Children Don’t Think Like Adults

A thinking infant

by David J. Lobina

A truism, perhaps, and certainly meant to draw attention, at least in one way. In reality, the title of this post is intended as a playful dig on last month’s post, which really should have been called Animals Don’t Think Like Humans!, even if in the event that post was about how some academics have gone about studying human thought via the study of the psychology of other animal species – and in that event, the post concentrated on the excesses of some philosophers. Similarly, this post will be about how some psychologists have gone about studying adult human thought via the study of the cognition of preverbal children (roughly, children under the age of 4), and about their excesses (the scholars’, in this case, not the children’s, though both go without saying). So, in a nutshell, this is Take 2: The Preverbal Case.

At first sight, the cognition of preverbal infants may be more relevant to the study of human thought than that of other animal species, these being human subjects, though not yet in possession of a natural language, but some caution must be exercised here too. No-one doubts that acquiring a natural language results in great cognitive benefits – a languageless mind is an impoverished mind. The issue here has to do with the sort of conclusions some scholars have derived about (adult) cognition on the basis of data from infant cognition.

The discussion will revolve around two questions: whether it is at all possible to neatly separate thinking and linguistic abilities in the cognition of infants and toddlers, and the question of whether it is possible to conclude that a specific conceptual ability depends upon acquiring a specific linguistic ability. It is in fact quite common to find claims to the effect that a specific conceptual ability depends on acquiring a specific language ability, and it is this particular conclusion that I will argue here ought to be resisted.

The overall issue is immediately complicated by the fact that, contrary to common perception, there is no such thing as a “preverbal” infant. Or at least not in any strict sense, for many aspects of language – certainly phonology, but many others too – start developing as early as birth, if not even earlier in the womb. What people presumably have in mind when they talk about preverbal infants is children who are yet to acquire the words of a language and thus are yet to produce language (words and combinations of words), a process that typically starts at 9-12 months of age and takes a number of years.

This is not an unreasonable position to hold, given that linguistic production is a pretty clear demonstration of mastery of language. However, if the study of language acquisition has taught us anything in the last thirty or so years is that linguistic comprehension precedes production by a number of years, and this changes what sort of data are obtainable at various stages of linguistic development. Indeed, psychologists have continued to lower the age at which children demonstrate this or that linguistic ability, mostly on account of employing perception-based experiments instead of (or at least in combination with) production-based ones, the latter the norm in the past.

What I mean by this is the following. If you run an experiment in which children are elicited to produce (that is, utter) this or that kind of sentence spontaneously, it is often the case that the age at which they are able to do so is not all that early in their lives, whereas if you run an experiment in which children are simply expected to listen to the same kind of sentences and then they undertake a simple task in which they can show whether they interpret the sentences appropriately or not, the typical result is that the age at which they manifest proper understanding of the relevant sentences tends to be much lower. Needless to say, this has important repercussions for theories of language and cognitive development.

Furthermore, linguistic knowledge develops in a rather systematic and gradual, therefore peculiar, way in the minds of children, and something along these lines is also true of many of the psychological processes we directly associate with thought and thinking, such as categorisation, inference, learning, and decision-making (recall that in previous posts I argued that having a thought, and thinking, involves some kind of fixation of belief; generating new beliefs, updating old ones, etc.). Some of these mental processes certainly surface in the minds of very young infants, and very often before they can be regarded as competent users of language – before, as mentioned, they have learned any words or the syntax of the language they are exposed to – but the connection between linguistic and cognitive development, if a robust distinction can be drawn at all, is not at all clear.

The literature offers many examples of these facts, from infants’ capacity to manipulate conceptual categories from the age of three months to the ability to entertain rather intricate inferences to do with causes, space, or motion before their second birthday – abilities that must surely be regarded as examples of thought. Such data have been taken as proof that at least some thinking is carried out in non-linguistic representations, and I am certainly not going to cast doubt on this claim, though I do wish to deny it the weight it usually carries.

My main point in this post, to be as clear as possible, will be that many aspects of infant cognition mature during ontogeny, and this makes identifying which properties of the mind are responsible for which behavioural data a rather muddled business.

The literature also offers many examples of accounts that sketch what I have just claimed to be so very hard to demonstrate. As a case in point, Jill de Villiers and colleagues have argued that the development of “theory of mind” abilities – the ability, that is, to “read” other people by ascribing mental states such as beliefs and desires to them – depends, to some extent, upon mastering the so-called self-embedded sentences typical of many languages – e.g., sentences such as I think that Michelle thinks that X, which contain a sentence inside a sentence (hence, the self-embedded epithet), and which are claimed to closely track the structure, or form, of the thought representations underlying theory of mind abilities.

Such proposals are usually phrased in terms of the ontogenetic milestones infants go through. In particular, theory of mind is said to be fully developed by age four, which is precisely the time at which children are said to master the corresponding linguistic structures (the self-embedded sentences), thus suggesting a rather close connection between this stage of language acquisition and the type of thinking that theory of mind involves.

