There are good reasons to suppose that the infant’s innate disposition for informational sensitivity is grounded on a representational mind. As Kim Sterelny (1991, p. 21) writes: «there can be no informational sensitivity without representation. To learn about the world and to use what we learn to act in new ways, we must be able to represent the world […]. Furthermore, we must make appropriate inferences from those representations». Representations must be inferentially linked to each other. In addition, the representations of information should have propositional form, just like regular beliefs.
The definition of belief in terms of propositional attitude is normally accepted. A propositional attitude is the mental state of having some stance canonically expressible in the form “S A that P”, where S picks out the individual possessing the mental state, A picks out the attitude, and P is a sentence expressing a proposition». Beliefs are essentially states that represent how things stand in the world. Therefore, a regular belief is a propositional attitude, which implies that it has, as a proposition, a specific meaning that can be formulated by a typical sentence in our belief-tasks context: “Sally knows that the ball is in the box”. Infants, like adults, expect that this must always be true for everyone. When Sally sees that the ball is in the box, then, according to infants Sally knows that the ball is in the box. Infants build their expectations from this premise.
Grasping propositional content does not require the possession of truth concept as such, nor the involvement of a linguistic dimension. In this case, language does not establish the rules that, following inferential and social paths, articulate the propositional contents of representations, which are in turn generated from the perceptual observation of worldly events involving social interactions.
The format of mental state representations presents a syntax similar to propositional attitude and composed by an agent, an attitude, and a content. These components can be changed flexibly in early mindreading operations as suggested by some evidence (Kampis et al. 2013; Kampis 2017) and described by belief-file theory. Belief-file is a theoretical construct aimed to describe «the core representational skeleton for online Theory of Mind reasoning» (Kovács 2016, 516), thus enabling to track and update one’s own representations about other people’s beliefs. Belief-file structure is flexibly articulated on three aspects: the agent as belief-holder, the belief-content, and the referent.
In infant mind, only two variables (the agent and the content) «can be modified independently from each other, which enables fast updating and modification of the elements» (Kampis 2017, 66), thereby allowing fast computation on any belief content. On the contrary, «the limit of the speed and effort of computing a particular mental state content would depend on the difficulty to calculate and represent such content itself» (Ibid.).
Furthermore, Kampis (2017) suggests that attitudes can also change in the sense that the prediction of an agent’s action rests on a preference for an object, accompanied by a belief about the same object. Kovács’ Belief-file structure fits well with Carruthers’ (2013) description of mindreading as a representational multi-componential process operating through interactions between a core attribution system and a planning system.
Following a developmental nativist account that I embrace in my book, it is reasonable to suppose the existence of a single mindreading system including propositional formats, attribution procedures, and prediction strategies. Infants and adults mindreading systems show a developmental discrepancy due to an enrichment of the contents that are represented, rather than to a sudden representational shift.
Among the metacognitive skills that infants and adults share, there is the aspectuality (belief attribution for dual identity objects), that would be processed spontaneously without involving an introspection machinery directed toward the content of metarepresentations. The infantile dimension of belief does not only include aspectuality but also normativity, in the sense that attribution procedures follow the transitional inferential rule that someone should believe something given that evidence.
In my opinion, the experimental works of Kampis and many other researchers illuminate the extraordinary infants’ capacities to make belief attributions to others. Young children show remarkable flexibility for attributing false beliefs about location, and for attributing other reality-incongruent epistemic states, like pretence, false perceptions, false information about identity.
However, someone advances that for all this kind of experimental settings there is a serious problem about reliability in terms of replication. Moreover, there are other experiments that seem to disconfirm (in part) some results in terms of infant metacognitive abilities.
Without any doubt, there is an objective difficulty to test cognitive abilities in young children in an ecological context. At the same time, it is very hard to deny the great number of results provided by so many different labs that show the extraordinary infant cognitive abilities.