Synopsis
In recent years there has been growing interest in uncovering evolutionary and lifetime factors that may contribute to the domestic dog’s (Canis lupus familiaris) success in anthropogenic environments. The readiness with which dogs form social attachments, their hyper-social focus, and social flexibility have all been areas of investigation. Prior research has demonstrated that many pet dogs form infant-caregiver type attachments toward human caretakers, even into adulthood. However, it is unknown if adult dogs form similar attachment bonds to other species, including cohabitant dogs, or if the dog–human relationship is unique in this respect. In the current study we used the Secure Base Test to evaluate behavioral indicators of stress reduction, proximity seeking and exploration, classifying dog–human and dog–dog dyads into attachment style categories. As in prior studies, we found that the majority of our dog–human dyads met the traditional criteria for infant–caregiver type attachment. However, the majority of dogs did not display this form of attachment toward cohabitant dog partners. Instead, behaviors observed in dog–dog relationships better matched attachment classifications described in human sibling attachment research. Overall, companion dogs were significantly less likely than human caretakers to elicit behaviors associated with attachment security in a focal dog. Dog–human attachment may play a distinct and important role in the success and resilience of adult dogs living in at least some anthropogenic environments. Bonds formed with other adult dogs, while important, likely serve a different function.
Introduction
Infant–caregiver attachment bonds, or filial attachments, play an important role in the survival and success of species whose offspring require care and provisioning after birth (Hennessy and Shair 2017). Attachment bonds promote proximity seeking and contact between offspring and their caretaker as well as separation distress and search behavior in the caretaker’s absence (Bowlby 1969; Ainsworth and Bell 1970). These bonds can also facilitate the “Secure Base Effect,” where the attachment figure’s presence reduces stress and facilitates exploration in novel or challenging circumstances allowing for important learning experiences (Bowlby 1969). From a comparative perspective, attachment bonds are an intriguing area of study as many different species and taxa form infant–caregiver attachments, including humans, and also because these attachment bonds can occur between both conspecific and heterospecific dyads (Bowlby 1969; Thielke and Udell 2020).
Domestic dogs are well known for their propensity to bond with other species, and recent evidence suggests that pet dog–human bonds often meet the criteria for infant–caregiver type attachment (Udell and Brubaker 2016). This has allowed dog–human bonds to be classified into attachment style categories using the same methodology and classification systems originally developed to study attachment bonds in primates (Secure Base Test [SBT]: Harlow 1959; Thielke and Udell 2017; Wanser and Udell 2019) and in human infants (Strange Situation Test [SST]: Ainsworth and Bell 1970; Schöberl et al. 2016). However, there is still much to learn about these relationships, predictors of different attachment outcomes, and also the influence of attachment style on other aspects of dog’s social behavior (Rehn and Keeling 2016). For example, while infant–caregiver type attachment bonds have been well documented (Schöberl et al. 2016; Thielke et al. 2017; Wanser and Udell 2019; Thielke and Udell 2020), it is not well understood why dogs maintain and continue to form infant–caregiver type attachments to humans into adulthood (Gácsi et al. 2001; Mariti et al. 2017). There is also little research exploring whether dogs exhibit this pattern of attachment to all social companions (independent of species or environment) or if this pattern of behavior is a special social adaptation to success in human homes.
There is some data to suggest that even in pet dog populations, this same pattern of adult bonding behavior may not occur with cohabitant dogs. For example, a study by Mariti et al. (2014) found that in an attachment test where the focal dog was temporarily left alone in an unfamiliar environment, the return of the partner dog did, on average, reduce some signs of stress compared with the alone phase. However, in the same test, these focal dogs showed stress reduction and even greater affiliative behavior toward a human stranger, making the nature of their attachment bond to the cohabitant dog unclear (Mariti et al. 2014). Another study by Mariti et al. (2017) posited that in contrast to unrelated cohabitant dogs, perhaps biological mother–offspring attachment bonds would endure if these related dogs remained in the same household into adulthood. While the bonds between the adult dogs and their biological mother did retain some elements of caregiver type attachment that appeared to extend beyond the relationship between adult dogs and unrelated cohabitant-dogs, they ultimately concluded that the relationships shared between even related adult dogs did not fit into the same category as attachment bonds previously described in the child–caregiver or dog–human bond literature (Savalli and Mariti 2020).
