Sorry, Magnus, but your "proof" merely begs the question. All you have done at this point is:

• asserted impossibility without derivation
• treated definitional existence as illegitimate by fiat
• accused others of fallacy and bad faith for not sharing your standards
• refused to specify what would count as proof

This is why the discussion keeps looping. If you want to move the discussion forward you need to either (1) derive (not assert) an actual contradiction within the accepted mathematical framework (per ) or (2) reject the standard framework and present a coherent alternative (e.g. intuitionism, finitism, non-classical logic, etc.).

As it stands, Banno has already shown that combining your premise (1) with transitivity, antisymmetry and the existence of infinite partitions leads to contradictions. At this point there is nothing of substance left to discuss.

I think this makes the disagreement very clear, and it turns on a specific claim you’re making: that it is logically impossible for the human mind to directly know how things are in a mind-external world, because everything we know comes through the senses. I agree entirely that all perceptual knowledge is sensory-mediated. But I don’t think mediation by the senses entails indirectness in the epistemic sense you’re assuming.

The inference you’re relying on is:
sensory mediation ⇒ only effects are directly known ⇒ causes can only be known by inference.
That inference is not a logical truth; it depends on a particular picture of perception as awareness only of inner effects from which outer causes must be inferred. The Direct Realist rejects that picture. On their view, perceptual awareness is a relation to the object itself via sensory capacities, not an awareness of an inner item from which the object is inferred as a cause.

So when I see a red screen, I am not directly aware of a mental effect and only indirectly aware of a wavelength. I am directly aware of the red screen as a mind-external object, even though my access to it is mediated by physiological processes. Those processes explain how perception occurs, but they are not what perception is of. Knowing that a wavelength and neural signals are involved is itself a further piece of knowledge, typically gained instrumentally and inferentially, but that doesn’t show that ordinary perception is awareness only of effects.

In short, the dispute is not about whether we use inference to explain causal chains — of course we do — but about whether perception itself is exhausted by awareness of inner effects. You take that to be a logical constraint; I take it to be a substantive philosophical thesis, and one the Direct Realist denies.

Thanks for the detailed reply. I think I now see fairly clearly where we diverge, and it’s not at the level of physiology, causal mediation, or even skepticism, but at a deeper metaphysical level about what counts as a feature of the world at all.

As I understand you, you’re assuming that any property defined in relation to human perceptual capacities collapses into a claim about perception rather than a claim about the world. On that assumption, statements like “the sky elicits blue-type responses under normal conditions” amount to nothing over and above claims about how humans experience the sky, and so the distinction I’ve been drawing between claims about experience and claims about the world simply disappears.

I reject that assumption. On my view, many genuine properties are relational without being mental or experiential in their subject matter. Properties like visibility, fragility, toxicity, solubility, or mass-relative-to-a-frame are all defined partly in relation to possible interactions or observers, but they are still properties of the world, with truth conditions fixed by how things are. I take ordinary color predicates to work in a similar way: they are world-involving, response-dependent properties, not reports about inner presentation.

This is why the two claims I’ve been distinguishing come apart for me. A claim about how humans experience the sky has its truth conditions in facts about experience. A claim about the sky’s standing in lawful relations to perceivers has its truth conditions in facts about the sky and those relations. If one denies that relational properties can be genuinely worldly, then of course that distinction collapses — but that is precisely the metaphysical constraint I’m resisting.

Once that difference is in view, I think it becomes clear why we’re talking past each other. Given your constraint, my position can only look like a terminological reshuffling. Given my rejection of that constraint, your insistence that everything here is “really about perception” looks like a substantive metaphysical narrowing of what the world can be like. At that point, the disagreement appears to be principled rather than clarificatory.

Shape as seen or shape as felt? — Michael

I’d say neither the look nor the feel of shape as such is identical to the mind-independent shape of an object. Shape is a structural property that can be accessed through different sensory modalities and at different scales, each of which presents only partial, resolution-bound aspects of that structure.

Molyneux-style results show that cross-modal access to the same structure is learned rather than innate, not that there is no shared object or that perception is indirect. Likewise, questions about “which scale is the real shape” rest on a false assumption that there must be a single privileged resolution. Shape descriptions are scale-relative but objective within a scale.

None of this requires that perceptual experience mirror shape as it is in the world, and none of it implies that perception proceeds via epistemic surrogates. It just means that perceptual access to structure is perspectival and modality-specific.

Then we're back to what I asked in this post (which I'll repeat below), which I don't think was addressed:

What's the difference between a bionic eye that is "integrated into perception such that judgments are still answerable to objects through ongoing interaction and correction" and a bionic eye that is "a surrogate whose adequacy depends on a generating process that stands in for the world"?

It just seems like there's a lot of special pleading here. — Michael

I think the reason this keeps sounding like special pleading is that you’re asking for a principled distinction I don’t think exists. On my view, there is no such thing as a physical process being an “epistemic intermediary” as opposed to a merely causal intermediary.

All perception—organic or bionic—involves deterministic transduction from the world to the nervous system. What makes something an epistemic intermediary is not its material constitution or causal role, but a theoretical decision to treat some inner item as what perception is of and as the standard against which correctness is assessed.

I reject that move. Perceptual error is explained by false world-directed judgment, not by mismatch with an inner surrogate. Once that assumption is dropped, the demand to distinguish epistemic from non-epistemic intermediaries simply dissolves.

