Tag Archives: flynn effect

The Divergence Problem: The Flynn Effect of Tree Ring Data

After a whitleblower posted secret emails detailing Climate Gate, some have talked about the “divergence problem.” The divergence problem, statistically, is the problem that the latent factor of temperature that is indicated by tree ring data, among other indicators, does not appear to be invariant. That is, it is not indicated equally well by its indicators as time goes on.

An almost identical problem in pschological measurement is the Flynn Effect, which plays havoc with our understanding of Raven’s Progressive Matrices as an indicator of general intelligence.

Liberals and leftists argue that the Flynn Effect is evidence that there are not heritable and/or racial differnces in general intelligence. The same liberals and leftists simply restate the fact that there is a “divergence problem,” with respect to indocators of temperature, and act as if it is not a big deal.

The reason is simple: global warming is a useful lie. Whether one wants to fight Russia, or merely enact your own puritanical wonderful on everyone else, lowering CO2 emissions is a greaet way to go about it.

Of course, I do not want to say that global warming is on the same bar as racial theorists. I am aware of no organized conspiracy of racial theorists on par with the University of East Anglia to squash dissenting voices. Likewise, the University of East Anglia’s destruction of its original data, combined with its poorly written and non-peer-reviewed computer models, is far more reckless and slipshod than, say, anything Hernstein or Murray ever put out.

The Hierachy of Intelligence(s)( Tests)

In the context of an attack on Howard Gardner’s theory of multiple intelligences (10 page PDF), David F. Lohman (of the University of Iowa) presents this visualization of the hierarchies of intelligence tests:


Hierarchy of IQ Tests

The closer to the center, the more general lintelligence (“g”) loads on the test. Some tests, such as identifying the correct endings of words, reading speed, or listening comprehension test “g” more indirectly than measures of verbal achievement, paper folding, or necessary arithemetic operations.

A good example is the Test Necessary Arithmetic Operations. This test was devised by Guilford to measure a specific cell in his Structure of the Intellect Model. Each item presents a short word problem. The examinee’s task is not to solve the problem, but to say which two operations she would use, and in what order. There are four operations: add, subtract, multiply, and divide. Thus, problems do not require advanced mathematics. Yet in the sample of over 100 Stanford undergraduates who were administered most of the tests in Figure 2, Necessary Arithmetic Operations had one of the highest loadings on the [fluid intelligence] factor (Marshalek, Lohman, & Snow, 1983).

I meant this because of Mark’s discussion of Dr. James Flynn on the Flynn Effect. Essentially, the Flynn effect explains the large-scale increase in measured general intelligence over the 20th century as reflecting increased society-wide patterns of practice on subtests. This implies two things: first, that tests should be renormalized every so often to make sure they still measure “g,” and not practice. Second, that ability improves with practice.

It strikes me that, when properly normalized, IQ measures something psychobioneurological… perhaps not working memory exactly, but something not too far apart from that concept.

This has implications for the heritability of IQ. Most obviously, the more environment changes, the more change in performance can be traced to the environment. (Of course, as environments become more similar, more of the variation in the population will be explained by environments.)