The replay button erases useful evidence. A B1 listener chooses option C, feels only 40 per cent sure, presses replay, and changes to A after hearing one familiar word. The final answer records A. It loses the uncertainty, the original interpretation, and the piece of sound that caused the change.
Before a second listen, ask learners to mark each answer high, medium, or low confidence and underline the words they believe they heard. The marks take under a minute. They turn replay from another chance to get points into a comparison between two interpretations.
Listening needs a visible process
Vandergrift and Tafaghodtari tested a metacognitive, process-based listening sequence over a semester with 106 learners of French. Fifty-nine learners used prediction, planning, monitoring, evaluation, and problem solving while 47 comparison learners heard the same texts the same number of times without that guided process. After controlling for initial differences, the experimental group performed better on the final comprehension measure, with larger gains among less skilled listeners. The 2010 study supports the full instructional sequence. it does not isolate confidence marking as the active ingredient.
That limit matters. A confidence column is a classroom adaptation of monitoring and evaluation, not a technique independently proven by this experiment. Its value is practical: it preserves a learner’s judgment at a precise moment, ready for discussion after the next listen.
Toapanta later compared three instructional approaches to L2 listening, including a metacognitive cycle, awareness raising through reflection and discussion, and vocabulary plus guiding questions. The 2022 intervention used pretests, posttests, and a delayed posttest. Results favoured instruction that developed metacognitive knowledge and regulation, with preliminary evidence that effects lasted. The study compares packages of instruction, so a teacher should resist claiming that any single worksheet feature caused the gains.
A 2023 meta-analytic structural equation modelling study found that metacognitive awareness explained only a small degree of L2 listening performance. Its authors also report moderation by participant type, test response format, and publication type. The meta-analysis puts language knowledge back in view: reflection cannot compensate for unknown vocabulary or speech that a learner cannot segment.
Confidence marks expose a problem. They do not solve it.
Separate uncertainty from error
Consider two wrong answers. Learner A marks high confidence because “fifteen” sounded clear, though the recording said “fifty.” Learner B marks low confidence and writes only “price?” beside the item. Both lose the test point. Their next work should differ. A needs a sound contrast and a reason to distrust the first parse. B may need another pass with the question focus made explicit.
Correct answers also split. High-confidence correct answers can leave the practice set. Low-confidence correct answers deserve review because guessing and fragile recognition look identical in the score column. This is especially useful for someone preparing for the 全民英檢 or TOEIC, where practice software often displays only right and wrong.
Use three labels, not a scale from 1 to 10. Fine distinctions create arithmetic without improving the conversation. High means “I can point to the words that support this.” Medium means “the overall message fits, though one part is missing.” Low means “I guessed or followed one isolated word.” Those definitions tie confidence to evidence.
Make the second listen answer a different question
After learners compare first answers, give the replay a target. One person listens for the time expression. another checks who performed the action. a third listens for the phrase connecting cause and result. This follows the logic in Choc’s article on why a second listen works harder after learners compare notes. Confidence marks sharpen the comparison by showing which claims are firm and which remain open.
After replay, learners use a different colour to change an answer, revise the supporting words, and mark confidence again. The productive discussion is about the movement: low to high after locating “unless”. high to low after noticing that the speaker corrected herself. high and wrong twice because “can” and “can’t” still collapse in the recording.
A transcript belongs at the end of this cycle. Ask learners to find the short stretch that settled the answer, then play that stretch once with text and once without it. Reading every line while listening can hide the actual bottleneck. The related article on English word boundaries explains why known words can disappear inside continuous speech.
Know when to skip the confidence column
Very easy extensive listening does not need item-by-item monitoring. The learner’s purpose may be following a story, and frequent ratings would break that experience. Confidence marking also adds little when the task is pure sound discrimination with immediate feedback. Use it for short, information-rich passages where interpretations can be compared.
The method can mislead anxious learners if teachers treat low confidence as a personal weakness. Calibration is the goal. A well-calibrated listener sometimes reports low confidence because the signal genuinely was ambiguous. Praise accurate uncertainty when the item contains a reduced form, an unfamiliar proper noun, or a correction midway through the sentence.
Keep the evidence phrase short. Learners should write “changed train,” “before Friday,” or “Sara corrected number,” rather than transcribing a sentence while the recording continues. The note is a retrieval cue for later discussion. If note-taking itself becomes the main difficulty, let learners mark a question number and add the phrase during a 20-second pause.
Teachers can aggregate the pattern without turning it into a score. Across four questions, a learner might have two high-confidence correct answers, one low-confidence correct answer, and one high-confidence error. The last category deserves attention first because a stable misperception is more likely than a random guess. After targeted practice, repeat the contrast in a new passage rather than replaying the old item until memory supplies the answer.
I prefer paper for the first attempt. It keeps the original mark visible after revision and prevents software from replacing one answer with another. Fold the page into four narrow columns: first answer, confidence, evidence heard, second answer. For tomorrow’s listening passage, use only four questions. The revealing entry may be the correct answer that remains labelled low.