As I said before, you can mean anything you like by "directness". I'm concerned with what it means in the context of the traditional dispute between direct and indirect realism, which I summarised here (which I'll repeat below), and which I also don't think was addressed: — Michael

And as I have said before, I'm rejecting a shared assumption (phenomenal mirroring) that the traditional framing is built on. I don't think direct realism requires taking on this assumption, but if you don't agree then that may be as far as we can go. I don't think re-litigating the traditional framing is likely to help move the discussion forward at this point.

I'm not calling your view Cartesian. I'm saying that the scenario with the visor and the screen functions like a Cartesian Theatre. This would clearly be indirect perception even though their perceptual judgement "there is a ship" is about an object in the world.

So your claim that "perception is cashed out entirely in terms of perceptual judgment, and perceptual judgments are about objects in the world ... [therefore perception is direct]" is a non sequitur. — Michael

Sorry, missed this somehow.

I don’t think there’s a non sequitur here once my notion of “directness” is kept in view.

On my view, “direct” and “indirect” perception are not distinguished independently of where epistemic answerability terminates. To say that perception is indirect just is to say that perceptual judgment is answerable, in the first instance, to an epistemic surrogate rather than to the world itself.

The visor-and-screen case counts as indirect precisely because it introduces such a surrogate: the subject’s epistemic access runs through an internally generated stand-in whose adequacy must be assessed. That is not true in ordinary perception, even though both cases involve world-directed judgments.

So the inference from “perceptual judgment is about objects in the world” to “perception is direct” is not meant to be a standalone logical step. It’s a definitional consequence of rejecting epistemic intermediaries altogether. If one insists on distinguishing directness from judgmental answerability, then we’re simply working with different explanatory primitives.

That’s an interesting scenario to consider. Here is how I would answer your questions:

(1) Is the scenario logically plausible?
Yes. There’s no contradiction in two populations having systematically inverted orientation mappings.

(2) Is it physically plausible?
In principle, yes (vestibular inversion, neural remapping, spaceflight, etc.). The details are contingent, but not conceptually incoherent.

(3) Is one group seeing the “correct” orientation?
No. Because orientation is not an absolute property of the world, there is no “correct” mapping independent of a frame.

(4) How would we determine correctness without begging the question?
I think I would say we wouldn’t, because correctness isn’t the right notion here. I think what matters is

• internal coherence,
• successful coordination with the environment,
• shared practices of action and correction.

To summarize, I would say that orientation is frame-relative in a way that shape is not. That doesn’t imply error or indirectness, just that “up/down” are relational predicates whose correctness is fixed within a shared frame, not absolutely. This makes orientation relational rather than illusory, much like “left/right” or “near/far.”

Thoughts?

This is a restatement of what I've said amount to the same thing? I can't see a response to what I've said there specifically. — AmadeusD

Identity is not comparison.

I'm not quite sure how you can make that claim: science tells us our mind cannot look at objects. Our eyes look at objects and our mind constructs images from sense-data. There is an unavoidable chasm between objects and our representations in this form. Can you explain what you mean in the above quote in light of this? — AmadeusD

What I mean is that causal mediation does not by itself settle what perception is of. Science tells us that perception is implemented by sense organs and neural processes; I agree entirely. But it does not follow from this that the object of perception must be an inner representation rather than a mind-external object.

Saying that the mind “constructs images from sense-data” is already a philosophical interpretation of the science, not something the science itself establishes. All that science requires is that perception depends on causal processes. It does not require that awareness terminates in sense-data or inner pictures rather than in the world itself.

So the “chasm” you’re describing is not something science forces on us; it’s the result of adopting a particular representationalist model of perception. My claim has been that rejecting naïve mirroring does not commit us to that model.

They are the same thing. — AmadeusD

No, they are not the same thing.

A claim about perceptual presentation is a claim about how experience is structured (e.g. “humans tend to experience the sky as blue”). A claim about the sky-as-related-to-perceivers is a claim about the world under certain conditions (e.g. “the sky has properties such that, under normal conditions, it elicits blue-type responses”).

Those differ quite clearly in terms of:

• subject matter (experience vs world),
• truth conditions (facts about perceivers vs facts about the sky),
• direction of explanation (mind → world vs world → mind).

Collapsing these distinctions is exactly what turns a claim about the world into a claim about experience, which is the move I’ve been resisting throughout.

But out of curiosity, would you make the same claims about shape and orientation (and other features of geometry) that you make above about colour? — Michael

That’s a fair question, and the short answer is: no, I wouldn’t treat shape and orientation in exactly the same way as colour — but I would reject naïve phenomenal mirroring for them as well.

Colour is plausibly response-dependent in a way that shape and orientation are not. Ordinary claims about shape and orientation track relatively stable, mind-independent structural features of objects — and that’s why geometrical error correction, measurement, and intersubjective agreement work the way they do.

But even in those cases, I don’t think truth requires that the phenomenal character of experience reproduce those properties as they are in the world. A judgment like “the ball is round” is true because the object has a certain spatial structure, not because a phenomenal roundness in experience mirrors a roundness in the object.

So the difference isn’t that colour judgments are non-realistic while shape judgments are naïvely realistic. The difference is that colour predicates are more tightly tied to perceptual response profiles, whereas shape predicates are tied to structural and relational features of objects. In neither case does perceptual truth require that properties be “directly present” in experience in the sense the naïve realist needs.

