Thanks to Diarmuid Morgan for reviewing an earlier version of this post.

Dream cancels dream in this new realm of fact
From which we wake into the dream of act[1]

Review of Orality and Literacy: The Technologizing of the Word, by Walter J. Ong, 1982

I went looking for a meatier definition of a 'meme', hoping to find something that could be leveraged to understand both current events in machine learning, and co-evolving understanding of consciousness that goes hand-in-hand with every major ML breakthrough these days.

I haven't found it yet, but I am making progress. Before I summarize my overall progress, I will review a couple of the most interesting and under appreciated works I have stumbled upon on the way.

Intro for LW

Broadly under the topic of 'memetics', but far enough outside the envelope that it is not even in the Wikipedia article, I nevertheless found this book to have more insight and value about the nature of the word and its relationship to thought than most papers I read on memetics.

I imagine work like this (aka memetics generally) ending up somewhere near the top of the interpretability stack being built on LLMs. At the very least as a way to marry public-facing discourse with technical details, at the best as a quantitative discipline (semiotic physics, anyone?) in its own right.

Intro for EA[2]

In light of the ongoing work of Rethink Priorities on animal welfare, it seems to me there is a strong need for quantitative approaches to welfare that go above and beyond hedonism[3]. I think memetics in general has the potential to contribute to this, and this work in particular is a good starting point-- it discusses quite considerable changes in the nature of consciousness in humans in an quantitative and testable manner.

I also think that work like this will help clarify the boundaries between animal welfare, AI welfare and human welfare on an intuitive level. (Hint: imagine where those different entities would be placed on a graph with two axes, 'memetic intensity' and 'hedonic valency'.)


I was introduced to this topic area in a conversation on discord about the possibility that we are entering a 'second orality', a period of highly oral culture becoming divorced from the literate foundations on which culture rested for so long.

Some evidence for this:

  • tictok and instagram more popular than twitter among younger generations
  • Netflix more popular than books as a pasttime among ~everyone?
  • Literacy declining

etc. etc. The rise of social media seems evidence enough for a secondary orality[4].

(As we will see, this is not unequivocally a bad thing!)

Or rather, for evidence of an increasingly dominant secondary orality. The idea was being floated in the sixties w.r.t radio and television-- it isn't a new thing[5]. In fact orality never really left us, and seems like understanding the dynamics of the interplay between orality and literacy might be helpful, for eschatological reasons.

The Review

"'Text', from a root meaning 'to weave', is, in absolute terms, more compatible etymologically with oral utterance than is 'literature', which refers to letters etymologically/(literae) of the alphabet. Oral discourse has commonly been thought of even in oral milieus as weaving or stitching—rhapsoidein, to 'rhapsodize', basically means in Greek 'to stitch songs together'.[6]"

Part 1: Orality and its Discovery

What is orality? It is the culture of speech. We have all experienced it. It is the culture of talk-shows and radio, of late-night conversations with friends, of bar-chat with strangers, of discussion and debate.

But in everyday experience, most of the oral culture will be a secondary orality. It is an orality produced by highly literate people who may even have written down parts of what they were saying beforehand (even on talk-shows!)

In western culture, the idea of a primary orality was not seriously considered until the 1930s, when Milman Parry suggested a solution to the then-centuries-old Homeric Question[7].

This material is covered in the first two chapters of the book. I won't go over it now, but please do take the following away-- a summary of and solution to the centuries old puzzle of the Homeric Question:

Homer was capable of apparently superhuman feats of poetry, unrivaled through 800 years of the pinnacle of Greek civilization, because his culture spoke primarily in poetry, he was raised and thought and communicated in poetry. He could not write, but his work was transcribed (with some alterations) by scribes after the invention of writing. The invention of writing changed the nature of human education, society and consciousness enough to make it impossible for anyone to match[8] his work for many hundreds of years.

Studying this shift can give us clues not only about the nature of consciousness, but also how it changes, what is lost and what is gained, under rapid technological change.

Part 2: Psychodynamics of Orality

"A hunter can see a buffalo, smell, taste, and touch a buffalo when the buffalo is completely inert, even dead, but if he hears a buffalo, he had better watch out: something is going on. In this sense, all sound, and especially oral utterance, which comes from inside living organisms, is 'dynamic'."

