Why does a bionic eye "function as an intermediary whose output purports to represent the environment and can be assessed as succeeding or failing" but an organic eye doesn't? — Michael

It turns on what kind of standard a system is answerable to. I'll try to explain what I mean by this:

You’re right that Jane’s eye can be assessed against physical and biological standards. We can say that it malfunctions if it fails to transduce wavelengths in the statistically normal way, or relative to how human eyes typically function. But that kind of assessment is not yet an intentional standard of correctness. A biological malfunction is a failure relative to a kind; it is not, by itself, a failure to accurately present the world. Even a normally functioning eye does not answer to some prior specification of how the environment is supposed to look for Jane. Rather, it fixes what “getting it right” means for her perceptually. That is what makes it constitutive.

By contrast, the visor in (4) is assessable as misrepresenting the environment even when it is functioning exactly as designed. That difference matters. The visor’s outputs purport to stand in for how the environment is perceptually available independently of it, which is why it makes sense to ask whether those outputs succeed or fail as presentations of the world. That question does not arise for Jane’s eye, not because it is organic, but because there is no independent intentional standard against which its mapping could be evaluated for her. Its mapping is identity-fixing, not performance-assessable.

This also explains why lawful covariation and consistency are insufficient to erase the distinction. An organic eye with an inverted mapping is not misrepresenting the strawberry; it determines what it is for strawberries to look any way at all for that subject. A visor with an inverted mapping can misrepresent even while covarying perfectly, because its role is instrumental rather than constitutive. The possibility of misrepresentation without malfunction is exactly what marks that difference.

So the asymmetry is not asserted but grounded: Jane’s visual system constitutes her perceptual capacities because it defines the intentional space within which perceptual correctness is possible at all. The visor intervenes on an already defined perceptual capacity, which is why its outputs are assessable as accurate or inaccurate in a way the eye’s are not.

I think your question here makes the issue as sharp as it can be. I'll try to clarify where I think the equivalence argument ultimately breaks down.

I’m happy to grant that there is nothing privileged about proteins, and that a genuinely bionic eye could, in principle, count as part of a subject’s perceptual system. Material composition does no work here. What matters is not what the system is made of, but the role the system plays in constituting perception for the subject.

The distinction I’m relying on is not between lawful vs. unlawful mappings, or between covarying vs. non-covarying systems. I agree that in all of your cases—(1) through (4)—there can be lawful, consistent causal mappings from strawberry to neural outcome. So counterfactual covariation by itself cannot mark the difference. The real distinction is this: whether the mapping in question is constitutive of what it is for the subject to perceive at all, or whether it is a substitutable, instrumental mapping that could be altered without redefining perception for that subject.

In (2), Jane’s visual system—eye, retina, and downstream neural processing—constitutes her perceptual capacities. Whatever the mapping from wavelength to neural state happens to be (even under inversion), that mapping is not something that stands in for perception; it is how objects are perceptually available to her. There is no further question of whether this mapping is “doing its job correctly,” because it is not functioning as an intermediary whose output purports to represent the environment. It defines what counts as seeing for Jane.

By contrast, in (4), the visor introduces a mapping that is not constitutive in this sense. Even if it covaries lawfully with the strawberry, it functions as a substitutable system that fixes perceptual outcomes independently of the strawberry’s own role in determining how it is perceived. That mapping could be changed, replaced, recalibrated, or removed without thereby redefining what it is for John to perceive at all. That is why it is intelligible to ask whether the visor is presenting the environment correctly. The mapping is instrumental rather than constitutive, and so its outputs are assessable as succeeding or failing as presentations of the world.

This is also why (4) aligns with (3) rather than with (2), even though (4) bypasses the eye entirely. The difference is not whether an image is interposed, nor where in the causal chain neural stimulation occurs, but whether the system in question defines perceptual access or intervenes by imposing a detachable mapping. A bionic eye could fall on either side of this divide, depending on whether it becomes constitutive of the subject’s perceptual capacities or remains an instrument that determines perceptual outcomes in place of object-governed availability.

So the equivalence argument only goes through if one assumes that any lawful causal mapping to neural states is sufficient to fix intentional structure. That assumption collapses constitutive and substitutable mappings into the same category. I reject that assumption. Once the distinction is in view, it becomes clear why (1) and (2) count as direct perception, while (3) and (4) do not—not because of biology, reliability, or phenomenology, but because of the different roles these systems play in individuating what perception is for the subject.

I think the remaining disagreement comes from running together three different questions:

(1) how persistence through time should be understood,
(2) how causation across time works, and
(3) what “direct” is supposed to contrast with in a theory of perception.

First, about persistence. On Presentism, to say that the Sun persists through change is not to say that past parts of the Sun still exist. It is to say that the present Sun stands in lawful causal continuity with earlier states. Persistence here is not identity-with-the-past, but continuity governed by physical laws. Losing one atom does not generate a new object because nothing in our best physical theories treats that loss as a boundary for objecthood. The absence of a sharp cutoff does not show that persistence is merely linguistic; it shows that persistence is a real-world phenomenon that our concepts track imprecisely.

On a Block Universe view, the Sun is a temporally extended physical process. Different temporal parts are different physical states, but they are unified by belonging to the same continuous spacetime process governed by physical law. The relevant commonality is not qualitative sameness at each moment, but participation in a single causal–spatiotemporal structure.