It is important to note that prior studies evaluating dog–dog attachments have primarily focused on averaged data (Mariti et al. 2014, 2017). Therefore, it is possible that previously observed differences in dog–dog and dog–human attachment could be rooted in different proportions of individual attachment styles represented within each population (e.g., Secure versus Ambivalent versus Avoidant), and not the presence or absence of an attachment bond. In addition, if dog–dog relationships cannot be classified in the same manner as dog–human relationships, then characterizing the kind of bond shared between dog–dog partners is an important step in fully understanding the nature of dog social behavior and bonding. However, to date, no study has directly compared attachment relationships that dogs share with their primary human caretakers to that of the relationship shared with a cohabitant dog. Therefore, in the current study, we evaluated attachment relationships between dog–human and dog–dog dyads living in the same household, attempting to classify individual attachment styles using a consistent methodology and attachment style classifications previously used in the infant–caregiver and dog–human attachment literature (Thielke et al. 2017).
Materials and methods
Ethical considerations
The study conducted was approved and in compliance with regulations set forth by Oregon State University’s Institutional Animal Care and Use Committee, protocol# 4837 and 2019-0013. Owner consent was obtained for each dog prior to involvement in the study.
Subjects
Forty-eight cohabitant dogs (24 dog pairs) were recruited for this study. Each pair of dogs had been living together for at least 4 months, ranging up to 13 years in the same household (4.20 years on average). Only one dog dyad was biologically related (siblings). The dogs varied in age from 1 to 16 years old (7.28 years old on average). See the Supplementary Material for a description of all participants. If more than two dogs, or more than one owner, lived in the household, we asked that the person considered the “primary caretaker” and two “bonded dogs” participate in the sessions.
Testing area
The testing area was an empty room 3.3 m by 4.2 m with a 1-m tape grid covering the entire floor to facilitate coding of proximity measurements during analysis. A video camera mounted on the ceiling captured video for the entire room during testing sessions. A second video camera on a tripod provided backup recordings for behavior and vocalizations. There were six clean toys placed at random in the room with two being ball toys, two squeaker toys, and two rope toys. The dogs had no previous interactions in this room prior to participating in this study.
Procedure
To assess attachment, we used a counterbalanced version of the SBT methodology, previously validated to analyze dog–human attachment relationships, with two potential attachment figures (Wanser and Udell 2019). Each participating unit consisted of two dogs and a primary human caretaker living in the same household. Each dog served as the focal dog during one visit, and as the potential attachment figure for the other dog on a second visit. The human caretaker was the same individual at both visits. Each testing session was 12 min in duration, and the two visits were scheduled at least 5 days apart. In prior work, we found that in the absence of major relationship changes (e.g., a targeted intervention aimed at improving the bond) attachment styles of dogs to humans remain robust across two attachment tests, even if testing occurred at the same location for both visits (Udell et al. Forthcoming 2021). The SBT was conducted as follows, except that for one dog in each household (half of the participants total) the human owner conditions came first, and for the other dog in each household the cohabitant–dog conditions came first (counterbalanced testing order).
Owner baseline (2 min): The owner and the focal dog entered the room together. The owner was allowed to freely interact with their dog. The cohabitant dog was not present.
Alone phase 1 (2 min): The owner was notified to exit the room, leaving the focal dog alone. The cohabitant dog was also not present.
Owner reunion (2 min): The owner re-entered the room with the focal dog. The owner was again allowed to freely interact with their dog as in baseline. The cohabitant dog was not present.
