Response to Hansen Wheat et al.: Additional analysis further supports the early emergence of cooperative communication in dogs compared to wolves raised with more human exposure

Wheat et al.’s response to Salomons et al. (2021) addressed two main questions: 1) are the dog puppies outperforming the wolf puppies at gesture comprehension before they go to a human raiser’s home (aged <9 weeks)? and 2) is the tendency to approach a stranger a better explanatory variable than species to predict success at gesture comprehension?

To address the first question, we start by comparing puppies still living with littermates (L puppies), who had not yet been sent individually to raisers’ homes, to same-age wolf pups with more human exposure (see supplemental information). L puppies (≤9 weeks, mean age 7.6 weeks, N=13) performed significantly above chance on the pointing task as a group (70.5% correct, binomial test, p = 0.0003778) and 10 out of 13 (76%) were correct on their first trial. Wolf puppies in the same age range (≤9 weeks, mean age 8.4 weeks, N=5) were not above chance (60% correct, binomial test p = 0.3616) with 3 of 5 wolves (60%) choosing correctly on their first trial. On the marker gesture the L puppies (N=13) performed significantly above chance as a group (85.9% correct, binomial test p = 6.118e-11) and all 13 (100%) were correct on the first marker trial. The five wolf puppies in the same age range were not above chance on the marker gesture (56.7% correct, binomial test p = .5847) and only 2 of the 5 (40%) wolves were correct on the first trial. At an individual level, when examining combined performance with the two gesture tasks, over half of the L puppies (7 out of 13) performed above chance (i.e., got ≥10 of the 12 trials correct), while no individual wolf (N=26) of any age performed above chance. Unfortunately, the small wolf sample at this age prevents a meaningful group comparison using inferential statistics. Instead, we compare our L puppies to our group of 12-week-old wolves that is the same size (N=13). As predicted by the directional hypothesis in the original paper, L puppies outperformed these older wolves on the pointing gesture task (dogs: mean = 70.5% correct, SD = 23.7%; wolves: mean 57.7% correct, SD = 14.6%; Welch Two Sample T-test, p=.04815, one tailed) and the marker gesture (dogs: mean = 85.9% correct, SD = 15.0%; wolves: mean = 56.4% correct, SD = 19.9%; Welch Two Sample T-test, p=1.8e-05, one tailed).

To further address the first question, we compare the performance of the L puppies to the older group of dog puppies already living individually in raiser’s homes at the time of testing (H puppies; N=18). L and H puppies did not significantly differ from each other on the pointing task (β age>9 weeks = 1.1561, SE = 0.7251, p = .1109). On the marker gesture, L puppies outperformed the H puppies (β age>9 weeks = −1.0028, SE = 0.4665, p = 0.0316). Figure 1 illustrates how puppies do not show significant increases in skill with age on either gesture task. Finally, Bray et al. (2021) tested 375 L puppies (mean age 8.5 weeks) from the same population studied in Salomons et al. (2021) with a similar pointing task. A significant portion of these subjects, many of whom had never lived in human homes, used a human pointing gesture on their very first trial (70%).

Fig. 1.

Correct number of choices made by the different aged subjects. Age and performance are only weakly related on either of the gesture comprehension tasks. Dashed line represents chance performance (50% correct). All points are jittered for visibility – see supplemental information for table of points. a) Pointing Gesture. Dog linear regression: y = 0.12x + 3.3, R² = 0.075; wolf linear regression: y = 0.043x + 3.2, R² = 0.0048. b) Marker gesture. Dog linear regression: y = −0.12x + 5.9, R² = 0.088; wolf linear regression: y = −0.031x + 3.8, R² = 0.0021.

Taken together, our new analysis further supports the hypothesis that the youngest dogs with the least amount of human exposure were already skilled at using human gestures and are more skilled than similar aged wolves with more human exposure. We also see no a priori rationale to justify Wheat and colleague’s comparison of the L puppies subsample to the entire wolf sample. This maximally confounds age and species in a way that our original and new analyses do not.

In regard to Wheat et al.’s second question, they state, “since gesture-following necessarily also involves approaching a stranger, this large difference in willingness to approach could account for the difference…between the two species”. Several design features in the original study argue against this interpretation. First, unlike the go-no go tasks used to assess temperament (e.g. stranger approach), both a correct or incorrect choice in the two-way object choice paradigm (used to test memory, gesture comprehension and controls) requires a subject to approach one of two bowls that are equidistance from the experimenter. Second, all subjects passed the memory test as a prerequisite to participating in the gesture tasks, which required the same approach behavior as the gesture tasks and on which the two species performed the same. Third, the body versus point control was designed as another test of this. In this control, subjects would have been most successful by avoiding the bowl closest to the experimenter. However, wolves did not avoid the experimenter and did not differ from dogs in their performance. Regarding the model comparisons, the variables of “species” and “willingness to approach stranger” strongly co-vary (as demonstrated by the results of the temperament test), and the models including these variables are indistinguishable using AIC scores (see supplemental analysis). Therefore, any separate contributions to the overall effect cannot be determined using model comparison with the current dataset.

Finally, as a supplemental measure, in response to the previously published suggestion that dogs are simply attracted to human hands and things they touch (Wynne et al., 2008), we coded whether our L puppies were attracted to the human hand or marker when choosing. L puppies only touched the human’s pointing hand or the marker before making a correct choice (touching the baited bowl) on 3.6% (2 out of 55) and 4.5% (3 out of 66) of trials respectively. No subject did so on their first trial with either gesture. Again, this supports the hypothesis that dog puppies recognize the cooperative-communicative nature of basic human gestures and their responses are not only due to attraction to human bodies, hands, or things humans have touched (see also Riedel et al., 2008).