The hormonal effects of owner-cat interactions on pet cats of different potential attachment styles

Abstract

Objectives:

The aim of the present study was to investigate how different owner–cat interaction types influence stress and owner-bonding in pet cats with different potential attachment styles, using salivary cortisol and oxytocin as physiological indicators.

Methods:

Twenty pet cats were classified into potential secure, insecure–anxious or insecure–avoidant attachment styles using the Secure Base Test. Owners engaged in four types of interactions with their cats at home on separate days. Saliva samples were collected before and after each interaction and analysed for cortisol and oxytocin concentrations. The effects of attachment style, interaction type and time on hormonal changes were examined.

Results:

Significant interactive effects of potential attachment style and time, as well as interaction type and time, were observed for both cortisol and oxytocin. Passive interactions were associated with increased oxytocin and decreased cortisol, whereas forced interactions showed an opposite pattern. Secure cats exhibited more positive hormonal responses to human interactions compared with insecure cats, who showed increased cortisol or limited hormonal changes. Insecure cats also showed higher basal oxytocin levels than secure cats.

Conclusions and relevance:

Human–cat interactions trigger changes in cortisol and oxytocin levels, depending on the interaction type and cat’s attachment style. These findings highlight the importance of considering individual cat attachment differences when promoting appropriate interaction strategies to improve cat welfare and human–cat relationships.

Introduction

Behavioural problems constitute major risk factors for cats being rehomed or relinquished.^{1 –3} Inappropriate human–cat interactions (e.g. restraint and uncomfortable petting) have been shown to cause stress ⁴ and increase aggressive behaviours in cats,^5,6 while implementing handling guidelines for humans increases affiliative behaviour and reduces aggression and signs of conflict in cats. ⁶ We recently identified that insecurely attached cats received more owner-forced interactions and exhibited greater behavioural issues. ⁷ These studies underscore the critical roles of appropriate owner–cat interactions in fostering secure cat–owner attachment and improved cat welfare.

Existing studies on human–cat interactions predominantly rely on observational metrics (e.g. frequency or duration of interactions and specific behaviours).^{8 –10} However, feline communicative behaviours can encompass multiple emotional states and be too subtle to detect for differences; hormonal measures have been suggested for additional validation. ¹⁰ Cortisol has been commonly used as a stress marker for welfare assessments in cats, although many factors (e.g. time of day, arousal state and individual differences) can influence its levels. ¹¹ In contrast, oxytocin can be used as a positive marker as it functions in facilitating social attachment. ¹² Canine research has indicated oxytocin-mediated positive feedback loops during affiliative owner–dog interactions,^13,14 but analogous research in cats remains scarce, potentially because of their underestimated social capacity and the challenges in sample collection. Meanwhile, secretion of oxytocin increases in response to stressful stimuli, and an anti-stress effect has been suggested based on evidence from human and rodent studies. ¹² One study found that cat urinary cortisol and oxytocin concentrations differed between social and non-social testing conditions with humans. ¹⁵ Questionnaire data from owners revealed that the frequency of tactile and auditory interactions, but not play and brushing or food interactions, positively correlated with cat urinary oxytocin concentrations. ¹⁶ However, urine samples may not reflect instant hormonal changes, while blood sample collection risks inducing stress in cats. ¹¹ By collecting saliva samples, we recently showed that securely attached cats who received fewer forced interactions from owners in daily life exhibited greater oxytocin release during the interactions compared with insecurely attached cats. ⁷ Yet, the direct connection between owner–cat interaction types and oxytocin and cortisol dynamics in cats remains unclear.

