Part-based and configural processing of owner's face in dogs

Abstract

Dogs exhibit characteristic looking patterns when looking at human faces but little is known about what the underlying cognitive mechanisms are and how much these are influenced by individual experience. In Experiment 1, seven dogs were trained in a simultaneous discrimination procedure to assess whether they could discriminate a) the owner's face parts (eyes, nose or mouth) presented in isolation and b) whole faces where the same parts were covered. Dogs discriminated all the three parts of the owner's face presented in isolation, but needed fewer sessions to reach the learning criterion for the eyes than for both nose and mouth. Moreover, covering the eyes region significantly disrupted face discriminability compared to the whole face condition while such difference was not found when the nose or mouth was hidden. In Experiment 2, dogs were presented with manipulated images of the owner's face (inverted, blurred, scrambled, grey-scale) to test the relative contribution of part-based and configural processing in the discrimination of human faces. Furthermore, by comparing the dogs enrolled in the previous experiment and seven 'naïve' dogs we examined if the relative contribution of part-based and configural processing was affected by dogs' experience with the face stimuli. Naïve dogs discriminated the owner only when configural information was provided, whereas expert dogs could discriminate the owner also when part-based processing was necessary. The present study provides the first evidence that dogs can discriminate isolated internal features of a human face and corroborate previous reports of salience of the eyes region for human face processing. Although the reliance on part-perception may be increased by specific experience, our findings suggest that human face discrimination by dogs relies mainly on configural rather than on part-based elaboration.

Figure 1. Experimental setting and detail of the touch screen.

Details of the experimental setting showing (A) a schematic representation of the experimental room, illustrating the position of the touch screen, the laptop, the automatic feeder, the water bowl and the owner's chair and (B) a photograph of the touch screen, with dimensions of the hemi-screens; the pictures shown in the screen exemplify the presentation of stimuli, in this case the whole, uncovered faces of the owner and the stranger.

Figure 2. Example of the stimuli used in Experiment 1.

Example of stimuli used in the Experiment 1, representing (A) face parts presented in isolation and (B) the whole face with single parts covered and with no parts covered (C).

Figure 3. Sessions needed to reach the learning criterion in the training phases of Experiment 1.

Mean ± SD number of sessions needed by dogs to reach the learning criterion when discriminating between (A) the owner's and stranger's face isolated parts and (B) faces with covered parts. Bonferroni corrected pairwise comparisons after generalized linear mixed model.

Figure 4. Example of the stimuli used in Experiment 2.

Example of stimuli used in Experiment 2, representing (A) the whole face in frontal view without manipulations and (B) the whole face with manipulations used in the test trials.

Figure 5. Choices in the test trials of Experiment 2 by expert and naïve dogs.

Mean ± SD number of owner's face choices done by expert and naïve dogs with each type of manipulation in test trials of Experiment 2. Error bars represent the standard deviation from mean.