Macaques really can use logical reasoning to solve puzzles
Macaques can use logical reasoning to think through tasks, a finding that adds to the growing evidence that animals don’t just make choices purely to maximise the reward they get.
Greg Jensen at Columbia University, US, and his colleagues tested the way macaques made what are known as transitive inferences. An example of a transitive inference is the understanding that if A comes before B, and B comes before C, then A must come before C.
Transitive inference is important for deductive reasoning and was long thought to be a hallmark of human intelligence. But in the last few decades scientists have found a range of animals that can complete tasks that appear to require transitive inference. This includes primates, rodents, birds and wasps.
There is, however, controversy over whether the animals’ success on certain tasks really shows they can use logic. The tasks invariably involve rewards, which may mean the animals are able to complete them based on associative learning – in other words, they are making choices in order to maximise their reward.
To investigate further, Jensen and his colleagues created an experiment that required macaques to learn a sequence of images, which included objects such as a zebra, a hot air balloon and some corn. They first trained the monkeys on the order of the images by showing them pairs of images that were adjacent in the sequence and rewarding them when they touched the image that came earlier. The monkeys had been on a water-restricted diet and the reward came in the form of drops of water. No water was given for an incorrect answer. It would have been possible for the monkeys to learn the sequence either by associative learning or by inference.
Next the researchers began showing the monkeys pairs of images, although this time they were not adjacent but from anywhere in the sequence. As in the training phase, the monkeys got a reward for correctly picking the image that came earlier in the sequence and received nothing if they were incorrect.
Jensen and his colleagues then weighted the rewards differently to see if macaques were simply conditioned by greater rewards, or if they had an abstract understanding of the logic behind the task.
In one condition, the earliest image was associated with six drops of water, the second image in the sequence was associated with five drops, and so on. In a second condition, the rewards were reversed, meaning the earliest image was associated with one drop of water, the next in the sequence was associated with two drops of water, and so on.
If the monkeys had no logical basis for their decision-making and were only driven toward options that had greater rewards, they should have performed poorly in the second condition.
But Jensen and his colleagues found in hundreds of trials that the monkeys could accurately identify which of the pair of images came earliest in the sequence in both conditions. The authors say this shows that the monkeys are driven both by rewards and logical reasoning.
Journal reference: Science Advances , DOI: 10.1126/sciadv.aaw2089