in a lying asleep position at T + 10 days and 3.95 times less at T > 10 days, although this difference was also not significant.

The time spent “lying on his belly, in alert” behavior was significantly higher at T + 10 days (584 ± 348 s [0; 3,129]) compared to T − 11 days (80 ± 51 s [0; 567]) (GLM: p = 0.001; z = 3.27), with a tendency to be significant according to post-hoc tests (Tukey: p = 0.083). This increase was identified on the second day in the presence of the new resident (d2 in Supplementary Material 2.B). Another tendency for significance was found between T + 10 days and T > 10 days, reflecting a decrease in time spent in this behavior at T > 10 days (85 ± 35 s [0; 257]) (Tukey: p = 0.059).

Although there was no significant interaction between the time spent in avoidance and the periods T − 11 and T + 10 days, Gizmo spent on average 2.8 times more time in avoidance at T + 10 days (with 2.5 times more occurrences of this behavior) (Figure 3). This increase in behavior was mainly detected on the first day of the presence of the new resident (d1 in Supplementary Material 2.A). However, a significant interaction between the time spent in avoidance and the period of observation was seen at T − 11 days (289 ± 287 s [0; 3,158]) and T > 10 days (8 ± 8 s [0; 66]) (GLM test: p = 0.011; z = −2.56), confirmed by post-hoc tests (Tukey post-hoc test: p = 0.004). Post-hoc tests also showed a significant difference between T + 10 days (813 ± 527 s [0; 5,400]) and T > 10 days (Tukey post-hoc test: p < 0.001) for avoidance behavior.

Concerning the areas frequented by Gizmo, no significant interaction effect was found between time spent in common or locker room, corridor, or cat tree(s)/carton box(es) and the three periods of observation (GLM tests: p > 0.05 and Tukey post-hoc tests: p > 0.05). However, at T − 11 days, before the new resident arrived, Gizmo occupied three areas of the unit: the living room, the corridor, and the outside. At T + 10 days, Gizmo occupied a fourth and new area: the residents’ bedrooms (with or without the resident), where he spent 11.4% of his time (Figure 4). This occupation was mainly detected on the first day of the presence of the new resident (d1 in Supplementary Material 2.C). A significant interaction between the time spent in bedrooms and the period of observation was seen at T − 11 days (1 ± 1 s [0; 5]) and T + 10 days (616 ± 534 s [0; 5,400]) (GLM test: p < 0.001; z = 4.20), confirmed by post-hoc tests (Tukey post-hoc test: p < 0.001). This significant effect was identified between periods T − 11 days and T > 10 days (915 ± 506 s [0; 4,031]) (GLM test: p < 0.001; z = 5.27), confirmed by post-hoc tests (Tukey post-hoc test: p < 0.001). Scan sampling provided additional information on bedroom occupancy ([number of scans where the cat was in the bedroom/total number of scans recorded during this period] * 100). At T + 10 days, 11% of the scans recorded indicated that Gizmo was present in residents’ bedrooms, and 90% of these records indicated that he was alone at that time. At T > 10 days, 21.1% of the scans recorded Gizmo in the bedrooms. However, 89.5% of these records indicated that he was present in the resident’s bedroom (R43) with her.

Eventually, a significant interaction between time spent outside and period tended to be measured between T − 11 days (323 ± 195 s [0; 2,194]) and T > 10 days (941 ± 296 s [0; 2,308]) (GLM test: p = 0.083; z = 1.73), but this difference was not confirmed by post-hoc comparisons (Tukey post-hoc test: p = 0.681).

A total of 977 behaviors emitted by Felix were recorded during the 45 h of observation. Felix performed 480 behaviors (a mean of 44 behaviors/day) before the new resident arrived (T − 11 days), 170 behaviors (17 behaviors/day) during the first period (T + 10 days), and 327 behaviors (37 behaviors/day) during the second period (T > 10 days) after the new resident arrived. Details of Felix’s general activity and general daily activity over the three periods are shown in Figure 5 and Supplementary Material 2.A, respectively.