There are three main reasons to doubt such a close connection, however. First, very often the supposedly complementary abilities do not always develop simultaneously at all. Second, age four may be a cognitive milestone tout court, given that all sorts of cognitive abilities surface by that age, linguistic and non-linguistic. And third, mastering the embedded sentences that apparently go hand-in-hand with theory of mind abilities does not at all mean that these sentences are the actual vehicles employed in order to ascribe mental states to others – or even that these sentences are directly involved in belief ascription in general.

Regarding the first two points, consider the following. The claim that theory of mind fully matures by age 4 fits well with the claim that theory of mind develops alongside apparently analogous linguistic abilities, which are seemingly properly understood by age 4. However, and as alluded to earlier, experiments that use comprehension/perception tasks paint a different picture of when in ontogeny children demonstrate a specific ability, and this applies in this case too. In particular, and whilst it seems to be well-established that self-embedded sentences are appropriately comprehended by age 4, as per the de Villiers study, though children only produce them reliably when they are older (see here), perception-based experiments show that even one-year-olds demonstrate a sophisticated understanding of false belief, a key component of theory of mind. Yet other components of theory of mind surface before age 4, in fact, and thus it would appear that theory of mind and the comprehension of self-embedded sentences do not quite proceed pari passu (let alone the production of such sentences). Thus, theory of mind starts developing at a time when the cognition of self-embedded sentences is further into the future of children’s development.

The third point is all-together different in nature, and perhaps more conclusive. Namely, from the fact that linguistic structures could be made to represent or describe thoughts, it does not follow that it is these very representations that the mind uses to do this thinking. Language could just reflect thought rather than being its vehicle. What’s more, if it can be shown that natural language is not the (main) medium of thought, as I have tried to show in previous posts, then any proposal suggesting that the development of some aspect of thought is dependent upon the acquisition of some property of language would be further undermined.

Complementarily, it also does not follow that by acquiring a natural language preverbal mental representations are replaced with linguistic ones in general cognition or thought. It is of course very possible, indeed probable, that the acquisition of a natural language changes the language of thought I have been banging on about in various ways – pre- and post-lexical thought are clearly different in nature – but this is a different matter altogether.

Another important factor is the apparent fact that the linguistic data children are exposed to do not wear their conceptual underpinnings on their sleeves, as it were. Children, in addition to having to assign the requisite structure to the linguistic input they receive, must also be able to conceptualise the linguistic data somehow – they must compute and represent the meaning and conceptual representations being express by a sentence – and this requires some sort of support. This is a crucial point, but it is also an old story, more prominently present in the claim that in order to acquire a natural language an infant must be in possession of a medium in which to represent the conceptual relations that natural languages typically express – again, they must make use of a language of thought!

I think this point needs to be simply assumed in discussions such as this, though it is important to stress that such a position does not preclude the possibility that acquiring a natural language probably provides new resources to an infant, perhaps offering a cognitive scaffolding of some sort, to allude to the letter but not quite to the spirit of a widely-used metaphor of Lev Vygotsky, he of inner speech fame.

The literature offers a panoply of possibilities in this regard; to mention but two: acquiring a language may allow for the representation of thoughts that would otherwise be too complex to entertain in the language of thought, the position defended by Jerry Fodor here and there; or it could instead be the case that language acquisition provides a way to augment the expressive power of non-linguistic cognition, as argued recently by the psychologist Susan Carey and others. But this is all for another post (or another lifetime, in fact!).

I have defended two claims in this post, then: that it is not possible to segregate linguistic and non-linguistic properties neatly in the mind of infants/toddlers, and that linguistic expressions/structures may not be the actual vehicles of thought. Put together, this makes unearthing properties of the language of thought through the study of infant cognition, though certainly a possibility worth exploring, a rather complicated and muddied affair. Infant cognition may provide a purer picture of the language of thought, and I for one am keen to get my teeth into Elizabeth Spelke’s multi-volume work on what babies know in this respect.

So, another rather narrow point in the end, much like last month, but I hope an important one anyway. After all, the take-home message is not unrelated to the “this language has this word or phrase that is sui generis and thus expresses an untranslatable thought or concept”, a common enough claim that sees conceptual differences in every linguistic distinction, but which happens to be false and in need of refutation and debunking every time it dares popping up. Let these posts be fodder for cocktail party conversations, at least!


With the new academic year upon us, to change tack, and starting with next month’s post, I shall start writing on more topical matters to do with language, philosophy, and psychology. Some of the topics in the pipeline including the sudden replacement of Kiev with Kyiv, the necessary “expropriation” of all rich people, the eternal dumbness of artificial intelligence (actually, of machine learning), the mental effects of uttering/writing the n- and other verboten words, and some others. Stay tuned!