Fluid vs crystallized intelligence: the difference, and what the age curves actually show

Fluid intelligence is the ability to reason and to solve new problems without help from what you already know. Crystallized intelligence is the knowledge and skill that you have learned, and your ability to use them. You use fluid intelligence when a road closure sends you onto streets that you do not know and you must find a way home. You use crystallized intelligence when you explain what a word means or answer a question from your course. On average, fluid abilities decline from early adulthood, while vocabulary grows into the 60s.

The psychologist Raymond Cattell published the two terms in 1943. This article gives the definitions from an open textbook and from the AP Psychology course. It then covers where the idea came from, what large studies show about age, and how the two types relate. It ends with a method to sort exam examples.

• Fluid intelligence (Gf) works on new problems. Crystallized intelligence (Gc) uses knowledge from experience and education. Typical tests are matrix puzzles for Gf and vocabulary for Gc.
• A 2016 history of the theory concludes that Cattell took the idea from the psychologist Donald Hebb, who called the two types "Intelligence A" and "Intelligence B".
• No single age is the peak. One large study found peaks from the late teens, for a speed task, to about 65, for vocabulary.
• A 2019 analysis found that studies that test the same people again made the decline look later, because people do better on tests that they have seen before.
• Working-memory training improves the trained task. A 2016 meta-analysis found no convincing transfer to measures of intelligence.

The difference in one table

The open textbook Psychology 2e by OpenStax defines crystallized intelligence as "acquired knowledge and the ability to retrieve it". It says that fluid intelligence "encompasses the ability to see complex relationships and solve problems". Its fluid example is the detour: "Navigating your way home after being detoured onto an unfamiliar route because of road construction would draw upon your fluid intelligence."

The College Board puts the pair in Topic 3.4 of its AP Psychology course and exam description, in the unit on development and learning. Essential knowledge 3.4.A.3 states: "Crystallized intelligence remains relatively stable through adulthood while fluid intelligence tends to wane as people age." That sentence is a summary. The section on age below shows what the data add to it.

Fluid intelligence: reasoning on new problems

Cattell's own definition, from his 1943 paper, is quoted in a 2016 history by Richard Brown. Fluid ability "has the character of a purely general ability to discriminate and perceive relations between any fundaments, new or old". In plain words, it is the ability to see how things relate: which pattern comes next, which rule links the items, which way leads home.

The key word is "general". Fluid intelligence is not knowledge of a subject. A 2008 study in PNAS by Susanne Jaeggi and colleagues gives the modern version of the definition: fluid intelligence "refers to the ability to reason and to solve new problems independently of previously acquired knowledge".

Tests of fluid intelligence try to remove the advantage of learned knowledge. Matrix-reasoning tests show a geometric design with a missing piece, and you choose the piece that completes it. Raven's Progressive Matrices is a well-known test of this type. Cattell also built non-verbal "culture-fair" tests for the same purpose. In 1963, he analysed the scores of 277 children in grades 7 and 8 on both culture-fair and traditional tests, and he found two general factors. He concluded that one factor "fits the crystallized ability factor measured in traditional intelligence tests and the other a fluid general ability measured in culture-fair intelligence tests".

Crystallized intelligence: knowledge you can use

Cattell defined crystallized ability as "discriminatory habits long established in a particular field, originally through the operation of fluid ability". Once you learn a skill or a fact, you do not need fresh insight to use it again. You only need to retrieve it. That is why OpenStax says that you use crystallized intelligence "all the time in your coursework" when you show that you have mastered the material.

Typical tests ask what you know. In the Wechsler Adult Intelligence Scale (WAIS), the subtests that call on learned knowledge include vocabulary, information, comprehension, arithmetic and similarities. Vocabulary and general facts are the kind of knowledge that memory researchers call semantic memory, the store of general knowledge.

No test measures only one type. Cattell wrote in 1943 that "intelligence tests test at all ages the combined resultants of fluid and crystallized ability". A vocabulary test needs some reasoning, and a matrix test needs you to understand the instructions.

Where the terms came from

Cattell is usually credited with the theory. Brown's 2016 paper in Frontiers in Human Neuroscience tells a more complicated story, based on letters in the McGill University Archives.