No significant interaction effect was measured between time spent in exploration, locomotion, feeding, or maintenance behavior by period (GLM tests: p > 0.05 and Tukey post-hoc tests: p > 0.05). In addition, no significant interaction was found between time spent in resting behaviors and periods (GLM tests: p > 0.05 and Tukey post-hoc tests: p > 0.05), but detailed observation of resting behavior indicated significant differences between periods (Supplementary Material 2.B). The time spent in standing behavior was significantly lower at T + 10 days (9 ± 3 s [0; 22]) compared to T − 11 days (70 ± 27 s [0; 197]) (GLM test: p = 0.006; z = −2.76), confirmed by the post-hoc test (Tukey post-hoc test: p = 0.016). The post-hoc tests also revealed a significant difference between T + 10 and T > 10 days, reflecting an increase in time spent in this behavior at T > 10 days (110 ± 40 s [0; 247]; Tukey post-hoc test: p = 0.03). Furthermore, the time spent in “lying awake, relaxed” behavior tended to differ between periods T − 11 days (788 ± 292 s [0; 2,455]) and T > 10 days (260 ± 156 s [0; 1,160]) (GLM test: p = 0.062; z = −1.86), but post-hoc comparisons did not confirm this tendency (Tukey post-hoc test: p = 0.374).

Moreover, a significant interaction effect between the time spent in avoidance and the period was seen between T − 11 days (79 ± 56 s [0; 452]) and T + 10 days (1 ± 1 s [0; 2]) (GLM test: p = 0.002; z = −3.14), confirmed by post-hoc tests (Tukey post-hoc test: p < 0.001) and reflecting a decrease of this behavior at T + 10 days. Post-hoc tests also revealed a significant difference between T + 10 and T > 10 days, reflecting an increase in this behavior at T > 10 days (76 ± 74 s [0; 517]; Tukey post-hoc test: p < 0.001). Finally, a significant interaction effect between the time spent in playing behavior and the period was defined between T − 11 days (175 ± 133 s [0; 1,059]) and T > 10 days (9 ± 7 s [0; 50]) (GLM test: p = 0.017; z = −2.37), confirmed by post-hoc tests (Tukey post-hoc test: p = 0.008) and reflecting a decrease of this behavior between these periods. An interaction effect also tended to be found between T − 11 days and T + 10 days (5 ± 5 s [0; 32]) (GLM test: p = 0.051; z = −1.95). This tendency became significant in post-hoc tests, reflecting a decrease in this behavior between T − 11 and T + 10 days (Tukey post-hoc test: p = 0.001). However, at T − 11 days, playing behavior was only obtained at d-2 and d-1, the 2 days preceding the arrival of the new resident (Supplementary Material 2.A).

Concerning the areas frequented by Felix in Figure 6, no significant interaction effect was found between time spent in the common or locker room, outside, bedroom, or cat tree(s)/carton box(es) by period (GLM tests: p > 0.05 and Tukey post-hoc tests: p > 0.05). However, an interaction effect between the time spent in the corridor and the period was detected between T − 11 days (275 ± 258 s [0; 2,848]) and T + 10 days (16 ± 11 s [0; 84]) (GLM test: p = 0.021; z = −2.31), but this difference was not confirmed by post-hoc tests (Tukey post-hoc test: p = 0.362). A significant interaction effect was also identified between periods T − 11 days and T > 10 days (62 ± 28 s [0; 210]) (GLM test: p = 0.016; z = −2.41), but post-hoc comparisons did not confirm this tendency toward significance (Tukey post-hoc test: p = 0.394).

Finally, although the time spent outside did not differ significantly between periods, Felix spent 1.77 times more time outside at T + 10 days than at T − 11 days. This was particularly the case on the second and third days (d2 and d3) after the arrival of the new resident (Supplementary Material 2.C).

The visualization and statistical results obtained for the social network “talking to/looking at each other” are described in Supplementary Material 3.

The innovative aspect of our study lies in its ethological approach, based on objective behavioral observation to assess cat welfare. Our method focuses on changes in natural behaviors (e.g., behavioral durations) in relation to the environment (e.g., spatial occupation). The cat’s use of space within indoor environments, often referred to as its “home range”, remains underexplored in the literature (Bernstein and Strack, 1996). However, our study provides useful contextual information for evaluating feline welfare, particularly in socio-medical environments.

