A typical IQ test will contain many subtests like Information (“What’s the capital of Turkey?”), Verbal analogies (“heavy is to football as tall is to?”), Comprehension (“Why do pizza restaurants put their name on the pizza box?”), Vocabulary (“What does the word rudimentary mean?”) and many more.
But how do we know these different subtests are measuring different functions. For example, the Information subtest could have the question, “How many people are in a couple?” and a vocabulary test could ask “What does the word ‘couple’ mean?”. Both these questions are asking essentially the same thing, yet depending on how they are worded, they could appear on different subtests.
Similarly, the question “Biden is to the United States as Trudeau is to?” might appear on a verbal analogy subtest as a measure of abstract reasoning, but if we worded “what country is Trudeau the leader of?” it would appear on the Information subtest as a measure of long-term memory, even though both versions of the question would correlate near perfectly.
Or the question “Why do we have ears?” could appear on either the Information subtest as a measure of general knowledge, or it could appear on the Comprehension subtest as a measure of common sense depending on what side of the bed Wechsler woke up on that morning.
So how do we know the different subtests in an IQ battery are actually measuring different cognitive functions and not just redundantly measuring the same thing in different ways?
Well, we would need to show that the different items in subtest A all correlated more with one another than they do with the items on any other subtest. A more sophisticated approach is factor analysis, a statistical technique which sees if a large number of variables can be explained by a smaller number of variables.
The original WAIS for example had 11 subtests, but factor analysis concluded that in addition to g (general intelligence) they all could be explained by just three factors: verbal, spatial and memory.
Of course by adding more subtests, you can increase the number of factors. For example Digit-Symbol (a measure of clerical speed) loaded on the memory factor in the original WAIS, but when they added more tests of rapid clerical work, a fourth factor dubbed processing speed emerged. The children’s Wechsler now measures five factors.
How many factors exist in the human mind? It’s a fascinating question because even though the number of subtests one can create is literally infinite, the number of factors is finite. In the biggest battery of tests I’ve ever heard of, 57 subtests were reduced to 19 factors, and these 19 factors were reduced to just four higher level factors.
In this way intelligence is kind of analogous to race. There are dozens of different ethnic groups (Italian, Polish, Brahman, Japanese, Sudanese etc) and these can be reduced to maybe one dozen clusters, which can perhaps be further reduced to just three races (Black, Caucasoid, and Mongoloid)?