Introduction: the meaning of 'metacognition' We assume that readers of this volume will by now have some familiarity with the sorts of paradigms that have been used to provide evidence of metacognition in non-human primates. In a common type of experiment (e.g. Smith et al. 2008), animals are trained to perform a primary task such as making a discrimination of some sort between categories (e.g. sparse versus dense) to achieve a favoured reward (either immediately, or after a delay; Couchman et al. 2010). After training, the animals are also provided with an 'opt out' response of some kind, which they tend to use in difficult cases where they are more likely to make (or have made) an incorrect judgement. Opting out generally either avoids the penalty that accompanies a mistaken answer (such as a timeout before there is another opportunity to obtain a reward), or guarantees a less-favoured reward. Such results are said to show that the animals are aware of their own uncertainty, especially since similar use of the opt-out response in humans is associated with self-attributions of uncertainty. We fully accept that this body of work, taken as a whole, cannot be explained in low-level associationist terms, as involving mere conditioned responses to stimuli. A great deal of careful experimentation has been done to demonstrate that this is not the case, and we are happy to embrace this conclusion (Beran et al. 2009 ; Couchman et al. 2010 ; Smith et al. 2010 ; Washburn et al. 2010). So it should be agreed that the animals have beliefs about the contingencies of the experiment and take executively-controlled decisions that depend on those beliefs (as well as having goals and other states like emotions, which some have been reluctant to attribute to animals; but see Panksepp 2005). However, to say that the animals' behaviour is fully cognitive and executively controlled is not yet to say that it is meta cognitive, in the sense in which this term is employed throughout cognitive and developmental psychology. For metacognition is generally defined as 'thinking about thinking' (Flavell 1979 ; Dunlosky and Metcalfe 2009), and therefore as involving metarepresentation. Moreover, metarepresentation in turn is understood to require a representation that represents another representation, or a mental state whose content represents, and is about , another mental state. This definition of 'metacognition' accords with the standard model for classifying and characterizing metacognitive processes in humans (Nelson and Narens 1990 ; see Fig. 5. 1), in which a metalevel monitors, represents, and controls the processes of object-level cognitive systems. Since those who study metacognitive processes in animals often cite this model with approval (e.g. Smith et al. 2003 , 2006 ; Couchman et al. 2010), we assume that is it some version of this architecture, or some of its components, that the animals in question are claimed to possess when they are said to have metacognitive capacities. And it should be noted that an important aspect of the ...