In the literature, domestic cat welfare can be assessed using questionnaire assessments carried out by the owners (Vojtkovská et al., 2020). Tools such as the Feline Quality of Life Measure (catQoL; Bijsmans et al., 2015; Tatlock et al., 2017; Henning et al., 2023) or the Cat HEalth and Wellbeing questionnaire (CHEW) (Freeman et al., 2016) are often used in domestic settings, but these rely on owner interpretation and may be biased due to attachment or over/underestimation of specific behaviors. More objective methods are based on recording health and behavior indicators, such as coat condition (visual assessment of grooming quality, cleanliness, and coat texture; Zito et al., 2019) or postural indicators (e.g., position of ears, tail, or head; cat stress score; Kessler and Turner, 1997). However, these methods also have certain limitations. A degraded body condition may reflect prolonged discomfort and fail to capture short-term stress responses. A cat can receive good quality care, be in good body condition, and still manifest abnormal behaviors (Adamelli et al., 2005). Furthermore, recording the cat’s body positions (e.g., ears, tail) requires that the observer know the individual well and be relatively close to the animal. Such proximity can impact the behavioral reactions of cats, which are particularly sensitive to human attention and eye contact (Pongrácz et al., 2019).

In our case, the arrival of the new resident led Felix and Gizmo to adopt different adaptive strategies. Indeed, Gizmo showed increased alertness and avoidance behaviors, which may indicate anxiety (Kry and Casey, 2007) and/or an effort to cope with a stressful situation (Carlstead et al., 1993; Rochlitz, 2005). When confronted with novel situations, cats often display more hiding behaviors and either passive (freezing) or active (distancing) avoidance (Uccheddu et al., 2022). Notably, Gizmo exhibited these behavioral changes in a space he had rarely used before: the unoccupied residents’ bedrooms. These behaviors were mainly observed during the initial days after the new resident’s arrival and continued up to 10 days. This shift in space use could be a strategy to avoid the new stimulus by seeking quieter areas of the unit (Vesque-Annear et al., 2024). Bernstein and Strack (1996) suggested that such spatial shifts may be associated with disruptions in cat–cat interactions; our findings suggest this may also apply to human–cat interactions. It is possible that, in the face of environmental change, cats seek out specific resources, such as familiar olfactory cues (e.g., those associated with a preferred human; Bernstein and Strack, 1996). This may apply to Gizmo, who spent more time in the bedroom of his privileged resident (R43), in her company, after the new arrival. This social proximity could have acted as a long-term positive reinforcer, supporting adaptation. This aligns with the findings of Vitale Shreve et al. (2017), who demonstrated that cats preferentially seek social interaction with a human rather than playing with a toy or accessing their preferred food. Thus, considering the individual social preferences of cats could facilitate adaptation to change in institutional settings.

Considering Felix, activity levels declined following the new resident’s arrival (i.e., less time spent standing, avoiding, or playing; Zhang et al., 2022). While the reduction in avoidant behaviors may seem counterintuitive, this change is certainly related to Felix’s occupation of the outside area (the unit’s terrace). Occupying this area could be considered as avoiding the environmental change taking place in the unit. However, since events outside the unit were not recorded, it is unclear whether this behavioral shift was linked to specific resource-seeking and which one (e.g., calm, food, stimulation, or social interaction with residents of the neighboring unit).

Although some behaviors appeared more stable after 10 days, the persistence of significant differences across periods—particularly for avoidance, play behavior, spatial distribution, and proximity patterns—shows that the cats had not yet fully returned to their patterns observed before the arrival of the new resident. Living in the presence of a conspecific may have facilitated their adaptation, as cats living with other cats have a better quality of life than cats living alone (Adamelli et al., 2005). This is consistent with the visualization of social networks, which illustrates the closeness between the two cats. Therefore, the period T > 10 days should be interpreted as a phase of relative stabilization rather than complete normalization, as several individual behaviors and dyadic proximity patterns continued to differ from those recorded before the arrival of the new resident. While this relative stabilization suggests a recovery in welfare, it highlights the importance of monitoring animals following environmental disruption (however, the duration of this monitoring period should be confirmed by future studies).