Second, about causation. You repeatedly say that on Presentism the past “no longer exists” and therefore cannot directly affect the present. But this conflates existence now with having causal efficacy. On Presentism, causal explanations are perfectly coherent: present states are effects of earlier states, even though those earlier states no longer exist. That is not indirect causation; it is just causation across time. Likewise, on a Block Universe view, causal relations are encoded in the spacetime structure itself. Nothing needs to “move” between moments for there to be causal dependence.

Third, and most importantly, about perception. When I say that perception is direct, I am not claiming that the past is perceived as past, or that temporal mediation is eliminated. I am denying that perception proceeds by inference from an inner surrogate. On Presentism, my perception is directly related to a presently existing physical state through which the mind-external object is perceptually available (for example, light now arriving), where that state is itself the lawful causal manifestation of the object. On a Block Universe view, my perception is directly related to a temporal part of a mind-external process. In neither case is the direct object of perception a mental item that stands in need of inference to reach the world.

This is why the regress point still matters. If the mere fact that a causal chain involves time were enough to make perception indirect, then your own claim that perception is “directly of something that exists in my present” would not stop the regress. That present item would itself be temporally conditioned, causally structured, and conceptually articulated, and so—by the same standard—would require a further intermediary. To halt the regress, something must count as non-inferentially present to the mind, and temporal mediation alone cannot disqualify it from playing that role.

So the core issue is not whether only the present exists, or whether time is block-like. It is whether “direct” means non-inferential openness to mind-external reality, or instead absence of temporal structure altogether. I reject the latter requirement, and without it, the arguments you’re pressing don’t force the indirect realist conclusion.

I think your equivalence argument is very helpful, because it shows exactly where the disagreement lies. I’m happy to grant that (1) and (2) are on a par: variation in neural mapping or phenomenal character by itself does not make perception indirect. Where I part ways with you is in the move from (2) to (4), and therefore in the further identification of (4) with (3).

Your argument assumes that if two systems produce the same proximal neural outcome—e.g. stimulation of the B neuron—then they must have the same intentional structure, and so must count equally as cases of direct perception. That assumption is precisely what I reject. Intentional content is not fixed by proximal neural causes alone, but by the functional role of the system within a larger perceptual apparatus and its standing relations to the environment.

In (2), Jane’s eye and visual system are her means of perceiving the strawberry. The neural outcome is part of the perceptual process itself, not something she perceives instead of the strawberry. Differences in mapping (including inversion) affect how the strawberry appears, but they do not introduce a new object of awareness or a new question about whether what is presented corresponds to the environment. The system is functionally integrated into perception in such a way that the strawberry itself remains the intentional object.

In (4), by contrast, the visor is not functioning as part of the subject’s perceptual system in that sense. It is an external device that intervenes on the subject’s neural states, fixing the perceptual outcome independently of the normal perceptual linkage between subject and object. Even though there is no intervening “image,” the visor still determines what the perceptual state is of by inserting itself into the causal–functional role that normally belongs to the visual system. That is why (4) is relevantly like (3), not like (2): in both cases, the visor fixes the intentional object of experience rather than merely enabling access to it.

So the difference I’m drawing is not about where in the causal chain neural stimulation occurs, nor about whether phenomenal character matches across cases. It’s about whether the system producing the perceptual state is functioning as the subject’s means of perceiving the object, or as an intervening device whose outputs stand in for the environment. In (2), the eye is part of the perceptual system through which the strawberry is perceived. In (3) and (4), the visor plays a different functional role: it intervenes on perceptual states in a way that makes it intelligible to ask whether its outputs succeed or fail as presentations of the environment.

If one insists that intentional structure must collapse entirely into causal structure, then your equivalence argument goes through. But that insistence is a substantive internalist thesis, not something forced on us by physics or by the bare facts of causal mediation.

In what sense has your visor "failed" to present the strawberry to you? And why do we not ask if my eyes have "failed" to present the strawberry to me? — Michael

First I want to say that I agree that it would be a mistake to say that particular wavelengths or phenomenal characters “succeed” or “fail” as presentations of a strawberry. Neither wavelengths nor raw phenomenal character, taken in isolation, have correctness conditions. The notion of success or failure I’m invoking is role-based, not phenomenal: it concerns whether a system’s outputs function as presentations of the environment in a way that is answerable to how things are beyond the subject.

This is where the visor differs from ordinary vision, even allowing for spectrum inversion. In the inversion case, we can grant that you and I have systematically different mappings from wavelength to phenomenal character, and even that we enjoy the same phenomenal character when looking at the strawberry. Still, neither of us is perceiving an image that stands in for the environment. We are perceiving the strawberry itself, through our visual systems, under whatever lawful mapping holds for each of us. That lawful variation does not introduce a further question about whether what is presented corresponds to the environment; it just is how the environment is perceptually available to that subject.

The visor changes this structure. The crucial difference is not that the visor is physical while the eye is biological, or that one is reliable and the other not. It is that we do not perceive the outputs of retinal processing and then perceive the world by way of them; we perceive the world through the eye. By contrast, the visor produces an image that is itself an object of visual experience—something we see, and which purports to show us the scene beyond. Because the visor’s image is experienced as a presentation of the environment, it becomes intelligible to ask whether it corresponds to how things actually are. That is the sense in which its outputs can succeed or fail as presentations.

