About 100 years ago, beginning with the testing of army recruits in 1917 in America, psychologists began to collect data that could be used to test theories of intelligence. They used the Stanford-Binet of 1916 and other tests that were the forerunners of Wechsler's IQ tests. These tests are an implicit response to a heuristic that has given advice to theory builders from that time to the present.

The heuristic of intelligence

I will summarize the content of that heuristic (for a full statement, see Flynn, 2009, pp. 53–4). It emphasizes the following traits.

1. Mental acuity: the ability to solve on-the-spot problems we have never encountered before. The problems deemed important vary, of course, from society to society.

2. Habits of mind: for example, the extent to which people are accustomed to use logic to analyze problems, and to deal with symbols increasingly detached from the concrete world.

3. Vocabulary, knowledge (including mathematics in our culture), and information: the more you have, the wider the range of problems you can attack.

4. Speed of information processing: the quicker one can assimilate data the better, particularly if problems must be solved within a time limit.

5. Rote memory and working memory: the larger the amount of relevant information (and conceptual rules) you can keep in mind, the better equipped you are to arrive at a solution.

Note that the whole content of this heuristic is cross-cultural in the sense that other societies prioritize problems unlike ourselves: Australian aborigines put mapping skills (to find water) well above the skill of using logic on abstractions so useful in formal education.

Wechsler thought within the context of modern Western society. He did not construct a theory of intelligence. He made fun of the Stanford-Binet for the cutting lines it used to separate levels of mental ability: gifted, high normal, low normal, mentally retarded. He noted that the scores all ended exactly in zero and said that this was astronomically improbable. All of Wechsler's new cutting lines ended in exact multiples of standard deviations, which was no less improbable.