Dabbling with Gibson

[From Bill Powers (930904.1500 MDT)]

Bruce Nevin (9308xx) --

Bruce, thanks for the article by T. Ryckman, " Is 'Ecological
Realism' Really Realism? Some implications of J.J. Gibson's
theory of perception."

Probably the most fascinating aspect of this semi-review article
is that I can't find a place for PCT epistemology in any of the 8
categories Ryckman lists, or in Gibson's. I can find some
connection to the first, REALISM, but only because as stated it
doesn't imply any relationship between "a fully determinate
external world" and our perceptions of that world. My position is
that we do have some evidence of such a world, supportable as
other scientific hypotheses are supported, but we can't prove it
and we can't state exactly how the world of perception is related
to it. The PCT argument for existence of a real external world
isn't a philosophical argument, but a factual one. Under certain
theoretical presumptions (justified by such evidence as the fact
that we must -- and can -- learn what actions are required in
order to control perceptions) we can make a theoretical argument
for an ordered external world. But the same theoretical argument
shows that it is unlikely to be isomorphic to the world we
experience. The degree to which we believe such arguments is, as
usual, dependent on the consistency with which this theoretical
stance conforms to evidence, not on philosophical universals.

The quotations from Gibson in this article have saved me a lot of
trouble obtaining references: they illustrate nicely the problems
I have had with Gibson. Consider this quote:

"When I assert that perception of the environment is direct, I
mean that it is not mediated by _retinal_ pictures, _neural_
pictures, or _mental_ pictures."

My usual response to such sayings is something like "That's very
interesting: how do you know that?" But my experience has been
that this sort of question is simply not understood by those who
argue in this manner. It took many years of reading philosophy
(no more often than necessary) to realize that Gibson's assertion
is meant to be taken as is, with no justification but the
listener's concurrence. If you concur, you're a Gibsonian; if you
don't, you get a different label. The assertion is true because
it was said. There is, in fact, no evidence either for it or
against it. The statement is not to be believed because of
evidence, but because of whatever plausibility the listener gets
out of it. That seems to be how this game is played.

Going to the content of the statement rather than its form, the
only way I know of to discuss it is in terms of observations and
theories. If perception is direct, but if it does not work in the
ways assumed by neurologists and the like (and control
theorists), then how DOES it work? Gibson goes so far as to say
that "sensory" nerves have been misnamed. If that is so, how
SHOULD they be named? What IS the role of the signals we observe
that seem to correlate with some aspects of external events and
some aspects of direct experience? Why does damage to such nerves
interfere with direct experience of the things with which the
nerves are taken to be associated? Gibson offers no model to
replace the ones he rejects. In effect he rejects the entire
modeling enterprise, preferring some other unnamed route to the

Do you recall the "mirror" diagram that Martin Taylor drew? In
this diagram, the hierarchy of perception was shown above a line
of symmetry, and below that line was a mirror image of the
perceptions in the hierarchy, showing how we project perceptions
into an external world. It seems to me that Gibson starts with
the mirror world in the environment, without considering that it
may simply reflect the organization of the perceptual systems at
many levels in his own head. Here's a quote from a discussion of
"ecological optics:"

"If _invariants_ of the energy flux at the receptors of an
organism exist, and if these invariants correspond to the
permanent properties of the environments, and if they are the
basis of the organism's perception of the environment instead of
the sensory data on which we have thought it based, then I think
there is new support for realism in epistemology as well as for a
new theory of perception in psychology."

Notice the questions begged starting in the first clause:
invariant, energy, flux, receptors, organism, properties,
environment. These are all some brain's concepts of an objective
reality, and in any epistemological argument we have to ask how
the brain (Gibson's brain) knows about them, too -- not just the
things to which they lead. Gibson talks about "light" and "solid
angles" and "surfaces," taking each one as too self-evidently
existent to be subject to epistemological questions. He seems to
exempt physicists from epistemological limitations: if physicists
say something exists, like photons, then by God they must really
exist. So if you can show that some perception depends on
photons, you have shown that the perception must have a real
counterpart, too. The unresolved epistemological questions in the
foundations of this sort of reasoning are just left in the dark
to gather mold.

Gibson doesn't seem to understand the concept of an invariant. An
invariant is created by applying a certain function to the
variables making up an input set. Given an input function,
invariants appear under some transformations of the input set.
That is, some actions on the environment alter the input set in a
way that does not alter the perception being generated by the
function. A perception is identified by the function that creates
such invariants. A perception can be controlled through actions
that alter the input set along trajectories NORMAL to the
trajectories of invariance -- i.e., actions that DO alter the
perception. That is how a perceptual variable is created.

The point that Gibson misses, or perhaps simply doesn't know, is
that the nature of an invariant is determined by the form of the
input function, not by the input set. If you alter the form of a
function, you redefine the operations under which a perception
will remain the same and those under which it will change. So
invariance per se is a function of the perceptual apparatus, not
of the environment.

Gibson does recognize that interaction with the environment is
essential. But I differ with his explanation for why this is
true. In a ballpark sense I agree with his concept of
affordances; it means nothing more than that the external world
does exist and does contain regularities. But I would place these
affordances at one more remove from experience than Gibson does.
Gibson says, for example, that the "layout of surfaces" affords
the derivation of visual invariances through interactions; I
would say that the nature of reality affords the possibility of
perceiving the world in terms of a "layout of surfaces."

I see "layout of surfaces" as being a constructed invariant in
itself. In most kinds of experience, there are many different
invariants that would serve equally well as a picture of reality.
In some, however, some ways of constructing invariances seem to
work better than others -- they afford the greatest
controllability of the resulting perceptions. To me this is
evidence for an independent reality -- the fact that there seem
to be "best ways" of perceiving things, in terms of effort
required for control. This still does not mean that these best
ways reveal isomorphisms of experience to reality, however.

'Nuff of that.



Bill P.

1 Like