Dog baseline (2 min): The owner was notified to exit the room. A research assistant then brought the cohabitant dog into the room with the focal dog and exited the room. Both dogs were left alone in the room together. The owner was not present.
Alone phase B (2 min): The research assistant entered the room briefly and took the cohabitant dog out of the room, leaving the focal dog alone.
Dog reunion (2 min): The research assistant returned the cohabitant dog into the room with the focal dog, and exited, leaving both dogs in the room together. The owner was not present.
Infant–caregiver attachment style classification
Behavior during each reunion phase of the SBT was used in attempt to classify the attachment style of a focal dog to their human partner and to their dog partner (Kier and Lewis 1993; Schöberl et al. 2016; Thielke et al. 2017). Two independent coders reviewed reunion phase videos for each dog’s SBT and categorized the dogs’ attachment styles based on patterns of behavior observed. Independent inter-rater reliability for holistic coding of attachment styles between dogs and their owner was 79% (binomial probability test, P < 0.001). Following independent evaluation, the two coders re-watched videos where they had selected different classifications and mutually agreed on the final classification assignment for that subject (Thielke et al. 2017). In cases where agreement could not be reached, the designation “Unclassifiable” was selected. Attachment style classifications taken from the prior literature (Thielke et al. 2017) were as follows:
Secure attachment: Little or no resistance to contact or interaction. Greeting behavior is active, open, and positive. Seeks proximity and is comforted upon reunion, returning to exploration or play.
Insecure avoidant attachment: May show little/no distress on departure. Little/no visible response to return, ignores/turns away but may not resist interaction altogether (e.g., rests or stands without bodily contact, out of reach or at a distance).
Insecure ambivalent attachment: Shows exaggerated proximity-seeking and clinging behavior but may struggle if held or blocked. Mixed persistent distress with efforts to maintain physical contact and/or physically intrusive behavior.
Disorganized attachment: Evidence of strong approach avoidance conflict or fear on reunion, for example, circling, hiding from sight, rapidly dashing away on reunion, and aimless wandering around the room. May show stereotypies on return (e.g., freezing or compulsive grooming). Lack of coherent strategy shown by contradictory behavior. Dissociation may be observed, that is, staring into space without apparent cause; still or frozen posture for at least 20 s (in the non-resting, non-sleeping dog).
Unclassifiable: Classifiers were unable to reach consensus on group placement for dogs from this classification category.
Data coding and statistical analysis
All attachment style categorizations were coded from video by two independent coders. In addition, the following duration and frequency measures were collected (all behaviors were coded by a primary coder and >35% were also coded by a second independent coder to evaluate inter-rater reliability):
Frequency of vocalizations: A single bark or bout of whining, howling, or vocal sound associated with movement of the focal dog’s mouth (IRR, R = 0.94, P < 0.0001).
Door proximity: Any part of the focal dog is within one or both of the two boxes labeled C or D directly in front of the main door (visual measure indicating area within 1 m of door) (IRR, R = 0.76, P < 0.00001).
Proximity: Any part of focal dog is within in at least one box directly adjacent to the owner/cohabitant dog (visual measure indicating area within 1 m). Only recorded for return phases (IRR, owner R = 0.95, P < 0.00001; cohabitant dog R = 0.97, P < 0.00001).
Due to video recording errors and attrition between sessions, data analysis was completed for 43 out of the 48 initially recruited dogs. Non-parametric statistics were used given non-normal distributions for all duration measures (Shapiro–Wilk P < 0.05). All statistical tests had an alpha level of 0.05.
