Michael Barkasi

philosopher and cognitive scientist

Could you be hallucinating right now?

Descartes famously claimed that “there are no definitive signs by which to distinguish being awake from being asleep” (Cress’s translation, Meditations on First Philosophy). There are a few possible interpretations of this claim, but an influential one is phenomenological. Dreams, and hallucinations of other sorts as well, do (or at least can) reproduce what it’s like as we use our senses, so that the experience you enjoy now as you (say) look around the room is one that could also be had while asleep.

To say that many philosophers (and psychologists, physicists, and random people off the street) have uncritically accepted this idea would be an understatement. For example, most people probably think the key premise behind The Matrix (the 1999 film) is realistic — they assume that there is a way to wire people up to a virtual simulation so that, from their private “inner” conscious perspective, it is for them just as it would be were they not in goo-filled pods, but physically out in the world.

Not everyone has bought into Descartes’ claim. John Austin, in Sense and Sensibilia, writes “we all know that dreams are throughout unlike waking experiences”. Surely Austin wrote the first three words with tongue in cheek, but a moment’s reflection should make his phenomenological point clear. What it’s like to dream is usually very different from what it’s like to be interacting with the world through your senses. For example, if you “see” an apple in a dream, the dreamed apple will usually appear less detailed — more “dim” — than an actual perceived apple. (Well-designed empirical studies bear out this point.) In general, dreams are often bizarre or “defective” in other ways that distinguish them from waking experience — e.g. dreamed objects may be achromatic, or have impossible confluences of colors.

Clinical hallucinations, like ones suffered by those with Charles Bonnet Syndrome, are also typically very different, phenomenologically, from successful perception. CBS patients, for instance, don’t have trouble distinguishing their hallucinations from what they actually see. Similarly, “seeing” things while on psychotropic drugs is not at all like actually seeing visual stimuli — the phenomenology is quite different. Sure, people often don’t know they’re hallucinating, but that’s not because their hallucinations faithfully reproduce perceptual experience. It’s because their hallucination is accompanied by cognitive malfunction (e.g. as with drugs or in schizophrenia) or cognitive suppression (e.g. as in dreams).

The point is that if you pay attention to real-life dreams and hallucinations, every example falls short of actually reproducing the phenomenology of perception. So, why are so many people so strongly convinced by Descartes’ claim? Further, what, exactly, are the (phenomenological) differences between hallucination and perception?

The nature of perception

If you ask a philosopher (or psychologist or neuroscientist) why they endorse Descartes’ claim, they’ll likely talk about how neural activity can be reproduced without the sensory stimulus. Whatever is physically happening in my head now (the electrochemical activity), as I look around the room, could be reproduced through direct electromagentic stimulation of my brain even if I was somewhere else entirely. In that case, my current perceptual experience would be reproduced, but I would be hallucinating — or so the argument goes.

The basic problem with this argument is that it presupposes a view of the nature of perceptual experience which itself is supported by Descartes’ claim (so, we have a vicious circle). Specifically, this argument works only if we assume that perceptual experiences are mental representations implemented in the spiking activity of the neural circuits connected to our sensory organs.

Now, many will throw up their hands and say (with exasperation) of course perceptual experiences are just mental representations implemented in the spiking activity of the neural circuits connected to our sensory organs! After all (they will say), studies in the lab show that perceptual experience can be manipulated by manipulating neural activity. They might also point out that what we experience can be “read off” brain scans by sophisticated decoders (e.g., see here and here).

The problem with this response is that this evidence is of essentially the same kind as what we had at the start. A fancy lab isn’t needed to know that experience can be manipulated by manipulating neural activity — psychotropic drugs will demonstrate that. To show that perceptual experiences are merely mental representations constructed in the brain, we need to show that they can be exhaustively reproduced in the absence of sensory stimuli — that is, we need to show Descartes’ claim! (No one, in a lab, has used artificial neural stimulation to exhaustively reproduce a perceptual experience; we’re a long way from that level of understanding and technology, if it’s possible at all.)

Worse yet, the dependence of perceptual experience on neural activity is compatible with an alternative view, one on which Descartes’ claim is false. The alternative view is that perceptual experience is a kind of revealing, or presentation, of sensory stimuli, enabled by the processing of information by our sensory neural circuits. As Benj Hellie and others have put it, on this presentation view, in perceptual experience sensory stimuli literally intrude into our private “inner” consciousness. So, for example, the apple I’m holding looks round because it is round, and that property becomes part of my experience as it’s revealed by my sensory processing (see also here, pp. 116-120).

Crucially, on this view, because perceptual phenomenology is literally shaped by stimuli themselves as they’re revealed to us, that phenomenology (the experience) can’t be fully reproduced in their absence (in a dream or hallucination). Because that revealing is facilitated by sensory neural processing, this view still fits the well-established experimental results showing that experience is modulated by, and dependent on, neural activity.

What’s missing from hallucinations

To summarize, Descartes’ claim isn’t supported by actual cases of hallucination, and while the mental representation view of perceptual experience would support Descartes’ claim, the argument for it presupposes the claim. So, a lot rides on Descartes’ claim. Is it true?

Well, as Melanie Rosen and others have pointed out, it doesn’t matter if dreams and hallucinations normally (or even always) fail to reproduce perceptual experiences. The question is whether they could exhaustively reproduce perceptual experience. Philosophers like Jesse Prinz and Bence Nanay (who endorse the mental representation view) try to make this possibility plausible by playing up the phenomenological similarities between hallucination and perception. Philosophers like Alva Noe and Bill Fish (who endorse the presentation view) play up the differences to make the possibility of perception-replicating hallucinations seem less likely.