That’s why rejecting naïve realism about colour doesn’t force indirect realism about perception more generally — and rejecting phenomenal mirroring about shape doesn’t amount to denying that objects have shapes.

Then I'll repeat what I said to Banno: I think the visor and its screen functions exactly like a Cartesian theatre (which is a strawman misrepresentation of indirect realism), and a Cartesian theatre is exactly the sort of thing that would qualify as indirect perception. So you've defined "direct realism" in such a way that even the strawman misrepresentation of indirect realism would count as direct realism. — Michael

Calling my view “Cartesian” doesn’t address the issue I’ve been pressing. The Cartesian Theatre is defined by the presence of an epistemic surrogate whose adequacy must be evaluated. My whole point has been that once phenomenal experience is not truth-apt, treating it as the “immediate object of perception” does no epistemic work. If that move reclassifies the traditional taxonomy, so be it—but that’s a consequence of rejecting phenomenal-first assumptions, not a reductio.

These seem to be the same thing? — AmadeusD

When I contrast mirroring with a judgment’s being correct or incorrect, I’m not redescribing the same relation. Mirroring posits a relation between mental items and worldly items; truth is a normative status of a judgment, not a relation between two objects. A judgment is answerable to how things are not by resembling the world, but by being correct or incorrect depending on how things are; when true, what is affirmed is identical with what is the case, without any mediating, internal mental replica.

Everything you say about variability, mediation, and scientific accounts of perception concerns causal dependence. I agree with all of that. Where I disagree is with the further step that treats causal mediation as implying epistemic mediation by inner representations. That step isn’t delivered by science.

It is a contradiction in terms, but I understand the second to actually mean "The sky is blue, as far as the HUman perceptual system tends to present" and that is obviously true. — AmadeusD

No, the claim is not just about how the human perceptual system presents things. It’s still a claim about the sky; namely the sky as it is in relation to the human perceptual system under normal conditions.

The point is just that ordinary truth doesn’t require predicates to correspond to simple intrinsic properties instantiated by objects. Rejecting naïve color metaphysics doesn’t make ordinary color judgments false—it just rejects a mirroring account of what makes them true.

The direct realist tries to avoid this by arguing that the sky appears blue because a) the sky is blue and b) the sky is directly present in experience. The indirect realist argues that this argument fails because (b) is false — Michael

I've granted that "blueness" is not a property of the sky, yet I maintain that "the sky is blue" is true. This sounds like a contradiction, but I don't think it is.

I would say that ordinary perceptual judgments like "the sky is blue" do not have to be interpreted in a naive way, but can be interpreted as something like "under normal viewing conditions, the sky systematically elicits blue-type visual responses in normal perceivers". This makes the claim objective, fallible, publicly assessable and non-projective. Nor does it require that the sky instantiate a phenomenal property as experienced. Many of the claims that people make ("the sun is rising", "that table is solid") can be cashed out in similar terms without resorting to naive realism.

And so we circle back to the example with the visors. The judgement "there is a ship" is a judgement about an object in the world, but it's still indirect perception. You seem to be conflating which things are the immediate objects of perception and which things our judgements are about. These are not the same thing. — Michael

It's not conflation, it's deflation. In the view I am defending, perception is cashed out entirely in terms of perceptual judgment, and perceptual judgments are about objects in the world. That’s not to deny that sensation causally mediates perception, only that it epistemically mediates it.

If you agree that phenomenal experience cannot be correct or incorrect, then the hypothesis that phenomenal experience is "what is directly seen" no longer explains error or motivates the skeptical worries you have presented. At that point, our disagreement reduces to whether experience is the direct object of perception or merely a mode of access. My point has been that the direct objects of perceptual judgments ("that's a ship") are objects in the world (ships), not phenomenal contents (redness as-seen, sourness as-tasted, etc). And this pretty much brings us full circle to where we landed a few posts back.

No, I'm saying that it's thing directly seen. From this we then make judgements about the world that can be correct or not. — Michael

I see what you're saying, but I think that the distinction you're making here is more terminological than substantive. As I understand your account, it requires that it is possible for there to be a mismatch between the phenomenal character of experience and the world. Understanding phenomenal experience as something that can succeed or fail to line up with how things are burdens it with a representational role that I would reject. That's a difference in how we locate epistemic mediation within the context of perception, not about whether judgments are made “from” experience,

Yes, I think this makes the divergence fully explicit now.

You’re treating phenomenal character as that which is assessed for correctness in the act of perception, whereas I’m treating judgments about the world as what are assessed, with phenomenal character merely causally occasioning those judgments.

The difference here concerns what we each take as epistemically basic. It may be that we've hit rock bottom on this issue, which is fine. Either way, I have enjoyed the discussion very much.

I think this is where we finally reach the deepest point of disagreement.

I reject the assumption that for veridical perception and hallucination to belong to a “common kind,” there must be a common object that is the immediate object of assessment. On my view, what is common is not an object, but a kind of epistemic activity: world-directed judgment undertaken from a perceptual standpoint.

In veridical perception, that judgment is answerable to objects in the environment and can be corrected by further interaction with them. In hallucination, the same kind of judgment is made, but it fails—there is no object that satisfies it. No inner surrogate is thereby promoted to the status of what is assessed; rather, the judgment is simply false.