In some cases, previous literature has found order effects within multi-phase SST attachment tests that could influence the balance between proximity seeking and exploration in the return phase of attachment tests. This may in part be due to findings that dogs’ exploration of the novel environment can decrease over the course of testing (Prato-Previde et al. 2003; Palestrini et al. 2005; Fallani et al. 2006; Palmer and Custance 2008; Rehn et al. 2013) along with some decreases in play behavior (Prato-Previde et al. 2003; Rehn et al. 2013). In the current study, we counterbalanced testing order to guard against such possible effects. However, after analysis, we found that testing order was not predictive of the proportion of time the dog spent seeking owner proximity in the 2-min return phase of our study (owner first order: median = 0.89, SD = 0.23; dog first: median = 0.90, SD = 0.16, Mann–Whitney U = 185, P = 0.52), nor was it predictive of dog proximity seeking in the return phase (owner first order: median = 0.78, SD = 0.25; dog first: median = 0.83, SD = 0.29, Mann–Whitney U = 156.5, P = 0.17). This may in part be due to the shorter duration of the SBT compared with the SST, and the fact that unlike in the original SST, in the SBT each reunion phase is proceeded by an alone phase (Wanser and Udell 2019).
Results and discussion
Attachment style analysis
Dog–human attachment: Out of 43 dogs, 42 were successfully classified into a dog–owner attachment style: 23 (55%) were classified as Securely attached to their owner, 19 (45%) were classified as Insecure Ambivalent, and 1 (2%) was scored as “unclassifiable.” Dog–human attachment data for both dogs living in the same household was available for 19 dog dyads (38 dogs total). Of these, 10 displayed the same style attachment to their human caretaker (e.g., both dogs had a secure attachment style), and for 9 pairs each dog exhibited a different attachment style toward the same human caregiver (e.g., one dog had a secure attachment while the other had an ambivalent attachment), suggesting that these attachment classifications were representative of unique bonds of each dog–human dyad (two-tail binomial test, P = 1.0).
Dog–dog attachment: In contrast to the one dog (2%) designated as “unclassifiable” in the dog–human attachment style analysis, the behaviors observed in the focal dog upon the companion dog’s return were not represented in the infant–caregiver attachment ethogram, with 41 dogs (95%) categorized as “unclassifiable” by both coders during the independent analysis. In many cases, dogs continued to search for the human owner (who was absent throughout the cohabitant dog conditions) instead of responding to the presence or absence of the cohabitant dog. This difference in ability to classify attachment style between dog–owner and dog-dog dyads was statistically significant (Fisher’s exact test, P < 0.001), suggesting that the caregiver–attachment ethogram traditionally used to evaluate human–dog bonds is likely not appropriate for describing adult dog–dog interactions. This is consistent with the findings of prior research (Mariti et al. 2017).
Similar outcomes have been described in the human literature when evaluating sibling attachment relationships (Stewart 1983). For example, when given a choice, siblings spend more time interacting with their parents than each other, leading researchers to believe that sibling dyads represent different interactional systems (Schrier 1965; Stewart 1983). However, upon separation from their mother, some infants have still been found to use their older siblings as a Secure Base to provide comfort and security to return to exploration even in the presence of a stranger (Stewart 1983). Since one aim of this study was to determine the nature of dog–dog bonds, and since our data suggested that the infant–caregiver model may be inappropriate, we adapted an ethogram from the sibling attachment literature, SPPIR—which stands for the classification categories Soothed, Partly Soothed, Playing, Ignoring, Resistant (Kier and Lewis 1993), to evaluate if dog–dog attachment behavior could be better classified using this system. SPPIR classification categories were defined as follows:
Soothed: Dog begins the return phase by crying/vocalizing and/or hovering near the door but ends the phase in joint play and/or relaxed behavior with the conspecific and no longer shows any signs of distress. (The dog may also engage in some solitary play.)
Partly soothed: Dog begins the return phase by showing distress (crying/vocalizing and/or hovering near the door), acknowledges the conspecific’s presence, and yet continues to periodically cry/vocalize and/or hover near the door during the phase and does not appear to be entirely soothed.
Playing: Dog begins the return phase in solitary play (with no signs of distress) and initiates interaction with the conspecific once the conspecific enters the room.
