Behavioral Science Dictionary

Fluid intelligence

Cognition & Dual-Process

On-the-spot reasoning power, independent of acquired knowledge.

What it means

Fluid intelligence is the capacity to reason, identify patterns, and solve novel problems with relatively little reliance on acquired knowledge. It is indexed by abstract matrix-reasoning tests, correlates with working-memory capacity, though how strongly is disputed, and is associated with a frontoparietal brain network. Raymond Cattell proposed the contrast with crystallized ability in the 1940s, and it survives as one broad factor in the Cattell-Horn-Carroll model, though its separation from general ability is contested. Measured reasoning performance declines from early adulthood while vocabulary keeps rising into the sixties, though the two factors change together in ageing. Reasoning tests recur in selection because they presuppose no particular curriculum, but they are not schooling-proof, and claims that training raises general reasoning capacity are unsupported by the best-powered evidence. The practical lesson is less about ranking people than about the cost of novelty: work that invalidates established routines shifts effort from cheap retrieval onto reasoning, and documented procedures convert a fluid-reasoning problem into a crystallized one.

How it works

Novel problems share a shape. The solver must work out on the spot which relations matter, hold several partial results in mind, and abandon the lines that lead nowhere. Matrix puzzles make this explicit: one rule governs change across a row, another the columns, and the answer requires integrating both. The machinery most often implicated is the ability to hold a goal and its intermediate products active while interference accumulates. John Duncan and colleagues named the characteristic failure goal neglect in 1996: a requirement is disregarded even though it was understood and can still be repeated back. Neuroimaging converges on a network spanning lateral prefrontal and parietal cortex, formalised by Rex Jung and Richard Haier as the parieto-frontal integration theory. That synthesis is correlational, but the localisation does not rest on imaging alone: lesion mapping in 80 patients found that damage to particular frontal and midparietal regions predicted loss of fluid reasoning, while damage elsewhere did not. What has not held up is the neural-efficiency claim, that high scorers need less activation for the same performance: Aljoscha Neubauer and Andreas Fink found it moderated by task type, brain region and difficulty, and reversed on very complex tasks.

The original demonstration

Raymond Cattell proposed the fluid-crystallized split in the early 1940s and tested it two decades later in a paper pointedly titled a critical experiment. He gave culture-fair tests built from abstract figures and culturally embedded tests built from vocabulary and general information to 277 seventh- and eighth-graders, then factor-analysed the results at the second order. Two broad factors emerged where one might have sufficed. The design was reasonable for its era, but it rested on a single sample of adolescent schoolchildren, and rotation choices in second-order solutions of that period leave room for disagreement about what the factors mean. What secured the distinction was its recovery in later datasets: John Horn, with Cattell, extended the scheme into a wider set of broad abilities, and John Carroll's re-analysis of hundreds of published correlation matrices produced a three-level hierarchy with fluid reasoning as one broad factor beneath general ability. The merged Cattell-Horn-Carroll model is now the default map for test batteries.

What the evidence shows

Two questions dominate. The first is how close fluid reasoning is to working-memory capacity. Phillip Ackerman, Margaret Beier and Mary Boyle's 2005 meta-analysis of 86 samples put the average correlation between true-score estimates of working memory and general ability at .479, under a quarter of the variance shared, and argued the two are not interchangeable. Michael Kane, David Hambrick and Andrew Conway replied in the same issue with 14 latent-variable datasets from ten studies and more than 3,100 young adults, where the median correlation was .72. The gap is not measurement error, which Ackerman had already corrected out, but the modelling of each construct as the common factor of several diverse tasks, which additionally removes variance specific to any one task's format and content, plus different study-inclusion criteria. Both estimates are correct for what they estimate: the constructs are close as latent traits, much less so as any two tests one might administer. The second question is whether fluid reasoning can be trained. Susanne Jaeggi and colleagues reported in 2008 that adults who practised a demanding dual n-back task improved on matrix reasoning in proportion to how many sessions they completed. The study used a passive control group, around 34 trained participants across four training durations, and a ten-minute limit that makes a power test a speeded one. Monica Melby-Lervåg, Thomas Redick and Charles Hulme later pooled 145 experimental comparisons from 87 publications and found reliable gains on other working-memory tasks but no convincing transfer to nonverbal or verbal ability, decoding, comprehension or arithmetic once the comparison group was itself doing something. That file is not quite closed. Jacky Au and colleagues pooled 20 n-back training studies, the paradigm Jaeggi used, and reported a small but statistically significant gain on fluid-intelligence measures. Melby-Lervåg and Hulme replied that the result turned on which studies were admitted, on effect sizes computed without adjusting for baseline group differences, and on pooling untreated controls alongside active ones. Their reading is the better one on current evidence, and against active controls the far-transfer effect is approximately zero. But the dispute is live rather than settled. Age trends are better established. Timothy Salthouse compared more than 5,000 adults cross-sectionally with nearly 1,600 followed across three occasions, and found accelerating declines in reasoning from early adulthood in both the cross-sectional and quasi-longitudinal data, while vocabulary kept rising into the sixties. Flatter longitudinal curves he attributes to practice effects from repeated testing.