That is why the Common Kind Claim does not force the conclusion you draw. Fallibility does not require that the immediate object of assessment be the same in success and failure. It requires only that the same kind of claim can succeed or fail.

This is also why the bionic vs organic distinction does no work here. I agree entirely that proteins are not privileged over silicon, and that both are deterministic transducers. But that shows only that causal mediation is ubiquitous. It does not show that perception involves an epistemic intermediary unless one assumes that error must always be explained by reference to an inner object.

So the dilemma you pose—either perception is indirect in both cases or direct in both cases—rests on an assumption I reject: that epistemic assessment must target an intermediary whenever perception can misfire. I deny that assumption. Judgments can be directly answerable to the world and still be wrong.

@Metaphysician Undercover @Magnus Anderson

It seems to me that this discussion keeps looping because the objection is being framed as an internal refutation of standard mathematical proofs, rather than as a foundational challenge to the notion of existence those proofs rely on.

Both of you have raised worries about the “doability” of bijection for infinite collections, which suggests a rejection of the identification of existence with formal definability and consistency. That’s a substantive philosophical position. But if that’s the objection, then it isn’t a matter of showing that the usual definitions lead to contradictions (they don’t), but of rejecting the underlying framework.

Put differently, the objection seems clearer if stated explicitly at the level of foundations, e.g.:

“I reject the identification of mathematical existence with formal definability. I require a constructive or modal account of possibility, and under that account I deny that completed infinite bijections exist.”

or

“I reject classical set theory in favor of a finitist or constructivist framework, where existence requires explicit construction.”

Framed that way, the disagreement would look less like an accusation about the failure of proof and more like a clash of foundational commitments, which is where I suspect the disagreement really belongs.

So why is this not also the case for the bionic eye? It simply replaces rod and cone cells with silicon. — Michael

It could be the case for a bionic eye — nothing I’ve said rules that out.

Simply replacing rods and cones with silicon does not by itself introduce an epistemic intermediary. What matters is not what the components are made of, but whether the system functions as part of the ordinary perceptual coupling with the world, or instead produces an output whose correctness must be assessed independently of that coupling.

If the bionic eye is integrated into perception such that judgments are still answerable to objects through ongoing interaction and correction — as with natural, transplanted, or lab-grown eyes — then there is no epistemic intermediary, and perception is direct in the sense I’m using.

The visor and nerve-stimulation cases differ because they interpose a surrogate whose adequacy depends on a generating process that stands in for the world, rather than being part of the perceptual relation itself.

So the distinction isn’t silicon vs biology, or artificial vs natural; it’s whether the device replaces part of the perceptual interface with the world, or replaces the world with an internal stand-in.

I agree entirely that judgment and reasoning depend on sensation in the sense you’re emphasizing. Without sensory experience, there would be nothing to judge about, and no reasoning could get started at all. I’m not denying that causal or developmental dependence.

The distinction I’m drawing is about epistemic role, not dependence. Sensations are conditions for the possibility of judgment, but they are not themselves reasons, premises, or justifications. That is why judgments, but not sensations, belong in the space of reasons. A sensation can prompt, occasion, or constrain a judgment, but it is the judgment that takes responsibility for saying how things are and can therefore be assessed as correct or incorrect.

So when I say that sensation does not “enter into” justification or inference, I don’t mean that judgments could exist without sensation. I mean that sensation does not function as a truth-apt item alongside judgments. The dependence you’re pointing to is real, but it doesn’t undermine the distinction I’m trying to mark.

The SDR says that they are directly cognizing the ship in the mind-external world, but if in the mind of the SDR there is no direct cognition of a weight of 10,000 tonnes, length of 200m, width of 25m and height of 30m, then what exactly is the SDR directly cognizing? The idea of a ship? — RussellA

I think what’s really at issue here is how we understand truth and directness. On my view, truth doesn’t consist in a resemblance or mirroring between what’s in the mind and what’s in the world, but in a judgment’s being correct or incorrect depending on how things are. That doesn’t require the ship’s properties to be present in the mind, only that the judgment be about the ship itself. Perception is direct in that sense: the object of perception is the mind-external ship, even though only some of its properties are perceptually available at any given time. Whether a judgment about the ship is true is settled by how the ship is, not by how closely something in the mind matches it. I realize this may sound like I’m simply assuming that judgments can be answerable to the world, but every account of truth has to take something as basic; here the difference is just whether one starts from mirroring relations or from the idea that judgments aim at getting things right about the world.

I would say that there is no relevant difference of the kind you are asking for — because the distinction I’m drawing is not about the material or biological status of the causal chain at all — but about the epistemic role it plays.

In ordinary perception — regardless of whether the eye is natural, transplanted, or artificially grown — one’s judgments are answerable to objects in a shared environment through ongoing interaction and correction, not to an internal signal whose adequacy must itself be evaluated.

I would say that the change doesn’t affect the point I was making.

Moving the interface from a screen to direct stimulation of the optic nerve changes the location of the causal mediation, not its epistemic role. In both cases, what the subject’s judgments are immediately answerable to is a generated input whose correctness depends on how it was produced, rather than to the objects themselves. That is the sense in which the perception is indirect.

I agree that humans ought to be continually making judgements …
Is not the normal use of the word “normative” a moral norm, such as “you ought not smoke”? — RussellA

Yes — in ordinary language, “normative” is often used for moral norms. But that is not the sense in play here.

