After reading, any written output task using the new words beat reading alone, at both immediate and delayed post-test. Prediction tasks beat question-answering, which beat summarising.
Study
Rassaei, E. (2017). Effects of three forms of reading-based output activity on L2 vocabulary learning. Language Teaching Research, 21(1), 76–95.
- Learners read a story containing pre-tested unknown words, glossed in L1 (Farsi).
- Four groups: summary / question-and-answer / prediction / reading only.
- Post-test and delayed post-test. Sample size not reported in the summary.
Findings
All three output groups > control at both tests. Prediction was significantly best at both points; Q&A came second. Explanations: Output Hypothesis (Swain), depth of processing (Craik & Lockhart; Laufer & Hulstijn’s Involvement Load is the closer construct (Claude)), retrieval frequency.
Classroom moves
- Follow every reading text with a short written task that requires the target words.
- Prefer tasks that make learners apply the words: predict what happens next, or transfer them to a new situation.
- Check with a delayed test, not only an immediate one.
- Gloss before the task.
Limits
Narrative texts only; a single study; context only lightly described. Non-fiction or evaluative tasks may behave differently.
Apply in my context (Claude)
- Prepare 2e reading pages end with comprehension questions. Add a 5-minute “what happens next / what would you do” written task using 4 target words. That is a cheap, evidence-based upgrade to every reading lesson.
- Homework-check decks can then test the same words a week later, which gives you the delayed post-test for free.