A brief introduction to the theory
This theory has been under development ever since my undergraduate studies in Linguistics and Pragmatics, where I studied Relevance Theory under Deirdre Wilson at University College London. It has been further informed by 15 years of working with children as a speech therapist, with about half of this time working primarily with children with Autism Spectrum Disorders. The final ingredient is introspection.
The theory (I’m toying with the name ‘Integrated Cognitive Development Theory’) adopts an ‘ages and stages’ approach, broadly taking after Piaget’s observation of distinct phases of early cognitive development. For the purposes of this summary of the theory, I will characterise these stages by how the child’s social domain is operating with their wider semantic system.
0-2 – Unintegrated social processing
The social domain and the wider semantic system co-activate, constructing parallel systems of understanding and action. These are analogous to one another in a way that isn’t yet apparent to the child.
2-7 – Integrated social processing
Represented mental environments in the child’s social domain are now treated as being representationally equivalent to the first order representation. The child can now contrast represented thoughts with their own thoughts. They also update the social domain with a truer ‘people processor’ that is informed by their unfolding awareness and understanding of themselves.
7+ – Recursive social processing
The child has now effectively realised that the second order contains a social domain, which itself contains representations that can be given equivalence with the first order representation.
Caveat
This is a developmental schedule, and as such it is an exercise in generalisation. Individual children can significantly vary from these averages due to genetic, environmental, and individual factors. If you pushed me, I would put the 16th and 84th percentiles at 18 and 24 months for the first transition, and 7.5 – 8.25 for the second.
Furthermore, the horizontal child will engage in passages of essentially integrated processing in certain contexts, and likewise the integrated child will engage in passages of essentially recursive processing.
Some definitions
I wanted to clarify my understanding of some key terms used in the summary above.
Social domain – I’ve gone through multiple names and characterisations already for this. I think the purest description of it is as a simple concept called OTHER PEOPLE, which just happens to have some specialised abilities (described below). Sensory experience from social sources is systematised to form this top-level concept, which contains subconcepts such as MUM and GRANDDAD. I’m not sure where to place ANIMALS with this understanding. I think it depends on whether the child was raised by wolves or not.
Wider semantic system – In my understanding, there are four important processors in the brain that all start out ‘speaking a different language’. One I call the child’s ‘sensory impression’. The child’s first job is to integrate this with his motor competence. He then needs to see both in the light of his emerging conceptual understanding. These motor, conceptual and sensory systems, now integrated, act together to help the child act upon his environment. They are ready to integrate with fourth processor, the social domain, which is off in the corner buzzing away with co-regulation hormones. They all want in on it!
Horizontal processing – The social domain is operating alongside the wider semantic system, but it is not operating in concert with it. The theoretical touchstone here is Fodor’s concept of informational encapsulation, but under this framework, this encapsulation is temporary. Indeed, the child, confronted by incompatible filetypes, will find that their course of action and thought is hijacked by a deep requirement to achieve integration.
First order representation – This is the child’s first order representation of the world, combining their systems of understanding and competence in all four domains. During integrated processing, I say (above) that the child updates their first order representation with reference to the second order representation, but at this stage this second order representation does not yet fully acknowledge the function of the social domain, so it can only operate fully on the motor, conceptual and sensory domains.
Linguistic and pragmatic processing under this framework
0-2 – Language is encapsulated
The social domain is built out of our accumulated sensory experience of other people, and the things they do. It bathes in hormones during co-regulation. It observes what will turn out to be social cues, and theorises about how these might have been linked with any subsequent action. It is building an understanding of motivation.
Some of the input is verbal. This is systematised, creating a lexicon linking phonological and semantic features, all within the social domain. The individual words and concepts in this lexicon will be activating simultaneously with the semantic avenues in the child’s wider thinking that they conventionally indicate, but at a conscious level, the child will not be aware of it.
In this age of horizontal representation, the way that words work – to summon certain favoured experiences for example – feels like magic to the child, since these lexical items are being activated within a module-like concept that they’re not really in control of.