Ignoring: Dog may or may not briefly acknowledge the conspecific’s presence, but mostly behaves during the return phase as if the conspecific was not present, engaging in solitary play or crying/vocalizing and/or hovering near the door.
Resistant: Dog becomes more irritable and rejecting during the return phase. This is shown by increased crying/vocalizations, resistance to contact and play, aggressing, active avoidance, fear, or stereotypies during the conspecific’s presence.
This coding system proved to be a better fit, with 42 of 43 (98%) dogs successfully classified into one of the sibling attachment style categories. When compared against the ratings of a second coder (for all dogs), inter-rater reliability was high with 83% agreement (agreed on 35/42 classifications independently, P < 0.00001). As with dog–owner attachment, we asked if dogs from the same household were more likely to share the same attachment style. Out of the 19 dog pairs where we had SPPIR data for both dogs, 8 pairs had the same SPPIR attachment style and 11 had a different attachment style suggesting that, as with dog–human attachment styles, individual attachment styles of the dog–dog dyads were not significantly linked (two-tail binomial probability, P = 0.65).
We then asked if there was evidence of a co-habitant dogs serving as a secure base for the focal dog. Even when using the SPPIR definitions, only one dog (out of 42) was classified as “Soothed,” the category associated with the secure base effect, upon the companion dog’s return (compared with 23/42 dogs displaying the secure base effect to their human caretaker). Therefore, dogs were still significantly less likely to exhibit evidence of the secure base effect (Fisher’s exact test, P < 0.0001). While the soothed category of the SPPIR best matches the criteria of the secure base effect in the human literature (Stewart 1983), the “partly soothed” category also required partial stress reduction and return to more relaxed behavior and therefore might suggest a partial secure base effect. Consequently, we re-ran this analysis including all dogs that were classified as soothed or partly soothed (13 dogs, 31%) upon return of their companion dog. However, even after expanding the criterion to include partly soothed dogs, significantly more dogs still displayed attachment security upon the return of their human caretaker compared with the return of a companion dog (Fisher’s exact test, P = 0.049). Therefore, while some dogs appeared to experience some benefits of a companion dog’s return in an unfamiliar situation, the return of a bonded owner had a significantly greater impact on stress reduction and behaviors associated with attachment security for more dogs. In fact, the majority of focal dogs either ignored (N = 26, 62%) or actively resisted (N = 3, 7%) their companion dog upon their return. In contrast, no dogs in this sample were classified as avoidant toward their human caretaker.
For a final comparison of evidence for the secure base effect in the presence of each potential attachment partner, the human reunion phase was also re-coded for the dog–human dyads using the SPPIR classification system by two independent coders. All 43 dogs were classified, however inter-rater agreement using this coding system was lower than when using the original dog–human attachment classifications with 67% matched classifications. This lends further support that the SPPIR may be a more appropriate coding strategy for relationships that fall in the sibling domain, whereas the traditional attachment classification system may better target behaviors observed in dog–human caretaker bonds. Nonetheless, inter-rater agreement was greater than expected by chance (Binomial test, P < 0.0001), therefore we next compared the number of dogs classified as “Soothed” by the return of their primary human caretaker versus their cohabitant dog partner using the SPPIR coding system. We found that significantly more, 33/43 (77%), dogs were classified as Soothed in the human-return phase, compared with only 1/42 (2%) dogs classified as Soothed during the cohabitant dog-return phase (Fisher’s exact test, P = 0.0072). In fact, 40/43 dogs were classified as Soothed or Partly Soothed at their owner’s return using the SPPIR, compared with 13/42 dogs upon their companion dog’s return (Fisher’s exact test, P = 0.0001), further supporting the finding that, on average, the return of a dog’s primary human caretaker resulted in a greater secure base effect than the return of the cohabitant dog companion.
