Monday, 17 May 2010

Reading Group - Heft (2001) on Gibson (Pt IIa)

Time for a break on affordances, and a quick check in with Heft. In the last chapter Heft laid out the key contributions Gibson made to the radical empiricist programme. In Chapter 4, Heft focuses on the specific consequences for Gibson of being a realist about relations, and continues to tie this to James via Holt. This post will cover the first half of that chapter.

Chapter 3 discussed Gibson's notion of the "mutuality between the knower and the thing known" (p.143). Chapter 4 now turns to "the context within which knowing processes transpire", i.e. the metaphysical landscape that serves as the basis for the act of perceiving. Simply put, this landscape is the rich web of relations between objects and events, which are real and perceivable. This web is the basis for the dynamic stability described by James and required for flexible but reliable behaviour.

Relations are stable
Relations between events and objects provide a stable framework within which perception can occur. These relations are not random; only certain objects/events can enter into a given relation, a given event may necessarily be preceded or followed by some other specific event, etc. The relations, while an indefinitely large set, are therefore not infinite in number and are thus constrained, making them potentially explanatory (this is a critical hurdle, and the idea that Gibsonian information was not sufficiently constrained was the initial dick move by the Establishment).

Relations are flexible
This web of relations is in constant flux, however, and the ever-changing global form of the web means behaviour is not locked into inflexible patterns. The flux is most readily apparent when you consider locomotion: objects and events come in and out of view, for instance, and your relation to these alters with every step.

(A note: This 'sweet spot' between stability and flexibility fits James' 'quasi-chaos' idea that these days gets formalised as 'edge-of-chaos' dynamics)

So, relations do the kind of work James wanted. He denied the two basic ways of getting relations between things (the empiricist/Hume account where they are contingencies you extract from experience, and the rationalist/Kant account where the structure is imposed by the mind or something similar) and makes them a primary part of experience. The reason for this is simple: relations are important, and no amount of processing will get you to them if they aren't part of your primary experience.

For Gibson, assuming relations are real makes direct perception possible, and provides perception with a rich source of potential knowledge, rich enough to explain behaviour. Gibson has answered his first question already: 'what is there to be perceived' is affordances. Now, relations gives him the answer to his next question: "what is it about the world that enables the perception of these things?" - the answer is ecological optics.

Sabrina has covered Gibson (1979, ch. 4) on this in detail so I won't repeat that summary. Suffice to say here that ecological optics is the study of the information potentially available in an ambient optic array to a moving observer. The proposed solution is that we have access to invariants over transformation, unchanging higher order relations between changing objects and events that specify those objects or events. Gibson's analysis solves a critical problem for traditional Helmholtzian representational theories, which assume that the information for an organism is the 2D retinal image, and that complex higher-order relations must be imposed by the contents of a representation: if the stimulus does not contain the information required for successful perception, then where does the content of the enriching representation come from? Both Gibson and James tackle this problem head on by discussing the classic example of the problem, depth perception. For Helmholtz, recovering the third dimension from a 2D retinal image was a serious problem facing the visual system that required help: both James and Gibson attempted to describe higher-order invariants that would allow the direct perception of depth, because they both felt Helmholt's problem was unsolvable as posed. Gibson had more success, because he had what James lacked: a theory of information.

If there is a specifying higher order variable and people detect it, then they have perceived the state of affairs being specified (not the light, nor the pattern in the light). This then also defines the empirical research programme for the ecological approach: step 1: identify the invariants in a given task, and step 2: establish whether organisms organise their behaviour with respect to these invariants. The former is very very hard, and frankly it's the empirical weak spot in the ecological approach. One of my personal aims is to improve the discipline with which we empirically engage with information (my personal best effort here is Wilson & Bingham, 2008) but really, it is very difficult, so while I think we need to do better, I can understand why we haven't yet.

Wilson, AD, & Bingham, GP (2008). Identifying the information for the visual perception of relative phase. Perception & Psychophysics, 70, 465-476. DOI:10.3758/PP.70.3.465

No comments:

Post a Comment