Metacognition in nonhumans: Methodological and theoretical issues in uncertainty monitoring

Is metacognition a uniquely human trait? Comparative studies suggest that the answer is no. To test metacognitive abilities in non-human animals, researchers have used a variety of methods. In the retrospective confidence judgment paradigm, animals have to provide a low or high confidence response following a first-order task. In the search for information behavioral paradigm, animals sometimes need to look for missing information in order to complete a task. In the uncertainty responding paradigm, animals can choose to perform trials varying difficulty, or they can choose to avoid a particular trial and move to a new one. Studies of metacognition in non-humans based on these paradigms show that animals can evaluate their meta-perceptual uncertainty. However, several methodological and theoretical objections have been raised against the possibility for non-speaking animals to be attributed metacognitive abilities. Animal subjects might be using associative mechanisms between exteroceptive stimuli rather than basing their judgments on evaluations of their own uncertainty. We address this criticism by summarizing recent studies in our laboratory using the uncertainty response method in which associative cues are minimized by making stimuli unpredictable, by dissociating specific responses from external feedback cues, and by using transfer tests.