Autism, Artificial Neural Nets, and Language

It would benefit people to learn how artificial neural nets (ANNs) work in order to better understand the autistic brain. One of the things I’ve noticed in my readings on autism is that the way autistic people learn strongly resembles the way one trains up ANNs. It takes many iterations of a perception for the autistic person to develop the concept, whereas with neurotypicals only one or two will do.

Neurotypical people learn language in a more language instinct-driven fashion. A word needs be repeated only a few times, and the person has it. But if you tried to train an ANN, you would find that you would need to train it up many more times, and more than that, the first thing it would do is simply repeat back exactly what you said to it. That is, it would engage in echolalia.

Language is thus structured differently in the autistic brain than in the neurotypical brain. The neurotypical brain has a deep grammar on which the details of a given language are hung. If autistic brains lack in certain kinds of instincts, like the language instinct, but still have enough complex network structure to build language, then we would expect less-than-typical structures in speaking. The unusual intonations of many autistic people likely derive from this fact as well, since every sentence is being constructed in a more mechanical way–giving us something like the voice of an ANN that learned language–rather than in the easier way generated by an instinct.

So language, in those who can learn it, is one of those many social behaviors neurotypical children learn automatically, but which requires direct instruction for autistic children.

Progress in Language Skills

Recently Daniel has been showing the degree to which he has been making progress in his language skills. Two areas in which he has shown improvement have been the initiation of conversations and the use of figurative language–two areas typically difficult for those on the spectrum.

Anna’s mother recently had a surgery, and we had talked about it a little bit around the house. When Anna’s mother came up for Thanksgiving, as soon as she sat down upon arrival, Daniel went up to her and asked her about her surgery, and then told her he was happy she was doing better. This was the first time Daniel had ever initiated a conversation. It wasn’t the first time he started talking to someone–but simply starting to talk to someone isn’t the same as properly initiating a conversation. Daniel will ask you questions about sharks or Star Wars or planets, but he’s not going to ask you about how your day went or, well, pretty much anything at all about you. So him asking his grandmother how her surgery went–and doing it on his own–was major.

The other one happened yesterday. I was having the kids clean up their rooms and the toys scattered throughout the house, and finally Daniel decided he was finished. He announced that if he continued cleaning, in 5 minutes he would blow up. I told him I wanted to see that, so keep cleaning for 6 minutes. He said, “I’m not going to actually blow up. It’s like saying I’m hungry enough to eat an elephant. If I did that, I’d die. Like the old lady who swallowed the horse.” I told him, “Oh, then, since you were just being hyperbolic, keep cleaning.” He gave a huge groan–which he does any time you “got” him–and he finished cleaning.

The significance here, of course, is that he not only used metaphorical language–hyperbole is of course metaphorical–and use it correctly, but was able to explain it. That doesn’t mean he’s now going to get that people are being metaphorical and will no longer stop taking people literally (I’m a poet, playwright, and fiction writer, dealing with metaphors all the time in my reading and writing, and I still take people literally when I shouldn’t), but he’s on the way to doing it less often. It’s important to understand that people don’t always “mean what they say” in this sense.

Daniel receives a variety of special education services at his elementary school, Arapaho Classical Magnet, including speech. A good speech teacher and a good special education program in general will get results, and I think we’re seeing some of those results in Daniel’s development of these skills. After all, among the foci of his ARD is being able to initiate conversations. That’s not something I have tried to work on with him, though I have tried to help him with non-literal language (something I’m better with because of my being a language artist, while I still struggle with initiating conversations outside of bombarding you with my interests). So I’m glad at the progress his teachers have made with him, and I think it’s important to give credit when it’s due. After all, our first instinct often seems to be to just complain when something goes wrong–and perhaps we need to focus more on giving credit when something goes right.

16p11.2

Autism is probably not a single thing, but is likely rather a variety of syndromes. This is possible because in complex systems like the developing brain, a variety of causes can have the same effects, and the same cause can result in different effects in its interactions with other causes.

Alterations in 16p11.2 have been identified with autism and most recently with specific neuroanatomic differences. Note, though, in their descriptions, that they identify low IQ and poor language skills. If this is typical of 16p11.2 variants, it can hardly explain those of us who are identified as being on the spectrum and yet having high IQs and even a degree of language mastery. The Intense World Theory version of autism probably does better with people such as my son and I, as well as many of the more gifted autistics. But that would imply at least two major divisions within autism that it may be a good idea to completely separate from each other.

