Expected results (as predicted by original researcher): 4-5 letters are easily recalled. More than that is difficult, but subjects will be aware that there are more letters even if unable to recall all the letters. Using familiar disyllabic words increases retention significantly. Most subjects can recall two disyllabic words. Bear in mind that time delays on the Internet prevent the current test from precisely duplicating the historical test. However, preliminary results from my test subjects show that the historical results are repeated and that this test seems accurate, at least with high-speed connections to the Internet.

The history of empirical psychology repeats itself. At the time of its introduction, tests such as this raised some issues that are essentially identical to our contemporary issues in cognitive science. For example:

1. What is an object of perception? When does perception become memory? Are memory images stored as visual data? Consider the evidence: 12 letters in words are perceived as easily as 4 random letters, but 6 random letters remains difficult. Does this mean there is stored, visually-formatted data, that is retrieved by a matching program? If so, this is evidence of STORED IMAGES. Many contemporary theorists seem to think there are stored images, and have introduced the concept of matching as an explanatory construct. (The idea goes back to Aristotle's "wax impression," and is repeated by Locke.) But is this simple experiment truly evidence in favor of a matching algorithm?

We must be careful to distinguish between visual perception and the interpretation of memory data. That is why experiements like this are important. (Students of cognitive science note: compare to the Shepard and Mezler experiment. See my critique of Shepard and Metzler.) This experiment deals with the borderline between those two cognitive modes. A common mistake is to extend explanatory models from one mode to the other.

For certain visual (not memory) data, this experiment may well indicate a matching program. However, if we are dealing with memory, the issue becomes clouded. In my dissertation, I argue against a matching program as a universalized, general, explanation of the recall of visual memory images. Be it noted, however, that this was a sore point with my readers, who did not accept my argument. For an extended discussion, see my Advanced Information

2. How are units of perception formed?

3. How precise are empirical techniques? What, precisely, do empirical techniques measure? To what extent are subjective variables, such as degree of attention and desire to do well important? The original investigator and designer of this experiment was motivated by these questions, and found (what was then) a surprising result.

4. What do empirical techniques measure? Do they measure, as it were, pure elements or algorithms present in the human cognitive system? As Pylyshyn has (correctly) argued, that is the goal of empirical experimental techniques -- but this goal is extremely hard to achieve. For example, Pylyshyn has argued the famous Shepard and Metzler experiment (also discussed at length in these pages) is NOT one that demonstrates the structure of elemental operations in the human cognitive system. This historically-based perceptual units test, although it seems simpler and more decisive, might also be subjected to the same sorts of critical appraisal.

5. Are there historical/cultural variables in the test? Answer: yes. Can you tell why?

Return to Home Page

Return to Introduction to Mental Images Page