Then came the test. Dogs were presented with a wide variety of never-before-seen dog faces paired against never-before-seen non-dog faces. As before, dogs had to approach the dog image and avoid the non-dog image to get a treat. This was no longer an easy feat as the dog images now captured dogs’ vast morphologic diversity in shape, color, size, head shape, ear position, you name it. On top of that, the dog images were now pared against a wide range of non-dog faces including human faces as well as domestic and wild mammals like cats, sheep, gerbils, cows, rabbits, reptiles, and birds, among others. Images were presented head-on (full face) or as a profile. Below are examples of faces dogs saw in the study:

The dogs prevailed! The nine subjects successfully identified “dog” from “non-dog” faces. Some dogs, like Babel, Bag, Cyane and Vodka, were able to do so quite quickly, taking few sessions to approach the required 10-out-of-12 dog images. Other dogs, like Bahia and Cusco, were slower on the pickup and took more sessions to identify “dog” from “non-dog” (dog subjects needed anywhere from 2 to 13 sessions to meet criteria). This is not to say, of course, that Bahia and Cusco don’t know a dog when they see one. The researchers highlight that a number of factors — like dog personality, learning styles and strategies, and motivation — can affect dog behavior and performance, particularly when it comes to this type of task.