Fifteen or Fifty? A Listening Item with Noise and Reduction
Cobalt Airlines is building a listening item for its gate-agent certification. The audio, recorded in a crew room with a ventilation hum, contains this stretch: “We’re looking at a fifteen percent increase in turnaround delays-that’s up from about twenty to twenty-three minutes-so you gotta log every gate change, not just the international ones.” A trial candidate reported hearing “fifty percent” and “you don’t gotta.” The item writer has the waveform, a transcript, and the speaker’s note that she stressed the second syllable of fifteen. Company policy says assessment items must test comprehension, not microphone luck, and any item with more than 20% mishearing in piloting must be revised or re-recorded. In the pilot, 9 of 30 candidates missed the number and 4 missed the obligation. The writer must decide whether the item is salvageable, what the correct answer key is, and whether to keep the noise as a deliberate distractor.
1. Resolve “fifteen” versus “fifty” using at least two independent cues.
Read the worked answer
The correct key is fifteen percent. The strongest independent cue is arithmetic: the speaker says the increase is “up from about twenty to twenty-three minutes.” A change from 20 to 23 is 3 minutes, which is 15% of 20, not 50%. The repair phrase “that’s up from …” functions as a definition of the earlier number, so it disambiguates the word even if the fricatives are partly masked.
The second cue is stress. The speaker reports stressing the second syllable: fif-TEEN. That stress pattern is typical of the teen number in contrast or careful speech; FIF-ty has initial stress. The ventilation hum can mask the /f/ and the final /n/, which is exactly how fifteen can be heard as fifty, but the stress and the arithmetic point the same way. A test taker should be able to use either cue; the item is stronger if the key can be recovered from context rather than from a single acoustic detail.
2. Explain the mishearing of “you gotta” as “you don’t gotta,” and state what the item should actually test.
Read the worked answer
Gotta is a reduced form of got to, meaning “must” or “have to.” In connected speech, the auxiliary before it is often reduced or lost: you’ve gotta can sound like you gotta, and the negative don’t is not present in the source. A listener who expects an explicit auxiliary may insert one, and the negative is a plausible repair because the following clause contains not just the international ones. That not belongs to the contrast (“every gate change, including domestic ones”), not to the obligation.
The item should test pragmatic comprehension: the agent must understand that logging every gate change is an obligation and that “not just the international ones” expands the scope rather than cancels the obligation. It should not test whether the candidate can hear a single reduced consonant. If the item is kept, the question should ask what the agent is required to do, and the distractors should include “log only international gate changes” and “no need to log gate changes,” not a spelling trap.
3. Given that 9 of 30 candidates missed the number, should the item be kept, revised, or re-recorded? Justify the decision under the company policy.
Read the worked answer
The item should be revised and re-recorded, not kept as is. Nine of 30 is 30%, which exceeds the 20% threshold. The policy says items must test comprehension, not microphone luck, and the pilot result is evidence that the acoustic environment is interfering with the construct. The fact that the number is recoverable from arithmetic does not rescue the current audio if a third of candidates never get far enough to use that cue.
The revision should preserve the comprehension demand while reducing accidental masking. Re-record the number with a clearer /f/ and a slightly slower fifteen percent, keep the ventilation hum at a lower level so it remains realistic but not dominant, and retain the repair phrase “up from twenty to twenty-three” because it is a legitimate contextual cue. Then re-pilot with a new group. Do not delete the noise entirely: the goal is to test listening in a realistic crew-room setting, not to create an anechoic studio item. If the re-recorded item still exceeds 20%, replace the number with one that is less vulnerable to /f/ masking.