By epistemic normativity I do not mean “humans ought to be continually making judgments.” I mean something closer to this: when we make judgments, we are implicitly adopting standards of correctness (e.g. truth, evidence, coherence, reasonableness).

In other words, judgment is normative because it is answerable to how things are. To judge at all is to commit oneself to being right or wrong, and to being accountable to reasons. That commitment is built into the act of judging; it is not a further moral obligation one may or may not take on. To reject those norms is not to judge differently, but to stop judging altogether.

And while these norms are indeed socially mediated, I would argue they are ultimately grounded in the subjectivity of the individual.

Surely, if we are looking to an authority, we would prefer an authority that cannot be wrong, such as the senses, rather than an authority that is more often than not wrong, such as a judgement. — RussellA

Here the contrast is misleading. Sensation is not “an authority that cannot be wrong”; it is not an authority at all, because it is not the kind of thing that can be right or wrong. To say that sensation is not truth-apt is to say it can be neither correct nor incorrect.

Judgments, by contrast, can be wrong — but that is precisely because they are the only things that can also be right. Error is not a defect relative to sensation; it is the price of intelligibility. Only what can be mistaken can be corrected, justified, or improved. That is why judgments — not sensations — belong in the space of reasons.

The problem is we give no authority to a judgement just because it is a judgement. We give authority to the content of a judgement. — RussellA

I think this shows that the word “authority” is doing more harm than good here, and that may be my fault for introducing it.

All I mean by saying that judgments have epistemic authority is that they are the locus of truth and falsity. Sensory qualities — redness-as-seen, pain-as-felt — are not the kinds of things to which truth or falsity apply at all. Judgments are. That is the only contrast I am trying to mark.

So the point is not that judgments are authoritative regardless of content, nor that every judgment deserves equal credibility. It is that only judgments, whatever their content, are even candidates for being assessed as correct or incorrect. Sensation constrains judgment, but it does not itself enter into justification or inference.

Compare this with words. "Dog" represents dogs. Yet, the word "dog" in itself, is not correct or incorrect. It simply is what it is. But, when placed in a larger context, for instance, pointing to an animal, and uttering "dog", then the word can correctly indicate the animal pointed to, or not. — hypericin

I would say that words are essentially representational: to be a word is to be a bearer of meaning or reference. And while I agree that context is required in order to fix a word’s conditions of use and meaning, context does not transform a word from a non-representational kind of thing to a representational one.

Smells are not like that. They can causally indicate ammonia, just as smoke can indicate fire, but the correctness conditions do not attach to the smell itself. They attach to the judgment made on the basis of it. When smell is misleading (hallucination, long covid, etc.), we don’t assess the smell as incorrect; we assess the judgment as mistaken or the sensory capacity as unreliable.

So even in context, phenomenal qualities are not what represents the world. They are a causal-enabling condition under which world-directed judgments acquire correctness conditions. That is the sense in which I deny epistemic mediation while fully granting causal mediation.

Hmm, this is not how I experience odor. The smell itself is what hits me first, viscerally and immediately. No introspection is needed. If the smell is a familiar one, I might identify it quickly, so quickly that it might even seem immediate. But if I haven't smelled that smell in a long time, it can take significant mental effort to identify it. Occasionally, I won't be able to at all, and I am left frustrated, wondering what that smell reminds me of.

Do you not relate to this? — hypericin

I relate to that description completely. But I think this brings out a distinction rather than a disagreement.

What you’re describing is immediate phenomenal awareness — the smell hitting you viscerally, prior to identification. I don’t deny that at all. What I’m denying is that this involves an introspective judgment in the epistemic sense.

An introspective judgment would be something like “I am having a sharp, acrid olfactory experience.” That’s a reflective, truth-apt claim about one’s experience. In ordinary cases, we don’t make that judgment first. We either make a world-directed judgment (“that’s ammonia,” “something smells off”), or we hesitate from making a judgment at all because we can’t yet place it.

I would say that the delay or effort you describe doesn’t show that we infer from an inner premise; it shows that perceptual judgment can be difficult, uncertain, or fail altogether. Phenomenal awareness can be immediate without functioning as an epistemic intermediary.

-------------------------------------------------------------

To sum up: it seems like the divergence between us can be captured by a single question: are phenomenal qualities representational vehicles whose correctness conditions are fixed in context, or are they non-representational causal conditions under which world-directed judgments acquire correctness conditions.

What do you think?

Cheers. Enjoy the weather.

I largely agree with the position you've been defending on this thread. The only significant divergence we have is the one we've discussed on another thread. Whereas you would say:

But our understanding is always, and already, an interpretation, so the "flower-as-it-really-is" or the "flower-in-itself" is already a nonsense. — Banno

I would say: "our understanding is always, and already an interpretation, and the notion of the 'flower-as-it-really-is' or the 'flower-in-itself' is a non-eliminable regulative ideal around which the act of inquiry itself is organized".

Another divergence is that I am perhaps more apt to treat talk of phenomenal experience as legitimate (though not epistemically foundational).

Outside of those two things, I don't find much to disagree with in what you say.

Thanks for clarifying!

Here is how I would approach each of the three propositions. I’ll try to reuse your examples so that we can better observe how our approaches differ.

(1) Phenomenal qualities represent aspects of the world.