This study compared four different owner–cat interaction types (e.g. active, passive, forced and no interaction) for their effects on cortisol and oxytocin release on pet cats in their home environment. Potential cat attachment style was included as another main factor in the study, which reflects individual differences in cat–owner bonding that might influence cat preference and tolerance for different owner interactive behaviours.^7,8 With behavioural tests, such as the Secure Base Test (SBT), recent studies have shown that pet dogs and cats can establish attachment with owners in a way that is similar to the parent–child relationship.^{8,17 –19} Based on cat performance during the reunion session of the SBT (i.e. whether they consider the owner as a secure base and display a balanced pattern between proximity-seeking and environmental exploration), their attachment to owners can be classified as a secure or insecure type (i.e. anxious, avoidant and, in rare cases, disorganised). ⁸ Caretaker’s characteristics (e.g. care-taking and interaction style) can exert major effects on the attachment behaviour of the individual who receives the care.^20,21 Meanwhile, attachment security formed at an early stage can have profound impacts on the social and emotional health of the individual later in life.^21,22 Insecure attachment has been positively associated with inappropriate owner–pet interaction and behavioural problems (e.g. separation anxiety, aggression) in dogs and cats.^7,23 As a result of insufficient validation of the SBT in cats and ongoing debate around attachment theory, we use ‘attachment’ merely as a shorthand for describing the cat responses in the test, which may instead reflect general social support effects or individual personality differences – a point we revisit in the discussion.

Materials and methods

Overview

The experimental protocol was approved by the Animal Experiment Ethics Review Committee of South China Agricultural University (approval number 2024f020). The human–cat interaction and saliva collection was carried out in owners’ homes and the SBT was conducted in the behavioural observation room of the Companion Animal Science Laboratory at South China Agricultural University.

Subjects

Between September and December 2024, 20 neutered and physically healthy pet cats of different breeds (nine male and 11 female, mean age 38.4 ± 17.6 months) were recruited from Guangzhou, China through social media platforms (e.g. WeChat). The cats recruited were aged at least 12 months and have lived with their owners for ≥4 months to allow for the formation of stable attachment. ⁸ Cat information and original data can be found in Tables S2 to S5 in the supplemental material.

Secure Base Test

Cat attachment style was assessed using the SBT, referring to the previous research method. ⁷ The classification criteria and additional behaviour validation are shown in Table S1 in the supplemental material. Among the 20 participant cats, seven (35%) were classified as secure attachment style, seven (35%) as insecure–anxious attachment style and six (30%) as insecure–avoidant attachment style. Attachment type was included as a dependent variable for grouping cats in subsequent experiments.

Human–cat interaction

The owners were instructed to engage in four types of interactions (i.e. no interaction, active interaction, forced interaction and passive interaction) with their cats on separate days with a 7-day interval between sessions. The order of experiencing the four interactions was controlled for each cat to prevent any two from receiving the same sequence. The interactions occurred in a room in the owner’s home, without any on-site intervention by the researchers. Cameras were set up in the room to cover the entire interaction area for recording the entire process. The definition of different interactions (Table 1) were adapted from previous studies.^6,9,10,24 Each interaction lasted for 10 mins.^7,25 Food and toys were excluded in the interaction process to minimise their impacts on hormones. ²⁶ All interactions were scheduled for the evening, when the owners returned home and had not been in contact with their cats for an extended period.

Table 1.

Different human–cat interactions.

Interaction type	Definition
No interaction	The owner does not initiate eye or physical contact with the cat and provides no response to the cat’s approach.
Passive interaction	The owner remains stationary with restricted movement, does not initiate eye or physical contact, but responds positively and appropriately (through talking, making eye contact and petting while avoiding caudal areas) when the cat approaches, then ceases interaction when the cat leaves.
Active interaction	The owner moves freely without restricting the cat, actively talks to the cat, and initiates both eye contact and appropriate physical contact with the cat.
Forced interaction	The owner interacts with the cat in ways that oppose its will or restrict its movement (e.g. picking up or hugging), inevitably touching less preferred body areas (e.g. caudal area).

Saliva collection and hormone assays

Sample collection and processing folowed the previous method in our laboratory. ⁷ Saliva samples were collected 30 mins before and 5 mins after the human–cat interactions, as well as 5 mins before and after the SBT. Samples were stored at –20°C until further analysis. Within 30 days after sample collection, the samples were analysed for oxytocin and cortisol, following the instructions of the commercial kits (MM-67503O1 and MM-67503O1; Jiangsu Enzyme Immunoassay). The intra- and inter-assay coefficients of variation were 2.97% and 5.43% for oxytocin, and 2.45% and 8.39% for cortisol, respectively.

