May 2021: Internalisation

Vygotsky on inner speech and the relationship between thought and word

I’ve scheduled this so that it goes out while I’m getting my first covid jab. Woohoo! Also Friday was my last day at work and I now have enormously more free time than I’m used to. ALSO also it’s finally been sunny the last few days after an incredibly rainy and cold May. Woohoo again!

In this newsletter I’m going to pick up where I left off with the last one, and cover the remaining chapter of Vygotsky’s Thought and Language, ‘Thought and Word’. This is the most interesting part of the book, but it’s also densely packed with ideas and I found the flow confusing. So if the notes read a bit like ‘and then Vygotsky said something else’, that’s why.

After that there’s a short section on other related things I’ve read this month, including Phil Agre’s ‘Writing and Representation’, and some very early stage thoughts on how I can use all this stuff I've been learning about writing to think about mathematics instead.

‘Thought and Word’ notes

This chapter transitions fast between arguments with other schools of psychology, experimental evidence, anecdotes and literary quotations. It’s hard to follow the thread of argument a lot of the time. I’ll do my best below but this is going to look quite disjointed.

Earlier theories

Vygotsky starts the chapter by dunking on two earlier theories of the relationship between thoughts and words. The first one is an associative theory where individual words and their meanings are linked, and these links can be strengthened by repeated association between the sound and the object the word stands for. There are obvious arguments against this, like that it applies well to concrete nouns but not so well to abstract ones, or other parts of speech. There’s also a more subtle one that Vygotsky gives, that the association between word and referent can change qualitatively over time. E.g. take this example from last newsletter, where a toddler calls puffed rice ‘popcorn’. For him ‘popcorn’ means… well, it’s hard to say without more evidence, but probably some kind of puffy crunchy thing. But over time he’ll pick up the commonly-agreed social definition, and ‘popcorn’ will mean the stuff that comes in boxes at the cinema.

The second one is some different theory the Gestalt psychologists had, I didn’t try to follow that one but the objections seemed to be similar.

Two planes of speech

Vygotsky is interested in a much more dynamic interplay between thought and word than these static word-to-meaning links:

... The relation of thought to word is not a thing but a process, a continual movement back and forth from thought to word and from word to thought. In that process, the relation of thought to word undergoes changes that themselves may be regarded as development in the functional sense. Thought is not merely expressed in words; it comes into existence through them.

He talks about the differences between what he calls ‘the two planes of speech’, the external phonetic part (what we hear) and the internal semantic part (what we mean).

He points out that in external speech, children move from saying individual words, to short phrases of a few words, to whole sentences. By contrast in meaning, ‘the first word of the child is a whole sentence’:

A child's thought, precisely because it is born as a dim, amorphous whole, must find expression in a single word. As his thought becomes more differentiated, the child is less apt to express it in single words, but constructs a composite whole. Conversely, progress in speech to the differentiated whole of a sentence helps the child's thoughts to progress from a homogeneous whole to well-defined parts.

He also discusses cases where the same set of words can have a different emphasis in meaning depending on the situation. One of his examples is the sentence ‘The clock fell’. If you found a broken clock and someone tells you this, you’re interested in learning that it fell. If you heard a loud thud in the next room and someone tells you the same sentence, you’re interested in learning that it was a clock.

After this there’s a really great example of cognitive coupling:

In one experiment, the children were told that in a game a dog would be called "cow." Here is a typical sample of questions and answers:

"Does a cow have horns?"


"But don't you remember that the cow is really a dog? Come now, does a dog have horns?"

"Sure, if it is a cow, if it's called cow, it has horns. That kind of dog has got to have little horns."

Piaget again

In the next section we’re back to dunking on Piaget, specifically his theory of ‘egocentric’ vs ‘social’ speech in children (roughly, ‘egocentric’ speech is like an out-loud running commentary to yourself, ‘social’ speech is the ‘normal’ back-and-forth kind where you’re attempting to communicate to others). This came up in earlier chapters, and I wrote a little bit about it in the last newsletter. He conveniently restates the differences:

Piaget believes that egocentric speech stems from the insufficient socialization of speech and that its only development is decrease and eventual death. Its culmination lies in the past. Inner speech is something new brought in from the outside along with socialization. We believe that egocentric speech stems from the insufficient individualization of primary social speech. Its culmination lies in the future. It develops into inner speech.

