Sunday, June 10, 2012

Specification and Perception - American Philosophy Perspective

Over on the PsychScientists blog, Andrew is trying to work through the importance of a theory of specification for a theory of perception. (So far, here, and here.) Specification, in this context, refers to the relationship between the many energy arrays we are constantly surrounded by... but for the sake of simplicity we usually just talk about how an object or event shapes ambient light. The topic is worthy of a lot of thought because, traditionally, one of the most important arguments for modern dualism is the argument that there is no specification capable of supporting perception - because, the argument goes, there is much ambiguity in the environmental support of perception, some additional process is needed to explain how we know the world. But if there is specification, then perception could take advantage of it, and there are all sorts of cool ripple effects this has through any theory of psychology. Basically, if there is specification, and if organisms do engage with these specifying patterns, then perception can explain an awful lot without need to reference other psychological processes. The possibility of this type of specification, therefore, should certainly be in the top 5 list of important things for psychologists to figure out... because it could rewrite the whole game.

The strongest modern proponent of their being specification, was Gibson, and he might be the only person in history with a real theory about specification (this started with the 1961 Ecological Optics paper, though the final version has some modifications from there). There were an intense series of debates about this back in the early '80's and a much more recent set of arguments by Withagen and Chemero and others. While I am still unsure about the current debates, I am convinced that the earlier arguments were confused because no one in the debates understood the context of Gibson in the lineage of philosophy that started with Pragmatism - neither ecological psychology's critics or its proponents understood. Gibson certainly knew he was in that lineage, but I am unsure how well he understood his connection to it, and Gibson died a few years before these fireworks started. Turvey, Shaw, and Mace (TSM), the new vanguard, did not have a strong connection to that tradition.

My understanding of Gibson is quite different than TSM, it connects strongly with the work of Charles Sanders Peirce, though my understanding of Gibson developed a decade before I was exposed to Peirce. I have been trying to engage a Peirce scholar in writing about the connection, which I think could clarify many issues in ecological psychology. What comes below is sketch of that project, with an emphasis on how it is relevant to the debate about specification.

Truth-seeking and Truisms
We need to start with a basic principle of truth-seeking and a truism. The first is a principle from Peirce, which I will not try to defend here, but would be happy to write more about later. The second is a truism from William James, though it could have come from others.

Truth-seeking: To have true knowledge of the world it is not crucial that you are never wrong, it is only crucial that you are always wrong in a way that allows you to self-correct.

Truism: All situations are different, and all differences are (in principle) detectable... because a (as William James pointed out) there is no difference that doesn't make a difference.

Building a Theory
Gibson's direct-perception argument is a limited instantiation of Peirce's truth-seeking mechanism. Peirce intended the mechanism as broadly as can be imagined, to cover the discovery of truth of the most basic facts of the world and the most derived ideas. Gibson had a more focused interested in the basic mysteries of perception, as commonly understood - how do I get along in the world that I see, pick up my coffee cup, walk through the room, play a game of catch, etc. For Gibson, acquainted with the long view of historic debates about perception, it would be a major accomplishment to show that direct perception was possible. He argued that the ambient energy in the environment specified the objects and events in the world around us. This was a major conceptual leap, the leap of recognizing the existence of a "stimulus ecology" in which specification must occur... but as much as it was a major leap, it is also just a rephrasing of our truism. The existence of ambient-energy-with-structure-specific-to-the-surrounding-world hadn't really been considered before, because, for historic reasons, philosophers of perception were obsessed with things that they knew were ambiguous (e.g. projective images). A stimulus ecology can be explored, ad infinitum. Once can move around objects, get closer to them, interact with them. If one can engage with this energy in the right way, one can know infallibly about the world around them.

There isn't really any reason to go any further into Gibson-land, in my opinion, if that first part can't be agreed upon. If you don't believe you are surrounded by lights and sounds and floating chemicals throughout your day to day life, go no further. If you believe that two objects could be different in a way that is in principle undetectable, because no means of interacting with the objects will lead them to affect their environment differently, go no further.

If you agree with those points, then we have the fundamentals of an argument for direct perception - we do not have an argument that organisms always perceive the world perfectly, but we have an argument that organisms could perceive the world correctly under the right circumstances, with no intervening mental stuff required. And that is all we mean by "direct". With that agreed upon, our inquiry into perception then shifts to relevant organism-environment processes on very different time-scales: evolutionary tuning, developmental tuning, "in the moment" exploration. These processes are relatively crude, but they have a lot of potential. 
  • Evolution favors accurate behavior; although it is just as happy for crude shortcuts most of the time, it still favors organisms with the abilities to be sensitive to the types of information that specify the objects and events relevant to their niche/Umwelt.
  • Development favors accurate behavior; it is highly dynamic process, and organisms are startlingly good at developing in ways that match the demands of their environment. 
  • Exploration is where those mechanisms converge in the interaction of organism and environment, where the ambient energy allows for the detection of object and event properties. 
At this point, the rest is just a practical discussion. Sure, organisms often base their behavior on non-specifying variables, because non-specifying variables are good enough. On the other hand, it is possible to determine when one is wrong and readjust. But isn't this perception of energy, an intermediary? No, this is this direct perception, because "to detect the unique consequences of a thing" is "to detect the thing" (again, based on the here-undefended pragmatism!).