The social domain is also the home of the ‘mirror neuron’, where our personal understanding of action sits, holding the wheel, with its clutch just under the biting point. This would place our articulatory capacity within the social domain as well.
Daily activities that are both emotionally engaging and meaningful for the child, stimulate the growth and the co-activation of dual schemata:
- In the social domain – a schema that encompasses and represents the words, feelings, intentions and actions of other people in respect of the activity
- In the wider system – a schema that encompasses and represents the child’s sensory impression, their ideas, and their understanding and use of objects in respect of the activity.
Language isn’t really interpreted at all, at this stage, but predictability makes it so. Expressive language becomes more organised as this stage progresses, but it is never fully motivated.
Eventually a tipping point is reached, and the pace of alignment of the ‘file types’ tops out. The way these two representations will relate to one another is fixed forever. This applies downward pressure on the first order representation, which now must concern itself with backwards compatibility in its interface with the social domain.
2-7 – Language is discovered
Subconcepts in the social domain, like MUM and DAD, are enriched by generalisations and understandings from the social domain as a whole, which itself has been enriched by its encounters with the wider semantic system.
The child also writes the concept ME in the social domain, probably the last major concept to be created there. This allows the child to see themselves in proportion with their overall understanding of typical human behaviour.
For our purposes, we will understand the social domain to be running a scratchpad of second order representations. This scratchpad has two functions:
- It ‘mind-reads’ – taking in social cues, words and actions, seeing these through the prism of the individual who is emanating them, and returning a probable mental environment, composed of emotional readings, understandings of intentionality, and beliefs. The child then ‘drops this representation down a level’, checking it against their own (first order) understanding, which they update as appropriate. The child, being alive now to the social cues of their activity partner, experiences a socially-mediated lexical and semantic explosion.
- At this stage, context is just something that arises during shared mental and physical activity. As far as the child is concerned, it cannot be directed or focused.
- Since the child does not recognise that a pair of interlocutors are able to mutually will a shift in the context of interpretation, the child’s interpretations are mechanical and literal.
- The child’s expressive language is motivated by the projection of desired mental states for their conversation partner on the scratchpad.
- The child must learn to stage these desired mental environments alongside their conversation partner’s actual current (assumed) mental environment.
- The child then navigates the semantic journey between the two representations. This results in the activation of lexical items and their associated motor speech features. Another piece of apparatus, likely outside of the social domain, with a sense of scope for the utterance at hand (and a secret desire to systematise motor routines), apprehends these activated lexical items and builds a phrase out of them.
- Initially, during this stage, processing limitations will prevent the easy staging of the two representations, forcing the child to ‘flick from one to the other’.
Whither the lexicon? Once the social domain finds a way to ‘talk to’ the wider semantic system, what happens to it? At an important level, the question doesn’t really make sense. The social domain and the wider semantic system are acting as one, inasmuch as they they serve to meet our linguistic needs (in normal functioning at least) so in a sense the function is distributed across both systems. Newly encountered lexical conventions will be physically located out in the wider semantic system, but at least until the age of 7, the child’s phonological processor and their motor speech capacity are still able to grow with the new experience that comes from hearing and saying these words.
The theory says that the earliest semantic formations, and the bulk of the child’s phonological and motor speech apparatus, will remain inside the social domain, which will go on to sustain the relationship with the ‘phrase-building’ apparatus. This seems to account for some linguistic presentations that we observe after stroke, and in dementia.
Towards the end of the stage, the child is conjuring with second order representations. While they are aware that someone else can have an intention towards them, they do not yet see it as an intention to direct their mental environment.