So the reason we do not ask whether your eyes have “failed” to present the strawberry is not that eyes are infallible, but that they do not introduce a detachable presentational layer whose outputs play the epistemic role of standing in for the world. The visor does. And that difference—between perceiving an object through a medium and perceiving an image that purports to present the object—is what introduces a normative dimension that causal covariance alone cannot supply.

Thanks for laying that out so clearly — I think this makes it obvious that we’re no longer talking past one another.

I agree with you on several points up front. I’m not trying to refute indirect realism, and I agree that IR can accommodate error, coordination, and action-guidance without collapsing into skepticism. I also agree that the epistemic situation is deeply underdetermined, and that “we can’t know with certainty” is a live philosophical outcome.

Before responding to your specific objections to experiential transparency, phenomenology of acquaintance, disjunctivism, and action-guidance, I want to flag why I haven’t engaged them directly. I’m not defending any of those positions. My claim is narrower: that the epistemic primacy of experience is itself a substantive theoretical commitment, not a neutral starting point. Whether or not disjunctivism or transparency succeed on their own terms is downstream of that more basic question.

Where I think we part company is over whether that underdetermination should be treated as a neutral default or as a philosophical choice. On the IR picture you’re defending, experience is effectively insulated from normative assessment: it can be causally aberrant, but it is not itself answerable to the world in a way that admits of correctness or incorrectness. Normativity enters only at the level of relations between experiences or theoretical overlays.

My move is to deny that insulation. I don’t think experience itself justifies our judgments, but neither do I think justification requires experiential comparison. The justificatory work is done by the norms governing world-directed judgment — norms shaped by, and answerable to, stable patterns of successful interaction with the world — with experience playing a causal, enabling role rather than an evidential one.

So I take your position to be coherent, but not compulsory. The disagreement, as I see it, is not about whether IR is viable, but about whether experience must be treated as epistemically primary in the first place.

I've been there myself, a long time ago. I agree with you: I was being chertitable. The post got some facts right, but it was wrong in all the ways that really matter. Take care.

Well, he at the least served as a poor example, showing us that the theory that there are two populations does not have a truth value. — Banno

Fair enough—I’m resisting the nudge to deny truth-value, but I’m happy to concede that Frank was a poor example either way. While I stop short of moving from “no epistemic role” to “no truth value,” the practical upshot is basically the same.

Good reply. — Banno

Cheers.

I think you’re right that we're hitting bedrock, but here are some additional thoughts for your consideration.

I’m not assuming direct realism in order to know that there is error. What I’m rejecting is the assumption — which I take to be doing a lot of work in the IR picture — that error must be identified by comparing experience with either a mind-independent phenomenal property or an inner experiential surrogate.

Returning to colourblindness: the basis for calling the judgment an error is not that the colourblind person’s experience fails to match mine, nor that it fails to match some phenomenal property instantiated by the object. The basis is that, within a shared practice of identifying and re-identifying objects across conditions, their judgments systematically fail to track features that figure in stable, publicly coordinated practices of correction and re-identification. That is an epistemic failure relative to those practices, not a phenomenal defect.

Importantly, this does not require assuming that colours are intrinsic properties of objects in a naïve realist sense. It only requires that our judgments are answerable to how things are in ways that admit of correction, stability, and disagreement. The colourblind person’s experience is not “wrong”; the experience is simply different. What can be wrong is the world-directed judgment, assessed within that normative context.

On the IR picture you gesture at, where there is no genuine error in perception outside of hallucination, I would say that this is not a neutral starting point but already a substantive philosophical commitment — one that insulates experience from normative assessment altogether, treating it as epistemically foundational rather than answerable to anything beyond itself. My move is to deny that insulation. It's not a leap so much as a refusal to grant that experience must be epistemically primary in the first place.

So the disagreement isn’t really about colourblindness as such. It’s about whether we think the notion of error belongs fundamentally to experience, or to the judgments we make about the world from within perceptual practices. I’m firmly in the latter camp.

You, EQV, just refuse to commit. :wink: — Banno

Poor Frank leaves the lab more confused than when he came in, but that's OK — he wasn't doing any philosophical work for us, and probably won’t be invited back. :wink:

I didn't say that it's about neural states. I'm saying that phenomenal experience is neural states (or emerges from them). — Michael

Thanks, that clarification helps, and I agree with more of what you say than perhaps my earlier wording suggested. I also do not claim that perceptual content is about neural states, nor that distal objects or their properties are literally present in neural activity. I also agree that there is no “real appearance” transmitted through space and into the brain. Where I think we still disagree is about whether causal covariance exhausts the intentional structure of perceptual states.

On your view, phenomenal character is self-standing, and the relation to distal objects is entirely causal. Given that, I agree that the visor looks continuous with ordinary perception: in both cases there is a neural state with a certain phenomenal character, caused in some way by the world. But that continuity is purchased by treating accuracy as non-fundamental—as a pragmatic gloss rather than a constitutive feature of perceptual content. If one is happy with that consequence, then that is a coherent internalist position; it is simply not one I accept.

I don’t think accuracy talk is optional in that sense. In the visor case, it matters whether the image on the screen corresponds to how things are in the environment, not because a “real appearance” is being compared to a copy, but because the perceptual state purports to present the environment itself—that is, it has correctness conditions that are not exhausted by its phenomenal character or causal history. The difference between an accurate visor and a misleading one is therefore not exhausted by differences in phenomenal character; it is a difference in how the state is answerable to the world.

