This study investigated dog’s responses to human social cues and the relationship between these responses and animal temperament. We found that social stimuli overshadowed nonsocial stimuli in most dogs, suggesting that the salience of social stimuli was relatively high. This was partly due to the fact that many subjects were likely to have received some form of training. In Japan, as of 2000, 58.5% of all dogs were estimated to live indoors (according to the Pet Food Manufacturers Association), which requires basic training for indoor maintenance. Therefore, many of the subjects likely had a strong history of reinforcement with human pointing, which may influence their preference for social stimuli as cues. However, while the social stimulus in this experiment was dynamic (pointing), the nonsocial stimulus was static, and there was a divergence in stimulus intensity. Therefore, social stimuli may be more salient than nonsocial stimuli and learning in response to both stimuli may vary. During the test phase, overshadowing (Pavlov, 1927), in which a more salient stimulus masks a less salient one, is likely to occur. For example, a high correct response rate when presented with a social stimulus is expected to produce low performance with a nonsocial stimulus. To create overshadowing by Pavlovian conditioning, it is common to compound different stimulus modalities, such as tone and light (Kahoe, 1982; Mackintosh, 1976); however, we presented the stimuli simultaneously in front of the dogs as visual stimuli. Given the cognitive resources of dogs, one of the stimuli (mainly nonsocial) may not have been initially recognized. Furthermore, although dogs may have poorer discriminative abilities for two-dimensional nonsocial visual stimuli, they are capable of discriminating pictures of human faces (Müller, Schmitt, Barber, Huber, 2015; Nagasawa, Murai, Mogi, & Kikusui, 2011) and visual stimuli with blue and yellow features (Kasparson, Badridze, & Maximov, 2013; Shinoda, Kosaki, Nagasawa, & Kikusui, 2022). Therefore, the discrimination of nonsocial stimuli was not a difficult task for the dogs.
Interestingly, two subjects showed higher correct response rates for nonsocial stimuli than for social stimuli. In one of these individuals (dog 19), neither stimulus differed by chance. This individual exhibited a 92% correct response rate in the final session of the training phase, which suggested that learning was achieved by using the compound stimulus as a cue. Therefore, it is possible that mutual overshadowing (Waldmann, 2001) occurred during training when the two stimuli were of similar salience. The other individual (dog 11) showed an orienting response to the social stimulus in the test phase (see the Supplementary video 1), but tended to respond to the opposite option. (This response tendency is discussed later with the C-BARQ results.) Dog 11 was an Italian greyhound, a sighthound breed that was bred primarily to track small animals such as hares (Täubert, Agena, & Simianer, 2007). Therefore, dog 11 may have had superior visual acuity and experienced nonsocial stimuli presented at eye level with greater salience than the other dog breeds. Alternatively, the effects of cue competition may be weakened for stimuli that are biologically important (Miller & Matute, 1998; Oberling, Bristol, Matute, & Miller, 2000). For example, there may be breed-specific biological preparedness, such that a shape printed on white paper is perceived as a small animal pattern and strongly attracts attention.
None of the C-BARQ items related to fear or aggression were related to correct responses. Overall, the results differ from the predictions of this experiment, as there was no association between the use of social stimuli and individual aggression and fear in dogs. However, interestingly, attachment scores showed a strong positive correlation with the correct response rate to nonsocial stimuli. That is, the dogs tended to prefer the social stimulus (human pointing), while dogs that displayed more attachment behavior toward their owners were more attentive to nonsocial stimuli. Unlike social stimuli, which may be experienced daily, the study subjects were likely naïve to the nonsocial stimulus in this study. In light of the C-BARQ fear scores and the fact that domestication is associated with reduced neophobia, it is unlikely that attention to the nonsocial stimulus was based on fear. Alternatively, individuals with strong attachment to the owners may be more attentive to various stimuli. Individuals that pay attention to various stimuli may present more "attachment behaviors" e.g., approaching their owners, at a higher frequency. In other words, it may be that individuals with traits that make them pay more attention to nonsocial stimuli often, rather than because they are strongly attached, are more likely to be evaluated by their owners as being more strongly attached. The two-stage hypothesis of Udell, Dorey, & Wynne (2010) explains that the extended socialization period of dogs due to domestication enhanced their social cognitive abilities toward humans because of the stronger conditioning influence on social signals such as human pointing. Given that the correlation between attachment scores and learning to nonsocial stimuli indicates an individual's attentiveness to various environmental stimuli, such individual characteristics may have contributed to dogs' increased contact or interaction with humans from their early ages. From here, reinforcement by responding to human pointing is likely to occur, and response generalization from owner to other humans (in this case, the experimenter) may enable the dogs to behave more and more appropriately in human society. In addition, the positive relationship with the owner through conditioning may have established the owner's role as a secure base for the dogs and encouraged exploration toward unfamiliar stimuli. A previous study has shown that the presence of other individuals of the same species facilitates exploration of novel objects in wolves and dogs (Moretti, Hentrup, Kotrschal, & Range, 2015), and it is possible that in the present study, the presence of the owners contributed to this facilitating effect (stronger exploration of responses to nonsocial stimuli). However, attachment scores and correct response rates to social stimuli are not significant, and there is no reliable explanation for the relationship between correct response rates to nonsocial stimuli and temperament, which needs to be clarified in the future.
Dog 11 also had the highest attachment score of the individuals tested (Table S1), but unlike most of the other individuals, it had a relatively low correct response rate to social stimuli compared to nonsocial stimuli. In addition, the stranger-directed aggression and stranger-directed fear scores were not high in this individual (Table S1), suggesting that this is not an aversion to the experimenter comprising the social stimuli. Furthermore, since Trainability was also not extremely low (Table S1), it is possible that this individual was not reinforced to obey non-owner pointing, or has been reinforced to respond in ways other than obeying. Alternatively, the strong attention to nonsocial stimuli may simply have inhibited relative attention to social stimuli.
Although the emotional reactivity hypothesis predicted that greater fear and aggression toward humans would be associated with less responsiveness to social cues, our results rather support the two-stage hypothesis and we propose that fear, aggression, and the ability to understand and follow pointing may be independently selectable. There are examples of seemingly similar social cognitive abilities that are also thought to vary independently; Sundman et al. (2018) used two Labrador retriever breeds (common or field type bred for hunting) and German shepherds to test how well they follow pointing and whether they exhibit human-oriented behavior in solvable tasks, expressed as a common social cognitive ability. Because there was little correlation between these abilities, they were hypothesized to have been selected as distinct traits. This suggests that these indicators (fear, aggression, and reactivity to pointing) evolved independently. Alternatively, the C-BARQ index may assume a different trait than fear and aggression, as would be assumed in domestication. Belyaev's (1979) experiments on fox farms considered the selection of individuals exerting low fear responses to humans placing their hands in the cages. The fear or aggression considered in the emotional reactivity hypothesis is likely to vary between wild animals and those habituated to daily interactions with humans, as the C-BARQ presumes. Since the present study only considered fully domesticated dogs that have undergone advanced artificial selection, we only discuss this variation as a matter of selection at the breeding stage rather than schematic domestication. Dog-only studies can only elucidate the potential effects of conscious domestication selection, as opposed to unconscious selection (Darwin 1859, 1868). Therefore, future comparative studies should consider either wild wolves or species that may have retained early domesticated canine temperaments, such as dingoes.