A blog about problems in the field of psychology and attempts to fix them.

Sunday, May 6, 2012

Two Ecological Psychologies - Continued

Continuing on my attempt to update Cutting's 1982 paper... two questions seemed of the most interest to blog readers 1) Why did Cutting write the original article? 2) What consequences of following Gibson's approach vs. the Connecticut approach?

The first question is more difficult. A lot was going on in the field at the time. All I can say for sure is that Cutting thought that Ecological Psychology was being moved away from Gibson's vision, and he thought that some aspects of the emerging approach were problematic (either because they undid desirable novelty of Gibson's approach, or because they resulted in tautology and related logical problems). It is worth noting, however, that regarding most of the differences, Cutting did not claim either approach was superior, only that they were different. He thought these differences would lead, presumably in the near future, to a splintering of the field. While I think most of Cutting's insights about the two emerging approaches were spot on, it has been thirty years, and the field is still together. Explaining why Cutting was wrong in that final prediction requires that we answer reader's second question.

---Off to California soon. I will be responding to comments, but am not sure about new posts over the next month.----


In Pragmatism, William James said
There can be no difference anywhere that doesn't make a difference elsewhere — no difference in abstract truth that doesn't express itself in a difference in concrete fact and in conduct consequent upon that fact, imposed on somebody, somehow, somewhere and somewhen.*

This is a showy way of stating Peirce's pragmatic principle that any two ideas which have all the same consequences are the same idea. Applied to the present situation, we might say:
There is no difference between the two ecological approaches, unless it alters how a scientist, interested in pursuing perceptual research, would conduct themselves.
If we take this seriously (which I do), then we must not care in the slightest about the differences between Gibson's approach and the Connecticut approach, unless they can be shown to make a difference elsewhere, i.e., outside the theoretical discussion itself. I think it does make a difference which approach we take, but I think the field is held together by the fact that in the course of any particular study, it makes little to no difference.

If you were following either approach, you would likely start to investigate a given problem in exactly the same manner:
  1. Pick a behavioral task, either one that is clearly a perception-action problem or one where people are tempted to use cognitive explanations. 
  2. Try to identify potential perceptual support for that task (higher-order patterns in ambient energy). 
  3. Try to show that people use that perceptual support to succeed in the task (by showing that their behavior varies as a function of the relevant patterns).
Oh, sure, there will be some differences in term usage, but the overall formula is the same. The field only uses a limited array of mathematical methods for step 2, so there will be no difference between the approaches there. There are also a limited number of ways to do step 3 - some observational methods, and a handful of experimental methods - so there will be no difference between the approaches there. While it must be admitted that there have been some differences in which types of task proponents of the two approaches have pursued most vigorously, so far as I can tell that is just historic accident. Except for the UConn crowd's recent forays into kinesthetic perception, I have no reason to think the two approaches would differ significantly in the types of problems they would eventually pursue. (In the kinesthetic case, I am still unsure what to make of it from Gibson's perspective.)

I believe that the differences between the schools is found between studies - as people react to the results of the last study, or as they pick a new topic of investigation.

Where to Start?

After admitting that, in the long run, the two approaches would likely investigate most of the same things... they do differ in the types of experiments they seem to start with. Followers of Gibson's approach likes to start with seemingly cognitive tasks (especially what Chemero called 'representationally hungry' tasks), with known illusions (e.g. the 'bent stick'), or with quite complex perceptual tasks (e.g. knowing what is gravitationally level while on a rocking ship). Followers of the Connecticut approach likes to start simpler perceptual tasks (e.g., fitting through a door, stepping over a gap, jumping to a given height). That said, once the topic is picked, the method of initial investigation is the same, and frankly many choices of topic are not dictated by investigators desires. Other important factors include adviser's desires, available funding, local equipment, and expectations about publishability. Thus, the differences are more visible later.

What if your study didn't work out?

A study could fail at step 2 or step 3. 

If you are following Gibson's approach, it is possible that perceptual support does not exist for the task in question - that is, a person might be stuck deciding what to do based on task irrelevant variables. Thus, if you fail to find a mathematically specifiable 'invariant' that could guide people's behavior, it might just be time to stop. It is also possible that perceptual support can be mathematically determined, but that people do not generally use that support. Thus, if you find that people fail to use the available invariants, it might be time to stop. Sometimes you can re-conceptulize the task in a way that is helpful. If that happens, admit that there is no perceptual support for the task as commonly understood, but show that there is support for a related task.

If you are following the Connecticut approach, then all behavior is guided by invariants, and invariants exist to control behavior in all possible tasks (note also the shift from 'guide' to 'control'). Thus, if you fail to find a mathematically specifiable 'invariant', you have not looked hard enough. Repeat, repeat, repeat, or admit that you are a failure. Eventually, let us suppose you come up with a candidate invariant. If experimental evidence shows that people do not use the invariant you found, then you are a failure once again. It is back to step two, to find another invariant. Sometimes you can re-conceptulize the task in a way that is helpful. If that happens, make sure to pretend that your new task answers any and all questions about the old task, as if nothing changed.

