Hostname: page-component-745bb68f8f-s22k5 Total loading time: 0 Render date: 2025-02-04T17:32:49.400Z Has data issue: false hasContentIssue false

General intelligence is an emerging property, not an evolutionary puzzle

Published online by Cambridge University Press:  15 August 2017

Franck Ramus*
Affiliation:
CNRS, Ecole Normale Supérieure, EHESS, PSL Research University, 75005 Paris, Francefranck.ramus@ens.frhttp://www.lscp.net/persons/ramus/en/

Abstract

Burkart et al. contend that general intelligence poses a major evolutionary puzzle. This assertion presupposes a reification of general intelligence – that is, assuming that it is one “thing” that must have been selected as such. However, viewing general intelligence as an emerging property of multiple cognitive abilities (each with their own selective advantage) requires no additional evolutionary explanation.

Type
Open Peer Commentary
Copyright
Copyright © Cambridge University Press 2017 

As the authors acknowledge, the concept of general intelligence is empirically grounded solely in the observation of positive correlations between all test scores, as reflected by a general factor termed g explaining a large share of variance in all tests (Spearman Reference Spearman1904). All other accounts are simply debatable interpretations or hypotheses attempting to relate g to some other cognitive or biological constructs. They run the risk of reifying what is primarily a statistical construct, and also of seriously confusing the search for an evolutionary explanation. For instance, Gottfredson's (Reference Gottfredson1997) definition of intelligence is little more than a scholarly formulation of the folk concept of intelligence, but offers no guarantee of matching psychometric g. Burkart et al. initially conflate g with executive functions, but this changes the nature of the problem. If general intelligence reduced to executive functions, then to the extent that each executive function offers a selective advantage, the evolution of general intelligence would not be a major puzzle. Similarly, general intelligence is also identified with domain-general cognitive processes, which is a different, and unnecessary, hypothesis as we will show. Furthermore, many putative domain-general cognitive functions turn out to be less general than they seem. For instance, there are separate working memory systems for verbal, visuospatial, and other modalities. Similarly, words such as inhibition and attention wrongly suggest unitary phenomena, whereas they are used to describe a host of distinct processes, none of which can be said to be truly domain-general, and none of which is an evolutionary puzzle. Finally, certain cognitive functions can serve domain-general purposes while having been selected for more specific adaptive value. This may be the case of language, which serves as a mediator across many cognitive functions, yet may have evolved for purely communicative purposes (Jackendoff Reference Jackendoff1999; Pinker & Bloom Reference Pinker and Bloom1990).

More generally, every attempt to reduce general intelligence to a single cognitive (processing speed, working memory, etc.) or biological (brain volume, nerve conduction velocity, etc.) construct has failed, each construct showing moderate correlation with g and being best described as simply one contributor to the g factor (e.g., Mackintosh Reference Mackintosh2011). Thus, trying to tackle the evolution of general intelligence by addressing the evolution of any of these constructs is a form of attribute substitution (Kahneman & Frederick Reference Kahneman, Frederick, Gilovich, Griffin and Kahneman2002).

Understanding the evolution of psychometric g requires understanding how it comes about. As early as 1916, Thomson (Reference Thomson1916) showed that it is sufficient to postulate underlying group factors that influence several tests to obtain a positive manifold without a general factor (see also Bartholomew et al. Reference Bartholomew, Deary and Lawn2009). Reframed in modern psychological terms, an elementary analysis of tests shows that no test is a pure measure of a cognitive function (or construct). The relationship between cognitive functions and test scores is many-to-many: Each test score is influenced by several cognitive functions, and each cognitive function influences several test scores (in the same direction). The latter observation suffices to explain that test scores are positively correlated. We submit that the logic of Thomson's bonds model is much more general, as it also applies to factors underlying cognitive functions. Indeed, each brain function or property (e.g., frontal gray matter volume, nerve conductance velocity, dopamine synthesis, etc.) influences several cognitive functions, thereby inducing intrinsic positive correlations between cognitive functions. One step further back, each gene expressed in the brain (e.g., genes that code for neurotrophic factors, transcription factors, and any molecule involved in neurotransmission) typically influences several brain functions and properties, thereby inducing positive correlations between them. In parallel, many environmental factors (e.g., nutrition, socioeconomic status, education, diseases, and so on) influence more than one brain or cognitive function, thereby inducing further correlations. Finally, van der Maas et al. (Reference van der Maas, Dolan, Grasman, Wicherts, Huizenga and Raijmakers2006) have shown that positive correlations between cognitive functions may emerge through mutual interactions in the course of cognitive development, even in the absence of intrinsic correlations. Thus, all of the factors underlying test performance are pleiotropic and conspire to produce positive correlations at all levels of description, hence the emergence of the positive manifold.

