Spatial ability can be partly trained, but the individual baseline gap persists

What it is

Spatial ability isn't fixed: practice and training reliably raise performance on tasks like mental rotation and spatial visualization — that finding itself holds up. But trainable does not mean levelable: training shifts most people's performance upward while largely preserving their relative ranking, so someone with a strong baseline typically remains ahead of someone with a weak baseline even after both receive the same training. Reading "spatial ability can be trained" as "training erases individual differences" over-extends the finding.

Why it happens

The improvement from practice comes mainly from two sources: strategy learning — practitioners learn more efficient mental operations, such as rotating only a figure's key features instead of reconstructing the whole shape — and task-specific familiarity with particular stimulus formats or test structures. Neither source expands the underlying processing capacity itself; both optimize how existing capacity is used. That is why training transfers well to similar tasks (the same type of rotation figures), called near transfer, but transfers much less to structurally different tasks (going from 2D rotation-figure training to real-world 3D wayfinding), where the evidence for far transfer is thin. Training shifts the whole distribution to the right, but the shape of that distribution — including the relative gap between individuals — changes only modestly.

Studying it

The standard paradigm is pretest-training-posttest with a control group: training materials commonly include action video games or purpose-built spatial-training software, with training durations ranging from a few hours to several weeks. The posttest should include both tasks similar to the training material (to measure near transfer) and structurally different tasks — rotation-figure types never seen during training, or a real-world wayfinding test (to measure far transfer). Key variables are training duration, the similarity between training material and the transfer test, and the time gap between the end of training and the posttest. The methodological point most often missed: most training studies report effect sizes concentrated in near transfer, while evidence for far transfer is thinner in both quantity and effect size, and most existing studies administer the posttest within weeks of training's end — data on whether gains persist longer than that is very limited. Treating "played a spatial-training game" as equivalent to "improved wayfinding ability" applies near-transfer evidence to a far-transfer claim.

Where it stops holding

Training effects may be larger during childhood and adolescence, when spatial ability is still developing; gains in adulthood persist but are typically smaller, with some evidence of further decline with age. Evidence on how long training gains last is limited — most existing studies use short posttest windows, so whether the benefit holds for months or longer is not well established, and a one-off training intervention should not be assumed to produce a lasting effect.

Applying it

• Don't treat "offer a spatial-ability training tutorial" as a fix for users struggling with a 3D interface or navigation task — the far-transfer evidence from training to a specific product task is itself thin, and a product's core path should never assume users will complete training and have their ability leveled up first.
• If a product does include an onboarding or tutorial flow, position it as helping users become familiar with that specific interface's operations, not as a way to raise their general spatial ability.
• Verification: to check whether a tutorial or onboarding flow actually improved a user's spatial comprehension, test with a new task that differs from the tutorial content (far transfer), not with a task of the same type as the tutorial material (near transfer) — the latter will almost always show "improvement" without telling you anything about performance beyond the tutorial's own context.

Related

• Same group: A11.10.1 Mental rotation ability shows stable individual differences that affect how fast people read 3D interfaces and maps · A11.10.2 Users with weaker spatial ability lean on landmarks and text cues rather than abstract direction indicators · A11.10.4 Navigation interfaces built on a high-spatial-ability assumption systematically slow some users down
• Adjacent: A11.04.3 Onboarding imposes a net cost on experts
• Search terms: spatial training · near transfer · far transfer