Data analysis

Data analyses were performed using the statistical software SPSS (version 26; IBM). Non-parametric analyses of cat behaviours for the classification of cat attachment style and parametric analyses of hormonal changes in SBT are shown in Figures S1 and S2 in the supplemental material. A generalised linear mixed model was applied to analyse the effects of time, cat attachment style and interaction type on salivary oxytocin and cortisol secretion in cats during human–cat interactions, with time (i.e. pre- and post-interaction), attachment style, interaction type and their interactions as fixed effects, and cat as a random effect. To remove the potential interference of baseline hormone levels, changes in oxytocin and cortisol (i.e. post-interaction value subtracting basal value) in cats with different attachment styles in response to various interaction patterns were further analysed using a generalised linear mixed model, with attachment style, interaction pattern and the interaction between the two as the main effects, and cat as the random effect.

Results

Effects of interaction type, potential attachment style and time on salivary hormones

Cortisol

Results were presented in the order of main effects, followed by two-way and/or three-way interactive effects. Effects that did not reach statistical significance or show a statistical trend are not reported in the Results section (the same applies to the ‘Effects of interaction type, potential attachment style and time on salivary hormones’ and ‘Interactive effects of potential attachment style and interaction patterns on the changes of salivary hormones in cats after human-cat interactions’ sections).

A main effect of interaction type on salivary cortisol in cats was observed (F_(3,136) = 13.33, P <0.001) (Figure 1(a)). Cortisol concentrations in forced interaction were, in general, higher than those in passive interaction (P = 0.002). There was a trend in the main time effect (F_(1,136) = 2.887, P = 0.092) (Figure 1(b)). An interactive effect between attachment style and time was also detected (Figure 1(c)). Salivary cortisol increased in cats with anxious attachment (P <0.01) but remained unchanged in cats with other attachment styles. There was also an interactive effect between the interaction type and time (F_(3,136) = 13.55, P <0.001) (Figure 1(d)). Specifically, the salivary cortisol concentrations of cats decreased after passive interaction (P <0.001) but increased after forced interaction (P <0.001).

Figure 1.

The main effect of (a) interaction type, (b) time and (c) the interaction between time and potential attachment style and (d) between time and interaction type on cat salivary cortisol concentration in human-cat interactions.

Note. #0.1<P <0.05; **P <0.01; ***P <0.001.

Oxytocin

There was a time effect on cat salivary oxytocin (F_(1,136) = 49.84, P <0.001) (Figure 2(a)), which significantly increased over time during human–cat interactions. Interaction type also had an effect (F_(3,136) = 13.33, P < 0.001) (Figure 2(b)). Specifically, oxytocin levels in forced interactions were significantly lower than those in passive interactions (P <0.05). The main effect of attachment style was also significant (F_(2,136) = 10.41, P <0.001) (Figure 2(c)). Oxytocin concentrations in the saliva of cats with secure attachment were lower than those of cats with anxious and avoidant attachment (P <0.001). The interaction between attachment style and time also influenced salivary oxytocin concentrations (F_(2,34) = 15.84, P <0.001) (Figure 2(d)). Cats with secure and insecure–anxious attachment styles, but not insecure–avoidant cats, exhibited an increase in salivary oxytocin (P <0.001). In addition, cats with secure attachment styles exhibited lower basal oxytocin levels compared with anxious-style cats (P <0.001) and avoidant-style cats (P <0.001). Interaction type and time also interacted on cat salivary oxytocin (F_(3,136) = 9.528, P <0.001) (Figure 2(e)). The increase in oxytocin levels in cats during human–cat interactions was observed in both passive (P <0.001) and active interactions (P <0.05), and a trend of increase was identified in no interaction (P = 0.07).

Figure 2.

The main effect of (a) time, (b) interaction type, (c) attachment style and (d) the interactive effect between time and potential attachment style, and (e) between time and interaction type on cat salivary oxytocin concentration in human–cat interactions.