From 1937 to 1939, Donald Hebb tested patients after brain surgery at the Montreal Neurological Institute. At the 1941 meeting of the American Psychological Association, he reported that "certain kinds of abilities are less affected by late than by early injury". In the same talk, Hebb proposed that intellectual development includes "(A) direct intellectual power, by neural maturation, and (B) the development of qualitative modifications of perception and behavior". He added that the first factor reaches "a peak somewhere around the beginning of adolescence, declining slowly thereafter".

Cattell spoke at the same meeting. Letters followed, and Hebb edited some sections of Cattell's 1943 paper. In that paper, Cattell wrote that "Hebb has independently stated very clearly what constitutes two thirds of the present theory". Brown concludes that "Cattell's" theory "is Hebb's theory of Intelligence A and Intelligence B, given another name and popularized by Cattell".

Cattell disagreed. In a 1984 autobiographical paper, as Brown quotes it, he wrote that the two men "had independently reached the same position, Hebb through physiology, I through factor analysis". So the question of credit is disputed.

The theory then waited. According to Cattell's student John Horn, Cattell "left it on the shelf" until the 1963 study with the 277 schoolchildren. Horn continued the work with Cattell, for example in their 1967 paper "Age differences in fluid and crystallized intelligence". Today, the two abilities sit inside a larger model. OpenStax calls the Cattell-Horn-Carroll (CHC) theory "the most comprehensive theory of intelligence to date". In CHC theory, general intelligence is at the top and narrow abilities are at the bottom. In the middle are broad abilities "such as fluid reasoning, short-term memory, and processing speed".

How each changes with age

The textbook pattern

The age pattern was part of the theory from the start. Cattell wrote in 1943 that fluid ability "increases until adolescence and then slowly declines". He also wrote that "in adult life, owing to the recession of fluid ability, the peaks of performance are determined by the crystallized abilities". The AP course gives the modern short form: crystallized intelligence stays relatively stable, and fluid intelligence tends to wane.

No single peak: Hartshorne and Germine (2015)

Joshua Hartshorne and Laura Germine tested the pattern with much finer age steps. In their 2015 study in Psychological Science, they combined data from 48,537 people who took tests online with the norm data of two Wechsler test batteries.

The broad pattern held. In the norm data, the five tasks that use learned knowledge peaked later than nearly every other task. But single tasks peaked at very different ages:

• Digit-symbol coding, a speed task, peaked in the late teens in the norm data.
• Digit span peaked in the early 20s in the norm data.
• In the online data, two working-memory tasks peaked at around 30.
• A widely used test of emotion perception showed a broad plateau between about 40 and 60.
• Vocabulary peaked at about 50 in the older norm data, and at about 65 in the online data.

The authors wrote that age-related change is "considerably more heterogeneous and complex than the fluid/crystalized intelligence distinction suggests". They concluded: "not only is there no age at which humans are performing at peak at all cognitive tasks, there may not be an age at which humans are at peak on most cognitive tasks."

The vocabulary peak also moved over time. The authors reanalysed vocabulary scores from 26,850 people tested between 1974 and 2012 in the General Social Survey. People tested in later years reached their peak at older ages. The authors concluded that, for vocabulary, the later peaks in newer samples are at least partly due to differences between generations. So the peak age for vocabulary is not fixed. It can change from one generation to the next.

Why retesting the same people made decline look later

There are two ways to study age. A cross-sectional study tests people of different ages once. A longitudinal study tests the same people again after some years. The two methods often disagree, and the timeline of decline depends on which one you trust.

Timothy Salthouse compared both methods in a 2019 paper in Psychology and Aging. He used cross-sectional data from more than 5,000 adults and longitudinal data from almost 1,600 adults tested three times, on average 5.9 years apart. Reasoning tests included Raven's Progressive Matrices. Vocabulary had four tests.

Each method has a weak point. In a cross-sectional study, older and younger people come from different generations, with different schooling. In a longitudinal study, people have seen the tests before, and practice raises their scores. To separate the two, Salthouse used a third design. He compared people from the same birth years who took the tests for the first time at different ages.

