Newsletter #001: Cognitive Ability
Independent intelligences as a degenerating research program; against IQ threshold effects; the IQ halo effect; and IQ fadeout
Theories of Independent Intelligences as a Lakatosian Research Program (2022) by Jonathan Egeland. The most common view among experts in the field of intelligence is that there is a single factor (the g factor) that emerges from the fact that all cognitively demanding tasks positively correlate. Other researchers such as Joy Paul Guilford, Howard Gardner, and Robert J. Sternberg propose alternative theories. These theories constitute auxiliary hypotheses defending the hard core that intelligence has a multidimensional structure. According to the philosopher of science Imre Lakatos, a research program is said to be progressive if it has theoretical progressiveness and empirical progressiveness, meaning it can make novel predictions and empirical evidence corroborates these predictions. Since the independent intelligences research program fails both, it is a degenerating research program. This paper can serve as an interesting introduction to the philosophy of Lakatos.
Excerpt from On the genetic architecture of intelligence and other quantitative traits (2014) by Steve Hsu. The Study of Mathematically Precocious Youth (SMPY) has been provided as evidence that cognitive ability does not have threshold effects as far as we can tell. See Gwern Branwen’s SMPY Bibliography. Even within the top 1%, being in the top quartile makes one much more likely to accomplish more. Hsu discusses this in the article above and provides this graphic.
I had seen this graphic previously. I also learned about the SMPY through Charles Murray’s book Human Diversity (2020). What I had not seen before that Hsu mentioned was the book The Making of a Scientist (1952) by Anne Roe which provided more evidence that cognitive ability does not have diminishing returns for eminence.
Harvard psychologist Anne Roe studied 64 randomly selected eminent scientists (ages roughly 40-50) in her 1952 book The Making of a Scientist. Among these scientists were physicists Luis Alvarez, Julian Schwinger, Wendell Furry, J.H. Van Vleck and Philip Morse, anthropologist Carleton Coon, psychologist B.F. Skinner, chemist Linus Pauling and geneticist Sewall Wright. Roe administered a high ceiling psychometric test to each scientist, obtaining median scores in both the mathematical and verbal categories in the +4 SD (better than 1 in 10k) range. Thus, randomly sampled eminent scientists were found to be far outliers even among research scientists.
The IQ Halo Effect by Gwern Branwen: Many people interested in IQ research have stumbled on Gwern Branwen’s blog. He has some fascinating content. The IQ Halo Effect is a bibliography of the relationship of intelligence with other desirable traits. An excellent summary chart is featured in a chapter of The Handbook of Intelligence (2015) by Tarmo Strenze. The table is correlations (r) between cognitive ability and measures of success that have been established through multiple studies (k) (Strenze, 2015, pp. 406).
“The environment in raising early intelligence: A meta-analysis of the fadeout effect” (2015) by John Protzko. This is a study that I have cited on occasion. It finds that we don’t know how to permanently and substantially increase a person’s cognitive ability. The gains that are had tend to fade out. I see this as a general property of the brain. It is tough to have permanent positive gains without consistency. People tend to revert to their natural emotional state, intelligence level, conscientiousness, etc.
Abstract Many theories about the role of the environment in raising IQ have been put forward. There has not been an equal effort, however, in experimentally testing these theories. In this paper, we test whether the role of the environment in raising IQ is bidirectional/reciprocal. We meta-analyze the evidence for the fadeout effect of IQ, determining whether interventions that raise IQ have sustained effects after they end. We analyze 7584 participants across 39 randomized controlled trials, using a mixed-effects analysis with growth curve modeling. We confirm that after an intervention raises intelligence the effects fade away. We further show this is because children in the experimental group lose their IQ advantage and not because those in the control groups catch up. These findings are inconsistent with a bidirectional/reciprocal model of interaction. We discuss explanations for the fadeout effect and posit a unidirectional–reactive model for the role of the environment in the development of intelligence.
Interesting study estimating intelligence from FMRI scans
https://www.biorxiv.org/content/10.1101/412056v1
First, this post convinced me to become an annual subscriber. Bravo!
Second, on the Protzko paper, which I skimmed during my lunch, I wonder if I'm interpreting it correctly as follows? It's not that environmental impacts are small, they in fact can be significant, but that they don't persist once the intervention ends. Thus, for example, Head Start programs have real, measurable benefits for children but those events fade once the child exits the program. If it were possible to keep the child in Head Start programs throughout their education, those effects would be significant the entire time? Therefore, the focus on environmental impacts, since your and I can't change our genetics, should be on persistent and maintainable environmental changes.
You might compare this to a diet. As long as you're on the diet, you will remain a healthy weight. As soon as you quit the diet, you will rapidly regain weight. Therefore, the most effective diets are not those that let you lose weight most quickly but those that are most maintainable, with the ideal diet becoming infinitely maintainable and basically a lifestyle change. Or, to use classical music as a pseudo-example, it works, but it has no permanent effect, therefore you should play classical music all the time.
Also, on the Egeland paper, I really like the layout of Lakatos' system, especially the empirical parts which focus on generating real predictions. But when I started to read into the paper, it felt like sniping at other articles and scholars. The whole idea of research being derived from a central thesis or only able to be generated from a single "hard core" felt very arbitrary. However, I'm far from a specialist on these matters. I'd appreciate anyone more familiar with the field telling me whether the Lakatos system and examples brought were really useful or just kind of cover to attack other theories.