Specification & Its Discontents

A topic that has been flying under the radar a little in Sabrina's language posts is the issue of specification. Sabrina's ecological analysis of language discusses information and what it means, but is not committed to the kind of law based account that is typically invoked in the perception-action literature. It can't - language can be used to talk about things in their absence, and it's not clear what kind of ecological laws might govern the connection between the speech event and it's meaning. Breaking specification has consequences, however, and has been a topic of some debate lately; the next few posts from me will be an in-depth look at the theoretical and empirical contributions to the debate as I try to come to terms with the idea and whether it can still support a direct theory of perception.

Gibson (and the key players who followed him, specifically Turvey, Shaw and Mace) insisted that specification was required to support direct perception. The lawfulness of the specification relation was the key defence against the early and strong attack on Gibson's theory from Fodor & Pylyshyn (1981; see Turvey, Shaw, Reed & Mace, 1981 for their response, and here for a brief summary of that argument). In essence, F&P think direct perception is the trivial claim that if you can perceive 'shoeness' you have directly perceived the shoe; if this what direct perception requires, it's clearly not what actually happens and thus perception is not direct. TSRM's reply is that not all properties can create information, and that only those created by an ecologically lawful process are candidate variables. This paper was the main reason the ecological approach wasn't killed off in 1981.

Recently, some authors (most notably Rob Withagen and more recently Tony Chemero as well) have been critiquing the concept of specification on theoretical and empirical grounds. Withagen and Chemero do not want to abandon Gibson, nor the directness of perception, but they feel that direct perception can occur without specification, and that relaxing this requirement will help ecological psychology account for a wider variety of behaviour. They also feel this will help make ecological psychology fit better with evolutionary biology: evolution favours things that work a little better, and doesn't demand access to 'the best' information for there to be an evolutionary advantage. Plus, finding specifying variables has so far proven really hard, so maybe they aren't all that common.

It may very well be the case that specification is not required for direct perception. Directness is a high bar, however, and anything less than specification will have to be able to work very hard to support it. If it can work, however, this could potentially be the bridge Sabrina is developing as she works on an ecological analysis of language, and could open the door to tackling all the hard problems representational cognitive psychologists like to throw at us and say 'mere perception can't handle this!'. But, on the grounds that a good theory gives us a reason to resist jumping ship at the first sign of trouble, I'm going to spend some time talking about this challenge and fighting it off until I can't any longer. 

The next post will lay out what specification is, and why Gibson, Turvey etc wanted such a high bar in the first place. I'll then review some of the papers suggesting specification is not required, critiquing the strength of the theoretical and empirical case against specification. At the end, I hope to draw some conclusions about how to take this research programme forward; it is potentially very useful, but it may simply not work for an ecological approach. This is going to get fairly technical, but I'll do my best to make it clear; these posts are me trying to come to terms with the argument and the data and I want them to serve as a useful resource.

Some reading I'll be discussing as I go along; feel free to read ahead :) If people are having trouble finding the papers I can create a public Dropbox folder containing the pdfs.

Tweet