This is why I don’t think the visor is just another causal conduit like light or reflection. Ordinary causal media do not introduce a layer whose outputs can succeed or fail as presentations of the environment. A visor does. Its outputs stand in normative relations to what is going on beyond the subject, even if the subject is unaware of the visor’s existence. That normative dimension is exactly what causal covariance alone cannot supply.

So the disagreement isn’t about whether appearances are transmitted, or whether phenomenal character is neural. It’s about whether perceptual states are merely causally covariant with the world, or whether they are constitutively world-answerable. If one denies the latter, then I agree the visor case collapses into ordinary perception—but only because one has already accepted a thoroughgoing internalism on which accuracy is not fundamental. That is the position I’m resisting.

Thanks, you've raised some good questions.

I’m not claiming that the mere fact that world-directed judgments can be true or false rules out inversion hypotheses, or renders them false. I’m happy to grant that spectrum inversion or other private aberrations remain metaphysically conceivable.

The claim is instead about explanatory role. Once truth and error are located at the level of world-directed judgment — governed by norms of use, correction, and responsiveness to the environment — inversion hypotheses no longer explain anything further about how perceptual judgments succeed or fail. They don’t add to our account of justification, error, or skepticism.

In particular, we don’t need to assume that colour is a property of objects or deny that assumption in order to make sense of perceptual error. Error arises when a judgment about the world fails to be satisfied by how things are, not when an inner experience mismatches an outer property.

Take colourblindness. We don’t identify error by checking whether the colourblind person’s experience matches ours or some phenomenal property in the object. We identify it through publicly accessible norms: stability across conditions, systematic correlations with wavelengths, successful coordination with others, and responsiveness to correction. The colourblind person’s experience is not incorrect — it’s simply different. What can be incorrect is the world-directed judgment when assessed within those shared practices.

Those norms are not arbitrary or merely conventional — they are shaped by, and answerable to, stable patterns of successful interaction with the world. But they are norms governing judgment, not standards for grading the intrinsic correctness of experience.

That’s why I say truth and error are located at the level of world-directed judgment. It’s not a leap so much as a refusal to start from the phenomenal-first picture that much of the traditional debate takes for granted. Once that picture is set aside, the notion of error no longer depends on assuming direct realism, but on the norms that govern our practices of saying how things are.

I agree entirely with the scientific picture you sketch: perceptual experience is realized in neural processes, and physics describes only particles, fields, and causal interactions, not colors, shapes, or appearances as intrinsic properties of objects. Where I disagree is with the inference you seem to draw from this. From the fact that phenomenal character is neurally realized, it does not follow that perceptual content is therefore about neural states rather than mind-external objects. That inference presupposes an internalist bridge principle—roughly, that the physical realization of a state fixes its intentional object—which is a substantive philosophical thesis, not a deliverance of science.

I reject that bridge principle. On my view, appearances are not intrinsic properties transmitted from object to perceiver, nor are they mental projections; they are relational ways objects are perceptually available to situated perceivers under specific conditions. This does not require that anything like an appearance be “carried” through space as a non-physical property. It requires only that perceptual states be individuated in part by their relations to mind-external objects—by the causal and counterfactual dependencies that link those states to the objects they are experiences of. In that sense, perceptual content is world-involving rather than internally bounded: what the state is about is constitutively tied to the object, not merely causally downstream of it. Science tells us how perceptual states are realized and transmitted; it does not by itself determine whether their content is world-involving or confined to the head. That question is exactly what separates internalism from externalism, and it cannot be settled by physics alone.

I agree that concepts are involved in perception, and that classification is norm-governed and interest-relative. But concept-involvement is not the same thing as perceiving concepts. The fact that there is no sharp, language-independent cutoff for when a Sun becomes a non-Sun, or a seed becomes a tree, shows that our classificatory practices are vague, not that there is nothing mind-external there, or that persistence through change is merely linguistic. Ontological continuity and conceptual boundaries are different issues.

Direct Realism does not require that the mind-external world itself “decide” when something counts as a Sun or a tree. It requires only that there be mind-external continuants with causal powers, and that perception be directly related to those continuants, even though the concepts under which we describe them are supplied by us. So when I say that the Sun I perceive is a temporally extended continuant, I am not claiming that “Sun” is an ontological category written into the fabric of the universe, but that perception is directly related to a real, persisting physical system rather than to an inner representation.

The Presentism/Block Universe distinction doesn’t change this. On Presentism, what I perceive is a presently existing continuant whose earlier state is made perceptually available by presently arriving light. On a Block Universe view, what I perceive is a temporal part of an extended object. Either way, the object of perception is mind-external, not something that exists only in language or concepts.

This is also why the regress point still stands. If temporal mediation or vagueness in classification were sufficient to make perception indirect, then all perception would be indirect—not only perception of mind-external objects, but even the “direct perception” of mental images or sense-data, since those too are temporally extended, causally conditioned, and conceptually classified. In that case, perception itself could never get off the ground, because every purported object of awareness would require a further epistemic intermediary, generating an infinite regress. Any account of perception must allow something to count as non-inferentially present to the mind, or explanation never begins.