To do this, concrete solutions based on the five pillars of a healthy feline environment (Ellis et al., 2013) can be proposed to facilitate the adaptation of cats when a new arrival occurs. In Gizmo’s case, increasing the presence of floor enrichment (e.g., carton boxes) in less frequented areas of the unit (e.g., residents’ bedrooms, locker rooms, caregivers’ office) could help him manage this new social environment (Kry and Casey, 2007) and reinforce his perception of control over the situation (Stella and Croney, 2016), thus increasing his chances of finding a safe place (pillar 1). In addition, it is important for caregivers to consider the social preferences of Gizmo (e.g., his relationship with resident R43). For example, by talking to this resident to involve her more in positive social interactions with this cat (e.g., playing with him, pillars 3 and 4) and preferably in her bedroom, a space with familiar and reassuring olfactory information (pillar 5). To ensure that both cats can maintain control over their environment, it would also be necessary to limit the movement of equipment or objects in the living room (e.g., moving armchairs and tables) that could be used for claw scratching or even as hiding places (under armchairs). This would allow them to maintain a constant location for the key resources that cats need (pillar 2). In Felix’s case, installing cat flaps could be considered, since this space seems essential to his welfare. However, access to the terrace is controlled by caregivers (i.e., using an access badge to open the door). Unrestricted access to this space would allow him to occupy the many safe hiding places in the vegetation on the terrace (pillar 1) in a space where olfactory information is better preserved (pillar 5). Indeed, this space, which is rarely visited by residents, is not cleaned daily by the establishment’s staff.

Finally, an important thing to do when adopting/placing a cat in a nursing home would be to systematically investigate their personality (Gartner and Weiss, 2013). According to Stella and Croney (2019), shy, calm, gentle, and fearful cats exhibit more hiding and feeding behaviors and take longer to approach humans than playful, active, and curious cats. Future studies should further investigate how personality traits affect feline welfare and adaptability in socio-medical environments.

Unlike dogs, which tend to quickly explore new situations, cats need more time to habituate (Uccheddu et al., 2022). This observation is supported by our findings. When the new resident arrived, cats had little or no social interaction with the members of the unit. In this specific socio-medical context, adapting to or initiating interaction with a new resident may be particularly complex. Indeed, cats prefer social interactions initiated by unknown humans who use visual and bimodal communication (de Mouzon and Leboucher, 2023). However, in our study, the living unit houses older adults with varied profiles and limited ability to communicate verbally and/or physically. The residents’ communication difficulties may have complicated the cats’ ability to take in visual information and impacted social interactions. This is consistent with the study by Mertens and Turner (1988), which highlighted that in a free situation, when a cat is in the presence of an unfamiliar human, the human is often the first to initiate interaction, which then facilitates feline social engagement. In this case, the new resident showed no particular interest in the cats and did not attempt to interact socially with them. The absence of mutual interest or ability to interact may have delayed the development of social interactions (Mertens and Turner, 1988). In our study, residents’ interest in cats was difficult to measure qualitatively using a questionnaire due to their limited verbal and physical abilities. Finally, the number of members in the household and therefore community life can impact a cat’s quality of life. Indeed, the most sociable cats, which do not exhibit abnormal behavior, live in small households (i.e., without children; Adamelli et al., 2005) whose members seem to show a stronger attachment (Stammbach and Turner, 1999). However, in a living unit with approximately 15 humans (residents and caregivers), the attention and time devoted to human social interactions certainly take priority over human–cat interactions. In addition, living in a community in a nursing home implies more constraints in caring for the animal (e.g., organizational constraints for overworked caregivers) and an environment that is less favorable to calm and routine (e.g., regular comings and goings of residents and caregivers, noisy environment, occasional visits from family members; Stammbach and Turner, 1999; Adamelli et al., 2005).

The research of Ines et al. (2021) defines five types of relationships between cats and their owners: the friendship, the codependent relationship, the open relationship, the remote relationship, and the casual relationship. In this study, we could assume that Gizmo has a friendly relationship with the members of this unit (i.e., a friendly cat that may consider his owner/privileged resident as a refuge, the latter having a strong emotional investment in the cat). This type of relationship suggests that the cat can live with a certain degree of independence from the owner and is more common in households with several cats.

