Ben Vermaercke compared human and rat performance on two forms of category-based learning. On one of them, the rodents trounced the homo sapiens.
The participants - 16 rats and 24 humans - were trained to recognise that certain patterns (stripes of light and dark, known as gratings) shown on a screen were the targets, while others were the distractors. The patterns were presented in pairs, and for the rats, if they followed the target pattern in a pair, this led them to the correct route (out of two) towards the safety of a platform in a water maze. For humans, choosing the target pattern simply led to presentation of a "correct" symbol - a green triangle pointing upwards; choosing the distractor pattern triggered a downward red triangle.
Through choosing the different patterns and receiving feedback, the rats and humans learned which patterns were targets and which were distractors. In one "rule based" version of the task, the targets and distractors always differed only along one dimension - either the frequency, or the orientation, of the light and dark stripes. In the other "information integration" version of the task, the targets differed from the distractors along both dimensions (frequency and orientation) simultaneously.
The key challenge occurred next, when the rats and humans entered the test phase, and attempted to generalise what they'd learned in the training phase to new pairs of patterns. The rats and humans performed similarly on the rule-based version of the task. However, when it came to the "information integration" version, the rats performed significantly better than the humans. This was because the humans' performance dipped in the "integrated information" version of the task, whereas the rats performed just as well at this version as they did on the rule-based version.
What was going on? In the version of the task where the target was distinguishable from the distractors along two dimensions simultaneously, the correct choice couldn't be identified based on a simple rule. But humans like to make conscious decisions and use explicit rules, even when this approach isn't optimal. It's for this reason that they struggled at this version of the task. Rats, in contrast, used an implicit similarity approach in both versions of the task (think of this as going with your gut, as to which pattern seemed most similar to the targets seen in training). This served the rodents fine in the "rule-based" version, and actually led them to beat us humans in the more complex information-integration version. In this latter version, the humans looked too hard for an explicit rule, and would likely have performed better if they'd gone with their instincts.
"We have shown that rats display superior generalisation performance in a generalisation context in which correct stimulus-response associations do not follow a dimension-based rule," the researchers said. "This is in line with the hypothesised competition in the human brain between an explicit, rule based system and in implicit category-learning system."
Vermaercke B, Cop E, Willems S, D'Hooge R, & Op de Beeck HP (2014). More complex brains are not always better: rats outperform humans in implicit category-based generalization by implementing a similarity-based strategy. Psychonomic bulletin & review, 21 (4), 1080-6 PMID: 24408657
Post written by Christian Jarrett (@psych_writer) for the BPS Research Digest.