Why Words Hurt
I helped to organize a colloquium by Dr. Jonathan Schooler, from the University of British Columbia. It was fascinating.

Schooler was talking about verbal overshadowing. Here's the basic scenario: Person A sees a bank robber, go to the police station, and participate in a line up, trying to identify the crook. Person B sees the same robbery, but gives a detailed verbal description to the police before participating in the line up. Contrary to what one might expect, person B is much less likely to correctly identify the crook--the Verbal Overshadowing Effect.

Another interesting test involved three type of wine taster: uneducated ones, novices, and experts. They were asked to try some wine, and then to describe it (a control group, of course, skipped this step) and then to pick the wine that was most similar from a group of wines. The uneducated taster had essentially no idea how to talk about wine, and when they were asked to do so it helped them at the end, but only a very small amount. The experts new exactly how to classify each wine, and verbalizing the process improved their results. Finally, the educated people knew could kind of describe the wine, but when they did it had a severe negative impact on their ability to "pick it out again" later.

These neat effects also work with taste, music, colors, and a whole bunch of other stuff. Intuition is another area that is effected. Schooler gave an example of an intuitive problem (a story about a coin dealer who sees a rare Summarian coin with an emperor on one side, and 544BC on the other, how did she know it was a fake coin?) and a non-intuitive problem (the example was a standard logic puzzle about "A says that B was not guilty. C says that A is lying. B says, blah blah"). He asked people to solve a bunch of these puzzles and to talk about their process as they were doing it, and the people doing the intuitive problems showed did much worse. (This is, again, compared to a control group who was not asked to talk out loud during the experiment).

The first idea was that verbalizing the experience essentially created a model of it, and this model takes priority over the actual experience. If the model is a poor one, it has a corresponding effect on performance. This was my intuition as well, but it turns out not to be the case. For one thing, the quality of the verbal description apparently does not influence the performance later--even a really good bank robber description still leads to poor odds of picking her out of a line up. Secondly, someone did a test where they showed the subjects a male face and a female face, and then asked them to describe one of them but not the other. It turned out that even though they did not describe the male face (for example) their recognition performance was just as bad as if they had. So clearly the verbalization does more than just "build a model", because they were never asked to verbalize one of the faces.

So Schooler's theory is that verbalizing causes a processing shift from left to right hemispheres of the brain. Verbalizing causes you to focus on the particulars, and not to process things holistically. Then, we you have to make the match, the specific processing doesn't help you as much.

This is were things get really cool. Cool as hell thing number one: someone did an experiment in which, after viewing and describing a face, the subject was given a series of figures to look at. The figures looked like this:

If they were told to answer based on the little letters (The first one is "E", the second is "F", the verbal overshadowing effect was more noticeable, and they were less likely to be able to identify the person later. However, if they were told to answer based on the bigger letters (the first one is "C" the second one is "D"), they were more likely to correctly identify the face later. In other words, police departments could get a higher rate of matches in line ups just by showing victims a page or two of these letter forms!

Even cooler thing number two: a similar setup, except that instead of looking at letters, participants were simple asked to either pull, or to push a table. Pushing the table, apparently, is "avoidance" behavior (I may have this bit wrong, but the idea is still there) and triggers the right hemisphere, resulting in more specific thinking and less accurate matches. Pulling the table is "approach" behavior, triggers the left hemisphere, and results in more accurate matching. How cool is that?!