Vygotsky contends that egocentric speech is still social in intent, and children who appear to be talking to themselves are still trying to communicate:

… Three- to five-year olds while playing together often speak only to themselves. What looks like a conversation turns out to be a collective monologue. But even such a monologue, being the most spectacular example of child "egocentrism," actually reveals the social engagement of the child's psyche. A collective monologue does not require either a purposive isolation or autism. Children who are participants of the collective monologue do believe that they communicate with each other. They believe that their thoughts, even those that are poorly expressed or unarticulated, belong to all participants.

He runs a series of experiments to test this. For example, he gets children to play with others who are deaf or speaking a foreign language, so that they have no expectation of being able to communicate, and notes that their level of egocentric speech drops dramatically. Or he has the children playing alone, or in a room with so much noise that their voice is drowned out, with similar results. There’s not much detail given, just a rough sketch of what they did.

Inner speech

Now we get to the interesting part on inner speech. This is not just normal verbal speech but ‘in the head’, it’s much more fragmented and incomplete:

… while in external speech thought is embodied in words, in inner speech words die as they bring forth thought. Inner speech is to a large extent thinking in pure meanings. It is a dynamic, shifting, unstable thing, fluttering between word and thought, the two more or less stable, more or less firmly delineated components of verbal thought.

This isn’t completely unique to inner speech, but a matter of degree: normal speech can be more or less disconnected and abbreviated, depending on context. If you already have a shared context with your audience, very little needs to be said:

… let us imagine that several people are waiting for a bus. No one will say, on seeing the bus approach, "The bus for which we are waiting is coming." The sentence is likely to be an abbreviated "Coming," or some such expression, because the subject is plain from the situation.

Vygotsky compares this to the situation with writing, which requires a much greater number of words to give context to an unknown reader (this part is interestingly similar to Ong’s Orality and Literacy). Inner speech is at the other extreme: you’re the speaker and the audience, and you have more or less all the context, so you don’t need to say much, just whatever handles will help trigger the nonverbal sense you want. There’s a nice comparison here with the concentration of meaning in central symbols in literary texts:

… Another excellent example is Gogol's Dead Souls. Originally, the title referred to dead serfs whose names had not yet been removed from the official lists and who could still be bought and sold as if they were alive. It is in this sense that the words are used throughout the book, which is built up around this traffic in the dead. But through their intimate relation with the work as a whole, these two words acquire a new significance, an infinitely broader sense. When we reach the end of the book, "dead souls" means to us not so much the defunct serfs as all the characters in the story who are alive physically but dead spiritually.

In inner speech, the phenomenon reaches its peak. A single word is so saturated with sense that, like the title Dead Souls, it becomes a concentrate of sense. To unfold it into overt speech, one would need a multitude of words.

Finally, Vygotsky makes some comparisons in the opposite direction, between inner speech and nonverbal thought, which is even more unarticulated and diffuse - for example, he talks about the case where the word for something will not come to mind, but there is still a nonverbal sense of its meaning. Nonverbal thought tends to have an all-at-once quality:

Thought, unlike speech, does not consist of separate units. When I wish to communicate the thought that today I saw a barefoot boy in a blue shirt running down the street, I do not see every item separately: the boy, the shirt, its blue color, his running, the absence of shoes. I conceive of all this in one thought, but I put it into separate words. A speaker often takes several minutes to disclose one thought. In his mind the whole thought is present at once. but in speech it has to be developed successively. A thought may be compared to a cloud shedding a shower of words.

