What the claim needs in order to be true
Pashler and colleagues set out what evidence would have to look like to support the learning-styles idea, and the requirement is specific. It is not enough to show that people have preferences, or that different people perform differently. You need a study that classifies learners by style, teaches the same material in two or more different ways, and finds that each group does best with its own matched method — an interaction, not a main effect. This is a testable design, and it has been run. The authors found almost no studies using it that produced the required interaction, alongside a great many studies with weaker designs that could not test the claim at all.
Why it persists anyway
The idea is appealing for reasons that have nothing to do with evidence. It flatters the learner, it explains past difficulty without blame, and it gives teachers something concrete to do. It also contains a true observation wrapped around a false inference: people genuinely do have preferences, and preferences do affect motivation and engagement. The unsupported step is that matching instruction to a preference produces more learning. Belief in the idea remains high among educators in many countries, which is worth knowing when you encounter it presented as settled.
What predicts learning instead
What the material demands matters more than what the learner prefers. Learning to identify birdsong needs audio; learning anatomy needs images; learning to weld needs a torch, and no preference changes any of that. Beyond the material, the techniques with the strongest evidence apply to everyone regardless of self-described style: retrieval practice, spaced practice, and interleaving where the subject supports it. Dunlosky and colleagues rate those in their highest utility category on exactly that basis — they work across ages, materials, and learners. Time spent identifying your style is better spent on a practice test.