Well... but that means there are a lot more admitted dead ends for the Gibson guys, right? Not really. There are many good research directions one could purse after the types of failure listed above: Maybe development is necessary? If you have an invariant, but your participants don't use it, give them a lot of experience with the task, and retest. (The Connecticut crowd generally won't do this, because it would require them to admit that the perception is imperfect at some point.) Maybe there is an invariant that can be detected by another species? Is there an animal that might be especially good at this sort of thing. (Like the rest of contemporary psychology, both approaches neglect comparative work. However, most comparative work that has been performed is in the spirit of Gibson's approach. Moving to animal work in this situation also requires the admittance that sometimes perception is imperfect.) Maybe there simply is not an invariant, or the invariant is not perceived - that is, maybe the task is not perceptual after all. This is not a failure per se, just an admittance that this particular is a problem for members of a different sub-discipline. If one was intrinsically interested in the task being investigated, it might be time to collaborate with non-ecological psychologists. (This conclusion, and the suggested collaboration, is heresy in the Connecticut approach.)

Thus, there is a big difference between Gibson's approach and the Connecticut approach it terms of what types of explanations they can offer for studies that do not find the desired results. This leads to different trajectories of follow up observations, moving the field in different directions. Among other things I personally care about, followers of Gibson's approach seem more likely to be interested in the process of perceptual development and in comparative work. (For practical purposes, James and Eleanor Gibson split up the problems of developed perception and developing perception, and it is a problem that the field has not been formally put back together.)

What if your study succeeds?

If your study succeeds, then you must determine what to do next. My vote is typically to find a new task and start from the beginning, but hardly anyone likes my investigate-a-million-things approach to research. So the real question is how to continue investigating a task after you have identified an invariant and shown that people use it.

If you are following Gibson's approach you want to keep things holistic - to always be investigating the organism-environment interaction. Most likely you will try to extend your current findings to highly related problems, and your options are pretty open: It might be nice to look for evidence in even more natural situations, or to involve participants you would expect to be more or less accurate. Maybe you will study the actual process of organism-environment interaction by watching a person perform the task to see how their movements are attuned. You might be interested in comparing free-moving performance with performance under constrained conditions. You might compare people's perception of affordances to the actual affordances (by having them both estimate their ability to do a task, and having them actually do a task), so that you can find error and perceptual bias. If you are really adventurous, you might move to investigate perceptual development or do comparative work. One weakness of Gibson's approach might well be that there is not a clear path for future investigation after initial successes.

If you are following the Connecticut approach, then you will probably start down one of the two obvious paths of investigation - further identify the object properties that permit the task to be performed (what they would call the 'affordance' or 'affordance properties') or further identify the organism properties that permit the task to be performed (what they would call the 'effectivity'). If we go down the former path, we will design a clever apparatus that allows us to manipulate object properties in ever finer detail to study ever more exactly when when participants (claim to) perceive the affordances or when they can perform the task. Only a minority of studies measure both, because the base assumptions of accuracy make that seem redundant. These studies can become infinitely detailed, but in doing so they move away from any obvious psychological question. If we go down the latter path, we will end up in a fog about what types of variables count as part of the "effectivity", but out of that fog we will probably end up with a series of simple measures of body size and strength. For example, if we are interested in your perception of whether or not you can fit through a doorway, we might measure your girth; if we are interested in your perception of whether or not you can jump over a gap, we might measure how far you can jump under various circumstances. These investigations can become infinitely detailed, but in doing so they move away from any obvious psychological question.

Trajectories of the field

One consequence of this difference in follow-up-study direction is that followers of the Connecticut approach have moved towards the more physiological sciences, and away from problems that would classically interest psychologists. Much of the best work in the field (and it is good work) is being done in Kinesiology, Sports Science, and Human Factors. I have nothing against detailed analysis of bodily movements, but I am concerned with the state of the field. Being a proponent of Gibson's approach, I see such work as peripheral to the field (something we should have some people working on), but not central to the field (something the majority of people should be working on). The heart of the field, in my point of view, should be investigating processes that are clearly psychological.

Well heck, why I am being so down on the Connecticut approach here? Who cares what is or is not a "psychological question"? Turvey sure doesn't.

True, but I do, and Gibson did. Gibson's approach was about solving problems in psychology. If the insights from his approach can be used to solve problems in other sciences, that is great. Such crossover happens in other sciences all the time. However, the desire to continuously return to core questions of psychology remains a difference between the two approaches.

Because of the way in which the Connecticut approach handles success and failure, there are areas of Gibsonian science that are clearly missing areas in the current Connecticut-dominated field. Perceptual development and comparative psychology have already been mentioned. Because of the increased emphasis on affordances within the Connecticut approach, there has also been a de-emphasis on the identification of specific invariants (e.g., we still have nothing resembling a general field of ecological optics), and a corresponding lack of attention to the movements of perception (with the exception of a very narrow range of studies looking the perception of length in a wielded rod).

(Preliminary) Conclusion

I hope that what is written above demonstrates that there is a difference between the approaches, i.e., that this stuff does 'make a difference' elsewhere. Alas, I began by admitting from the start that the two approaches would begin to investigate a given topic in the same manner. This base similarity means that investigations by followers from one approach are perfectly understandable to followers of the other approach. There is no Kuhn-esque incommensurability problem, no dramatic moment where the difference is obvious and painful. Thus, I don't know if I have demonstrated the particular types of differences that readers might care about.


* It would not be much of a stretch to morph the first part of James's quote into a more directly Gibsonian claim: "That there are no objects or events that do not, in some way, structure ambient energy in a manner specific to the distinct properties of those objects and events." But that discussion should probably be saved for later.

1 comment:

  1. The more approcahes, the more new colcusions could be summerized afterall. Every question in the psychological field can not be explored by a single perspective only0 the variety of approcahes gives the rightest explanation.