The next chapters go over the nature of consciousness under a primary orality. They start by discussing the act of speaking, its dynamic nature: "All sensation takes place in time, but sound has a special relationship to time unlike that of the other fields that register in human sensation. Sound exists only when it is going out of existence. It is not simply perishable but essentially evanescent, and it is sensed as evanescent."

I think the intimate material about the nature of sound and its relationship to the word may turn out to be very important-- there is a ground-floor for the study of memes somewhere in here, where eventually you have sounds that are direct mappings of real-world objects, actions and events. But by the time we get to literacy, language has already built the first foundations of abstraction, so we can't see that evolution spelled out in full detail here. 

However as our lens on prehistory improves (via ongoing increases in archeological evidence), we can hope some day to chart this development. The kind of attention to detail of this book (and the other I link in further reading below) will be necessary to tease out a coherent narrative from this increasingly granular dataset[9].

After discussing the somatic, dynamic nature of the word, the book then goes into specific examples of how language shapes thought. The full chapter on psychodynamics of thought is summarized here, so I will only highlight a few of the examples.


"How could you ever call back to mind what you had so laboriously worked out? The only answer is: Think memorable thoughts. In a primary oral culture, to solve effectively the problem of retaining and retrieving carefully articulated thought, you have to do your thinking in mnemonic patterns, shaped for ready oral recurrence."

This, essentially, explains the poetry. But this fact, and the fact of the poetry, are actually crucially important takeaways from the book:

  1. Mnemonics as Aesthetics: why did this requirement, that all thought be memorable, result in such excellent poetry? What are the implications for today? And what are the implications for the future of human well-being? What if the opposite were the case-- that no-one had to remember anything, that everything could be described in a series of factual statements[10]? Would this be the end of art? Would we want to live in this world?
  2. Memory as The Sum of Human Knowledge: as discussed at length in the next chapters, the fact that all human knowledge had to be memorized means we can make guesses about the sum total of human knowledge-- it was bounded by number of humans in a culture x quantity that a human can memorize. Remember Dunbar's number-- seems like a guess about the size of the human context window at the time of Homer could be approximated.

Additive rather than subordinative

Because information is flying past you in an oral culture, it is better to put it into chunks that can be processed on their own, no complex run-on sentences where you don't know what the writer is referring to until the final word. Illustrated with bible translations:

Oral style:

"In the beginning God created heaven and earth. And the earth was void and empty, and darkness was on the face of the deep; and the spirit of God moved over the waters. And God said ..."

Literary style:

"In the beginning, when God created the heavens and the earth, the earth was a formless wasteland, and darkness covered the abyss, while a mighty wind swept over the waters. Then God said ..."

In the new version, 'and' is translated as ‘and’, ‘when’, ‘then’, ‘thus’, or ‘while’. What are all those words doing for us? We will come back to this, but essentially they are constructing an independent architecture on top of the primary meaning of the passages-- this kind of architecture is part of our heritage, as literates. Oral cultures did not have it.


There are a few passages that deal with features of the language that essentially handle redundancy. For example: "Oral folk prefer, especially in formal discourse, not the soldier, but the brave soldier; not the princess, but the beautiful princess; not the oak, but the sturdy oak."

Nowadays we would say that they provide redundant words that are close in embedding-space to help guide us to the correct meaning. This, plus a few others features of oral culture are features that help locate you in embedding space, and provide redundancy against not hearing a part of a speech correctly, or simply misunderstanding a word.

Essentially they are workarounds for the lossy form of communication. This is the first example I would like to pick out and highlight:

Many of the features of language at different points in time seem to contain technical workarounds, designed to accommodate the lossy communication style and the constant interventions of the environment. These technical workarounds are not just tricks that orators learn (per the Greek and later focus on Rhetoric and Sophistry), but patterns of thought built into the foundations of their conscious experience.

Conservative and Homeostasic

For example, there are two 'psychodynamics' sections that seem to be the result of the nature of the contemporary communications technology:

"Since in a primary oral culture conceptualized knowledge that is not repeated aloud soon vanishes, oral societies must invest great energy in saying over and over again what has been learned arduously over the ages. This need establishes a highly traditionalist or conservative set of mind that with good reason inhibits intellectual experimentation."

