Changing the cue beat repeating the same prompt — and learners guessed the opposite

What the researchers did

Butowska-Buczyńska, Kliś, Zawadzka, and Hanczakowski ran seven experiments. In six of them, Polish speakers with no Finnish learned 40 Finnish words. Each word sat inside a sentence that hinted at its meaning (for example, a sentence about sweeping the lattia). Across five practice cycles, that sentence was either always the same or always different. Practice was retrieval — trying to produce the translation — sometimes with feedback, sometimes without.

They compared this to restudy (reading the translation instead of recalling it), and they compared longer gaps between repetitions with almost-massed mini-blocks. One experiment added a 24-hour delay. Another collected judgments of how well items were learned.

The seventh experiment switched materials: English speakers watched five short geology lecture clips, then answered three questions per concept that were either repeated or varied, always with feedback. The final test used new transfer questions, not the practice items.

What they found

• Variable cues beat constant cues on a later test without the practice sentences — for example d = 0.81 without feedback and d = 0.67 with feedback in the first pair of experiments (n = 31 each).
• Retrieval beat restudy more clearly when cues varied. On an immediate test, retrieval vs restudy was not statistically significant with constant cues (d = 0.20) but was with varied cues (d = 0.50). After 24 hours, both conditions showed a retrieval advantage, still larger when cues varied (d = 0.61 vs 0.25).
• Spacing helped more with varied cues (d = 0.87) than with constant cues (d = 0.52).
• Learners’ ratings went the other way: after practice they judged constant cues as more effective (d = 0.61 for item ratings; d = 0.73 after the test). In the lecture study, 20 of 38 still picked constant questions as best even after seeing transfer results (d = 0.53 for varied practice).

What this means for learners and educators

• If you already use spaced recall, changing the wording of the prompt (new example sentence, new question angle) may add more than replaying the same card text.
• Feeling fluent during practice is a weak signal here: identical cues got easier fast, which looked like progress but transferred less well to a bare word or a new question.
• Teachers and deck authors can write several cues for the same fact rather than one canonical prompt. Students rarely invent extra questions themselves, so the design work sits with materials, not willpower.

Limitations and what we don't know yet

Most experiments used Finnish–Polish vocabulary in a lab-style task, not a semester-long course. Sample sizes per experiment were modest (about 30–80). Feedback was often needed so both conditions saw the correct answer equally often, which mixes “successful recall” with “encoding the correction.” Interactions among spacing, retrieval, and variation still need classroom tests with authentic homework. The metacognitive mismatch may make students push back on varied questions even when scores improve.

FAQ