Perception and the Representative Design of Psychological Experiments

Perception and the Representative Design of Psychological Experiments

I’ve been reading a bit of Egon Brunswick’s work lately. A psychologist from the 1940s-50s. Some really solid ideas about perception. This particular book touches upon a bunch of things I’ve been thinking about lately. Perception. Statistics. Communication. It hits it all.

Here he is on perception.

In view of all this, one is reminded of Thurstone’s remark that perception is based on insufficient evidence, or of William James’s saying that perception is of probable things. Looking at cues of limited trustworthiness under the aspect of biological adjustment rather than under that of physiological mediation […] their ecological validities, defining more specifically their inherent potentialities as representatives of a more distal factor, should be mirrored within the response system of the organism by the actual effectiveness of the cue in establishing a specific reaction focused on the more distal variable. This effectiveness, or degree of utilization (or impression value, or subjective weight, or response eliciting power) of a cue may be small or entirely absent, even in the face of considerable ecological validity; or it may surpass it, as it probably does in many of the social stereotypes formed in our culture, although its share of distal functional validity cannot consistently exceed its ecological validity. In a perceptually well-adjusted organism or species, however, the rank order of utilization in what may be called the “or-assembly” of cues, or the “cue family hierarchy,” should be the same as the order of their ecological validity.

He has a really long-winded way of saying that people place weight on cues separately from their validity in the environment. I like this “rank order of utilization in the cue family hierarchy” concept. In well-adjusted people, they place emphasis on cues in tandem with their validity. An interesting way to think about arguments, not as either side being right or wrong, but rather ranking cues differently. Overrated vs. underrated instead of right vs. wrong.

He also writes about the generalizability of psychological experiments.

What, then, are the possibilities of applying results of an experiment to new situations by means of generalizing inference? The question can be answered in the same general way as it is customarily answered in the statistics of individual differences. Everyone knows that encountering, say, a wife who is taller than her husband (or, to come back to an example cited above, a railroad president who looks like a criminal) does not justify the inference that wives (or railroad presidents) are always, or are overwhelmingly, taller than their husbands (or indistinguishable from criminals). What the instances mentioned do demonstrate, however, is the fact that it is possible for a wife to be taller than her husband (or that not all railroad presidents can be told from criminals). They thus enable us to make minimum statements of the form “at least in certain individual cases . . . “ Experiments in the biological and social sciences are often formally analogous to the instances referred to above, by virtue of the fact that they do demonstrate a mere possibility; this time, however, specification is made by adding “at least under certain conditions … ,”

This fits with all the concepts developed in “Fooled by Randomness”.

Here is his bit on communication. A concise and timely passage, especially as he references Claude Shannon’s seminal work which came out only 6 years prior.

Whenever the “capacity” of a channel is less than the richness of variability of the source from which it accepts messages, the channel is “overloaded.” In this case no code will reduce the error frequency as low as one may please. Shannon and Weaver point out that regardless of how clever one is with the coding process, it will always be true that after the signal is received there remains some undesirable (noise) uncertainty about what the message was.

The crux of organismic adjustment which we have studied in this book may be rephrased in quite a similar way: distal perceptual and behavioral mediation must, in the nature of things, in the general case rely on overloaded channels, and the limited dependability of all achievement mechanisms is a result of this overloading. And we must further note that at least part of the trouble lies with the overloading and noise in the external rather than the internal medium.

According to Shannon and Weaver, the chances of error can be decreased by “redundancy,” however. Redundancy may be exemplified by, but is by no means restricted to, verbal repetitiveness. When there is noise there is some advantage in not using a coding process that eliminates all the redundancy, for the remaining redundancy helps combat the uncertainty of transmission.

And finally, for all their shortcomings, perception heuristics win.

The ability of perception to organize information into cognitively parsimonious units and subunits in overviewing the situation is unmatched among psychological functions. Its speed and richness make up for much of the shortcomings in strategy and attainment.