Does IQ Decline With Age? A Decade-by-Decade Cognitive Trajectory

IQ Career Lab is a cognitive assessment platform that measures intelligence across five domains and matches your cognitive profile to high-fit career paths.

Does IQ decline with age? The honest answer in two sentences

Some abilities decline starting in your 20s. Other abilities keep rising into your 50s and 60s. The question that matters is not "does IQ decline" but "which subskills decline, which hold, and which still rise on my decade?"

The popular framing collapses a 12-component cognitive profile into a single curve and then anchors that curve at age 25. That framing is wrong in a measurable way. Below is the decade-by-decade picture, drawn from five primary sources covering more than 170,000 participants combined [Hartshorne 2015] [Salthouse 2019] [Schaie 1994] [Singh-Manoux 2012] [Seblova 2020].

If you arrived at this article worried about decline after 35 or 45, the most useful thing to know first is that the worry is half earned and half the wrong frame. The earned half: speed-loaded fluid reasoning does decline, and the slope is real. The wrong half: that decline is one variable in a profile of about twelve, and the variables that pay the most in midlife knowledge work are not the ones that fall first.

The chart below combines three methodologies plus a labor-market overlay, and we have not seen another published synthesis combining all three plus a labor-market overlay at this granularity. Cross-sectional data covering N=48,537 [Hartshorne 2015] maps where each subskill peaks across a population at one point in time. Longitudinal VCAP on N=1,598 [Salthouse 2019] tracks the same people. Cohort-sequential data from the Seattle Longitudinal Study, N≈6,000 [Schaie 1994], blends both methods and is generally considered the gold standard despite being expensive and rare. Each method captures something the others miss. Each has a known bias.

The labor-market overlay is BLS Current Population Survey 2024 data on over 11 million knowledge workers (11.15M across the 11 occupation codes summed). We added it because the question most readers actually want answered is not abstract: "Am I past my peak for my actual job?" The honest answer depends entirely on which job.

The unified asynchronous-peak chart: no single peak age

The single most important fact about cognitive aging is that different abilities peak at different times. As one large cross-sectional analysis put it (Hartshorne & Germine, 2015): "Considerable heterogeneity in when cognitive abilities peak: some abilities peak around high school graduation; some plateau in early adulthood, beginning to decline in subjects' 30s; and still others do not peak until subjects reach their 40s or later."

That is the headline. The chart below plots peak age (midpoint of the published range) for each subskill; the highlighted bars on the right are the abilities that keep rising into the 40s, 50s, and 60s, the half of the picture the popular story leaves out.

The detailed table beneath the chart adds source, sample size, function shape, and methodology for each row.

The first six rows are the rows that scare people. The last six get left out of the popular story almost every time, and that omission is doing real damage to how midlife professionals interpret their own cognitive trajectory. Both columns are real. Each row pulls from a primary source with a sample size north of 1,500 [Hartshorne 2015] [Salthouse 2019] [Schaie 1994].

When does fluid intelligence peak?

Fluid intelligence peaks somewhere between 25 and 29 in cross-sectional data and shows a longer plateau into the first half of the 30s in longitudinal data. The headline number is 25-29; the reality is a long, gentle slope rather than a cliff. Decline through the 40s is small, with longitudinal slopes still positive on un-speeded measures.

Matrix-reasoning curves hit a maximum around 25-29 and begin a slow decline through the 30s (Hartshorne & Germine, 2015). Longitudinal reasoning slope stays positive through the 40s at +0.019 z/year (Salthouse, 2019, p<.01), turning flat in the 50s and negative only in the 60s and beyond, where slope drops to -0.041 z/year by the 80s.

If you would like to see which of your individual subskills are still rising versus declining, the IQ Career Lab assessment reports them as separate scores rather than collapsing them into a single "IQ."

What's the difference between fluid and crystallized intelligence?

Fluid intelligence (Gf) is the ability to solve novel problems without prior knowledge: pattern matrices, spatial rotations, working-memory tasks. Crystallized intelligence (Gc) is the accumulated knowledge and verbal judgment built up across a lifetime: vocabulary, verbal meaning, professional pattern libraries. The two have opposite age trajectories.

Gf is what makes a 25-year-old a fast learner. Gc is what makes a 55-year-old executive's first instinct correct more often than not. Gf peaks in the late 20s then declines on a gentle slope, while Gc keeps rising into the 60s. For a fuller career-stage breakdown, see our deep dive on fluid versus crystallized intelligence as a career strategy.

Why knowledge work peaks in the late 40s and 50s

This is the part that the popular narrative gets wrong. The fluid-crystallized split is not just visible in lab data; it shows up in the actual labor market. Below is the BLS Current Population Survey 2024 occupation-by-age table for 11 high-knowledge-work codes.

