Wednesday, November 10, 2010

Runeson's Variables of High Informational Value

Now, if the theory of physics cannot be claimed to have monopoly on descriptions of "what is really there" , there is no longer any reason to assume that the perceptual systems must necessarily begin by registering what is basic to physics. On the contrary, we should expect perceptual mechanisms which directly register variables of high informational value to the perceiver. (Runeson, 1977, p. 173).
Why should we expect perceptual mechanisms that directly register variables of high informational value to the perceiver?  I'm not sure exactly what "variables of high informational value" are, but maybe they are like affordances, such as something like an apple.  Maybe the reason that we do not directly register affordances is that evolutionary constraints have made this impossible.  (This is a theme I've mentioned before.)  The early visual system (think back to fish) might have been constrained to detect light.  So, the best that we can do today is detect things like apples by detecting the light that they reflect.  We can detect apples and other fruits in trees by detecting their color.  The shift from the rejection of reductionism to physics to the adoption of directly registering variable of high informational value to the perceiver seems to me hasty.

Runeson, S. (2008).  "On the possibility of "smart" perceptual mechanisms".  Scandinavian Journal of Psychology, Volume 18, Issue 1, pages 172–179, September 1977


  1. OK, first of all 'variables of high informational value' are not 'like affordances'. You should know the difference by now, we've been banging it around long enough: affordances are properties of the world, information is the spatial-temporal pattern in energy arrays which specifies the affordance. You've got to keep world and optics distinct, they are related but not identical.

    Why should we expect perceptual mechanisms that directly register variables of high informational value to the perceiver?
    The argument has always been that direct perception the only way to reliably get to informative variables. The only patterns that persist and specify are higher-order, and these are available for direct detection. The elements from which these higher order variables are composed are either noisy or not available as individuals (they might be units of physics but not units of information).

    And who's telling evolutionary just-so stories now? You're in spherical cow territory unless you can establish a positive case for your suggestions.

  2. "variables of high informational value" sort of sound like maybe light intensities. They vary and could be of high informational value to a perceiver. Only light intensities are physical variables. So, that sounds wrong.

    Is information a variable of high informational value? That sounds wrong too.

    So, I was guessing by process of elimination that he must be talking about affordances.

    Actually, the apple thing is fanciful, but I think the evolution of primate color vision is commonly linked to frugivory.

  3. Instances of Gibsonian information are 'variables of high informational value' - that's why Gibson called them information! It's not a tautology, if that's what you're worried about.

    Light intensities are variables of only low informational value; they vary too wildly to specify anything or to be informative about anything much at all.