If someone is still trying to challenge the direct perception argument, there should only really be two lines of argument remaining open to them:
  1. "That's all fine and good, but in practice, you will never find an organism attuned to the correct variables."
  2. "Hey, back up a second, infinite uniqueness isn't really the issue here!"
I can draw out the ensuing debates, but I it would be a big distraction here. For now, lets just say that the debates lead to exactly the types of disagreement points from which people can actually do science... which isn't a bad thing.

 Back to Peirce
My point here is to say that if people try to fit Gibson's ideas into the framework of continental philosophy, then there are going to be problems, and you are going to end up in places where the ideas cannot be defended. Gibson's work fits into a different framework, one that challenges the starting place of its opponents on such a fundamental level that trying to argue things at the level of Gibson's theory of perception is to start the argument way, way, way ahead of where the actual disagreement lies. Example: In the continental tradition, you only need one illusory phenomenon to show that your opponent can never be 100% sure of anything perception tells him. In the pragmatist tradition, the fact that your argument is premised upon already knowing that the phenomenon-at-issue is illusory clearly indicates the potential for correct knowledge.

In the pragmatist tradition, the crucial factor is to be able to move towards the truth, to adjust one's self toward ever increasing accuracy. Gibson articulated a detailed mechanism that allows organisms to do this, at least within a clearly demarcated set of circumstances: The interaction of organisms with objects and events on an ecological scale of analysis, as mediated by their perceptual systems. Though this is less ambitious than Peirce, the limitation allowed for rapid advancement of the conceptual infrastructure needed to support the argument.


These ideas need to be drawn out more thoroughly, because I think they will help clarify both what ecological psychologists are up to, and why there is such a strong tradition of debates in which people seem to talk past each other. Hopefully this is making sense so far. If there are any Peirce scholars out there who think this sounds interesting, please comment!


  1. Hi Eric, we've talked some time ago, I've asked you if there's anything that applied behavior analysis might adopt from the ecological approach in order to increase its effectiveness in the topics of (1) verbal behavior, or (2) behavior modification in educational or clinical settings. I'm interested in those issues, and in the possibility of an intertheoretic communication between behavior analytic and ecological approaches. Which articles would you recommend me to explore, given these interests?

  2. MC, sorry for the delay... Summer is hectic, and I am working on a pair of massive posts (and several publications that desperately need submitting). The "verbal behavior" question is harder, because ecological psychology does not have a well developed approach to verbal behavior yet (though several people are working on it). The main function in an educational or clinical setting is offering alternative ways to conceive the relevant "stimuli" and the processes that the properly behaving organism is engaged in. This would help guide intervention, which could use standard behavior mod methods.

    To get your feat wet, I would recommend two readings:

    First, Ed Morris's review of Harry Heft's book "Ecological Psychology in Context". I think Ed does a good job summarizing the book, and noting ecological psychology's potential for close connection with modern behaviorism. The challenges Ed notes are valid, but a better understanding of Gibson-in-relation-to-Holt resolves most of them.

    Second, my paper on an ecological social psychology, the main points of which are summarized here.

  3. Thanks for your answer, in fact I've readen those two papers some months ago, and I've also readen Tonneau's chapter in the book you've edited about Holt's new realism, but I don't see how they could answer my questions. There're some evident affinities between Ecological Psychology and Behavior Analysis (e.g. Skinner, Kantor, Rachlin, Tonneau), but: is it possible to find more concrete (methodological or theoretical) relations? You said the issue of verbal behavior is already being explored: what is being proposed so far? You said the alternative ways to conceive stimuli and behaviors might help to guide interventions: how might they help, for example? Can we get beyond a general claim about their philosophical affinities?

    1. Fran├žois TonneauAugust 11, 2012 at 7:24 AM

      MC, it is only now that I see your interesting query. Due, among other factors, to the unfortunate habit of behavior analysts of worshipping operant reinforcement to the neglect of virtually everything else, there is almost nothing to answer your question. This is most unfortunate, but this is also a good opportunity for _you_ to build the needed bridges!

      To answer your question, I can think of only one short paper on discrimination training:

      Soraci, S. A., Carlin, M. T., Sharp, D. L., Franks, J. J., Vye, N., & Bransford, J. D. (1992). It's what's up front that counts: Some thoughts on the information given. Experimental Analysis of Human Behavior Bulletin, 10, 26-30.

      that you may find of interest.

      About matching to sample training, they write, p. 26: "The complexity of visual structure present in the stimulus array, especially that involving interstimulus relationships (e.g., salience, disparity, geometrical properties, etc.), has been in large part ignored."

      And they give concrete examples. Unfortunately I have not kept a copy of this paper, which may be hard to find.

  4. Francois,
    Thanks for weighing in!

    I have recently had the pleasure of meeting Ed Morris, and he has put me in touch with Max Jones, a behavior analyst in Australia who is very interested in getting ecological psychology into the ABA discussion. In particular, he thinks it is crucial to distinguish affordances from discriminative stimuli. Still working on a post that MC might find more satisfying, but life is getting in the way.