- seCtIon FIVe: BoRDeRLAnDs
visual perception. But then we have to ask: what makes the higher-order state
conscious? Why is there any what-it’s-likeness to those laughing faces? And we
cannot invoke a further, third-order state, for people aren’t conscious of being
conscious of a hallucination – they are conscious of the experience itself (Pret-
tyman, 2012). One riposte is simply that ‘the higher-order state happens to be
the right kind of awareness’ – the kind that we call phenomenal consciousness
(Brown, 2012). But this still leaves us asking ‘why?’
Dan Dennett begins Consciousness Explained (1991) with ‘Prelude: How are
hallucinations possible?’, intending this to prepare the ground for his multiple
drafts theory of consciousness. He proposes what would now be called a pre-
dictive-processing account, based on ‘generate-and-test’ theories of perception:
perceptual hypotheses based on expectations and interests are constantly cre-
ated and either confirmed or disconfirmed by the sensory input. This cyclical
process of generate-and-test produces a model of the world that is constantly
being updated but relies on having sufficient sensory input. When deprived of
meaningful input, the data-driven part of the hypothesis-generating system
lowers its threshold for noise. This means the answers coming back from the test-
ing-and-confirmation part make little sense, and it goes into a random cycle of
confirmation and disconfirmation. The result is hallucinations based on what the
system already knows about, whether that is the simplest of geometric designs
or highly detailed hallucinations produced by anxious expectation followed by
chance confirmation.
This account fits with much of what we already know about hallucinations: that
they are common during sensory deprivation, are induced by drugs that increase
noise through cortical disinhibition and other effects, and are often elaborated
into complex forms from simple beginnings.How does this help with Dennett’s theory of consciousness? If hallucination is
a phenomenon of prediction and interpretation, the key point is that ‘the only
work the brain must do is whatever it takes to assuage epistemic hunger’ (1991, p.
16). Sensory systems are seen not as providing a picture or representation of the
world that ‘enters consciousness’ or is watched by the audience in the Cartesian
theatre, but as continually asking multiple questions, checking against the input,
and acting on the responses. This implies that a principled reality/imagination
distinction is not required.
Predictive-processing accounts have made further progress with understanding
hallucinations. One suggestion is that psychosis involves a breakdown in normal
predictive processing. For example, people with schizophrenia fail to see many
common illusions and are not susceptible to the McGurk effect, and in binocular
rivalry their rate of switching between the two images is much slower than aver-
age, all implying a failure of predictive processing (Wilkinson, 2014).More generally, in this way of understanding perception, ‘your conscious percept
is determined by the overall hypothesis that your brain has adopted in order
to minimise prediction error’ (Wilkinson, 2014, p. 148). In this way, predictive-
processing accounts do away with any mystery about where the content of hallu-
cinations comes from, and obliterate any strict dividing lines between perceptions,
illusions, and hallucinations; they are all phenomena in which the brain selects the‘perceptions are
hypotheses [. . .] –
like the predictive
hypotheses of science’
(Gregory, 1966/1997, p. 10)