Two Kinds of Thinking
Raymond Cattell proposed in 1943 that general cognitive ability splits into two components that behave differently enough to be worth separating.
Fluid intelligence (Gf) is reasoning applied to something new: finding structure in unfamiliar material, without relying on anything you were taught.
Crystallised intelligence (Gc) is what you have accumulated: vocabulary, factual knowledge, learned procedures, professional judgement built from experience.
Why the split was proposed at all
Because the numbers demanded it. Measures of reasoning and measures of knowledge correlate — but not tightly enough to be the same thing, and they move in opposite directions across the lifespan.
A single-factor model cannot explain a sixty-year-old who reasons more slowly than they did at twenty-five while judging their field far better.
Which One Does Abstract Reasoning Measure?
Fluid, and about as cleanly as any practical instrument manages.
The design is explicit about it. Abstract items use shapes, arrows and arrangements that refer to nothing — precisely so that nothing you know can help you, and only the reasoning is left exposed.
What the test removes, and why
- Language. Removes vocabulary and reading advantage, and lets one form run across countries.
- Arithmetic. Removes taught procedure, which would otherwise measure schooling.
- Recognisable objects. Removes cultural familiarity with the depicted things.
- Prior exposure. The rules are inferred on the spot; there is no syllabus to have covered.
What remains is close to the definition of Gf: extract a rule from novel material, hold it, apply it.
How Clean Is the Separation?
Cleaner than for any other common format, and not perfect.
Two channels of learned advantage survive. Familiarity with the format is one — knowing that rotation, count, shading and position are the usual dimensions is knowledge, and it is worth marks.
Schooling is the other. An education spent reading diagrams, graphs and abstract notation trains exactly the perceptual habits these items reward.
The evidence that this leakage is real
The Flynn effect is the strongest single piece of it. Scores rose substantially across the twentieth century, and the gains were largest on the abstract instruments — the ones built to be least sensitive to environment.
A pure measure of a stable capacity should not have drifted that far within a few generations. Something learned is being measured alongside the reasoning.
Where Crystallised Intelligence Gets Measured Instead
Not on abstract tests — which is why a battery uses several.
- Vocabulary and verbal comprehension. The classic Gc subtests, and the most stable across a lifetime.
- General knowledge. Directly cumulative, directly a product of exposure.
- Verbal reasoning as employers use it. Comprehension and inference over real text, so reading skill is part of what is scored.
- Numerical reasoning. A hybrid — the reasoning is fluid, but percentages and ratios are learned procedure.
Does Fluid Intelligence Decline With Age?
On average, yes, and this is one of the better-documented findings in cognitive ageing.
Fluid measures typically peak in the twenties or early thirties and decline gradually afterwards, with the rate increasing in later decades. Crystallised measures hold steady or continue rising well into the sixties and beyond.
The qualifications that matter
- Cross-sectional studies overstate it. Comparing seventy-year-olds with twenty-year-olds also compares two different educational cohorts. Schaie's Seattle Longitudinal Study, which followed the same people over decades, found gentler declines than snapshot comparisons implied.
- Speed does much of the work. A large share of measured fluid decline tracks processing speed rather than reasoning capacity, and heavily timed tests exaggerate it.
- The variation between people is large. Group averages describe trajectories, not individuals, and individual paths differ enormously.
Why the decline is often invisible at work
Because crystallised ability compensates. An experienced professional recognises the situation instead of reasoning it out from scratch — pattern-matching against decades of cases, where a novice must derive.
Expertise is stored fluid reasoning, and it does not decline on the same schedule.
Is Fluid Intelligence Trainable?
This is the most contested question in the area, and the honest answer has three parts.
What definitely improves
Test scores. Familiarity with the item families, a fixed search order through the usual dimensions, and better time discipline all produce measurable gains — and none of them touch the underlying ability.
What was claimed
A 2008 study by Jaeggi and colleagues reported that training on a demanding working-memory task (dual n-back) transferred to improvements in fluid reasoning. It was a striking claim, because transfer is exactly what cognitive training usually fails to produce.
What replication found
Larger and better-controlled attempts have found the effect much weaker or absent, and meta-analyses have repeatedly pointed at design problems in the encouraging studies — particularly control groups that did nothing, leaving expectancy effects unaccounted for.
The defensible position today: expect real gains on the test from familiarity and strategy, and treat any claim of a large durable increase in fluid ability itself as unproven.
Which One Predicts Success?
Both, in different places, and the distinction is practically useful.
- Fluid ability predicts learning rate. It matters most when the material is new — early in a degree, early in a role, at any genuine change of field.
- Crystallised ability predicts current performance. It matters most where the job is applying accumulated judgement to familiar problems.
This is why graduate schemes test fluid reasoning while senior hiring interrogates track record. They are selecting for different things, and both are behaving sensibly.
Reading Your Own Abstract Score
What you have is an estimate of one component — reasoning on novel material, under whatever conditions you happened to take it in.
It says nothing about what you know, and the two are genuinely separable. Strong fluid reasoning with a thin knowledge base learns fast and delivers little yet; deep expertise with average fluid reasoning delivers a great deal within its domain and finds a change of field harder.
Most working lives run on the second. Testing regimes mostly measure the first, and knowing which is which is worth more than the score.