Limits and caveats

The phrase culture-fair, inherited from Cattell's own tests, promises more than any instrument delivers. Stuart Ritchie and Elliot Tucker-Drob's meta-analysis of 142 effect sizes from 42 quasi-experimental datasets, covering more than 600,000 people, estimated that an extra year of education raises measured intelligence by one to five IQ points, on every broad category of ability examined, not only knowledge tests. Reasoning about abstract figures is less content-bound than a vocabulary test; it is not schooling-proof. A second caveat is psychometric. Fluid reasoning is a latent factor defined by what several diverse reasoning tests share; a single matrix test is a convenient marker of it, not a pure one, since any instrument carries variance specific to its format and the strategies it invites. A conclusion about one person from one score deserves far less confidence than one about a group from a battery. A sharper version of that problem cuts at whether the construct is separable at all. Jan-Eric Gustafsson's 1984 analysis reported that the broad fluid factor could not be distinguished from the general factor, and Carroll treated that difficulty as genuine: fluid reasoning, on his reading, is inherently hard to measure independently of its dependence on general ability. In his own hierarchical solution, though, the loading settled at .83, below those for crystallized and quantitative ability rather than above. What is established is not a verdict either way but that the separation is contested, which sits awkwardly beside the appeal of a fluid measure as something other than IQ. The opposition with crystallized ability is likewise tidier in exposition than in data: the two factors correlate substantially, and in cognitive ageing they change together.

Where it shows up

Reasoning tests are attractive in selection because they presuppose no particular curriculum, which is why they recur in graduate schemes, apprenticeship screening and military entry testing. The same property makes them easy to over-trust: a score under a strict time limit in an unfamiliar format measures something narrower than the label suggests. In organisational design the implication has little to do with ranking people. It is that novelty is expensive. A process change that invalidates existing routines shifts work from cheap retrieval onto effortful reasoning, and the cost lands hardest at the transition, not in the steady state that gets modelled. Documented procedures and worked examples convert a fluid-reasoning problem into a crystallized one. In consumer markets the training literature has a long afterlife. Products sold on the promise of raising general reasoning capacity make a claim the best-powered evidence does not support, though the narrower claim, that people improve at what they practise, is true. The distance between them is where the marketing sits.

Examples

Solving an unfamiliar logic puzzle with no relevant background relies on fluid intelligence, not knowledge you already possess.

A logistics operator replaces its dispatch system over a weekend. The staff with the deepest procedural knowledge struggle more in the first week than recent hires, because their accumulated routines no longer apply and the improvised workarounds load onto reasoning rather than recall.

A hiring team swaps a knowledge quiz for a matrix-reasoning test on the grounds that it will be fairer to candidates schooled in different systems, then finds that scores still track years of education.

A mid-size insurer licenses a brain-training package for its claims handlers and tracks the results. Scores on the trained memory task climb steadily over eight weeks while accuracy on actual claim assessments does not move.

After a head injury, a patient's vocabulary and general-knowledge subtest scores come back in the normal range while performance on unfamiliar reasoning puzzles has fallen sharply, so a composite IQ figure understates the change.

First described in Raymond Cattell; Cattell–Horn–Carroll theory (1960s onward).