So the disagreement isn’t about whether concepts are involved (they obviously are), but about whether perception is fundamentally a relation to mind-external reality, or instead a relation to inner items plus inference.

Allow me to apologize if my previous replies came off as an attempt to ridicule you. That was not my intention.

I see that what I've said so far has not convinced you. That's understandable. That said, I'm not sure I have the ability to express my critique any more clearly than I already have. I say that not in an attempt to blame you for misunderstanding me, but more as an acknowledgement of my own limitations in that regard. I still stand by my arguments, but I'm not sure how to productively move the discussion forward from here. Thanks.

It's not clear to me what you mean by perception...
Could you clarify? — Michael

That's a fair question, and I think the disagreement turns on a few distinctions that are easy to blur, so I'll try to make them explicit.

By perception I mean a non-inferential sensory openness by which an object is presented to a subject. By judgment I mean the act of affirming or denying that things are a certain way (“the apple exists now,” “the apple is red”). Perception is not itself a judgment, but it constrains judgment; inference is a further step where one belief is formed on the basis of others. So when I say perception does not proceed by inference from a surrogate, I mean that awareness of the apple is not achieved by reasoning from awareness of something else to the apple. When I say error lies in judgment rather than perception, I mean that perceptual presentation can remain world-anchored even when the judgment formed on its basis is false.

This helps with the slow-light apple case. If the apple disintegrates before the light reaches me, then the judgment “the apple exists now” is false. But that does not mean what is present to perception is a mental item or a memory. What is present is the apple’s visible presence at my location, carried by light. The light is not a third object perceived instead of the apple; it is the means by which the apple makes itself perceptually available across space and time. I am seeing the apple, but not the apple-as-it-exists-now. The mistake is one of temporal indexing at the level of judgment, not a loss of perceptual contact with the world.

This also clarifies the visor case. The point is not that the subject must consciously assess the visor’s accuracy. The point is structural: the epistemic warrant for beliefs about the distal object depends constitutively on the visor’s reliability. A visor produces a representation—an image whose correspondence to the scene is a further fact beyond what is perceptually given. Even if the subject is unaware of the visor, their access to the object is mediated by something whose correctness matters for warrant. That is why the perception is indirect.

This is what distinguishes visors from ordinary causal media like light, windows, or mirrors. Windows and mirrors can distort, but such distortions are typically perceptually available as distortions: a tinted window looks tinted, a curved mirror looks curved. They do not introduce a representational layer whose fidelity must be independently assessed in order for the object to be perceptually present. By contrast, a visor can systematically misrepresent without any perceptual cue that it is doing so. The difference is one of epistemic role, not degree of distortion.

That is why the question “when does mediation stop?” has no answer in terms of speed, distance, or number of causal links. Directness is not defeated by more mediation, but by a change in kind—from causal conduits that transmit an object’s own appearance to representational systems whose accuracy must be relied upon. Ordinary light propagation, reflection, and refraction do not play the latter role; visors, screens, and instruments do.

So the distinction I’m drawing is not a truism that would also accommodate a Cartesian theatre. The contrast is not between mediated and unmediated, but between non-inferential presentation of an object and awareness that depends on the correctness of a representational intermediary. That is the sense in which perception can be direct without being instantaneous, infallible, or free of causal structure.

Quite so. But then one has to explain what a hallucination of a dagger is, if not a mental image. That's not easy, because most people are absolutely sure that, like Macbeth, they see a dagger that is not there. Hence, a dagger-like object. Illusions like the bent stick are easy - we can demonstrate that the stick in water should look as if is bent - it's an actual physical phenomenon. At the moment, I'm inclined to just say that Macbeth is behaving as if he can see a dagger, and believes he is seeing a dagger - but there is no dagger and hence no perception of a dagger. — Ludwig V

That’s more or less the approach I take as well. On my view, hallucination involves mental imagery together with a false judgment that something mind-external is being perceived. There is imagery and belief-like commitment, but no perceptual relation to an object.

I think your slow-light apple case is a very good stress test, and it helps clarify what “direct” can and can’t mean.

If we build “direct perception” to require strict simultaneity — the object must exist at the very time of the perceptual experience — then your conclusion follows. With light at 1 m/s, after the apple disintegrates I would still have an experience as of an intact apple, and it would indeed be odd to say I am directly seeing something that does not exist now. But that shows that the simultaneity requirement is doing the work; it is not forced by the ordinary contrast between direct and indirect perception.

On the view I’m defending, “direct” does not mean instantaneous or unmediated by delay. It means that perception does not proceed by inference from an inner surrogate. In your case, what is present to perception is not a mental intermediary, but a worldly manifestation of the apple itself — its visible presence at my location. The light that carries this presence is not a numerically distinct object perceived instead of the apple; it is the means by which the apple is perceptually available across space and time.

So am I directly seeing the apple? The right answer, I think, is: I am seeing the apple, but not the apple-as-it-exists-now. I am perceptually related to the apple as it was at the relevant emission time, via its causal presence reaching me now. That is not “seeing a non-existent object” in the sense that would imply illusion or imagination. The error, if there is one, lies in the judgment “the apple exists now”, not in the perceptual relation itself.