Note that, according to the explanation given previously, the positive manifold can arise in an entirely modular mind (because modules selected for different purposes nevertheless have to share underlying factors), and therefore there is no antagonism between modularity and general intelligence. Furthermore, the very same pleiotropic mechanisms are at work in other species and, therefore, readily explain that a g factor can be measured in nonhuman primates, rodents, and probably all organisms with a nervous system. Finally, in the speciation process, genes that progressively diverge between two populations influence more than one brain and cognitive function; therefore, the two populations are bound to eventually differ in more than one brain and cognitive function. This directly predicts that performance in different tests should covary across species, or what the authors term G. Thus, all of the evidence that the authors gather in support of a reified notion of general intelligence is more parsimoniously explained by the pleiotropy of the underlying factors, within and across species. The “independent evolution of large numbers of modules instead of general intelligence” is not “particularly difficult to reconcile with interspecific findings of G” (sect. 2.5, para. 5); it directly follows from an understanding of what modules are made of: the same building blocks, shared between species.

There is, therefore, no need to postulate that the positive manifold reflects one particular cognitive function or one brain function, whose evolution would require a special explanation. The positive manifold emerges spontaneously from the pleiotropy of all of the underlying factors. Only these underlying factors require an evolutionary explanation. It is indeed very interesting to inquire about the evolution of genes involved in brain development and function, the evolution of brain functions and properties, and the evolution of cognitive functions. If there is any brain or cognitive function whose evolution is a major puzzle, then it should be identified and studied as such. However, this is not the case for general intelligence, which does not reduce to a single brain or cognitive function, and whose evolution follows directly from that of the underlying biological, cognitive, and environmental factors.

References

Bartholomew, D. J., Deary, I. J. & Lawn, M. (2009) A new lease of life for Thomson's bonds model of intelligence. Psychological Review 116:567–79. Available at: https://doi.org/10.1037/a0016262.Google Scholar
Gottfredson, L. S. (1997) Mainstream science on intelligence: An editorial with 52 signatories, history, and bibliography. Intelligence 24:1323.Google Scholar
Jackendoff, R. (1999) Possible stages in the evolution of the language capacity. Trends in Cognitive Science 3:272–79.Google Scholar
Kahneman, D. & Frederick, S. (2002) Representativeness revisited: Attribute substitution in intuitive judgment. In: Heuristics and Biases: The Psychology of Intuitive Judgment, ed. Gilovich, T., Griffin, D. & Kahneman, D., pp. 4981. Cambridge University Press.Google Scholar
Mackintosh, N. J. (2011) IQ and human intelligence (2nd ed.). Oxford University Press.Google Scholar
Pinker, S. & Bloom, P. (1990) Natural language and natural selection. Behavioral and Brain Sciences 13:707–84.Google Scholar
Spearman, C. (1904) “General intelligence,” objectively determined and measured. The American Journal of Psychology 15:201–92.Google Scholar
Thomson, G. H. (1916) A hierarchy without a general factor. British Journal of Psychology 8:271–81.Google Scholar
van der Maas, H. L. J., Dolan, C. V, Grasman, R. P. P. P., Wicherts, J. M., Huizenga, H. M. & Raijmakers, M. E. J. (2006) A dynamical model of general intelligence: The positive manifold of intelligence by mutualism. Psychological Review 113(4):842–61. doi: 10.1037/0033-295X.113.4.842.CrossRefGoogle ScholarPubMed