Thinking Without Words (and Other Heresies)

2–3 minutes

Evelina Fedorenko has been committing a quiet but persistent act of vandalism against one of modernity’s favourite assumptions: that thought and language are basically the same thing, or at least inseparable housemates who share a fridge and argue about milk. They’re not.

Audio: NotebookLM summary podcast of this content.

Her fMRI work shows something both banal and scandalous. Linguistic processing and high-level reasoning live in different neural neighbourhoods. When you switch language on, the ‘language network’ lights up. When you do hard thinking without words, it doesn’t. The brain, it turns out, is not secretly narrating your life in subtitles.

This matters because an entire philosophical industry has been built on the idea that language is thought. Or worse: that thought depends on language for its very existence. That if you can’t say it, you can’t think it. A comforting story, especially for people whose entire self-worth is tied up in saying things.

Now watch two chess players in deep play. No talking. No inner monologue helpfully whispering, ‘Ah yes, now I shall execute a queenside fork’. Just pattern recognition, spatial anticipation, constraint satisfaction, and forward simulation. If language turns up at all, it does so later, like a press officer arriving after the battle to explain what really happened.

Video: Dina Belenkaya plays chess.

Language here is not the engine. It’s the after-action report. The temptation is always to reverse the order. We notice that people can describe their reasoning, and we infer that the description must have caused the reasoning. This is the same mistake we make everywhere else: confusing narration with mechanism, explanation with origin, story with structure.

Fedorenko’s findings don’t tell us that language is useless. They tell us something more irritating: language is a post hoc technology. A powerful one, yes. Essential for coordination, teaching, justification, and institutional life. But not the thing doing the actual work when the work is being done. Thought happens. Language tidies up afterwards.

Which leaves us with an awkward conclusion modern philosophy has spent centuries trying to avoid. The mind is not a well-ordered library of propositions. It’s a workshop. Messy, embodied, improvisational. Language is the clipboard, not the hands. And the clipboard, however beautifully formatted, never lifted a chess piece in its life.

As for me, I’ve long noticed that when I play a game like Sudoku, I notice the number missing from the pattern before any counting or naming occurs. The ‘it must be a 3’ only happens after I make the move.

Cognitive Processing Flow Model

The Cognitive Process Flow Model illustrates how we process the phenomenal world. It’s reductionist and is missing aspects because it is just a back-of-the-napkin sketch. I created it because I uttered, “I can model it for you”. And so I did.

EDIT: I’ve updated the model slightly as the article head image, but the copy content refers to the first draft.

My response was to a person making the claim, that all you need to facts and logic prevails. Rather than restate the argument, I’ll just walk through the diagramme.

There’s meta information to set it up. We are subjective entities in the world. We have a sense-perception apparatus as we exist in it. Countless events occur in this world. We recognise only a few of them within our limited range, though technology expands this range in various ways.

Most of us interact in the world. Some are less ambulatory, so the world visits them. Some have sense-perception deficits whilst others have cognitive deficits. My point is not to capture every edge and corner case. This is just a generalised model.

It starts with an event. Events occur ceaselessly. In our small portion of the world and elsewhere. For the purpose of the model, the first thing that happens is an event catches our attention. We might notice a shape, a colour, or a movement; we might hear a sound, smell an aroma, feel a sensation, or taste something.

A pre-emotion, pre-logic function serves to process these available inputs. Perhaps, you hear a report on anthropogenic climate change or read something about a political candidate. This emotional filter will police sensory inputs and unconsciously or preconsciously determine if you will react to the initial stimulus. If not, you’ll continue in an attention-seeking loop. Not that kind of attention-seeking.

As my dialogue was about the presentation of facts, our next stop will be logical evaluation. Does this make sense to us, or can we otherwise make it? This is a process in itself. I’ll assume here that it requires little elaboration. Instead, I’ll focus on the operating environment.

Our logical processes are coloured by past experiences and tainted by cognitive biases and deficits. We may also trigger the calling of additional facts through past experiences or the current engagement.

We’ll process these fragments and reach some logical conclusion. But we’re not done. We take this intermediate conclusion and run it through more emotional processing. Cognitive biases come back into play. If the event conforms with your past experiences and cognitive biases, we may run it through a cognitive dissonance routine. To be honest, this probably is part of the emotional reconciliation process, but I’ve drawn it here, so I’ll let it be. In this case, it’s just a filter. If it happens to conform to our belief system, it will pass unfettered; otherwise, it will be squared with our beliefs. Again, this leads me to believe it’s a subcomponent of emotional reconciliation. I’ll update the chart later.

In any case, we’ll end at Final Acceptance. This acceptance may be that we accept or reject the logic, but we arrive at an opinion that gets catalogued with the rest of them. Some may be elevated to facts or truths in the epistemological hierarchy. Although an end marker is identified, it’s really a wait state for the next event. Rinse and repeat until death.

I’ll update this presently. Be on the lookout. It could include more dimensions and interactions, but that might have to wait until version 3.

Meantime, does this feel right to you? Did it even get your attention?

An Example: Anthropogenic Climate Change

Let’s wrap up with an example. I’ll use climate change. An article comes into your attention field, and you have an interest in these things, so it passes through the emotional filter. If your propensity for these articles is high, it might race to the next stage.

You read the article, and it contains some facts—rather, it contains claims for evaluation. To do this, you’ll recall past experiences and cognitive biases are always lying in wait. You may have to look for new facts to add to the mix. These will have to take a similar route past your attention gatekeeper and emotional sidekick.

If you are already predisposed that climate change is a hoax, these facts will filter through that lens—or vice versa.

When all of this is resolved, you’ll have arrived at a conclusion—perhaps we’ll call it a proto-conclusion. It hasn’t been set yet.

You are still going to introspect emotionally and decide if this is a position you want to hold. Perhaps, you feel that climate change is a hoax but this doesn’t jive with that position. Here, you’ll either accept these facts and flip a bit to a sceptical believer or cognitive dissonance will kick in and ensure your sense of the world isn’t thrown off kilter. You may update your belief system to include this datum for future assessments.

Now we are ready for final acceptance. You can now express your established opinion. If the net event is to counter that acceptance, rinse and repeat ad infinitum.

Man versus Machine

Human-designed systems seem to need a central orchestration mechanism—similar to the cognitive homunculus-observer construct substance dualists can’t seem to escape—, where consciousness (for want of a better name) is more likely the result of an asynchronous web with the brain operating as a predictive difference and categorisation engine rather than the perceived cognitive coalescence we attempt to model. Until we unblock our binary fixedness, we’ll continue to fall short. Not even quantum computing will get us there if we can’t escape our own cognitive limitations in this regard. Until then, this error-correcting mechanism will be as close to an approximation of an approximation that we can hope for.

The net-input function of this machine learning algorithm operates as a heuristic for human cognition. Human-created processes can’t seem to create this decoupled, asynchronous heuristic process, instead ending up with something that looks more like a railway switching terminal.

Cover photo: Railroad tracks stretch toward Chicago’s skyline at Metra’s A2 switching station on March 29, 2019. (Antonio Perez/Chicago Tribune); story