John Sweller kept finding students who could follow a worked example and then fail the problem printed under it. The material was not too hard for them. The format was spending their working memory on bookkeeping before they reached the part that needed thinking.
The split he ended up with is the useful part. Some of the effort is what the task actually needs, and five prices really do have to be compared. The rest is what the layout added, and that share is the only one anybody gets to delete.
Below, nine rounds of one task. Five parts, five prices, click the one that costs least. Four rounds print each price on its part's own row. Four rounds put the prices in a separate block, keyed by letter, in an order that never matches the list. One round at the start is a warm-up and is thrown away.
The two layouts carry the same characters. The letter keys are printed in both, so the split version is not simply more to read — it is the same marks, moved. The rows you click are the same size and in the same place either way, and the question is on screen before the round starts, so reading it is outside the timed window.
What this cannot separate is carrying from travelling. The split layout makes your eyes cross the screen as well as makes you hold a letter while they do. One click cannot tell those two apart. Both of them are the layout's doing rather than the task's, which is the claim being made, and it is a coarser claim than it looks.
This one needs JavaScript. It times the same task in two layouts, and there is nothing to time without it.
What the split layout is doing
Nothing exotic. A legend under a chart, or an error summary at the top of a form whose fields are two screens further down. Each one asks somebody to carry a value across a gap that did not have to be there.
The timing runs on the correct rounds only. A wrong answer is a different thing happening, so it is counted as an error and kept out of the medians. If three or more of the eight went wrong, the page stops comparing times altogether — a run where the task was not done has nothing to say about how long doing it takes.
Four rounds a layout is a coarse instrument. A gap under 400 ms, or under a tenth of the faster median, is not reported as a difference. Rounds under 300 ms are thrown out as clicks that started before the screen did, and rounds over 20 seconds are thrown out as somebody looking away.
What it costs you
Extraneous load is the only kind you can remove. The comparison is the task and it is not negotiable. The letter you had to hold while you found the row is negotiable, and it was put there by whoever drew the screen.
Nobody can tell you which kind they paid. The two feel identical from the inside and both come back as "confusing" in a research session. The person is right that something was hard. They have no way of knowing which part of it you were free to delete.
Distance is load. Two things that get used together and are drawn apart make a person hold one of them in their head for the length of the gap. The gap is usually a layout decision, sometimes a scroll, and occasionally a second screen.
Removing load and removing information are different moves. Folding six fields behind a disclosure makes the first screen quieter and makes somebody remember what they saw on it. That is the load in a new place. It might still be the right call. It is not a deletion.