On the other hand, it is more difficult to determine the relationship that Felix has with the members of the unit. His frequent visits to the outdoor area and quite possibly to the neighboring unit complicate our interpretation of the bonds he has with humans in the nursing home. However, we could assume that this cat has a casual relationship with the members of this unit (i.e., a friendly cat that does not seek long-term closeness with humans, prefers the outdoors to busy homes, may be absent for long periods of time, and may visit other homes).

However, given that Felix’s behavior has not been observed outside (i.e., on the terrace and most likely in the neighboring unit), it cannot be ruled out that this cat has an open relationship with the two units he visits. This relationship is described as independent and distant but getting along well with humans without needing to be close to them.

Social network analysis revealed that, unlike proximity or communicative interactions, physical contact interactions were more affected by the arrival of the new resident. Although their overall frequency remained stable, the identity of the interaction partners changed. This reorganization may be explained by the socio-medical context of the nursing home, where physical contact between individuals occurs mainly during direct care, such as washing, feeding, or mobility assistance (Garcia et al., 2011; Vesque-Annear et al., 2024). Before the arrival of the new resident, the unit had 11 residents and 3 caregivers on site per day. The integration of a new resident may have altered the distribution of care responsibilities, resulting in new resident–caregiver pairings and increased demands from the new resident, who was himself facing a period of transition. Additionally, such changes may have disrupted the social dynamics among residents. Older adults in vulnerable situations often rely on stable routines to cope with change (Bouisson, 2002), and some may respond by engaging in social interactions (Hunter and Gillen, 2009).

Despite its innovative approach, our study had limitations. Our research team could not control the timing of the new resident’s arrival. As a result, the observational periods were determined post-hoc, according to data availability. A more detailed analysis might have been possible by subdividing each period further for better differentiation of behavioral changes over time. Some behavioral adjustments occurred rapidly after the arrival of the new resident, and a finer temporal resolution might have captured these short−term dynamics more precisely. Although 3−day intervals would have been ideal for identifying immediate shifts, the small number of events per interval prevented robust statistical modeling. Future studies with denser sampling or continuous monitoring could apply finer temporal windows to better characterize short−term changes. Alternative approaches such as rolling−window analyses or Bayesian change−point detection may also provide powerful tools for detecting rapid behavioral transitions when larger datasets are available.

However, the observation periods provided important information. Indeed, it would be relevant to conduct future research involving continuous and longer observations throughout the day. For example, it would be interesting to observe the natural activity patterns of the two cats during crepuscular periods. Lastly, the study focused on only two cats, an advantage ethically (i.e., focusing on the animal as an individual), but a constraint methodologically (i.e., limited replicability and reproducibility). It appears necessary to repeat and continue studies on this subject, as little data remain in scientific literature. Finally, the permanent presence of animals in such environments remains underexplored, with only a few documented cases (Dosa, 2010; Tournier et al., 2020). However, further studies are needed, particularly considering new French legislative measures guaranteeing older people the right to be admitted to nursing homes with their pets. These new measures suggest that issues relating to wellbeing, quality of life, and the cohabitation of humans and animals in this context will become increasingly common.

The admission of a new resident set off an interconnected chain of changes linking feline behavior, space use, and human routine. Gizmo responded with more on-alert and avoidance states and relocated to the residents’ bedrooms, zones of low traffic and familiar odors. Felix, by contrast, decreased standing, avoiding, and playing inside while markedly increasing time outside on the terrace. These divergent spatial choices reshaped when and where caregivers and residents encountered the cats, and the altered human movements immediately fed back on feline options—a co-adaptive loop rather than a linear stimulus—response pattern.

Social-network matrices echoed this dynamic. The total volume of physical contacts held steady, yet partners were temporarily reassigned, showing that the system absorbed disturbance by flexibly reallocating ties. This plasticity without loss preceded the 10-day back to normal in behavior and the stabilization of the contact network, suggesting that network reshuffling can flag welfare disturbances earlier than body posture or health scores.

Seen through the One Welfare agenda, cats act as stakeholders whose welfare co-determines the quality of human care. Effective policy should pair i) architectural niches with graded levels of control—bedrooms, quiet rooms, terraces—with ii) participatory welfare audits that track key behaviors, spatial occupation, and contact patterns. By embedding objective behavioral observation and network snapshots into daily practice, nursing homes can move from isolated checklists to the active stewardship of a multispecies care ecosystem.