False. The sharp, pungent, acrid scent of ammonia as-smelled does not, in itself, represent anything in the world. Neither does redness as-seen, loudness as-heard, or sour as-tasted. A representation is something that can be assessed for correctness, truth or fidelity. Raw sensory qualities are not the kinds of things that can be correct or incorrect; they simply are what they are.

(2) Ordinary perceptual judgments are judgments about phenomenal qualities.

False. Ordinary perceptual judgments are about things in the world (“that rag smells of ammonia”), not phenomenal qualities (“there’s a sharp, pungent, acrid scent in my olfactory map”). The former are typically referred to as “perception”, the latter as “introspection”. Introspection is second-order, reflective and derivative with respect to ordinary perception.

(3) Our knowledge of the world is inferred from such (introspective) judgments.

False. We do not ordinarily infer perceptual judgments from introspective judgments. Rather, perceptual judgments are epistemically primary, and introspective judgments are appealed to only in reflective, corrective, or explanatory contexts. And even then, introspective judgments are typically not used to justify perceptual judgments, but only to help reason about (or explain) anomalous cases (e.g. uncertainty, disagreement, illusion).

As you can see, we approach and answer these questions in significantly different ways. What do you think of this?

I don’t dispute either of the points you raise. Yes, we experience phenomenal character, and yes, the looks, sounds, smells, and feels involved in perception are properties of the nervous system. I also wouldn’t deny that perceptual judgment is causally mediated by phenomenal character.

But the issue isn’t causal mediation; it’s epistemic mediation. Consider the following commitments:

(1) Phenomenal qualities represent aspects of the world.

(2) Ordinary perceptual judgments are judgments about phenomenal qualities.

(3) Our knowledge of the world is inferred from such perceptual judgments.

One influential way of understanding epistemic mediation in the indirect realist tradition involves accepting commitments like these. I reject all three. For that reason, I reject the claim that perceptual judgment is epistemically mediated in the traditional sense. That’s also why I think the photograph analogy misleads: it tacitly presupposes at least one of these commitments, whereas my view denies them.

Out of curiosity, which of three propositions above would accept, if any? Does the distinction between causal and epistemic mediation as laid out above make sense to you, or would you qualify it in some way? I’d be interested to get your thoughts.

Thanks for contributing to the discussion with a thoughtful reply. I thought I would chime in since this overlaps with so many of the same issues I've been discussing with @Michael and @RussellA.

The photograph case you raised is an interesting test case because I think that it subtly presupposes exactly what is at issue. The photograph itself is what is perceptually present, and the person is not. That’s why it’s natural to say the person is only perceived indirectly. But in ordinary perception there isn’t an analogous surrogate object that stands in for the ship in that way; at least, that is the point at issue. The ship itself is what our judgments are about, and it's what also constrains our beliefs over time.

That’s why I’m hesitant to say that the “primitives of perception are hallucinations of the brain.” That description already assumes that phenomenal character functions like a photograph—i.e. as the thing perceived instead of the object—whereas both @Banno’s point (if I'm understanding him correctly) and my own have been that phenomenal character causally constrains perception without being its direct object.

I agree that rejecting naïve realism is mandatory, and that causal mediation alone doesn’t settle the issue. But I don’t think the photograph case shows that perception must be indirect in the sense of being mediated by inner surrogates rather than being answerable to the world itself.

I think there’s a subtle but important shift in your reply that ends up missing the point I was making.

My claim was not that single judgments are reliable, infallible, or likely to be correct. Epistemic authority is not a matter of probability, reliability over isolated cases, or confidence in one-off judgments. It concerns what kind of act is even eligible to be assessed as correct or incorrect at all.

Sensation, as you agree, is not truth-apt. Judgments are. That difference is not about likelihood of error; it is about logical role. Only truth-apt acts can be wrong, and only what can be wrong can be corrected, justified, or criticized. That is the sense in which epistemic authority “resides” in judgment rather than sensation.

So when you say that neither a single sensory experience nor a single judgment has epistemic authority, I agree — if “authority” is taken to mean certainty or high probability. But that is not the sense at issue. The point is that only judgment participates in the space of reasons at all, even when judgments are tentative, revisable, or likely to be false.

Likewise, the normativity I’m invoking is not the moral norm “you ought to judge,” nor the psychological commitment to judging. It is the epistemic normativity built into judgment itself: judgments answer to how things are, can succeed or fail, and can be revised in light of further reasons. Sensory experience constrains this process, but it does not enter it as a premise.

So I don’t think we need to “look elsewhere” for epistemic authority. We just need to distinguish authority from certainty, and normativity from reliability. Epistemic authority lies in judgment because judgment alone is answerable to truth — even when, and especially when, it turns out to be wrong.

Yes — that’s a good way of putting it, and I agree. I didn’t mean to suggest bijection is foreign to finite counting, only that when we move to infinity the remainder-based cues we rely on in finite cases stop being reliable, even though the underlying correspondence idea remains.

This is where I think a crucial distinction is getting lost.

The normativity I’m talking about is not a property of the content judged, but of the act of judging. The proposition “if I see orange, then the screen is orange” is entirely descriptive. What is normative is taking it to be true — committing oneself to its correctness and taking responsibility if it turns out to be false.

That’s why comparisons with laws or moral rules don’t quite apply. Legal and moral norms are prescriptive and externally grounded. Epistemic normativity is neither enforced nor derived; it is imminent and constitutive. To judge at all is to place oneself under standards of truth, justification, and error. No additional rule, institution, or innate principle is required.