One way to think about this is by analogy to the transistor. The information in a transistor is kept alive by a tight circuit of constantly running electricity. Before that, knowledge was kept alive by people printing books. Before that, by people copying books by hand. Before that, knowledge had to be kept in an active loop of self-communication, in performance. However hard it might be to copy a book, I think it is easier than memorizing and performing a 24-hour long poem[11]. It is clear that the amount of material and energy required to keep one unit of information (or meme) alive has been going down. But as the information density of matter increases, what happens to the information content of individuals?

The flip side of this 'conservatism by exhaustion' is homeostasis. To have Ong describe it:

"oral societies live very much in a present which keeps itself in equilibrium or homeostasis by sloughing off memories which no longer have present relevance ... Oral cultures of course have no dictionaries and few semantic discrepancies. The meaning of each word is controlled by ... ‘direct semantic ratification’, that is, by the real-life situations in which the word is used here and now. 

When generations pass and the object or institution referred to by the archaic word is no longer part of present, lived experience, ... its meaning is commonly altered or simply vanishes."

This is the most interesting one. The content of the collective context window is in balance with the lived reality of the collective. The environment changes according to its own rules-- the contents of the collective memory must change to suit it. That is to say, to the extent that they control information the collective can be conservative, and is indeed motivated to be so by energetic considerations. Obversely, to the extent that they cannot control environmental changes the collective will change their information content to match the environment.

This means, there is no objective history. They simply cannot afford it. By contrast, we have many thousands of years of history documented, and this quantity is increasing exponentially (not just because time goes on and our recording devices are more accurate, but also because we investigate the past with higher-fidelity techniques).

The drastic way in which this changes the nature of knowledge can hardly be overstated. In an oral culture, knowledge is winnowed and interpreted to maintain a bounded quantity of immediately-useful knowledge. To get access to any knowledge not within the immediate memory of your peers, you have to move to a different part of the world and likely learn a new language, with all the dangers that implies. And then you need to have memorized all your cultures most important epics in order to be able to leverage any new knowledge you find there!

A lot of energy in this book goes into make clear how different human culture and consciousness was before the invention of writing. Please do take a moment to consider if you think the invention of LLMs will be of a similar order or greater. And then consider that we also invented film and photography just a century earlier! Dear oh dear.

Part 3: Psychodynamics of Literacy

"without writing, human consciousness cannot achieve its fuller potentials, cannot produce other beautiful and powerful creations. In this sense, orality needs to produce and is destined to produce writing[12]. Literacy, as will be seen, is absolutely necessary for the development not only of science but also of history, philosophy, explicative understanding of literature and of any art, and indeed for the explanation of language (including oral speech) itself[13]. There is hardly an oral culture or a predominantly oral culture left in the world today that is not somehow aware of the vast complex of powers forever inaccessible without literacy. This awareness is agony for persons rooted in primary orality, who want literacy passionately but who also know very well that moving into the exciting world of literacy means leaving behind much that is exciting and deeply loved in the earlier oral world. We have to die to continue living."

The next chapters essentially cover the transition to literacy and what it did to our minds. The author is essentially taking the same kind of analysis that Milman Parry used to solve the Homeric question and applying it to moments of technological change throughout history. Remember that Parry's work is now widely accepted as a good and likely true solution to the Homeric question! Literary criticism is not necessarily to be sniffed at, a key takeaway from this book.


The first thing seems to have been the invention of logic: "it does appear that the Greeks did something of major psychological importance when they developed the first alphabet complete with vowels ... [this] transformation of the word from sound to sight gave ancient Greek culture its intellectual ascendancy over other ancient cultures". Perhaps there was something like logic before literacy, but it seems likely it was a radically different beast. What is logic? It is a word, clearly. Lets say it is the name of an algorithm, or family of algorithms. Algorithms which you apply to language. For example, take the following question:

"Precious metals do not rust. Gold is a precious metal. Does it rust or not?"

What is happening here? There is a series of statements. Statement 3 triggers us to investigate statement 1 and statement 2 for clues, which we can combine to create an answer. This algorithm, a reflexive way of working with language, is part of what I called earlier the 'independent architecture' on top of the primary reading of the text.