Note. ^#0.1 < P < 0.05; *P <0.05; ***P <0.001.

Interactive effects of potential attachment style and interaction patterns on the changes of salivary hormones in cats after human–cat interactions

Cortisol

The main effect of interaction type and attachment type was significant on the change of cat salivary cortisol after human–cat interactions (F_(3,76) = 28.04, P <0.001; F_(2,78) = 8.47, P <0.001) (Figure 3(a) and (b)). There was also an interactive effect between interaction type and attachment style on the change of salivary cortisol (F_(6,73) = 2.32, P <0.05) (Figure 3(c) and (d)). During forced interactions, cats with anxious and avoidant attachment exhibited greater increases in salivary cortisol concentrations compared with securely attached cats (P <0.001). During active interaction, cats with anxious and avoidant attachment both exhibited greater increases in salivary cortisol concentration compared with securely attached cats (P = 0.02). The secure cats showed a decrease of cortisol when experiencing passive and active interactions than forced interaction (P <0.05). Both insecurely attached cats exhibited significantly greater increase of salivary cortisol in forced and active interactions than passive and no interaction.

Figure 3.

Main effect of (a) interaction type and (b) potential attachment style and (c, d) the interactive effect of interaction type and potential attachment style on the changes in salivary cortisol and the main effect of (e) interaction type and (f) attachment style on the changes in salivary oxytocin of cats after human–cat interactions.

Note. *P <0.05; **P <0.01; ***P <0.001.

Oxytocin

The main effect of interaction type and attachment type was significant on the change of cat salivary oxytocin after human–cat interactions (F_(3,76) = 8.99, P <0.001; F_(2,78) = 10.71, P <0.001) (Figure 3(e) and (f)).

Discussion

The SBT, although validated for evaluating dog–owner attachment, lacks a stranger control condition compared with the original Strange Situation test. ¹⁷ Consequently, the findings may reflect general social reliance on humans rather than specific attachment to the owner, potentially confounding attachment effects with cat personality traits and offering alternative explanations for some results. We address this concern early in the discussion to remind readers that it may influence how the findings are interpreted and how the subsequent discussion is read.

Potential cat attachment style and hormonal changes in the SBT

All cats showed a decrease in cortisol levels during the SBT (Figures S2 and S3 in the supplemental material). The high pre-test stress in the cats was likely due to travel, ²⁷ which subsequently declined as the testing proceeded. The securely and anxiously attached cats might effectively reduce their stress response with the release of oxytocin through social buffering, ²⁸ which depends on an individual’s active participation and positive perception of social interactions. ²⁹ Secure and anxious cats actively seek reassurance from their owners through the frequent exhibition of owner-proximity behaviour in the reunion session of the SBT, resulting in elevated oxytocin release. In contrast, for avoidant cats, their active avoidance of owners and suppression of emotional responses likely facilitated their habituation to the testing environment. This adaptive process, however, may have simultaneously contributed to their reduction in oxytocin levels, a mechanism consistent with the physiological desensitisation observed in avoidant humans under stress. ⁵

Hormonal changes in cats under different interactions

To our knowledge, this study is the first to directly evaluate the emotional valence of different types of owner–cat interactions in cats through the measurement of instant changes of salivary cortisol and oxytocin. Former studies often relied on urinary samples, which might not reflect instant hormonal changes,^15,16 and did not differentiate between different types of social interaction but employed correlation analysis between observed or reported interactive behaviours and hormones.^7,16 In our study, forced interactions inhibited oxytocin release and promoted stress responses in cats, indicating a negative valence of this type of interaction. By contrast, other interactions, especially the passive interaction, resulted in an increase of oxytocin and a decrease in cortisol, suggesting its effectiveness as a positively valenced interaction in inducing the relaxation response.^24,30,31 The results support the notion from previous reviews that the effect of social contact on oxytocin release and stress regulation depends on the nature of the interactions.^14,32,33 We also provided supportive evidence for the current cat interaction guidelines proposed by professionals, where forced interactions involving, for example, restricted movement or touching less preferred body areas, should be avoided and cat-initiated passive interactions are encouraged.^6,9,10,24