The results of this third design were similar to the cross-sectional results, not to the longitudinal ones. In the longitudinal data, memory and reasoning scores went up until the 60s. In the other two designs, they went down at all ages. Salthouse concluded that, at least until about age 65, longitudinal results "are distorted by prior test experience".

His summary of normal ageing has three parts:

• Speed declines in a nearly straight line from early adulthood.
• Memory and reasoning decline, and the decline becomes faster after about 65.
• Vocabulary increases until the 60s, in all three types of comparison.

Vocabulary scores were even "higher among adults in their 70's and 80's than among those in their 20's, 30's, and 40's". These are averages across many people. They do not predict the path of one person.

How the two are related

Cattell's definition already contains a claim about how the two types relate: crystallized ability forms "originally through the operation of fluid ability". Later researchers call this view investment theory. In it, fluid ability drives the growth of crystallized knowledge.

Recent studies give this view only partial support. Rogier Kievit and colleagues studied 785 people aged 14 to 25, most of them at two time points, in a 2017 study in Psychological Science. They compared investment theory with two other models. A mutualism model fitted the data best. In that model, abilities support each other: people with higher vocabulary scores later gained more in matrix reasoning, "and vice versa".

A smaller 2024 study by Vsevolod Scherrer and colleagues tested 341 German students in grades 7 to 9, one year apart. In their report in the Journal of Intelligence, earlier fluid intelligence did not predict the growth of crystallized intelligence, "contrary to theoretical assumptions". The authors note that one year is a short interval. These two studies do not support a simple one-way flow from fluid to crystallized intelligence.

Can you train fluid intelligence?

Crystallized intelligence grows when you learn, because it is learned knowledge. Fluid intelligence is the harder case.

In 2008, Jaeggi and colleagues reported that training on a demanding working-memory task transferred to fluid-intelligence tests. They also reported that more training gave a larger gain.

Later studies did not confirm it. In 2013, Thomas Redick and colleagues gave young adults 20 sessions of training on an adaptive dual n-back task, which is a working-memory task. A second group practised a visual search task as an active placebo, and a third group had no contact. Both trained groups improved on their own task. But there was "no positive transfer to any of the cognitive ability tests", which included tests of fluid and crystallized intelligence.

A 2016 meta-analysis by Monica Melby-Lervåg, Redick and Charles Hulme pooled 87 publications with 145 experimental comparisons. Working-memory training gave short-term gains on working-memory tests. For nonverbal ability, verbal ability, reading and arithmetic, the authors found "no convincing evidence of any reliable improvements" when the training group was compared with a control group that did a different activity.

Classifying exam scenarios

Two questions sort most examples:

1. 1.Does the task need knowledge or a skill that the person learned before?
2. 2.If not, is the problem new, so that the person must reason it out?

If the answer to the first question is yes, the example is crystallized intelligence. If the task is new and needs reasoning, the example is fluid intelligence. Speed is not the cue. A fast answer from memory is still crystallized. In CHC theory, processing speed is a separate broad ability.

The fourth row comes from sample question 15 in the AP course description. It describes 90-year-old Mrs. Glass, who chooses activities that she can still do well, which shows that her crystallized intelligence works. The answer key gives the "name that song" game. The wrong options include a trivia game about current movies and a new video game. Both need recent knowledge or new learning, not old knowledge.

How to keep the two terms apart

The names give a memory aid. A fluid has no fixed shape, and fluid intelligence works on problems of any shape, new or old. A crystal grows slowly and keeps its form, and crystallized intelligence is knowledge that builds up over years and stays. If you can solve it without having learned anything special, it is fluid. If you need what you learned, it is crystallized.

The vocabulary of this topic is a small set of related terms: fluid intelligence, crystallized intelligence, Gf, Gc, investment theory, mutualism, CHC theory, cross-sectional study and longitudinal study. In Psychology World, you look up a term and save it to one of your own lists. Then you drill the list as flashcards, by typing the answer, or as multiple-choice questions once the list has four or more terms. The four steps of the learning loop and what each study mode does are on the home page. For study methods with strong evidence behind them, read our guide to memorising psychology terminology.