Read that number twice. The unweighted mean of the median ages across 11 knowledge-worker SOC codes in the United States [BLS 2024] is 45.6 years old [BLS 2024]. That is not an outlier; it is the central tendency. Median worker age in an occupation reflects market sorting, selection, and tenure dynamics — it is not a direct measurement of cognitive peak alignment, but a market-equilibrium signal that the role rewards what people of that age tend to bring. The roles cluster predictably:

• Speed-heavy roles cluster younger: software developers (38.6), operations research analysts (40.2), financial analysts (40.2). These are the roles where speeded fluid reasoning matters most, and the median worker sits well before the 45-49 decline-onset window.
• Verbal-judgment roles cluster much later: lawyers (46.4), postsecondary teachers (49.2), chief executives (51.6), judges (54.2). These roles weight crystallized verbal judgment over speeded fluid reasoning, and the median worker sits squarely inside Schaie's verbal-meaning peak window of 53-60.

This is not coincidence. The labor market is pricing the cognitive trajectory in real time. Roles that pay for speed clear out their median worker by the late 30s. Roles that pay for accumulated judgment keep their median worker through the first half of the 50s and beyond.

What does this mean for an individual reader? It means the answer to "Am I past my cognitive peak?" depends entirely on what you are paid to do. A 47-year-old software engineer is past the median age for their occupation; a 47-year-old corporate lawyer is sitting on the median. Same age, opposite story. The thing that varies is not the brain. It is the cognitive profile that the role rewards.

A 2025 Intelligence paper puts the upper bound on this synthesis (Gignac & Zajenkowski, 2025). Their Combined Personality and Functional Intelligence (CPFI) index pools nine cognitive and personality constructs into a single weighted measure of "overall functioning." Both their models put the peak at 55 to 60, older than the peak on any single cognitive measure considered alone.

This is also why we pair the IQ assessment with the Personality Test on this site: the late-midlife peak is half cognitive and half personality. The Gignac CPFI synthesis captures both.

A critical disclosure: Gignac & Zajenkowski's 55-60 figure is a weighted synthesis index combining nine cognitive and personality constructs. It is a [MODEL] of overall functioning, not a [MEASURED] peak in any single test. We list it on the chart as the directional ceiling on the same gradient that Hartshorne, Salthouse, and Schaie produce, not as evidence equivalent to those primary measurements. Gignac himself agrees that fluid IQ peaks around age 20. The CPFI is a different question than "when does fluid g peak."

When cognitive decline becomes detectable: 45-49

The cleanest published numbers on when decline begins come from Whitehall II. That prospective cohort followed 7,390 British civil servants across 10 years [Singh-Manoux 2012] [Whitehall 2012]. Reasoning declined -3.6% in both men [95% CI -4.1, -3.0] and women [95% CI -4.6, -2.7] in the 45-49 age band [Singh-Manoux 2012]. Vocabulary moved the opposite direction in the same cohort: +0.7% in men [95% CI +0.4, +1.1] and +0.8% in women [95% CI +0.2, +1.4] [Whitehall 2012].

That is the gold-standard answer to "when does decline begin." It is detectable at 45-49, and the size is modest: about a third of a standard deviation per decade for the speeded reasoning measure. Two important caveats come with that headline.

First, the 45-49 onset is the Whitehall II UK civil-servant figure; US population samples (HRS, MIDUS) show the same direction but a smoother, more continuous mid-life decline rather than a sharp 45-49 knot. Knowledge-worker readers are closer to Whitehall II's selection than to HRS's general population, which is why we use it as the anchor, but readers should not generalize the -3.6%/decade figure [Whitehall 2012] to all US adults at the 45-49 boundary.

Second, the AH4-I "reasoning" instrument used in Whitehall II is heavily speeded (Singh-Manoux et al., 2012). It captures speed-loaded fluid g, not pure reasoning. Longitudinal VCAP data on un-speeded reasoning placed the inflection much later, in the 60s and beyond (Salthouse, 2019), and the 10-year follow-up overlaps two test administrations, so practice and retest effects bias the slope upward. The observed declines therefore understate the true age-related drop on speed-loaded measures and may overstate the drop on un-speeded reasoning. The headline decline is real. The exact size depends on the instrument.

The decade-slope picture from VCAP longitudinal data adds nuance (Salthouse, 2019). Reading the table by domain rather than by age, speed is the lone domain with negative slopes from the 30s onward (-0.022 z/year by the 50s). Memory and reasoning do not turn negative in any large way until the 60s and 70s; reasoning slope is still +0.019 z/year in the 40s before flipping to -0.041 z/year by the 80s. Vocabulary stays at a small positive through the 50s. The popular collapse of all four into "IQ" obscures every interesting pattern.