So when I say judgment is normative, I don’t mean that it issues an “ought” in the moral sense, or that it is governed by conventions. I mean that to judge is to take a stance that can succeed or fail — that can be correct or incorrect — and that this answerability is what distinguishes judgment from mere sensation or description.

That is also why sensory experience, while indispensable, cannot itself function as an inferential premise. Sensation is not the kind of thing that can be right or wrong. Judgment is. And that difference is where epistemic authority resides.

Thanks for laying this out so clearly. Unfortunately, I think a couple of confusions have arisen regarding my position. Let me try to clarify.

First, when I speak of normativity, I am not talking about moral norms (e.g. “evil is bad”), but epistemic normativity: truth, falsity, correctness, and justification. To make a judgment is to take on a set of epistemic responsibilities. That normativity is constitutive of judgment, not something inferred from experience or imposed by the will, and it is independent of any moral “ought”.

Second, I don’t think the rationalism/empiricism contrast maps cleanly onto the disagreement between us. I’m not claiming substantive knowledge of the world independent of experience, as classical rationalists did. But neither am I claiming that experience supplies inferential premises from which all other knowledge is derived, as classical empiricists did. What I reject is empiricist foundationalism: the idea that non-conceptual sensory states can function as epistemic grounds or premises.

On my view, experience is indispensable, but it does not do epistemic work by representing or grounding inference. It constrains judgment non-inferentially, by situating inquiry and correcting it, while judgment alone bears epistemic authority. That places my position outside the traditional rationalist/empiricist divide rather than on the rationalist side of it. My position is best described as post-Kantian with a contemporary anaytic twist, and is heavily indebted to thinkers like Peirce, Sellars and McDowell.

This is why the “direction of fit” framing doesn’t quite apply. I’m not saying the mind legislates how the world must be, nor that sensation ought to match reality. Sensation itself is non-normative. The act of judgment is intrinsically normative, not because it is independent of experience, but because it is truth-apt.

So in the orange-screen case: I’m not saying that because there is an orange screen I ought to see orange, nor that seeing orange licenses an inference. Rather, when I judge “the screen is orange,” that judgment is assessable for truth, and experience constrains it without functioning as a premise.

That, I think, is where our views genuinely diverge: whether sensory mediation itself does epistemic work by representing and grounding inference, or whether judgment alone is epistemically authoritative while experience constrains it non-inferentially.

I think this helps clarify where we diverge. You’re treating “number of elements” as a notion whose inferential rules must be fixed by finite counting, and on that assumption the infinite case does look contradictory. Mathematics takes a different route: it treats counting as the finite implementation of a more general notion of size, and allows the implementation to change when counting no longer applies. I think has done a fine job of showing the inconsistencies that arise if we don't.

That’s why defining a bijection counts as establishing existence in this context, and why reindexing in Hilbert’s Hotel isn’t seen as pretence. At that point the disagreement isn’t about technique, but about whether such revisions are legitimate at all.

I see what you’re getting at, and I agree that bijection strictly extends our ability to reason about size — especially once infinities are in play. In that sense it’s an enrichment, not a rival notion.

The small point I was gesturing at is that, while the bijection criterion agrees with the remainder-based notion on all finite cases, it does so by no longer treating “having a proper remainder” as decisive for size comparison. That inferential role is preserved extensionally for finite sets, but it no longer has the same explanatory force once we move to the infinite case.

So I’m not suggesting that anything correct is lost in the finite domain — only that some intuitive cues we rely on there stop doing the work we expect of them when the concept is refined for a broader domain.

Cheers. I largely agree with you that formal language is not something alien to natural language, but a tightening of it — making explicit commitments and inferential roles that are often left implicit in ordinary use.

Where I’d add a small nuance is that the act of tightening isn’t always neutral. In refining a concept, we sometimes preserve certain inferential roles while deliberately abandoning others that no longer serve the new domain. In the case of “size”, the move to formal language preserves comparability and transitivity for infinite collections, but it does so by dropping the remainder-based role that functions perfectly well in the finite case.

So I don’t see formal language as distinct from natural language so much as selectively continuous with it: a refinement that’s purpose-driven rather than a mere sharpening of everything we already mean.

I agree that this is now a question about how the debate has been traditionally framed, not about skepticism or justification.

I don’t deny that many historical direct realists tied direct perception to naïve colour primitivism, nor that Locke-style views are the canonical contrast class. In that sense, my view is revisionary with respect to the traditional dialectic.

What I reject is the claim that phenomenal mirroring is constitutive of directness rather than a substantive thesis adopted by particular direct realists. The falsity of naïve colour primitivism shows that the world is not phenomenally as it appears, but it does not by itself show that perception must proceed via inner representations standing in for the world.

When I say that perception relates us directly to mind-independent objects, I do not mean that any causal or informational link counts as direct. The visor case is instructive precisely because it introduces an epistemic intermediary whose outputs are the immediate objects of assessment. By contrast, in ordinary perception, our judgments are answerable to objects themselves within a shared public environment, not to internal surrogates whose accuracy must be inferred.

So the disagreement isn’t about whether my usage matches a traditional definition — I’m happy to grant that it doesn’t. It’s about whether we should inherit phenomenal mirroring as a constraint on realism about perception at all. That is the assumption I’ve been rejecting throughout.

