Neural correlates of metacognition: A critical perspective on current tasks

- Several concerns have been raised about experimental tasks used to assess metacognition in animals.
- Post-decision wagering and uncertain option tasks can be solved by simple mechanisms that do not involve metacognition.
- The X pattern, usually associated with confidence can be produced by a simple difficulty estimation mechanism.
- The neural activity recorded during an uncertain option task can be reproduced by a simple multiple-choice decision model.

Humans have a remarkable ability to reflect upon their behavior and mental processes, a capacity known as metacognition. Recent neurophysiological experiments have attempted to elucidate the neural correlates of metacognition in other species. Despite this increased attention, there is still no operational definition of metacognition and the ability of behavioral tasks to reflect metacognition is the subject of debate. The most widely used task for studying metacognition in animals, the uncertain-option task, has been criticized because it can be solved by simple associative mechanisms. Here we propose a broad perspective that generalizes those critiques to another task, post-decision wagering. Moreover, we extend this critical view to account for recent neurophysiological evidence. We argue these tasks are simple enough that any animal could solve them using very simple mechanisms such as sensory-motor associations. In this case, it is impossible to know whether all animals are metacognitive, or if the tasks are simply not appropriate. Therefore, we suggest using better defined concepts until a suitable task for metacognition is available.