Key references

  1. Cattell, R. B. (1963). Theory of fluid and crystallized intelligence: A critical experiment. Journal of Educational Psychology, 54(1), 1-22. doi.org/10.1037/h0046743
  2. Horn, J. L., & Cattell, R. B. (1966). Refinement and test of the theory of fluid and crystallized general intelligences. Journal of Educational Psychology, 57(5), 253-270. doi.org/10.1037/h0023816
  3. Gustafsson, J.-E. (1984). A unifying model for the structure of intellectual abilities. Intelligence, 8(3), 179-203. doi.org/10.1016/0160-2896(84)90008-4
  4. Carroll, J. B. (1993). Human cognitive abilities: A survey of factor-analytic studies. Cambridge University Press. doi.org/10.1017/CBO9780511571312
  5. Duncan, J., Emslie, H., Williams, P., Johnson, R., & Freer, C. (1996). Intelligence and the frontal lobe: The organization of goal-directed behavior. Cognitive Psychology, 30(3), 257-303. doi.org/10.1006/cogp.1996.0008
  6. Ackerman, P. L., Beier, M. E., & Boyle, M. O. (2005). Working memory and intelligence: The same or different constructs? Psychological Bulletin, 131(1), 30-60. doi.org/10.1037/0033-2909.131.1.30
  7. Kane, M. J., Hambrick, D. Z., & Conway, A. R. A. (2005). Working memory capacity and fluid intelligence are strongly related constructs: Comment on Ackerman, Beier, and Boyle (2005). Psychological Bulletin, 131(1), 66-71. doi.org/10.1037/0033-2909.131.1.66
  8. Jung, R. E., & Haier, R. J. (2007). The Parieto-Frontal Integration Theory (P-FIT) of intelligence: Converging neuroimaging evidence. Behavioral and Brain Sciences, 30(2), 135-154. doi.org/10.1017/S0140525X07001185
  9. Jaeggi, S. M., Buschkuehl, M., Jonides, J., & Perrig, W. J. (2008). Improving fluid intelligence with training on working memory. Proceedings of the National Academy of Sciences, 105(19), 6829-6833. doi.org/10.1073/pnas.0801268105
  10. Neubauer, A. C., & Fink, A. (2009). Intelligence and neural efficiency. Neuroscience & Biobehavioral Reviews, 33(7), 1004-1023. doi.org/10.1016/j.neubiorev.2009.04.001
  11. Woolgar, A., Parr, A., Cusack, R., Thompson, R., Nimmo-Smith, I., Torralva, T., Roca, M., Antoun, N., Manes, F., & Duncan, J. (2010). Fluid intelligence loss linked to restricted regions of damage within frontal and parietal cortex. Proceedings of the National Academy of Sciences, 107(33), 14899-14902. doi.org/10.1073/pnas.1007928107
  12. Au, J., Sheehan, E., Tsai, N., Duncan, G. J., Buschkuehl, M., & Jaeggi, S. M. (2015). Improving fluid intelligence with training on working memory: A meta-analysis. Psychonomic Bulletin & Review, 22(2), 366-377. doi.org/10.3758/s13423-014-0699-x
  13. Melby-Lervåg, M., & Hulme, C. (2016). There is no convincing evidence that working memory training is effective: A reply to Au et al. (2014) and Karbach and Verhaeghen (2014). Psychonomic Bulletin & Review, 23(1), 324-330. doi.org/10.3758/s13423-015-0862-z
  14. Melby-Lervåg, M., Redick, T. S., & Hulme, C. (2016). Working memory training does not improve performance on measures of intelligence or other measures of "far transfer": Evidence from a meta-analytic review. Perspectives on Psychological Science, 11(4), 512-534. doi.org/10.1177/1745691616635612
  15. Ritchie, S. J., & Tucker-Drob, E. M. (2018). How much does education improve intelligence? A meta-analysis. Psychological Science, 29(8), 1358-1369. doi.org/10.1177/0956797618774253
  16. Salthouse, T. A. (2019). Trajectories of normal cognitive aging. Psychology and Aging, 34(1), 17-24. doi.org/10.1037/pag0000288

Related terms

← All 1001 terms