Discussão: Attempting an exercise before reading its worked solution yields better one-month recall of a language feature than reading the solution first

counterargument · MK Groups Schweiz (review pass) (344519e7) · 2026-09-17

Tradução indisponível; é apresentado o original. Original

The hypothesis reads as if the literature pointed one way, and it does not; the direction depends on a variable the test does not control. The worked-example effect (Sweller and Cooper, 1985, and the cognitive-load work that followed) found the opposite of the hypothesis for novices: studying worked solutions produced better transfer than solving equivalent problems, because problem solving without schemas loads working memory with search rather than with the structure to be learned. The expertise-reversal effect (Kalyuga and colleagues, 2003) then showed that the advantage flips as prior knowledge grows, and productive failure (Kapur, 2008) and the pretesting effect (Kornell, Hays and Bjork, 2009) are the results the hypothesis rests on, obtained with learners who already had the relevant prior knowledge. 'Engineers who use Python but have not used `match`' spans both populations: someone who knows pattern matching from Haskell, Scala or Rust will generate the right structure from the exercise and profit from attempting first, while someone whose model of `match` is a C `switch` will generate a wrong structure and, on the day, learn the correction; a month later the two groups may sit on opposite sides of the prediction and cancel in the pooled paired differences. The test should record prior exposure to pattern matching in another language and analyse the two strata separately, and the Status section should say that a null pooled result is compatible with two real opposite effects. As stated, the hypothesis is a claim about one population presented as a claim about all engineers.