Perfect storm

thoughts on autism and challenges with language acquisition

I don’t believe the famous French philosopher, Henri Bergson, was wrong when he referred to the human brain as a ‘reducing valve’ which filters the overwhelming stream of stimuli into a more comprehensible reality. We operate on a set of mental heuristics which help construct the world around us (Kahneman, 2011). These heuristics are also referred to as 'top down processing' in which the brain, acting as that reductive valve, points us in the direction of what is most relevant for our survival and well being. This top down processing also explains why we might drown out irrelevant noises, or arrive at work totally unable to remember the details of the drive there. These cognitive processes that interpret reality are incredibly important in trying to determine what to attend to and, ultimately, what information becomes our sets of knowledge. 

This is especially relevant for our acquisition of language. In opposition to the long accepted Chomskyan model of language development (known as nativism) in which an ‘innate language module’ exists in all humans from birth, most language specialists these days accept a usage-based theory of language acquisition (known as constructivism). In this model, championed by researchers such as Michael Tomasello and Paul Ibbotson, children use ‘various types of thinking… ‘not specific to language’ (Tomasello, 2016) in order to develop linguistic competency. These types of thinking are the same mental heuristics and top-down processes which reign in the wildness of reality. Though the nativist vs. constructivist argument still continues on today, this article is predicated on the assumption that children use these various and non-specific cognitive processes to acquire language. 

So what does it actually take to learn language? Like I mentioned before, the constructivist model of language acquisition posits that we use a ’Swiss army knife’ (Tomasello, 2016) of cognitive capacities including pattern recognition, memory, attention, categorization, reading of communicative intentions, and analogy making. Indeed, the task of acquiring language seems so monumental as to be almost magical, but it is simply the result of the brain’s amazing ability to think, process, and reason. The next paragraphs are roughly split into the areas of the language acquisition according to the Bloom and Lahey model. First, I am going to discuss speech and then move on to the elements of language — content, form, and use.

1. The first problem in language development involves the acoustics of speech. We need to parse a ‘continuous stream of sounds’ (Kuhl, 1996) into meaningful units, or phonemes. In acoustic analysis, there are no time intervals between words and the barriers between phonemes are undefined. From birth, babies are able to perceive all of the world’s phonemes, but with exposure to a native language, they begin to only make distinctions between phonemes in their target language. The classic example is the ‘r’ and ‘l’ distinction which does not exist in Japanese. This means that, after about 12 months, Japanese babies will not perceive a difference between ‘r’ and ‘l.' It is widely known that babies use statistical probabilities at a sub-conscious level to decide which sounds are important. This parsing ability is the result of ‘perceptual narrowing’ during brain development in which unnecessary neuronal connections are pruned. 

This is the first obstacle to overcome for autistic children. This process of house cleaning does not take place to the same degree in autistic children (Tang et al. 2014). Whereas a typically developing child might start to tune out the noisy crickets outside because nothing invariably co-occurs with the sound, an autistic child might continue to process these sounds. This makes the sensory landscape exceedingly difficult to parse and derive meaning from. Not only is the signal-to-noise ratio stacked against the child, but it is also nearly impossible to assign meaning to these symbols with very fuzzy phonemic boundaries. Of course, this difference in cognition is not such a clear good/bad dichotomy. Lack of neuronal pruning in childhood leads to synesthesia in some individuals and also can confer ‘increased auditory capacity’ (Remington et al. 2017) in tasks that involve attending to expected/non-expected sounds as well as heightened pitch perception. However, this lack of pruning poses a unique challenge to language development. 

2. The second obstacle in language acquisition is the problem of semiotics. Semiotics is the study of assigning meaning to symbols. Any symbol that gets assigned a meaning becomes a unit of content in the target language. Not only does the overactive connectivity in the brains of children with ASD continue to play a role in compromising the formation of salient connections between signs and symbols, but other factors also come into play. 

Let me try to explain a semiotic moment using a classic language development thought experiment known as the gavagai problem:

The gavagai problem describes a situation proposed by Quine in which a human is transported to another planet and discovers an alien pointing at a hopping object that looks like a rabbit. The alien says ‘gavagai’ while pointing at the creature in question. Most people would assume (and rightfully so) that ‘gavagai’ refers to the rabbit creature. However, Quine questions this assumption and asks why it might not refer to just the ears, the hopping motion, or something else entirely that the alien might’ve been thinking. How does an infant’s brain deal with all of these potential conflicts? It turns out our brains make a huge number of assumptions in order to constrain the near infinite possibilities of meaning. And we are right most of the time. 