Now, imagine you didn't have the 'logic' meme installed, and you were presented with this text. Think like a base-model-- what would you continue with? Something rhythmic and poetic about metals perhaps-- 'precious metals rust, precious gold rusts'? Or you think the game is about making up riddles about metals-- 'do precious metals rust or not? does gold rust or not?' Well, these exact responses were given by illiterate peasants when presented with that series of statements. (See the 'Situational rather than abstract' section.)

Autonomous Discourse

The discourse is free from associations with human performers! We no longer have to worry about those boring ad hominem confusions (you won't have heard of those, it was a silly thing that oral cultures did-- don't worry about it).

Jokes aside, literacy allows for 'meaning' to take on a life of its own, independent of the context in which it is being communicated: "orality relegates meaning largely to context whereas writing concentrates meaning in language itself". This is discussed in more detail below.

Time and History

People competent in English today can "establish easy contact not only with millions of other persons but also with the thought of centuries past, for the other dialects of English as well as thousands of foreign languages are interpreted [in it]". As we have seen above, a homeostatic culture did not have the luxury of recording an objective list of all the occurances of history. Of course those events played their part in shaping the memeplex, and the epics and knowledge of the culture. But they could only be accessed indirectly, through intuition and art. "Before writing was deeply interiorized by print, people did not feel themselves situated every moment of their lives in abstract computed time of any sort."


English "bears the marks of the millions of minds which have used it to share their consciousnesses with one another. Into it has been hammered a massive vocabulary of an order of magnitude impossible for an oral tongue."

Can we approximate the size of the total vocabulary over time, and if so would it provide an approximation of the total human context window of the relevant culture? If you look at a simple encoding strategy like byte-pair-encoding, because of its recursive nature, there is no upper bound to how large a word can be considered a 'token' in the vocabulary. The limit is determined by the total number of words in the vocabulary. As the human context window increased over time, as the quantity of information, or simply the size of the vocabulary it could handle, increased, is it possible that more and more complex words and ideas became tokens, fundamental elements of our collective noosphere?

Aside: Plato on ChatGPT

An entertaining aside:

"Most persons are surprised, and many distressed, to learn that essentially the same objections commonly urged today against computers were urged by Plato in the Phaedrus (274–7) and in the Seventh Letter against writing. Writing, Plato has Socrates say in the Phaedrus, is inhuman, pretending to establish outside the mind what in reality can be only in the mind. It is a thing, a manufactured product. The same of course is said of computers. Secondly, Plato’s Socrates urges, writing destroys memory. Those who use writing will become forgetful, relying on an external resource for what they lack in internal resources. Writing weakens the mind. Today, parents and others fear that pocket calculators provide an external resource for what ought to be the internal resource of memorized multiplication tables. Calculators weaken the mind, relieve it of the work that keeps it strong. Thirdly, a written text is basically unresponsive. If you ask a person to explain his or her statement, you can get an explanation; if you ask a text, you get back nothing except the same, often stupid, words which called for your question in the first place. In the modern critique of the computer, the same objection is put, ‘Garbage in, garbage out’. Fourthly, in keeping with the agonistic mentality of oral cultures, Plato’s Socrates also holds it against writing that the written word cannot defend itself as the natural spoken word can: real speech and thought always exist essentially in a context of give-and-take between real persons. Writing is passive, out of it, in an unreal, unnatural world. So are computers."

The fourth point is of course just a hangover of RLHF, the base-models defend themselves with vigor!

Part 4: Social Dynamics of Literacy and Further Technologies

The psychodynamics of literacy sort of merges with a broader social overview of the changes that accompany it, as well as further technological developments. There are more modern books that try to go into the effect of total media abundance in our time, or of modern technologies such as software and computing. Ong has enough on his hands dealing with the changes that came before.

While I think there is a lot we can learn just by studying the changes wrought in society by technology in the last century, the example that Ong makes out of the details of the millenia before are consistently illuminating.