Behavior analysis
Proximity seeking
During the transition to alone phases of the SBT, 42 out of 43 dogs orientated toward and/or followed both the owner and companion dog to the door as they exited. There was not a significant difference in the average proportion of time focal dogs spent in proximity to the door after the owner versus the cohabitant dog exited the room when left alone (Wilcoxon signed-rank test: Alone phase [owner], M = 0.69, SD = 0.25; Alone phase [dog], M = 0.70, SD = 0.26, W = 149.5, P = 0.22). This suggests that the absence of both the owner and dog triggered following and search behavior in the focal dog associated with mild separation distress.
However, on average, during the return phases of the SBT, focal dogs spent significantly more time in proximity to their human owner (82%) compared with that of the cohabitant dog (71%) (Wilcoxon signed-rank test, owner proximity M = 0.82, SD = 0.20; dog proximity M = 0.71, SD = 0.27, W = 248, P = 0.048).
Vocalizations: Vocalizing is another behavior often displayed following an attachment figure’s departure, and is thought to be an effort to reestablish proximity, or to protest separation from the attachment figure, by signaling to the caregiver the individual’s distress (Ainsworth and Bell, 1970; Mariti et al., 2014). Complete vocalization data were recorded and audible across all sessions for 24 dogs.
The average frequency of all vocalizations (whining, barking, howling) did not differ significantly when comparing the alone phase following the departure of the cohabitant dog to the alone phase following the departure of the human caretaker (Wilcoxon signed-rank test: alone after owner baseline, M = 28.75, SD = 27.72, after dog baseline, M = 33.42, SD =25.76, W = 97.5, P = 0.13). In combination with the door proximity data, these findings would suggest that dogs experienced similar levels of separation distress when left alone, independent of whether the immediate individual exiting was the human caretaker or companion dog.
Attachment theory predicts that in the case of a secure attachment, the secure base effect should effectively reduce the separation distress exhibited during the alone phase upon the return of the attachment figure. Therefore, we also compared the frequency of vocalizations exhibited by dogs in the return phase for the dog partner versus the human partner. During the return phase, dogs vocalized significantly more following the return of the cohabitant dog, compared with the return of the owner (Wilcoxon signed-rank test: after owner return, M = 8.17, SD = 14.40, after dog return, M = 31.21, SD = 27.22, W = 16.5, P = 0.0001).
To evaluate the impact of cohabitant dog/owner return on stress reduction we compared the frequency of vocalizations exhibited in the alone phase versus the return phase for each dog–dog and human–dog pair. Dogs showed a statistically significant decrease in vocalizations upon the return of their owner following the alone phase (Wilcoxon signed-rank test, alone phase [owner] M = 28.75, SD = 27.72, owner return M = 8.17, SD = 14.40, W = 7.5, P < 0.0001). Conversely, we did not observe a significant decrease in vocalizations upon the return of the cohabitant dog after the alone phase (Wilcoxon signed-rank test, alone phase [cohabitant dog] M = 33.42, SD = 25.76, cohabitant dog return M = 31.21, SD = 27.22, W = 125, P = 0.48) (see Fig. 1).
Average vocalizations of the focal dog across alone and return phases of the SBT when tested with the cohabitant dog or the human owner. Error bars indicate SEM.
Conclusions and future directions
This study provides further evidence that the kind of attachment relationship formed between many pet adult-dogs and humans (infant–caregiver type attachment) is different from the kind of relationship shared between unrelated cohabitant dogs living in the same household. While our study did include one pair of sibling dogs, these dogs also did not share an infant–caregiver type attachment bond (were unclassifiable). While dogs have been found to prefer companionship from another dog rather than being alone (Fredericson 1952; Pettijohn et al. 1977; Mariti et al. 2014), the findings of the current study suggest that the primary human caregiver is more likely to provide attachment security than a companion dog. In our study, dogs also sought the proximity of this human companion for a greater proportion of time and continued to search for their human companion when left in an unfamiliar place, even when a cohabitant dog was present. This may in part be due to the fact that pet dogs typically remain dependent on human care, protection, and provisioning even as they age. The extension of infant–caregiver attachment patterns could represent an effective social strategy for dogs, amplifying their success under certain environmental conditions (Udell and Brubaker 2016; Savalli and Mariti 2020). However, it would also follow that cohabitant dogs would share bonds appropriate to the role they play in each other’s lives, and the dog–dog dynamic within a household is typically quite different from that of the dog–human caregiver relationship.
