Notes: The Language Instinct by Steven Pinker

What exactly is mentalese

Janky, unedited, written to help myself retain and understand better.

The first few chapters were extremely interesting. Learned a couple of things:

• Mentalese - the language in which you have thoughts, is distinct from spoken language. You can have a concept for an idea even if you don’t have the word for it. The prospect of a Newspeak which modifies the ontology available to your mind by excising regime disfavored words is hogwosh.

  • This feels intuitively true. I often have a lot of difficulty writing down a thought which is perfectly clear in my head, and I have to keep iterating on the phrasing and wording (diffusion model like) in order to get the mentalese onto paper. And there’s a bunch of other evidence that one can do complex thinking without spoken language:

    • Babies can understand many concepts (like numbers) without having learned any words yet.

    • Monkeys can keep track of complex social relationships (obviously without language)

    • Scientists/novelists often claim that their ideas first occurred to them as a physical intuition before they translated them into language (Einstein being the paradigmatic example).

  • But I was frustrated that Pinker didn’t answer the obvious next question one might have: what exactly is this mentalese in which all our thoughts are communicated? Does it have a grammar? Are there things you can express in mentalese but not spoken language? How about the other way around? How valid is it to analogize mentalese to the latent space/residual steam representations in neural networks? What relationship does mentalese have to images? When I meditate and observe all my thoughts going by, am I looking at mentalese? Because often the thoughts feel like English.

    • I guess I shouldn’t complain too much, since Pinker has another book literally titled The Stuff of Thought. I haven’t read it yet, but based on the title, I’m assuming it’ll answer a lot of my questions.

• Language ability is not the same thing as general cognitive ability, and it’s possible to have one but not the other.

  • There are many disorders which lead to a person being born clinically retarded but with their language abilities intact. Examples include Williams syndrome and aphasia. If I were a deep learning skeptic, I’d mention these as analogies to make the case that LLM can have superficial fluency with language while lacking true intelligence.

  • Conversely, one can have normal intelligence but lack facility with language. The textbook example is aphasia, where someone who suffers a stroke or injury which destroy the part of their brain responsible for producing or understanding language, but leaves them capable of reasoning, thinking, and the rest.

    • Pinker doesn’t mention the following since this discovery was made after the book was published, but in 1998 scientists found the FOXP2 gene that helps give many animals (including humans) their verbal abilities. When humans are born with a mutation on this gene, they have normal range cognitive abilities (and have no deformities in the actual muscles associated with speech), but for some reason they cannot use grammar and complex language.

  • If language and g are shipped in separate packaged, why do they tend to be so strongly correlated most of the time? After all, people tend to score in the same quartile on their Reading and Math SATs.

• Language is intrinsic to humans. I’m just gonna list out a bunch of evidence that suggests so:

  • We’ve never encountered a tribe that didn’t have language.

  • Slave holders in some Caribbean slave societies imported slaves from different ethnicities, so that the slaves couldn’t organize a revolt. In order to communicate with each other, the adult slaves developed a combined pidgin - a half language which is a mishmash of the ones of offer with an incomplete grammar. But when children are brought up in these societies, every single time they spontaneously develop a creole out of that pidgin - a full language with complete grammar.

  • In the 70s, students at a school for deaf students in Nicaragua were being taught to communicate with Spanish and lip reading. Obviously, this made it difficult for the deaf students to communicate with each other. The students developed a sign pidgin and finally a creolized sign language with complex grammar, verb agreement, and all the other facets of a full language. When American linguists visited, they realized that a totally new sign language separate from American Sign Language had been spontaneously developed by these students. Today, it’s called Nicaraguan Sign Language.