Education predicts level, not rate of decline

The most-told story about how to slow cognitive aging is "stay educated." That story is largely wrong on the rate question, and right on a different question. The distinction is the part the popular framing keeps blurring.

A 2020 meta-analysis pooled 92 longitudinal cohort studies covering more than 109,000 adults (Seblova et al., 2020). Their pooled mean estimates failed the test for significance.

The episodic-memory upper bound is +0.004 SD/decade with a 95% CI of -0.003 to +0.012 [Seblova 2020]. That is a tight CI, not an underpowered null. The conclusion is robust: education predicts what level you are at when aging begins, not how fast you decline thereafter.

The strongest published counter-finding reports a positive education × age slope of β = 0.112, 95% CI 0.056 to 0.170 [Lovden 2018] (Lövdén et al., 2018, PMC6211246). The 2020 meta-analysis pooled Lövdén alongside 91 other studies and still found the cross-domain effect negligible [Seblova 2020]. Lövdén is real; it is also one paper among many, and the meta-analytic verdict is null.

The takeaway is not "education does not matter." The version of the claim doing most of the popular work, "stay in school to keep your fluid reasoning sharp at 60," is not what the data supports. The version that is supported is more specific: education raises your level, may delay clinical dementia, and does not change your normal-aging slope. For more on what the malleability literature does and does not show, see our scientific consensus piece on whether you can increase your IQ score.

Can you keep IQ high as you age?

Two parts to the answer. You can preserve crystallized abilities at low cost, and you can slow but not stop fluid decline; the interventions with real evidence are physical exercise, cardiovascular health, sleep quality, and continued cognitive demand at work. None of these reverse aging on the underlying speeded measures.

The interventions that the popular literature oversells include far-transfer brain training (a null result in most 2020+ meta-analyses) and supplements (mixed at best). Near-transfer training shows small effects (g = 0.27-0.89, wide CIs); far-transfer to general fluid intelligence is null.

Skill acquisition into your 40s and beyond does protect specific abilities; for the protective-skill-learning angle, see neuroplasticity after 40 and learning new skills to protect IQ.

Do IQ tests account for age?

Yes. Standardized IQ tests like the WAIS-IV report age-normed scores: your raw performance is compared with others in your age band. A 55-year-old scoring 115 has performed at the 84th percentile relative to other 55-year-olds, not relative to 25-year-olds. Raw scores fall on speeded subtests; the scaled score adjusts for that.

This is why a 60-year-old can have a higher IQ than they did at 30 even with slower raw performance: the test grades on the appropriate curve. If you are using an age-adjusted instrument, you are not competing against your younger self in any meaningful sense.

What this means if you are 45 and worried

The data does not say your best years are behind you. It says different abilities peak at different times across a 40-year span, and the abilities that are paid the most in midlife knowledge work tend to peak in the second half of that span.

If you are a Career Pivoter wondering whether you are too old to make a move, the BLS data says no: the median person making lateral moves into senior individual-contributor roles is your age.

If you are an Overqualified Worker frustrated that the work has gotten too easy, the issue is rarely fading horsepower. The more common cause is a role that no longer matches your cognitive profile. For more on what that mismatch looks like and why it tracks with high processing speed, see our breakdown of processing speed versus working memory and our piece on why high-IQ individuals burn out in middle management.

The reframe that matters: your specific subskill profile is not "your IQ at 45." It is twelve numbers, and they are moving in different directions on different timelines. Some are still rising. Some have plateaued. A few are slowly declining. Knowing which is which is the difference between a generic story about decline and a usable picture of your actual cognitive position.

A note on the chart, the methods, and what we owe the reader

We labeled every row of the subskill table with its source and method [Hartshorne 2015] [Salthouse 2019] [Schaie 1994] because mixing cross-sectional, longitudinal, and cohort-sequential data without disclosure is the most common methodological sin in popular cognitive-aging coverage. That said, single-curve summaries serve a useful purpose for non-specialist audiences who need a one-paragraph answer; their flaw is not in existing but in being treated as the full story. We labeled the Gignac CPFI as [MODEL] because it is a weighting index, not a measurement. We used Singh-Manoux's exact CIs and reported Salthouse's slopes only with their published p<.01 asterisks [Salthouse 2019] because Salthouse's Table 2 does not publish 95% confidence intervals; constructing them after the fact would have been false precision.

The chart that survives all those disclosures is still the headline finding of this article. Cognitive ability is asynchronous. The popular collapse of "IQ" into a single curve hides the actual story. The actual story has rises and falls and plateaus, all happening simultaneously across different abilities, and the labor market is already paying for the version of the story that the popular framing leaves out.