What you experience when you hallucinate

Recently I published a paper on hallucination. Since the paper is long and technical in ways that obscure what I take to be its key insights, I wanted to write up those insights in a more accessible (and succinct) form.

Often, when we have normal perceptual experiences, it appears to us that our experience is of some particular object. For example, as I now look over this pencil I’m holding, my experience seems to present me not only with an array of colors, but with an object instancing various properties.

Hallucinations, at least often, have the same character: They seem to present us with particular objects. For example, in the “vanishing ball illusion”, many people “see” the ball thrown up on the last toss, when in reality the ball is palmed. That is, they have a visual experience as of a particular ball moving upwards through the air, despite there being no ball that’s actually been thrown.

Intuitively, we would describe this as an experience of something that’s “not really there” — you seem to see a ball moving through the air, but (at that moment) there is no ball that’s reflecting light into your eyes. So, does that mean that the thing you experience doesn’t exist? Should we describe this hallucination as an experience of nothing? The standard philosophical move is to say “yes”; on the standard account, when you hallucinate, there does not exist a thing which you are experiencing.

Of course (to continue with the above example), the ball that was thrown on the first few tosses still exists, but on the fake toss that ball is no longer interacting with your visual system, and so (on the standard view) it cannot be the thing you’re experiencing.

While, at first glance, this seems to be a reasonable way to describe hallucination, it turns out to cause all sorts of theoretical problems. Even on a first pass, the very formulation of the view isn’t obviously coherent: We’re essentially saying that when you have a hallucinatory experience of an object you don’t actually have a hallucinatory experience of an object (since there is no object which you experience).

The most popular way to resolve this tension is by construing experiences as representations. The idea is that when you have an experience, you’re in some mental state that represents the sensory stimulus causing it. These states are taken to have content of the form P(x), where “P” is whatever properties your sensory state attributes to the stimulus causing it, and “x” is the stimulus. When I look at my pencil, my state represents (and so I experience) my pencil, since my pencil is the stimulus causing the state. When you hallucinate the ball in the vanishing ball illusion, your state ends up not representing any particular object, since there’s no stimulus causing it. But since you still end up tokening a representation with content “P(x)”, you still (seem, to yourself) to have an experience of a particular object, even though there’s no particular object which you experience.

I’m skeptical that this representationalist approach actually works to explain the particularity of hallucinations, but let’s set that worry aside. The bigger issue is that the whole discussion, thus far, goes wrong by assuming that there’s no stimulus in the hallucination case. It is correct that, on the fake toss, the ball is no longer concurrently stimulating your visual system, but that doesn’t mean that your visual system is no longer engaged in interaction with the ball. It could be that your visual system fills in the missing input (on the fake toss) with a memory of the ball from previous tosses. In effect, your memory of the previous tosses affords you a causal link with the past-instances of the ball you did perceive, and your visual system’s use of that memory to fill in missing concurrent input amounts to just the same kind of interaction that happens in normal cases of perception.

That, anyway, was the idea that lead me to write my paper. I thought (and still think) that many instances of hallucination actually involve sensory interaction with stimuli — just not concurrent interaction with present stimuli. Instead, thanks to memory traces and our sensory system’s attempts to exploit those traces to carry out current sensory processing, it’s interaction with past-perceived stimuli. Since object-directed experiences are (so is widely assumed) experiences of the objects with which we interact through our sensory systems when we enjoy those experiences, it would follow that hallucinations are experiences of those recalled past-perceived stimuli.

I’ve been informed, since publishing my paper, that other people have also suggested that hallucinations are experience of the past (see here, here, and here). There are a few important ways my work goes beyond these prior statements of the idea.

First, I fill out the idea with a lot more precise and detail. In the prior instances, the idea is merely floated as a suggestion, in just a line or maybe a page. I afford sustained discussion.

Second, prior suggestions have all come as defenses of “relational” views of experience, according to which to have a perceptual experience of a stimulus is just to stand in a relation of awareness to it. Thus, the move has a post-hoc feel to it, or seems unmotivated apart from as a defense of a controversial theory of experience. Instead, I motivate the claim in a neural way that doesn’t presuppose any background theory of experience. I point out that hallucination just as much involves sensory interaction (with the past) as normal perception involves sensory interaction (with the present), and build from there.

Third, because I’m more careful to develop the view in a principled way, I restrict my claim to only certain cases of hallucination. The idea that hallucination involves experience of the past only makes sense for those cases of hallucination which do, in fact, involve sensory interaction with past-perceived stimuli (and this is not every case of hallucination). Prior statements of the view suggest that it may apply universally, to all cases of hallucination, or at best aren’t concerned to delineate the cases. But when you unpack what independent reason we have for thinking hallucination is experience of the past, those reasons don’t obviously apply to all cases.

Fourth (and finally), to the extent that prior attempts did try to motivate the claim, it was via minimal reference to how there are causal links between past-perceived stimuli and present hallucinations. But just because a hallucination I’m currently suffering may have a past-perceived stimulus as a causal antecedent doesn’t mean it’s an experience of that stimulus. (After all, thanks to how perceptual learning and synaptic plasticity works, current perceptual experiences have past-perceived stimuli as causal antecedents, but aren’t experiences of those past stimuli.) What’s needed, I think, to make the connection between past-perceived stimuli and current hallucinations is the idea that the subject is engaged in genuine sensory interaction with those past-perceived stimuli during the hallucination. Perhaps the key, truly original, insight of my paper was to (a) see how hallucination sometimes involves robust sensory interaction with the past, and (b) see how this stronger claim is needed to argue that object-directed hallucinations are experiences of past-perceived stimuli.