Information-Theoretic Adverbialism

ABSTRACT

Adverbialism is the view that to have a conscious perceptual experience is to be consciously experiencing in a certain way, and that this way is not to be understood in relational or representational terms. We might compare what it is for a conscious being to be experiencing in a certain way with what it is for a string to be vibrating in a certain way. This paper makes a new case for adverbialism by appealing to the fact that we can pick out ways of experiencing by treating them as information-bearing signals. In both presenting and defending the view, it will be crucial that information is not the same as representational content, but can still concern objects and the properties of those objects. The resulting information-theoretic adverbialism can answer or deflect the most influential objections to adverbialism.

KEYWORDS:

• adverbialism
• perception
• information
• Consciousness
• representationalism

Disclosure Statement

No potential conflict of interest was reported by the authors.

Notes

1 In addition to Tye [1984], adverbialism was advocated by Ducasse [1942], Chisholm [1957], Aune [1967], Cornman [1971], and Sellars [1975].

2 One might also defend an information-theoretic adverbialism about unconscious perceptual experience alongside the view that I am defending in this paper. The fact that conscious and unconscious perceptual experiences would be realised by different sorts of states might help to explain some of the functional differences between them, such as differences in degree of cognitive availability, while still allowing unconscious perceptual states to carry the same sort of information as conscious ones do, and explaining why they both count as perceptual states.

3 Of course it is possible for a representationalist to make use of a naturalistic account of representation of the sort that Millikan [1984] and Dretske [1995], for example, have offered. But Millikan and Dretske would not hold that tree rings represent the age of a tree, even on their naturalistic understandings of representation. Those rings nevertheless carry natural information.

4 I do not mean to tie my adverbialism to Scarantino’s account in particular. But it is useful to have a worked-out view to help fix ideas, and Scarantino’s emphasis on learning and background data fits nicely with other aspects of my account.

5 On his account, the informational content of a signal L with respect to a set of states (S₁, S₂, … , S_n) and background data d, is given by a set of the following sorts of claims, where i ranges from 1 to n:

C_i: S_i is incrementally supported to degree log[p(S_i | (L & d)) / p(S_i | d)] and overall supported to degree p(S_i | (L & d)).

Here p(x) is the probability of x, and p(x | y) is the probability of x, given y. Incremental support is a matter of changes to the probability of a state.

6 I have replaced Atick’s term ‘representation’ with ‘signal’ in the quoted passage, because that is what he actually means.

7 See Peacocke [1984: 373] for the ‘prime’ terminology, and Block [2007: 73] for the ‘the quale that I get when’ terminology.

8 See Singh and Anderson [2002: 507–8].

9 For such a claim, see Byrne [2003: 645].

10 Alternately, all of the terms in Tye’s account might be clear, in which case the mystery will be how that could possibly be a conscious experience.

11 As Cohen [2010: 106] claims, ‘it is generally conceded that [the many properties problem] is a very serious problem for adverbialism.’

12 The information-theoretic strategy also deals, in precisely the same way, with the many relations problem presented by Dinges [2015].

13 Tye [1984] gives it this label, but attributes the problem to Sellars [1975].

14 I frame the problem in terms of after-images because this is how Tye [1984: 199] presents it. This introduces some complexities, but they are not relevant for present purposes.

15 Indeed, considerations of transparency are part of what pushed Tye to reject his earlier adverbial views.