Hormonal changes during different human–cat interactions in cats of different potential attachment style

Insecure cats exhibited a high basal oxytocin level before the SBT and human–cat interactions than secure cats. A similar result was observed in our previous study. ⁷ The varied basal oxytocin might result from altered stress regulation in cats of different attachment styles. A human study showed that basal oxytocin levels were higher in participants experiencing more stress in their relationships and were associated with elevated cortisol levels. ³⁴ High basal oxytocin levels observed in anxiously attached humans were associated with hypersensitivity to social cues. ³⁵ Cats with insecure attachment were also shown to exhibit more behavioural issues, indicating potential experience of chronic stress.^7,36 In our study, even though baseline cortisol concentrations were comparable in cats with different attachment styles, insecure cats experiencing chronic stress might compensate by continuously hyper-activating their oxytocin system to maintain stress levels similar to those of secure cats, thereby exhibiting higher baseline oxytocin levels.^34,37 The low baseline oxytocin in secure cats instead indicated a more efficient (i.e. context-specific) and dynamic regulatory capacity of their oxytocin system. Consistent with this, securely attached mothers, even with lower basal levels, were shown to secrete more oxytocin when interacting with their infants, and the extent of oxytocin release was positively correlated with the strength of the mother–infant bond. ³⁸

Regardless of interaction type, cats with insecure attachment styles exhibited increased cortisol after interacting with owners. Oxytocin increased after interactions in both secure and anxious cats, but not in avoidant cats. The results indicate that cats’ attachment styles influenced their stress and emotional responses to owner-involved interactions, as further stratified by the significant interaction effect between interaction and attachment style on changes in cortisol concentrations. Hormonal responses to both passive interaction and no interaction followed similar trends in cats with different attachment styles, although the extent of the responses varied; however, differences between secure and insecure cats were amplified in response to forced and active interactions. Secure cats tolerated forced interaction better than insecure cats, as reflected by a much lower increase in cortisol and generally greater oxytocin secretion, again suggesting more effective stress buffering through oxytocin-mediated parasympathetic activation.^31,38,39 In contrast, insecure cats showed significant cortisol increase and divergent oxytocin release patterns (i.e. increase for anxious cats and decrease for avoidant cats) during forced interaction, potentially owing to dysregulated hypothalamic–pituitary–adrenal (HPA) axis activity and/or the compensatory oxytocin-mediated stress alleviation.^40,41 The non-coercive, owner-led active interactions were perceived as similar to passive interactions by secure cats, while insecure cats perceived them more as forced and intrusive, reflecting in the increased cortisol release. This may reflect a compromised perception of autonomy in insecurely attached cats (i.e. when the interaction was not initiated by them, as in the case of active interaction, they view the interaction as forced and their HPA axis was activated). ³⁵

Limitations

One limitation of this study is the relatively small sample size. During recruitment, we aimed to balance the number of cats across attachment styles; however, collecting a sufficient number of avoidant cats took considerable time because of their low prevalence in the population. ⁸ In addition, interactions with owners in the home environment, even in the case of forced interactions, may not have induced sustained stress in cats. Cortisol levels can be influenced by a general arousal state, while oxytocin may increase alongside activation of the stress response.^11,12 Therefore, hormonal fluctuations observed in response to different types of interactions may not reliably reflect the emotional valance of those interactions.

Conclusions

The fact that the attachment styles of cats impact their emotional response to different human-involved interactions highlights the importance of tailoring owner interactive strategies to meet the cat’s specific social needs. The findings of the current study provide evidence for developing interactive approaches better suited to individual cat populations. Passive interactions, where cats control the interaction pattern, remain the safest social strategies for owner–cat interactions. For cats with insecure attachment styles, owner-controlled active interactions, even though not coercive, should be avoided in addition to forced interactions. Future research should include larger sample sizes, incorporate additional emotional biomarkers (e.g. beta-endorphin, serotonin), and examine the association between attachment styles and personality traits in cats to further validate the present findings.