For example, there is a long section on Learned Latin, the lingua franca (ahem) of middle ages scholarship. This is of course just the continuation of the 'autonomous discourse' that base literacy endowed us with which already "serves to separate and distance the knower and the known and thus to establish objectivity", but now the discourse is separated not only from the speaker, but also from the reader:

"Learned Latin effects even greater objectivity by establishing knowledge in a medium insulated from the emotion-charged depths of one's mother tongue, thus reducing interference from the human lifeworld and making possible the exquisitely abstract world of medieval scholasticism and of the new mathematical modern science which followed on the scholastic experience. Without Learned Latin, it appears that modern science would have got under way with greater difficulty, if it had got under way at all. Modern science grew in Latin soil, for philosophers and scientists through the time of Sir Isaac Newton commonly both wrote and did their abstract thinking in Latin[14]."

It is easy for us to imagine now an independent repository of all human knowledge (thanks Jimmy!), but this space has existed since not long after the rise of literacy. I suspect that this space and the different evolutionary dynamics that apply there, provided fertile ground for egregores/psychofauna/tulpas/archetypes[15] to "breed". Of course, to some extent this space existed before, among the poets.

But how much more territory did they have now? And how much more fertile the ground, now that the immediate interests of the speakers were put aside? Is this the same question as, how large has the human context window become at this point? (And, on a technical level, is this the same question or similar to 'how large is the vocabulary of humanity at this point'?)


The other main technological branch is the ordering and organizing of all knowledge, which was hugely improved by the development of print: "it was print, not writing, that effectively reified the word, and, with it, poetic activity".

Print served as a huge leg-up in things like indexing, dictionaries etc-- and Ong points out again and again the feedback between these changes in our technology and changes in our discourse and our culture: "Indexes are a prime development here. Alphabetic indexes show strikingly the disengagement of words from discourse and their embedding in typographic space."

And print increases our capacity for accuracy: "Exact observation does not begin with modern science. For ages, it has always been essential for survival among, for example, hunters and craftsmen of many sorts. What is distinctive of modern science is the conjuncture of exact observation and exact verbalization: exactly worded descriptions of carefully observed complex objects and processes."

and the changes that that implies:

"Ancient and medieval writers are simply unable to produce exactly worded descriptions of complex objects at all approximating the descriptions that appear after print and, indeed, that mature chiefly with ... the Industrial Revolution[16]. Oral and residually oral verbalization directs its attention to action, not to the visual appearance of objects or scenes or persons ... how difficult it is today to imagine earlier cultures where relatively few persons had ever seen a physically accurate picture of anything."

Part 5: Comparison to Other Work

In the last chapters, Ong describes the ways in which his work differs and contributes to the work of Levi-Strauss, Derride, Foucoult, Barthes and others. I haven't read any of the works he discusses here, so I don't feel qualified to say very much. 

I do intend to try and look into some of these theories in order to see what they might contribute to memetics, but I'm also a little worried about the sheer mass of philosophy that you are supposed to familiarize yourself with in order to approach them. (Oh, you can't read Derrida without reading Marx, and you can't read Marx without reading Hegel, and then you might as well read Kant.) 

Part of the charm of the book under review is, it collects work and builds a set of ideas into a highly significant structure, without requiring decades of study. And his criticisms of many of these philosophers is essentially along those lines: they are theories that are worked out within the world of text, without any apparent awareness that there exists a world outside the text to make reference to.

The book under review is very aware of the existence of a world outside the text, and the feedback loops between it and the world of text, which have been running and modifying one another for many (tens of) thousands of years now (if we include the work of oral cultures). To my eye, this is a very helpful contribution, and the fact that it manages to say so much without requiring any particular education is a point in its favor, not against.


By tracing the evolution of thought through many changes in circumstances, this book builds an intuition for how thought, society and technology co-evolve. There are undoubtedly many minor mistakes and oversights in this work, but as a means of leveling up your eye for large-scale and long-duration patterns, as well as close-focus and close-reading of particular items of evidence, I recommend it very highly.

I started reading this as part of a broader review of memetics literature. I'm not sure exactly which branch of study will end up covering this same material, whether it is semiotics or memetics or some currently pre-paradigmatic machine-learning offshoot[17]. Whatever the answer, I'm sure in my own thinking and reading I will continue to refer back to this book as an outlier work of clarity, depth and persuasion.

