Gedanken Experiment - Alphabetization

All math and no psychology make Don a dull boy. Not that I'm a psychologist or anything silly like that, but I've taken enough grad courses in the Cognitive Psych department to have developed an appreciation for my epistemologically talented colleagues. Anyway, it's nice to have a common ground with the Humanities kids during our lofty discourses at Ivy League coffee shops.

Along with the torrent of Dynamical Systems, Hidden Markov Models, and Mean Shift Algorithms with which I've been grappling, the not-so-small matter of earning my paycheck has also been demanding my attention. I am a TA, and there are homeworks to grade!

Where there is grading, there is alphabetization. Where there is alphabetization, there is boredom. Soul-numbing boredom. Somewhat inevitably, this led me to our latest installation of Gedanken Experiments.

We all know how alphabetization works. There is a surprisingly long Wikipedia article about it, because there are nuances add complexity if you care to deal with them. However, in essence, alphabetization is a simple sorting process that follows very simple rules.

I wonder, though, if we approach this sorting in a less straightforward manner. Do we subconsciously preprocess the words? Do we put letters into fractions of the alphabet? Most of all, is there any way to tease this information out of some experiment?

I contend that yes, although this is purely conjecture. In fact, a quick lit review reveals that no work has ever been done on this in the entire history of psychology, which is surely impossible. Or maybe I'm just that innovative. Feel free to scoop me on this one, kids.

Let's assume that we tend to divide the alphabet into groups. Let's say that those groups are A-I, J-P, S-Z. If, in the middle of a sorting task, we come across the name Franklin, then we will first put that name into the 1st group, after which we find its proper location within that group. The end result is the proper ordering of all of the items, but this subconscious strategy might speed the process up, maybe.

One experiment that could test this hypothesis is the sorting of some number of words - 52, for argument's sake. We can choose 26 of those words so that each of the 26 letters of the alphabet is represented, producing a uniform distribution across the alphabet. We can then choose the remaining 26 words to be uniformly distributed (U), or we can create a condition in which they are clustered (C). Subjects would be timed as they sort.

If it turns out that our assumption is correct, that the process of alphabetization involves a categorical preprocessing, then the U condition should produce similar times over repeated trials (within, not across, subjects), while the C conditions should be slower - or faster! - depending on the location of the clusters.

Let me explain that last statement. If the clusters are located in the center of our natural groupings - for example, if the clusters are around F and U and the groups are A-I, J-P, and S-Z - then this should be slower than the uniform distribution. The preprocessing won't help, because the clusters are far away from the borders between groups.

On the other hand, if the clusters are located at the borders - for example, if the clusters are around I and P - then the preprocessing will have an effect on the grouping, and alphabetization would be facilitated. This should be much faster.

The experiment would have to be run dozens of times, varying the number and location of the clusters. It would take forever, and it would piss off your subjects. Worst of all, I hardly think this could be a smoking gun, even if it produces results as expected... I can already think of a handful of objections to my claims as proposed above.

If you do end up running this experiment, please let me know what you find. You can just leave a comment here, or you can contact me at:

YouWastedYourTime [at] theninjadon [dot] awesome