However, the relationships dogs form with specific humans can also vary. Therefore, it should not be assumed that the behavior directed toward a primary human caretaker would be equivalent across all bonded humans in a dog’s life. For example, attachment bonds between dogs and adult human owners are often different from those shared between the same dog and a child in the household (Wanser and Udell 2019). Pet dog–owner bonds have also been found to differ from dog–shelter caretaker bonds (Thielke and Udell 2020). In other words, understanding the broader context of the bonds dogs form, not only across different relationships with humans, but also with other dogs and other animal companions is important to our understanding the function and quality of these relationships (Savalli and Mariti 2020). While attachment bonds to a single individual are often very stable (Udell et al. forthcoming 2021), a single individual may have a variety of bonds to different individuals that vary in type or style. Therefore, the observed differences in this study may have less to do with whether the attachment figure was a human or a dog, and more to do with the functional role of that each individual plays within the focal dog’s life.
While infant–caregiver attachment may not explain adult dog–dog relationships, dogs in our study still spent roughly 70% of the time in each other’s proximity, exhibited separation distress upon the cohabitant dog’s departure, and roughly a quarter of dogs showed signs of decreased stress upon the cohabitant dog’s return, indicators that this relationship is also important, albeit different from the human–dog relationship. Ultimately, we found that the behaviors displayed during the dog–dog attachment test better aligned with definitions used to describe sibling relationships (rather than the infant–caregiver attachment classifications). Style of attachment also varied between dog dyads, and it is possible that dogs may form different styles of attachment to different dogs, just as with different humans, for example, when more than two dogs live in the same household. Interestingly, in the human literature, older siblings have a higher potential to serve as a secondary attachment figure to their younger sibling if they reliably responded to infant distress in extreme circumstances (Teti and Ablard 1989). Given the resemblances between sibling attachment and our current findings, it would be interesting to investigate if cohabitant dogs that are more behaviorally responsive to the focal dog’s stress cues are also more likely to “Soothe” or “Partly soothe” the focal dog more effectively. Other factors, such as the amount of time two dogs typically spend together or apart, may also be interesting to evaluate in future studies.
More research is needed to understand factors that shape and influence dog attachment relationships, and how attachment bonds can vary across different individuals within the focal animal’s life, in general. However, the current findings highlight a critical need for more research that considers dog attachment to species beyond humans (including attachment bonds with other dogs). Future work in this area should also consider dogs beyond pets, including working dogs, free-ranging dogs, and dogs in shelters, to provide a broader view of how attachment relationships—including those that persist into adulthood—may fit into the behavioral ecology and success of dogs as a species.
Author contributions
All authors contributed to the study design, data analysis, interpretation, and manuscript preparation. N.S. led the data collection. N. S. & M.A.R.U. wrote the final version of the manuscript.
From the symposium “Biology’s best friend: Bridging disciplinary gaps to advance canine science” presented at the virtual annual meeting of the Society for Integrative and Comparative Biology, January 3–7, 2021.
Acknowledgments
The authors would like to thank the following research assistants who assisted in data collection and/or behavioral coding: Champayne Master, Vivienne Ruiz, Hallie Shean, Kaia, Holly Duvall, Prema Nissinen, and Diana Fajer. They are also grateful to those who volunteered their dog to participate in this study.
Funding
During this project, M.A.R.U.’s research salary was partially supported by the USDA National Institute of Food and Agriculture, Multistate Project NC 1029, Applied Animal Behavior and Welfare.
Supplementary data
Supplementary data are available at ICB online.