7+ – Language is mature
The child has spent a lot of time querying their partner’s mental environment, during shared activity and conversation. They come to the realisation that their partner is in possession of the same social hardware as them, running the same brand of second order representations. They discover that if they drop this hazy representation down two levels, it can interact productively with their own first order representation. Indeed it is a full referent of that representation
They are now ready to use their understanding of a speaker’s intention for them to guide their interpretation. Language processing shifts from the mechanical to the dynamic. The child, as listener, is able to make a running guess at the speaker’s idea of what they (the child) should be understanding from the speaker’s utterance. They then use this as a reference point to guide an iterative process of utterance interpretation. They are now integrating the third order representation (the child’s understanding of the speaker’s anticipation of the child’s reaction) with their own wider mental environment.
Specifically, the child posits an array of presumed mental states that they can imagine the speaker motivating and intending. The child ‘competes’ these against one another (based on a quick reckoning of the balance of cognitive effects that each intended interpretation would derive against the processing cost involved in getting there), enriching all of the representations as they go (as appropriate).
They then attribute the victorious representation to the speaker – his representation of the child’s mental environment after his words (and any extralinguistic information) have had their effect. Then, most importantly, the child uses this double-attributed mental state to update their thoughts and understandings.
(This ‘cost / benefit death match’ is hosted when the child attributes a second order representation as well).
In summary then, from the age of seven, the child is able to construct an anticipatory model of the speaker’s intended mental state for them, post-utterance. With this upgrade, they are operating on the same discourse level as the adults around them.
Full utterance interpretation under the ICDT
This explanation seems to me more elegant than the second communicative principle of relevance, because there is no need to make reference to the ‘communicator’s abilities and preferences’, since both the communicator’s abilities and preferences, and their understanding of the listener’s abilities and preferences, are natural constraints that derive from the manner in which these mental environments are being metarepresented (i.e., by fallible, quirky humans) and attributed (i.e., to fallible, quirky humans).
Full utterance interpretation involves the listener representing the speaker’s representation of how their words should land, and then using this representation to guide the way that their semantic system responds to the words they are hearing. This iterative process results in interpretations whose referents are fully defined and identified, and where all possible inferences are noted and tested.
Now that the child understands that carefully chosen words can be intended to direct a listener’s mental environment down any avenue at all, anything that both parties know to exist within the shared context of interpretation can be invoked alongside the shifting interpretation.
In the same way as we can invoke a shift in the context, we can invoke a shift in voice, by choosing a form of expression that cannot reasonably be attributed to us. In these situations, a character is invoked alongside the context, and the representation attributed to them, inviting the listener to consider how it must be to go about ones day in such a way. A whole world of contextual deftness and attributional nuance is the result.
This model suggests that utterance interpretation is not about ‘listening to words’. Instead, it is a deep, recursive process of attributing and assessing mental states, where we understand each other by constructing and evaluating layered mental representations of each other’s anticipated responses. This recursive mental modelling enables nuanced, efficient communication.
Coda
The way that the theory accounts for development in ASD is outside of the scope of this summary. If you’re interested in this area, you could take a look at this AI-generated summary of the wider theory. There is also extensive discussion about cognitive development in ASD in the original statement of the theory here. If you read these, you will note that I keep changing my terms. ‘Social domain’, SE and OP are intended to be synonymous, as are the terms ‘wider semantic system’ and SM. The best characterisation of these terms is in the present summary, but I’m sure I’ll keep developing it.
The theory needs a great deal of work in all areas. I’m keen to delve more into how third order representations might be involved in the process of motivating and tailoring utterances. The ‘downward pressure effect’ is something that I want to investigate as well. I am keen to further characterise its effect on the first order ‘filetype’ as a result of the major integration at 2. I also imagine that the effect applies somehow at the age of 7, as a result of the double-attributed mental states needing to find a way to maintain a relationship with the first order representation. How, exactly, I couldn’t yet say.
[Edit note – 28/02/24 – I altered the names of the phases of cognition from ‘horizontal processing’ to ‘unintegrated social processing’, from ‘integrated processing’ to ‘integrated social processing’, and from ‘recursive processing’ to ‘recursive social processing’. I also slightly expanded the caveat about the timing of the transitions. I want to put a lot more caveats in – this statement of the theory is now out of date. A more recent statement can be found here].
