The authors used a setup in which the dog was presented with an array of four stimuli, three identical (e.g. 3 rubber ducks) and one that was odd (e.g. a rubber fire truck). He was initially trained on 20 different stimulus sets, all containing novel stimuli. If he chose the odd stimulus then he got a tasty treat, if he chose one of the matching stimuli then he didn’t. The authors were very careful to ensure that he was learning what they wanted him to learn and not using spurious cues from the experimenters or following the scent of the reward.
“One challenging element of the project was controlling for confounding cues. We wanted to make sure our results would not be due to a “Clever Hans effect”. It was also vital to control for odor cues.”
was a clever horse who was thought to be able to do arithmetic. However, he was just really good at picking up cues from the people setting him the question. If the questioner did not know the answer then his performance dropped to 6%. You have to be particularly careful of these issues when working with dogs as they impressively good at using human cues.
Batman learned pretty fast and once he was performing at 90% correct on the training stimuli, he was then given some critical tests. In the first set of tests he was present with the same stimuli that had been used during training but the object that had previously been numerous was now odd (e.g. 3 rubber fire trucks vs 1 rubber duck). If the dog was responding on the basis of previous reinforcement history of the specific stimulus then he should choose one of the fire trucks. However, if he had learned the relational rule then he should choose the duck. And he did, around 80% of the time. In the second set of tests he was presented with entirely new stimuli that he had never seen before and again performed well, choosing the odd object 70% of the time.
The findings reveal that Batman was able to learn about the relations between stimuli. This has important implications in our understanding of dog cognition.
“Our findings help to narrow the gap between what is known about the intelligence of man’s best friend and man. The ability for an individual to be able to identify sameness or difference of several different objects has been considered a prerequisite for more advanced forms of reasoning including understanding conservation and class inclusion. Our subject learned object oddity quite easily which indicates that this level of cognitive abstraction was well within his ability.”
This study provides the first evidence that a dog can discriminate between a stimulus array on the basis of oddity, a skill which is considered to be cognitively complex because it requires an understanding of the relations between the stimuli. This is an exciting step forward in our understanding of canine cognition.