This last section of the chapter has a particular confusing flow. There’s a discussion of subtext, where there is some unspoken meaning behind the words:

Just as one sentence may express different thoughts, one thought may be expressed in different sentences. For instance, "The clock fell," in answer to the question "Why did the clock stop?" could mean, "It is not my fault that the clock is out of order; it fell" The same thought, self-justification, could take, among others, the form "It is not my habit to touch other people's things. I was just dusting here."

and then a kind of summing up. I’ll leave it here:

The connection between thought and word, however, is neither preformed nor constant. It emerges in the course of development, and itself evolves. To the biblical "In the beginning was the Word" Goethe makes Faust reply, "In the beginning was the deed.' The intent here is to detract from the value of the word, but we can accept this version if we emphasize it differently: in the beginning was the deed. The word was not the beginning — action was there first; it is the end of development, crowning the deed.

Writing and Representation

I happened to read a couple of other relevant things through the month. Via twitter I learned about Mike Travers’s book chapter with Marc Davis, ‘A Brief Overview of the Narrative Intelligence Reading Group’. This has a whole bunch of interesting looking links. The one I ended up reading was Phil Agre’s Writing and Representation. I think I must have read this before, some of the examples look familiar, but I got a lot more out of it this time. Looking at the references, it’s striking how much overlap there is with what I’ve been reading in the last few years: Vygotsky’s there (different book though), also Ong’s Orality and Literacy, Latour’s ‘Visualisation and Cognition’ paper, stuff by Derrida and Norris… whatever rabbit hole this is, I seem to already be some way down it.

This time I was most interested in the ‘Writing as bad and good metaphor for representation’ section, which is partly about the differences between writing and speech:

Roy Harris (1987), among others, has argued that ideas about representation in philosophy and linguistics have been biased by writing. He observes that these fields have emphasized those aspects of human utterances that appear in a conventional written representation. One might read in a textbook a sentence such as,

Suppose that John says to Mary, "Please close the window".

and this sentence will be taken to specify some hypothetical event. We do not normally wonder, and only rarely are we told, about several aspects of John's action:

  • his tone of voice,

  • his articulation of the various phonemes,

  • the shape of his intonation,

  • the timing of the various elements within the utterance,

  • the timing of his utterance relative to other actions and events,

  • whether he and Mary have a history of interactions over this window,

  • his position relative to Mary and the window,

  • his posture,

  • his gestures,

  • his facial expression,

  • the direction of his gaze,

  • whether and when he has caught Mary's gaze.

(For the horrors of trying to make written notations of these things, see Atkinson and Heritage (1984) or Levinson (1983) for an introduction to Jefferson's notation system used in conversation analysis.)

I’m still very interested in whether any of this cluster of ideas about the differences between writing and speech and thought can be useful in thinking about mathematics… it’s a vaguely similar setup with marks on on paper used to stabilise and extend our limited abilities to think about quantity and shape unaided, but there are obvious differences (no real ‘speech’ layer, for one, and much more unforgiving rules for combining symbols than in normal writing).

I don’t have anything clever to say about this yet but there are a few very half-baked thoughts I want to start writing up. I also want to do more reading in this area, whatever this area is - it seems to fall down some hole between fields so I don’t even know where the best place to look is. I’m planning to finally read the rest of Catarina Dutilh Novaes’s Formal Languages in Logic, the most relevant thing I currently know about. I should also read at least some history of numerical systems - David MacIver suggested Numbers and the Making of Us as an option. Other suggestions are welcome!

Next month

I’ve got a lot of free time now! So hopefully I can do more than normal. Along with the writing-and-mathematics stuff I talked about in the section above, I’ve got a half-written blog post in the style of these old ‘crackpot time’ posts, basically an update on the various physics things I’ve done in the last couple of years, so I’ll finish that.

I’ve just started fiddling around with note-taking in Logseq, one of the many Roam-like competitors that are cropping up. I liked the look of this one because it stores everything locally in markdown files on my filesystem, and it’s fast, no waiting for it to load. It’s pretty minimalist and I’m sure there are lots of features that it’s missing, but all I really want for now is [[whatever these links are called]] and it has those. I’m going to try making some book notes in there, and also add links to my existing writing to see if I spot any interesting associations that I’m currently missing.

I’m also going to try at least one more research speedrun, as I’m building up quite a list of questions. I’m planning to do one at the end of this week, hopefully as a sort of group event - fill in this poll to fix a time if you’re interested!