This also answers the non-arbitrary cutoff worry. There is no threshold speed or distance at which perception suddenly flips from direct to indirect, because directness is not a function of causal delay. Delay determines which temporal aspect of the object is perceptually available; it does not introduce an epistemic intermediary. The relevant contrast is between perception as non-inferential openness to the world and cognition that proceeds by inference from a representation.

Finally, when I distinguish proximal stimulation from the intentional object of perception, this is not a retreat to indirect realism. The proximal stimulation is not something we perceive instead of the object; it is how the object makes itself perceptually available within the physical world’s causal structure. That distinction allows us to acknowledge causal mediation without collapsing perception into awareness of inner or outer surrogates.

It is not my intention to obscure the facts. I am engaging honestly with you - and in good faith - even if it may not seem like it to you.

Here are the facts as I understand them:

The formal definition I provided to you (or similar variation) is the one you will find in many of the standard textbooks on Real Analysis, Set Theory and Discrete Mathematics that discuss countably infinite sets. This is why it confuses me when you say that you don't believe that this is the standard formal definition of "countably infinite".

Likewise, and for the same reason, I am also confused by your insistence that the definitional existence of a bijection requires that the bijection be temporally or procedurally executable. Within the global mathematics community it is commonly understood and accepted that procedural execution is not a requirement for definitional existence. This is why you will not find such a requirement listed in the aforementioned textbooks. This is also why I previously stated that adding this requirement would amount to something like an external constructivist critique of the dominant paradigm.

I hope that this helps clarify my perspective on this. I understand that you may not agree with the criticisms that I have offered, but they are based on sincere and honest confusion regarding your claims, given my current understanding of academic mathematics. I am certainly open to being mistaken on these points, but it's currently hard to see how given that these are fairly basic observations about how mathematics is currently done. Thanks.

Thanks — this is a very clear statement of your position, and it helps isolate where we disagree.

I agree that perception is causally mediated and temporally downstream, and that in cases like astronomy our perceptual access depends on events in the past. I also agree that if objects are individuated strictly as momentary temporal instantiations, then the Sun-at-t is not identical to the Sun-at-t–8 minutes, and that no relation can obtain to what does not exist.

Where I disagree is with the inference you draw from this. I do not take the objects of perception to be momentary temporal stages. On my view, mind-external objects are temporally extended continuants that persist through change. The fact that the Sun is continually changing does not entail that it is a numerically different object at each instant in the sense required to break perceptual reference.

I also reject the claim that temporal mediation entails that the object of perception must be a present mental item. The causal chain explains how perception occurs, not what perception is of. That the light emitted earlier makes perception possible does not entail that what is perceived is a memory, an illusion, or an inner surrogate. It shows only that perceptual access is finite and temporally indexed.

This is why I think the Caesar and crime-scene analogies mislead. We deny direct knowledge of Caesar not because he is in the past, but because our access is symbolic, testimonial, and inferential. By contrast, perceptual access to the Sun or a ship is sensory and causal, not mediated by beliefs or descriptions. Temporal distance alone does not make knowledge indirect; mode of access does.

Finally, I think your conclusion overgeneralizes in a way that undermines Indirect Realism itself. If temporal mediation and non-simultaneity were sufficient to make perception indirect, then all perception would be indirect — not only perception of mind-external objects, but even the perception of mental images or sense-data, since those too are causally and temporally mediated. In that case, perception itself could never get off the ground, because every purported object of awareness would require a further epistemic intermediary, generating an infinite regress.

So while I agree that a relation cannot obtain to a non-existent object as such, I deny that this forces the conclusion that the object of perception must be a present mental item. The disagreement now seems to be about ontology — whether objects are momentary temporal stages or persisting continuants — rather than about logic or semantics.

If you re-read my reply carefully you will see that I did not say that mathematicians do not use the word "capable", but that they use it in a different way.

"A is countable" means "∃f such that f is a bijection between A and ℕ". That's it. There is nothing procedural in this definition. That was the point I was trying to make.

Thanks, that’s a fair question — but I think it slightly mislocates the point I was making.

I’m not claiming that the mere fact that world-directed judgments can be true or false rules out inversion hypotheses, or renders them false. I’m happy to grant that spectrum inversion or other private aberrations remain metaphysically conceivable.

The claim is instead about explanatory role. Once truth and error are located at the level of world-directed judgment, inversion hypotheses no longer explain anything further about how perceptual judgments succeed or fail. They don’t add to our account of justification, error, or skepticism.

In particular, we don’t need to assume that colour is a property of objects or deny that assumption in order to make sense of perceptual error. Error arises when a judgment about the world fails to be satisfied by how things are, not when an inner experience mismatches an outer property.

So the point isn’t that inversion is impossible or incoherent, but that it’s explanatorily idle with respect to the epistemic issues under discussion — even if it remains metaphysically possible.

I thought it might be interesting to interject here since I see my position as being wedged between @Banno's and @Richard B's on the one hand, and @Michael's on the other.

I’m broadly sympathetic to the spirit of Banno's and Richard's replies here, but I wouldn’t go quite as far as saying these inversion scenarios are outright incoherent or fail to be truth-apt.

I’m happy to grant that scenarios involving inverted neural realizations or inverted experiential mappings are logically and even physically conceivable. Where I part company with Michael is in what follows from that conceivability. I don’t think the mere possibility of private experiential differences that make no difference to judgment, action, or correction does any epistemic work.