Daniel’s Literal Interpretations

Sometimes Daniel’s literalism can result in some funny situations.

____________________________________

Just the other day, Anna puckered her lips and told Daniel, “Give me a smack.”

The minute I heard it, I knew exactly what was going to happen. And sure enough, he got a quizzical look on his face, then lifted his hand in the air…

Fortunately, Anna also realized what she had said and caught his hand in time. She laughed and told him, “No, I meant give me a kiss.”

Daniel responded, “Well why don’t you just say what you mean?”

______________________________________

Every week Daniel has homework. One week the homework was to create a coin. I read the instructions to Daniel exactly as written: “Create a coin and put your face on it.”

So Daniel drew a circle on the paper, then laid his face in the middle of the circle and said, “I don’t know how this is going to work.”

______________________________________

One day Daniel had a caterpillar on his t-shirt. We all got in the van to go somewhere, and he didn’t want to leave the caterpillar behind. Because it was nice, we rolled our window down. Melina told Daniel, “Roll up your window. The caterpillar is going to fly out.”

“No it’s not!” Daniel said. “It doesn’t have wings!”

______________________________________

Last Christmas, Anna sang part of Mariah Carey’s Christmas song to Daniel, “All I want for Christmas . . . is you!”

Daniel gave her his quizzical look and said after a few seconds, “So . . . you want a Daniel statue?”

______________________________________

Needless to say, he also tends to take teasing literally and seriously. The good news is that more and more he’s starting to ask me, “Are you joking?” And just the other day I made some ridiculous claim as a joke, and my daughter said, “You’re lying,” but Daniel defended me and said, “No, sometimes daddy’s joking.” So it seems my relentless challenging of his literalism is starting to work. It will take a while, but metaphors, figures of speech, and jokes will eventually make sense to him. And if he’s like me, he’ll come to find them pretty fascinating.

GABA Receptor and Synaptic Pruning

Recent research suggests a role for GABA receptor in synaptic pruning. Autism (and schizophrenia) are often associated with a lack of synaptic pruning, meaning neurons are more active, with positive feedback dominating.

GABA is associated with negative feedback, meaning the brain slows down to a steady-state. Glutamine is similarly associated with negative feedback. Glutamate is associated with positive feedback. All of these are neurotransmitters. More, they are closely related to each other, and can be biochemically derived from each other.

This suggests a few potential pathways to autism. If there is a problem with the GABA receptor, you would not get enough pruning. But if there is not enough GABA being produced, you would have the same effect. A mutation on either the GABA receptor protein or on one of the enzymes associated with GABA production could have pretty much the same result.

Neurons with unpruned dendritic spines get more input than do those properly pruned. The more input a neuron (or other complex system) has, the more is acts as though there is positive feedback. Indeed, it can result in increasing cycles, driving more input. In essence the brain becomes more hyperactive, at least until a physical limit is reached, at which point the system crashes, cycling down.

The result is a more active brain that may have some difficulty learning new things, but which may at the same time show exceptional abilities because of the higher activity. While the senses themselves won’t show increased activity at the source, you would see increased activity in the brain, resulting in the sensory overload associated with autism. One would even expect a certain degree of “phantom” sensory information–as we see with schizophrenia. Indeed, this association between autism and schizophrenia (which I keep coming across in different ways) does suggest that the old categorization of autism with schizophrenia meant that the researchers at the time were on to something.

Also, unpruned dendritic spines is a feature of a child’s brain before they turn two (more or less). The fewer pruned dendritic spines (and less cell death of neurons, which also occurs around the age of two, in conjunction with the pruned dendritic spines) there is, the more an autistic person will act like they are two years old, perhaps even younger. This can explain the neotenous features of autism, even among those of us who are considered to be only moderately autistic. And if the brain is kept in a pre-verbal state by being kept in an even younger state than that of a two-year-old, it can go a long way to helping us understand why there are nonverbal autistics.

The Language Theory of Mind

Once while watching one of his favorite cartoons, Daniel said to me, “He’s afraid he’s going to get in trouble, but he knows it wasn’t him who did it.”

I’m not sure one could possibly have a better statement of theory of mind: “he knows it wasn’t him.”

For those who think people with autism do not have Theory of Mind, I present that as Exhibit A. After all, it requires ToM to understand that a person/character both knows that someone else thinks he did something, and that he knows he didn’t do what the other thinks he did.

The cartoon in question is slapstick with no language use at all. And yet, Daniel managed to figure out who knew what about whom, including the necessary recursiveness of knowing what he himself had or had not done.