The hall, the laptop and the 3.5-mark drop
Fourteen learners aged nine and ten sat a mock listening paper in the school hall on a Friday afternoon. The recording was played through a laptop's internal speakers on a table at the front. The hall has a tiled floor and a glass roof; a hand-clap takes about six seconds to die away, and the group was spread over six rows, with the back row roughly twelve metres from the laptop. The week before, the same recording had been played in the classroom through two small speakers, in a carpeted room, to the same group; the class mean was 11 out of 20. In the hall the mean was 7.5. Errors clustered on items containing numbers (thirteen/thirty, fourteen/forty), on names spelt aloud, and on two items where a speaker corrects himself mid-sentence. Three of the four strongest listeners dropped five or more marks; the weakest two scored exactly what they had scored before. One parent has emailed asking why her daughter 'has got worse'.
1. Decide whether the drop should be read as a proficiency problem or as a test-condition artefact, and state what further evidence you would collect before reporting anything to anyone.
Read the worked answer
Artefact, on the balance of evidence - but not yet confirmed.
Three things point that way:
- The errors are not spread evenly. They cluster exactly where the acoustic signal carries the answer: minimal-pair numbers, names spelt letter by letter, and self-corrections where the second word has to override the first. A learner who has stopped understanding loses content words too, not just signal-critical ones.
- The loss is largest at the top. Strong listeners had the most to lose from degraded input; the weakest two did not move, which looks like a floor effect rather than a plateau.
- The interval is one week. A genuine 3.5-mark regression in seven days is not a plausible proficiency story.
Before reporting, I would re-run the same paper in the classroom with good speakers, ideally in a different item order to avoid practice effects, and compare item by item rather than totals. Only items that stay wrong in good conditions are evidence about the learners.
2. Write what you would say to the parent and what you would record in the learners' files. State explicitly what you would not say, and why.
Read the worked answer
I would not write 'her listening has got worse', and I would not quote 7.5 as if it measured her. That number measured a room.
To the parent: the facts, plainly - two runs of the same paper one week apart, in different rooms, with different equipment; the hall run was affected by echo and speaker quality; the pattern of errors matches that explanation; we are re-running it in the classroom next week and will send that result. No apology for the child's ability, no promise that she will 'improve', no comparison with other children.
In the record: 'Mock listening 2 - hall, laptop speakers, ~6 s reverberation, back row 12 m from source. Scores depressed across the top of the group; error pattern consistent with signal degradation. Not valid for progress tracking. Re-run scheduled.'
That note matters because the next teacher will otherwise read 7.5/20 as a baseline and teach to a deficit that does not exist.
3. List three changes to how mocks are run next term, each tied to a specific feature of this episode.
Read the worked answer
- Fix the audio chain, not the children. External speaker on a stand at seated ear height, and a colleague sits in the back row and checks a test item before the group comes in. The 13/30, 14/40 and spelt-name errors are precisely what reverberation destroys; a two-minute room check would have caught it.
- Hold seating constant. Same room, same distance, same seats for every mock, so totals are comparable across sessions. Here the back row was 12 m away and the front row about 2 m - those children did not sit the same test.
- Change one variable at a time, and mark by item type. Never change room and equipment in the same run. Report number/name items separately from gist items, so that the next unexplained drop can be diagnosed rather than debated.