In particular, I don’t think such scenarios motivate skepticism, indirect realism, or the introduction of epistemic intermediaries. Even if two subjects differed in their neural realizations or phenomenal character while making the same world-directed judgments, all that would show is multiple realizability at the causal level, not that perception is mediated by inner surrogates or that perceptual justification is undermined.

So my view sits between the two positions on offer here: I don’t want to deny the coherence of these scenarios altogether, but I do want to deny that they carry the philosophical weight Michael wants them to carry. Once truth and error are located at the level of world-directed judgment, private inversion possibilities become explanatorily idle, even if they remain metaphysically conceivable.

Thanks for the clarification. I think it shows how much ground we may actually agree on. But I don’t think the temporal argument you’re introducing does the work you want it to do.

From the fact that perception is causally mediated and temporally downstream, it does not follow that the object perceived no longer exists, nor that what is perceived is a memory or an illusion. Temporal priority in a causal explanation does not turn perceptual awareness into awareness of the past in the sense relevant to memory or illusion. If it did, then all perception—including the sensory contents the Indirect Realist treats as directly known—would collapse into illusion as well.

When I see a ship, the light reflected from it may have been emitted a fraction of a second earlier, but the ship itself has not thereby ceased to exist, nor has my awareness become memory-like. The causal story explains how perception occurs; it does not determine what perception is of. Conflating causal mediation with indirect awareness is precisely the move the Direct Realist rejects.

So at this point, the disagreement is no longer about logic or semantics, but about whether temporal causation entails that the object of perception must be a present mental item rather than a mind-external object. I don’t think that entailment holds, and if it did, it would undermine perceptual realism of any kind, not just Direct Realism.

Well said.

Nothing is capable of being put into one-to-one correspondence with all of the positive integers. — Metaphysician Undercover

I will attempt to clarify once more for the sake of the thread.

This statement of yours is neither a theorem, nor a definition nor a logical consequence of anything from within the formal system. This is a philosophical assertion grounded in a procedural interpretation of "capable" that is foreign to the mathematics. All you are saying here is that the impossibility follows from your definition of "capable", and that you think your definition is the right definition. This is an external critique. At no point have you derived a contradiction from within the system. Therefore, nothing you have said so far justifies the claim that the system is inconsistent.

I apologize if this comes off as rude, but this has been spelled out multiple times now from multiple different users. I think that if we still can't agree, then we have probably reached a principled stopping point that no further clarification is likely to resolve.

It all depends on how one defines "countable" — jgill

Exactly. "Countable" means something very specific within the formalism. The critique provided amounts to a rejection of that notion, not a derivation of contradiction from within the system.

But you haven't derived your conclusion from the axiom “I exist”. You have simply defined existence itself as relation to your Window, and then ruled out other Windows on the basis of that definition. Given that ontology, symmetry is excluded by stipulation. But there is no need to accept this ontology, and there is nothing in the axiom "I exist" that forces it. So the contradiction you describe is conditional on your metaphysical definitions. If those definitions are rejected (and I do reject them), then the contradiction never surfaces.

Cheers. :up:

That's an interesting pivot. At this point, I think the disagreement is no longer about logic or indexicals. You’re explicitly adopting an ontology on which existence itself is defined by relation to a unique Window, and nothing exists independently of it. Given that assumption, symmetry is ruled out by stipulation. But that assumption is precisely what I reject, and nothing in the logical facts about first-person perspective forces it. So the contradiction you describe is conditional on that ontology, not a consequence of logic itself.

Nice. It looks like you've noticed the pressure point, however, I don't think your proposed solution evades the problem. You’re right that relativizing “You” undermines the argument, but replacing “your world” with “The World” doesn’t fix that, because “The World” is still being defined indexically as the world in which you are You. That just reintroduces the same subject parameter under a different name.

Unless “The World” can be specified independently of the very first-person perspective it is supposed to ground rather than being fixed by it, the argument remains circular. Capitalizing “World” doesn’t turn a subject-relative fact into a global one. As it stands, “You is global because it is true in The World” and “This is The World because I am You in it” mutually define one another.

If this still seems unclear or incorrect on my part, no worries. We may have reached the point where we're simply talking past one another.

Saying “in your reality only your perspective is first-person” is exactly the token-indexical point, not a denial of it. Once you relativize first-person facts to a perspective (“in your reality / in my world”), global absoluteness is gone.

This would become clear if the argument were to be formalized. At the beginning of the argument you treat first-person perspective like a predicate that takes a subject and a perspective as parameters. The decisive moment in your argument is where you introduce the notion of the Window. This is where you absolutize the predicate by dropping the subject parameter, thereby equivocating on the meaning of "first-person perspective".

Grammatical person does not track metaphysical kind: the fact that I must refer to your perspective in the third person does not make your perspective third-person simpliciter. It is first-person for you.

The non-triviality of self-location (“who am I?”) does not turn indexical facts into world-level constants. No contradiction arises unless one assumes—without argument—that first-person must be a single global slot. That assumption, not logic, is still doing all the work.

As SolarWind posted his thesis 5 years ago, I had been thinking about this issue also way before that. Here I used LLMs as a tool to help formalize the thesis. — bizso09

Fair enough. However, I must say that as I look through your conversation with Gemini I see the familiar pattern playing out where (in my opinion) the LLM treats your assertions largely as stipulations rather than pausing to assess whether the key inference actually follows.