For example, if — as suggested above — the differences between hallucinations and perceptual experiences come down to mere differences in detail or “vividness”, then (as Clark points out) these phenomenal differences can (plausibly) be explained by neural differences. Sure (the argument goes), actual dreams and hallucinations are less detailed and less vivid than perceptual experiences, but that’s because in real-life cases the neurally constructed mental representation is less detailed, and the neural activity itself is weaker.

If real-life dreams and hallucinations fell short of replicating perceptual experiences merely because of a lack of detail or “vividness” (a term that’s probably meaningless), this explanation would be a compelling way to support Descartes’ claim. But I think there’s another way in which dreams and hallucinations fail to reproduce fully the phenomenology of perceptual experience — a way that doesn’t allow for an easy explanation by proponents of the mental representation view.

What’s missing from real-life cases of dreams and hallucinations, but there in perceptual experiences, is the feeling of presence. By “the feeling of presence”, I mean to capture a loose collection of phenomenological qualities. These qualities are discussed by Mohan Matthen and Alva Noe, although here I’m not strictly following either’s view.

To get a grip on the feeling of presence, consider what it’s like to look over an object you’re holding. Here are the phenomenological features of the experience associated with the feeling of presence:

  1. The object (and your seen hand) looks to be in the same space as you; that is, you see the object from a spatial perspective, and that spatial perspective seems to be your perspective — your perspective on the space containing yourself and it.
  2. The object looks to be “ergonomically significant”; that is, it looks as if it’s available to be grasped, squeezed, manipulated, etc. If someone took the object from you and threw it at your head, you’d feel the urge to duck. In other words, as the object is presented in your visual experience, it appears relevant for bodily actions.
  3. The look, or appearance, of the object seems to depend on its position relative to you. For example, as you direct your eyes to different parts of the object, you bring different parts of it into view. As you turn the object in your hand, you bring different parts of it into view.
  4. The object seems to have an unlimited amount of detail which can potentially be brought into view, with the right movements or shifts of attention. For example, while you may not see every scratch or surface defect, or color variation, at once, as you shift your attention around you bring these into view.

Now, it’s plausible that dreams (and hallucinations) can reproduce some of these features. For example, dreams seem to involve (1), what we might call the feeling of immersive presence (see here, here, and here).

But it’s much less plausible that dreams and hallucinations involve some of the features lower on this list. What I propose is that dreams and hallucinations lack (4), the sense that there is an unlimited amount to detail accessible via the right movements. Instead, I propose, dreams and hallucinations present themselves as “complete”. What is presented in a dream or hallucination is exhausted by what shows up in phenomenal consciousness in the moment. In contrast, right now, as you look at a nearby object, you may only immediately experience a few details directly, but your experience at the same time presents an unlimited wealth of detail that’s accessible with the right move or body movements.

In this way, dreams and hallucinations are much more like physical images (paintings, film-printed photos, illuminated screens), or videos (sequences of images). Although you may not be able to take in all the detail of an image all at once, the image has all its detail on display (at once). (Similarly, each frame of a video has all of its detail on display at once.) The image is, literally, a temporally static snapshot. According to the proposal I’m making, dreams and hallucinations are very much like images in this respect. What’s presented in dreams and hallucinations is exhausted by what they explicitly present in any given moment — any “snapshot” of what’s in consciousness at a moment will exhaust what’s in the experience. But normal perceptual experiences, facilitated by sensory interaction with the environment, are different. What’s presented in perceptual experiences is not exhausted by what you take in at a glance; instead, these experiences present an unlimited wealth of detail accessible via the right body and eye movements.

Now, an important question is whether this phenomenal difference can be explained by a neural difference. Ultimately the success of this proposal will turn on that question, but note that there’s no simple or obvious proposal a proponent of the mental representation approach can make. For example, it’s not merely a difference in the level or amount of represented detail, so it won’t work to propose that the neural representations constituting dreams and hallucinations just involve less represented detail. The difference is somehow in the temporal profile of the experience: the detail shows up “all at once” in the one case, while the detail is “revealed” through sensorimotor exploration in the other case.

Conclusion

Could you be hallucinating right now? More specifically, could your current perceptual experience be reproduced in a dream or hallucination? Well, maybe. But it’s not obvious either way. We have no good evidence, thus far, one way or the other. Certainly, the evidence is not conclusive. If you couldn’t be hallucinating right now, what’s the difference? What is it about your current perceptual experience which couldn’t be reproduced in a hallucinatory experience? I propose it has something to do with the sense of presence and the temporal profile of the experiences. Hallucinations never (and can’t) reproduce the full, rich feeling of presence, and they are image-like insofar as their content is exhausted by what shows up, in any one moment, in consciousness.

If this proposal is right, one way to put the matter is that purely brain-generated experiences are restricted to a certain image-like, snapshot character. The brain, at best, is able to produce the phenomenological equivalent of static snapshots. (Here “static” may be a bad term, as these experiences may be more akin to video, with movement; still, at each moment in a video, all the detail is there.) To enjoy the rich experience as of a sensorily accessible environment, you may need to actually be engaged with that environment through a sensory system.

Could you be hallucinating right now?