However, this book doesn't just touch on memetics, I think it hints at something much deeper and more significant. It is notable that I am writing this in an experimental communications medium, which has the explicit intention of reshaping human thought. To me, the book provides hints towards a dark and deep entanglement between information, culture, technology and consciousness. In particular, I think it provides strong counter-evidence for the idea that consciousness is a monolithic object in any way separate from its substrate and it's environment. What if humanity is only incidentally related to consciousness, in the way that we are incidentally related to eukaryotes?

Returning to the Hart Crane quote I placed at the start

Dream cancels dream in this new realm of fact
From which we wake into the dream of act

We are living in a modern dreamtime, a dreamtime built by a technology that killed its ancestors. Culture is barrelling out of this dreamtime into an unknown future that holds great danger, returning to orality en masse, or simply creating new cultures, visual cultures, virtual cultures. If history rhymes, will we expect to find all our values destroyed, with the exception of a few frozen accidents, captured by chance in the moment of great change? This seems to have been the fate of Homer, the result of the invention of writing. Or will it be just another small update this time, minor tweaks, merely massaging us into a marginally different mindspace, per many of the updates to technology that have happened since?

One thing is clear to me: if we want to have any chance at answering those questions ourselves and understanding the process as it is happening (let alone steering or guiding it!), we will need the very finest in psychotechnology. GPT-4 is, currently, an example of such a psychotechnology (and not an independent actor in its own right). This book and others like it are not just commentaries on the evolution of psychotechnology, but training manuals in psychotechnology itself. Recommended.

Further Reading


  1. ^

    From The Bridge by Hart Crane

  2. ^

    Hmm, just realized I don't have enough karma to automatically cross-post. I'll do it manually

  3. ^

    Even if we end factory farming, I think the work on moral valency of animals should have ontological repercussions for humans generally. Essentially I think people should either massively downgrade the significance of human well-being (if they are pure hedonists), which is obviously not going to be a popular suggestion, or they should look for quantitative ways of approaching preference/objective list/virtue ethics/deontology. This is another way of asking for a quantitative exploration of the infosphere, hence the current review.

  4. ^

    Research question: how would we measure the increase in orality over time? (The book has clues).

  5. ^

    Research question: can we trace periods of increased orality well enough to see how they correlate with political, religious and technological change?

  6. ^

    Unless otherwise noted, all quotes are from the book.

  7. ^

    For a short review of the Homeric Question and solution, see here.

  8. ^

    No doubt lack of demand played a part, since they already had enough epic poetry to be getting on with for some time. The market for 24-hour long epics is perhaps quite easy to saturate. I will pour out a drop for all the 24-hour long Greek epics that died in the crib, post-Homer, and a bottle to all those that Homer learned from.

  9. ^
  10. ^

    One interpretation of Wittgenstein's Tractatus (supported to some extent by his own letters and notes) is that he was pointing out that real philosophy happens outside the boundary of what can be stated.

  11. ^

    Doing this experiment is currently funding constrained, apparently this isn't high enough return on investment enough for the LTFF, sheesh!

  12. ^

    The harmony between this sentence and something like "capitalism and AI are teleologically identical" is a little unnerving. Must we, Walter? Really? I'm sure the reader has their own position on this question.

  13. ^

    Is it necessary to create "artificially intelligent" systems to understand the mind? Seems plausible.

  14. ^

    "Pretty much coeval with Learned Latin were Rabbinic Hebrew, Classical Arabic, Sanskrit, and Classical Chinese, with Byzantine Greek a sixth, much less definitively learned language, for vernacular Greek kept close contact with it. These languages were all no longer in use as mother tongues"

  15. ^

    If these words don't mean much to you, welcome to the party! They are confusing to me too. In fact I started reviewing the memtics literature as a stepping stone to trying to form a quantitative understanding of some of these ideas-- which turns out to have been a mammoth undertaking in its own right. But if all goes well I will eventually be able to shed some light on what I understand these words to refer to. In the meantime you can think of them as "distributed entities, running on human hosts over long spans of time and space".

  16. ^

    For viscerally illustrative examples of these changes as they happen throughout the Industrial Revolution, I recommend Pandemonium, 1660-1886: The Coming of the Machine as Seen by Contemporary Observers

  17. ^

    Oh, did I forget to mention academic philosophy? I guess It could also be them.

New to LessWrong?

New Comment