As for your reply above, I would say that you are not deriving the singularity or absoluteness of “You”; you are simply stipulating it. From the fact that only one perspective is this one, it does not follow that only one perspective is first-person.

The inference “if another perspective were first-person, it would be You” is invalid; it confuses token uniqueness with category membership. Other perspectives are not You, but that does not make them third-person simpliciter.

No contradiction arises unless you assume, without argument, that “first-person” must be a single global slot. That assumption, not logic, is doing all the work. Absent an argument for that assumption, there is no contradiction to resolve.

Interesting. I think this nicely illustrates why we should not uncritically accept the output of LLMs when discussing philosophical topics (or anything else for that matter).

You keep sliding from token uniqueness (“only one perspective is this one”) to global uniqueness (“only one perspective exists”). From the fact that I can unequivocally tell who I am, it does not follow that there is exactly one first-person perspective in reality.

When I say “I am Esse” and you say “I am OP,” we are not asserting competing world-level facts. These are token-indexical truths with different centers. No meta-world or selector is required, and no contradiction arises unless you assume—without argument—that “first-person” must be a single global slot.

What forces solipsism or dialetheism in your reasoning is not logic, but the insistence on that unsupported premise.

I still think the argument equivocates between the necessity of first-person reference for asking questions and the existence of a unique, world-level subject. “You” is token-indexical, not an absolute global fact, so no contradiction arises when multiple subjects each truthfully say “I am You.”

Appealing to a “total perspective” doesn’t help here because a total description of the world does not amount to a single perspective. To get that, one would have to posit an additional subject that experiences all perspectives first-personally, which changes the ontology. Either that subject has its own first-person standpoint (in which case it is just another “I”), or it doesn’t (in which case it cannot ground a unique global “You”).

The problem is clear. The mathematicians in this forum refuse to accept the refutation, though it is very sound. — Metaphysician Undercover

I’m afraid it’s not, and I’ll try to clarify why.

All you’ve claimed so far is that mathematicians are working with a notion of infinity that you don’t accept, and you’ve given some philosophical reasons for rejecting it. That’s a legitimate philosophical position.

The problem is that this is a philosophical objection, not a mathematical one, and as such it doesn’t justify the claim that the mathematical notion of infinity is contradictory. The mathematical definition is perfectly sound relative to the formal system in which it is embedded.

By analogy: suppose we’re playing a game of Chess and, on your turn, you legally move your queen from d1 to a4. Suppose I respond to your move by saying: “that move doesn’t make sense because in real life kings are more powerful than queens and so only kings should be able to move like that”. That may be a fine external critique of the rules of Chess, but I haven’t thereby shown your move to be illegal. Given the established rules, it was a perfectly valid move.

Likewise, your objection to the mathematical notion of infinity is a meta-level objection. It doesn’t undermine the internal coherence of mathematics as it is standardly practiced. At most, it shows that the standard mathematical notion of infinity conflicts with your own metaphysical views.

If you wanted mathematicians to take this challenge seriously as mathematics, it would require proposing an alternative formal framework built around your accepted notion of infinity and showing that it does at least as much mathematical work as the existing one. As things stand, no such reason has been given for abandoning the standard definition.

I'll leave it at that.

Neat puzzle.

I think the apparent contradiction hinges on the fact that the puzzle quietly slides between two different levels of description: impersonal/third-person description and indexical/first person description.

These are not competing descriptions of the same kind. They answer different questions.

• The impersonal description answers: What exists?
  
• The indexical description answers: Which perspective is mine?

Once you keep those apart, the apparent contradiction dissolves.

The impersonal facts about the situation don't change, the only thing that changes is the perspective that is occupied. "You" are not an extra object over and above Alice, Bob, etc., but rather an indexical that shifts across perspectives.

No contradiction arises unless you mistakenly demand that indexical facts must be reducible to non-indexical ones.

I think this is a helpful clarification, but I want to push back on one point. The inference you’re calling “logical” is not on the same footing as the laws of identity, non-contradiction, or excluded middle. Those are formal constraints on any intelligible discourse whatsoever. By contrast, the premise “the mind is only directly aware of the senses” is not a law of logic; it is a substantive epistemological thesis.

The argument you give is valid if one accepts that premise, but that is exactly what the Direct Realist denies. The dispute is therefore not about whether one accepts logic, but about whether one accepts a particular account of what perceptual awareness consists in. Rejecting that premise is no more a rejection of logic than rejecting sense-datum theory or representationalism would be.

To put it another way: the claim that sensory mediation entails awareness only of inner effects is not logically forced. It is a philosophical interpretation of perception. The Direct Realist’s alternative claim is that perceptual awareness is a relation to mind-external objects via sensory capacities, not an awareness of sensory items from which external causes must be inferred. That difference is not something logic alone can settle.

So I agree with you that if someone simply takes it as a basic truth that the mind can only ever be directly aware of sensory items, then no argument will move them. But that cuts both ways. What’s at issue here is not acceptance or rejection of logic, but which epistemological starting point one finds more compelling.

It looks like we've circled back to the starting point again, which is fine. I think this shows that we still have a disconnect at the level of foundational epistemic commitments. Your response attempts to push the discussion back into the traditional framing, whereas my view rejects that framing. It seems like we've hit bedrock here.