Can intelligence be changed?

I often hear both researchers and laypersons confidently state that an individual’s intelligence cannot be changed. This has always been an opinion that surprises me. The firm conviction with which the opinion is expressed by some people is perhaps even more surprising. It is certainly an opinion that I do not think is supported by the scientific literature.

The modern scientific literature on intelligence dates back to the discovery that differences between people in performance on tests of cognitive abilities (e.g., memory, reasoning) are positively associated (Spearman, 1925). That is, people who perform good on a test of a particular cognitive ability (e.g., associative memory) also tend to perform good on tests of other abilities (e.g., deductive reasoning). This common aspect of differences between people in cognitive performance is typically referred to as general intelligence or general cognitive ability (g).

In the addition to the general differences between individuals, there are also broad cognitive abilities (e.g., memory) that have their own unique differences between people. These broad cognitive abilities can further be broken down into narrow cognitive abilities (e.g., associative memory, word fluency). In other words, there are exceptions to the rule that people tend to perform equally well across cognitive tests and these exceptions are systematically organized (Carroll, 1993). For example, most individuals do fairly average on the tests of most abilities (e.g., memory, processing speed, auditory processing), but may perform unexpectedly good (or, bad) on all of the tests of one broad ability (e.g., visual processing).

General crystallized intelligence (gc) and general fluid intelligence (gf) are perhaps the most fundamental broad cognitive abilities (Cattell, 1943; Hebb, 1942). Fluid intelligence refers to abilities needed to solve problems that cannot be solved by retrieving the solution from long-term memory. The ability includes for example various types of logical reasoning. Crystallized intelligence is made up from abilities that rely on the use knowledge (e.g., vocabulary, general and specialized knowledge).

There are thus obviously aspects of intelligence that can be changed: Crystallized intelligence is an aspect of intelligence that by definition can be improved through learning. The more knowledge you can acquire and use, the more crystallized intelligence you will possess. Crystallized intelligence is also not in any way a trivial aspect of intelligence. The advanced use of knowledge and the acquisition of complex skills are things that make us humans adapt so well to life on this planet. Thus, sweeping statements that intelligence cannot be changed are obviously wrong.

However, when people claim that intelligence cannot be altered, they probably think more on how nifty and clever people are rather than on how knowledgeable and skilled they are. That is, fluid intelligence is probably what most people have in mind when they think of intelligence. It turns out that there is strong empirical evidence that also fluid intelligence can be changed. Let us take a few basic examples of robust and replicable empirical findings that should be enough to substantiate this claim.

Performance of test of fluid intelligence increases rapidly during childhood and declines in aging (Horn, 1965; McArdle et al., 2002). There are systematic moment-to-moment and day-to-day fluctuations in fluid cognitive performance (Rabbitt et al., 2001; Schmiedek et al., 2013). Thus, fluid intelligence is not in any way a constant within people over time.

Fluid intelligence test scores have systematically and substantially increased in many parts of the world during the 20th century (Flynn, 1984). That is, at least until recently, each new generation of humans have been performing better than the previous generation on tests of fluid cognitive ability. Researcher disagree on how to interpret these changes, but it is quite clear that genetic factors cannot explain them, and that there are environmental factors that could (e.g., the substantial rise in length of schooling).

Schooling has substantial effects on fluid intelligence test performance (Ceci, 1991) relative to other common alternative activities (e.g., work). Historical policies directly targeting education (e.g., increases in length of compulsory schooling) have for example been used to reveal such effects. This is a clever methodological trick because several of these reforms were implemented little by little and quite randomly across geographic areas in some countries. This allows for studying the effects of these reforms on intelligence in an almost experimental way. A recent meta-analysis of studies employing this design showed sizable and trustworthy effects of schooling on both crystallized and fluid intelligence (Ritchie & Tucker-Drob, 2018). Other even stronger designs, such as randomized controlled studies of early childhood educational interventions (e.g., entry to preschool or “Head Start” centers), also show robust and quite considerable immediate effects on how well children do on tests of fluid cognitive ability (e.g., Bailey et al., 2017).

I know that researchers quibble over the mechanisms explaining these effects on intelligence. Some of these disagreements are related different views on how the cognitive abilities and the hierarchical model of intelligence should be interpreted. Researchers also disagree on what does and does not affect some or the other aspect of intelligence. But, the categorical proposition that human intelligence is immutable is not backed up by scientific evidence.