Harman on Patterns and Harms

Quentin Meillassoux’s  coinage ‘correlationism’ refers to any position which holds that thought can only think the relation between thought and its object; never the object as an absolute without relation to cognition. Kantianism and transcendental phenomenology are correlationisms. They conceive objectivity as an invariant structure of possible experience. Husserl conceives physical objectivity in terms of a subjective synthesis of experiences and anticipations which never give the thing in its totality. So for Husserl, there could no thought which lacked this temporal structure, just as there could be no thing that was not a correlated with it.

Graham Harman has recently cited Daniel Dennett’s patternist position (Dennett 1991) as well as Ladyman and Ross’ (L&R) ‘Rainforest Realism’ as examples of correlationism.

Here’s Graham from his discussion of Everything Must Go:

The sense in which real patterns are “real” is extremely weak in the book (as in Dennett’s article “Real Patterns” where the term seems to have been coined). They are real only because they are the most efficient descriptions, such that neither finer-grained nor coarser-grained descriptions of the same phenomenon allow us as much generalizing power. They are real only in comparison with false patterns that can be eliminated in the authors’ opinion (of which sensory qualia is the only unadulterated example given in the book).

And this is, I am sorry to say, a clearly correlationist position. These intermediate objects of the special sciences exist only for humans and animals who encounter them. The book holds that a mountain cannot even be said to be composed of smaller rocks, nor can patterns be linked to their causal ancestors either.

There are, arguably, problems with the patternist position. For example, I’m not convinced that it can deal with singularity or novelty or the reality of ‘noise’. There are also cases – e.g. in cognitive science – where functional decomposition may have more eliminativist implications than Dennett or L&R allow. Be this as  it may, I don’t think correlationism is one of the liabilities of patternism. Or rather if Dennett’s position is ‘correlationist for the reason that Harman gives, then Harman’s position is just as correlationist as Dennett’s and for the same reasons. If so, we better start being more nuanced in our metaphysical invective.

According to Dennett and L&R a pattern exists ‘is real’ if the compression algorithm required to encode it requires a smaller number of bits than ‘bit string’ representation of the entire data set in which the pattern resides (Dennett 1991, 34).

Both Dennett and L&R argue that compressibility is a non-observer dependent property of patterns – roughly a measure of the computational resources required to produce an encoding of the pattern (Ladyman and Ross, 202). An encoding can be thought of a as a string of ‘yes’ and ‘no’ answers (bits) formally isomorphic to the pattern itself (Collier 1999, 6).

The correlationist claims that any object of thought is constituted by a noetic relation to a possible thinker. For example, for Husserl the physical thing is the unity for consciousness of an open-ended multiplicity of perspectives (Husserl 1970, 64; Roden 2006, 81). To think of a physical object just is – according to Husserl – to think of this unity for a subject (See also Brassier 2010).

However, the mapping from pattern to string is not noetic but, as has been stated, purely formal; requiring only a one-one mapping from pattern to encoding string.

Such mappings or equivalences must also obtain within Harman’s universe. For example, let’s suppose there are Harmanesque withdrawn objects – call them  ‘Harms’.

If there are Harms we can know one thing about them: they are discrete and unified. We know this ex hypothesi because Harman says they are:

[The] reason we can’t make direct but partial contact with objects is that objects are essentially unified, and if you make contact with the parts of the unity then it’s not a contact with the unity itself.

If there are discrete Harms formal mappings of just the kind that obtain between encoding strings and patterns will abound in the Harmiverse. For example, if there are countably many Harms there is a  one-to-one mapping from the natural numbers onto all Harms. If there are uncountably many Harms – if the cardinality of the set of Harms is greater than that of the natural or rational numbers – then there is no such mapping. There are also one-to-one mappings between discrete Harms.

In conclusion, the formal relations between a pattern and its encoding string are of the same kind as must obtain between Harms – assuming that there are such. So if a mapping from pattern to bit string is noetic enough to be a ‘correlation’ in Meillassoux’s sense, so must any instance of Harm-Harm mapping. If Dennett is a correlationist, so is Harman.


Brassier, Ray (2010) ‘Concepts and Objects’, in Levi Bryant, Nick Srnicek and Graham Harman (eds) The Speculative Turn: Continental Materialism and Realism (Melbourne: Re.press), pp. 47-65.

Collier, John (1999). ‘Causation is the Transfer of Information’, in Howard Sankey (ed) Causation, Natural Laws and Explanation (Dordrecht: Kluwer, 1999)

Dennett 1991, ‘Real Patterns’, Journal of Philosophy 88: 27-51.

Husserl, E (1970),  The Crisis of the European Sciences and Transcendental Phenomenology, trans. David Carr (Evanston: Northwestern University Press).

Ladyman James, Ross Don, (2007), Every Thing Must Go: Metaphysics Naturalized, Oxford: Oxford University Press.

Meillassoux, Q. (2006) After Finitude: An Essay on the Necessity of Contingency, Ray Brassier (trans.). New York: Continuum.

Roden, D. (2006). Naturalising deconstruction, Continental Philosophy Review 38(1-2): 71-88.


5 thoughts on “Harman on Patterns and Harms

  1. Interesting slant, and one I have some sympathy for. I’ve thought for a long while though that the C-word is today’s equivalent of ‘logocentric,’ and that it may prove as hard to shuffle off as metaphysics proved to deconstruction (everyone denouncing their immediate predecessor as still ‘metaphysical.’) In any case, the mathematical compressibility argument does veer toward a purely abstract definition of ‘real’– not so unlike Badiou’s equation of ontology and mathematics; but the “pattern” half of the definition reminds me of nothing so much as Wittgenstein’s “logical form” in the T L-P. Anyway, thanks for this– it’s made me want to go have a closer look at the Dennett essay.

  2. Hi Skoliast! Thanks for the feedback. My argument doesn’t presuppose one rigorous interpretation of ‘correlationism’ – I just took Kant and Husserl as prototype correlationists here – and I’m not satisfied that I’ve got one to hand. I merely aimed to show that Harman’s complaint is misdirected.

    It is entirely possible that Harman’s real beef with Dennett et al. is that their theories are ‘logocentric’ in a sense: i.e. that they assume that the real is structurally encodable, though not necessarily by us.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s