Years ago I spent forty minutes explaining to a junior engineer why one version of a function was wrong and the other was fine. I drew it. I ran both. He nodded at every step, and at the end he said, honestly and without embarrassment: "They look the same to me."
They did look the same to him. He was paying attention, and he was clever. He had no category to put the difference in, so his eye slid off it, the way your eye slides off a word in a language you do not read.
I think about that conversation every time someone tells me a training programme "covered" something.
The colour nobody can describe
There is a colour called Olo. It is produced by stimulating a narrow class of retinal cones directly, and it sits outside the range ordinary vision delivers. People who have seen it call it impossibly saturated, then immediately apologise for the description.
Everybody reaches for "intense teal" and everybody says, in the same breath, that this is wrong.
A more careful sentence would change nothing. The listener has no bucket to file the thing under, so every analogy lands in the nearest existing bucket and gets quietly bent to fit. The receiver is missing the category.
Eyeballs make it vivid, but missing the category is the normal condition of anyone learning anything.
Every field is full of Olo
Expertise, in large part, is having categories other people do not have. The expert sees a different number of things in the same view.
Field | Looks identical to a beginner | What the expert sees |
|---|---|---|
Code review |
| Sequential I/O vs. parallel: 40 requests at 200ms each is 8 seconds or 0.2 |
Arabic phonology | ح and ه | Pharyngeal vs. glottal. Different words entirely |
Interface design | 16px vs. 20px of padding | A broken vertical rhythm that makes the whole page feel cheap |
Statistics | p = 0.04 vs. p = 0.06 | Almost nothing |
Operations | 200ms p50 vs. 200ms p99 | A healthy service vs. one that is failing for your best customers |
Every one of those pairs is a thing a beginner can look straight at, twice, and register as one thing.
So when someone says the course "covered" it, ask what it covered. You can cover a distinction the learner cannot perceive. It happens constantly. The material was correct, the delivery was clear, the quiz was passed, and the category never formed.
You cannot explain a category into existence
Most programmes get this wrong, and it is an honest mistake, because explaining is what teaching feels like.
You learn to hear ح versus ه by hearing the pair, hundreds of times, guessing, and being told immediately whether you were right. Wine tasters, radiologists, and chess players are all built the same way. The research calls it perceptual learning; in practice it is embarrassingly simple:
Show pairs. One instance teaches nothing. The difference is the object of study, so the pair is the unit.
Make them commit before the answer. A guess creates a prediction, and a wrong prediction is what makes the correction stick. Recognition without commitment does nothing.
Give feedback in seconds. The signal has to arrive while the guess is still in working memory.
Do hundreds of short reps. Ten minutes a day beats a full-day workshop by a wide margin.
It is closer to a drill than a lecture, and it is unglamorous enough that it rarely survives contact with a training budget.
Nobody puts "we did 300 flashcards" in the annual report.
Next time you are building a course, or reviewing one, find the pairs. Ask which two things your learners currently see as one thing.
The Olo write-up is here, and it is worth five minutes.
One email when we publish. Research, product decisions, and what teams report back.