Point taken. I agree that once infinite collections are treated as completed totalities, the intuitive, remainder-based concept of size becomes inconsistent, and your examples make that very clear. I don’t think the intuitive concept is incoherent as such — it’s well-behaved in the finite case — but I agree that Magnus’s attempt to generalize it to infinite collections fails.

Your reply nicely clarifies the remaining disagreement.

You say the minor “premise” “I am seeing orange” is not truth-apt but can still function in an inference. I don’t think that position is stable. Inference is a normative relation between propositions, and only truth-apt contents can play that role. If “I am seeing orange” is not truth-apt, then it is not a premise at all, and the conclusion cannot be inferred from it.

This is why I distinguish constraint from grounding. Sensory contents constrain inquiry by occasioning and shaping judgment, but they do not function as inferential grounds alongside judgments. The epistemic work is done entirely at the level of judgment.

This also bears on representation. To say that “the orange I see in my mind represents the screen being orange in the world” reintroduces normativity at the level of sensation. Representation can succeed or fail, and once sensation represents, it is no longer non-normative. That is precisely the move I’m resisting.

Finally, rejecting empiricist derivations of norms does not imply Idealism. Epistemic norms are conditions for the possibility of inquiry, not constituents of reality. To say that judgment is norm-governed independently of experience is not to say the world is mental, but that knowing has irreducible normative structure.

So the disagreement is not about whether the senses mediate our contact with the world — I agree they do — but about whether that mediation itself does epistemic work by representing and grounding inference, or whether judgment alone bears epistemic authority while experience constrains it non-inferentially. That is where we part ways.

Thanks — that clarifies your position.

I don’t deny that many historical formulations of direct realism build in the conditional you cite: if perception is direct, then the world is phenomenally as it appears. Nor do I deny that naïve colour primitivism is false.

What I reject is treating that conditional as definitive of direct perception rather than as a substantive thesis adopted by particular theorists. The falsity of naïve colour primitivism shows that the world is not phenomenally structured as it appears, but that conclusion is neutral on whether perception relates us directly to mind-independent objects or only via inner intermediaries.

Your argument depends on defining direct perception in terms of phenomenal mirroring. My resistance has been to that definition. As I’m using the term, directness concerns whether perceptual judgment is answerable to the world itself, not whether phenomenal character reproduces the qualitative character of reality.

So I don’t dispute your historical reconstruction or the science. I’m disputing whether we should inherit that conditional as a constraint on how the problem must be framed.

I agree that everything I described — coherence, responsiveness to further experience, success in inquiry — could in principle occur in a hallucination or brains-in-a-vat scenario.

But I don’t think that concession supports the conclusion you want to draw.

The mere possibility of global deception does not by itself show that perception is indirect, nor that the world is not as it appears. It only shows that perceptual judgment is fallible — something I have never denied.

The step I think you’re taking for granted is this: that if perception were direct, then massive and systematic error would be impossible. But I don’t see why that should be accepted. Directness is a claim about the kind of relation perception bears to the world, not a claim about epistemic guarantees or immunity to skepticism.

In other words, the fact that we could be wrong about everything does not entail that our beliefs are mediated by representations standing in for the world, rather than being judgments answerable to the world itself. Fallibility and answerability are compatible.

That’s why I don’t take brains-in-a-vat scenarios to motivate indirect realism. They motivate epistemic humility, not a particular metaphysics of perception. To get from “global error is possible” to “perception is indirect” you need an additional premise — that direct perception would rule out such error — and that premise is exactly what I reject.

So the disagreement here isn’t about whether skeptical scenarios are conceivable. It’s about whether conceivability alone licenses conclusions about the structure of perception. I don’t think it does.

This is helpful, because now the skeptical pressure you’re worried about is fully explicit.

I think the crucial step where we diverge is here: you’re assuming that for an ordinary perceptual belief like “there is a ship” to be justified, I must also be justified in believing something like “my current experience is not hallucinatory” — i.e. that I must first justify a claim about how my experience relates to reality before I’m entitled to make any claims about the world.

I reject that requirement.

It helps to distinguish carefully between phenomenal character itself and claims about how things appear. Phenomenal character is not truth-apt; it does not assert anything, and so it is not something that can be correct or incorrect. Claims about how things appear (e.g. “it looks orange to me”) are truth-apt, but they are reflective, second-order claims that arise for special purposes — disagreement, error-checking, theory-building — not as the epistemic basis of ordinary perceptual judgment.

Ordinary perceptual beliefs are not justified by first establishing that one’s experience is veridical rather than hallucinatory. In normal cases, one does not infer “there is a ship” from the premise “it appears that there is a ship.” One simply judges that there is a ship, and that judgment is assessed over time by its coherence with other judgments, its responsiveness to further experience, and its success or failure in inquiry.

This is why the indistinguishability of hallucination and so-called “veridical” experience does not generate skepticism on my view. That skeptical pressure only arises if we assume that perceptual justification requires antecedent justification of claims about how things appear. I deny that assumption.

Hallucinations matter epistemically when they function as defeaters within inquiry, not as a standing possibility that must be ruled out in advance in order for any perceptual belief to be justified at all.

So the disagreement isn’t about whether hallucinations are possible or whether phenomenal character can be misleading. It’s about whether justification for ordinary beliefs about the world depends on first justifying claims about appearance. That is the assumption I’ve been challenging throughout.