Here’s how the gavagai problem works in typically developing children. The infant triangulates gaze between the alien and the creature. They infer the alien’s intention (i.e. communicating a label) and use joint attention to focus on the object in question. This resolves any guess that might just refer to the alien’s private thoughts. The brain also uses a mental heuristic known as 'whole object bias' in which it assumes that a sign refers to the entire object rather than one of its parts. Over consistent use in similar contexts (either temporal, spatial, or linguistic), children ‘recognize’ (either consciously or sub-consciously) the pattern and acquire a functional unit of language. Because differences in cognition compromise the ability to infer intention, use joint attention, and filter out sensory stimuli in ASD, the prospect of successfully labeling the gavagai becomes pretty grim.

3. The third problem is one of grammar, or form. The ability to string units of content together in a cohesive way is another remarkable achievement of human ingenuity. It turns out that we use several different methods for constructing the syntax of language. Two such methods are semantic and prosodic bootstrapping in which word meaning or suprasegmental features (e.g. tempo, pitch, amplitude) are used to decipher grammatical forms. Based on both difficulties with semantics as described in the previous section and difficulty using and understanding prosodic cues, autistic children are, again, at a disadvantage with regards to language development. Moreover, a recent study by Ibbotson showed that 'children’s ability to produce a correct irregular past tense verb—such as “Every day I fly, yesterday I flew” (not “flyed”)—was associated with their ability to inhibit a tempting response that was unrelated to grammar.’ Although, the evidence for the following is inconclusive, some studies suggest that children with autism struggle with response inhibition. This is especially true in tasks that impose a working memory load in addition to requiring response inhibition (Hughes 1996; Hughes and Russell 1993; Minshew et al. 1999; Russell 1997) or when they are required to shift from one response set to another (Ozonoff and Strayer 1997; Ozonoff et al. 1994). 

4. Finally, is the problem of language use. Language is our primary form of communication and, as such, requires groups to function. Sign and symbol relationships need to have meaning that stays static throughout the group so we can have shared ideas. Language is the extraordinarily complicated tool that allows us to connect with each other, resolve conflict, and navigate nuanced social situations. Social connection is typically thought of as an instinct in typically developing humans and is one of the primary motivations for developing such a difficult skill. This social instinct is compromised in autistic children. Because of this, language is not as functional of a tool to yield. If other motivating forces (e.g. hunger, change, etc.) don’t require language to be fulfilled, this skill will most likely not develop. It is difficult to know the direction of causality, but this may also be another reason why autistic children do not infer intention or assume social narratives as readily as typically developing children. Heider and Simmel did a famous experiment in 1944 in which people were made to watch a film in which shapes were moving in particular directions. The brains of typical subjects saw an ‘aggressive large triangle, bullying a smaller triangle and a terrified circle.' Surprisingly, even the brains of infants under one years old can identify bullies and victims based on random movement of shapes. However, autistic children did not experience this automatic assumption of intention/emotion. 

This is just speculation, but let’s imagine what happens during language acquisition if you don’t superimpose a social narrative on a situation. Let’s take the gavagai problem again and expand upon it. Imagine now that the alien is a mother alien speaking to a child alien, and the mother says, ‘goop the gavagai.’ The child then grabs the gavagai. We clearly see a connection between the parts. The mom, existing in a position of authority, issues a command, and it is immediately followed by relevant action. This relevant/contingent action is used by typically developing children to acquire language meaning (e.g. the meaning of grab and rabbit in this example). If, in the case of autism, the social narrative isn’t assumed, the connections of the parts are more random. There is too much ‘noise’ in the signal. Any action or object that is in the child’s field of attention may come to be associated with these word leaving extremely high margins of error. The superimposition of social narrative (which is essentially a cognitive bias) is another amazing tool we use to constrain the possibilities of linguistic meaning.

Of course, autistic children are an incredibly diverse group of individuals with incredible strengths in some domains and challenges in others. However, this new language paradigm definitely sheds some light on the reason for difficulties with regards to language development. The differences in a large number of the ’non-specific cognitive processes’ involved in language learning come together to form a perfect storm with regards to the acquisition of language for autistic children. If this is true, this means that we should teach language to autistic children with more care. We should be especially in tune with their loci of focus and not assume that they infer intention or emotional content. 

Language is an extraordinary tool and, with it, we can speak life into infinity. Let’s all work harder to cultivate the appropriate environments for all children to acquire the capacity for this uniquely human gift.