qCLS Operator Guide

Categorical loudness scaling in the browser. Everything runs locally; no participant data leaves the machine.

1. Before each session

Calibrate. Click Calibrate for a continuous 1 kHz tone at 60 dB SPL and set the hardware to match. Click again to stop. Every level, every contour and the audiogram prior inherit this step.
Set Max Output. The loudest level that will ever be presented. Default 100 dB SPL. Lower it for a listener with reduced tolerance.
Enter the participant details and stimulus settings.
Optionally enter the audiogram (see below).

2. Safety

Nothing is presented above Max Output. Whatever is typed is clamped to the 110 dB SPL equipment limit, and the field locks once a run starts.

The working ceiling also drops 5 dB automatically every time the listener responds "Too Loud", in both phases, and never rises again during a run. A final check sits immediately before the level becomes sound.

Stimulus selection targets the boundary it is least certain about, and the top category sits near the top of the range, so an unbounded run would drift toward loud levels rather than away from them. The ceiling and the backoff are what prevent that.

3. The audiogram (optional)

Left blank, the procedure behaves exactly as it would without this feature. Filled in, it starts the listener from a prior matched to their hearing rather than from a population average, worth about 0.4 dB on average.

Nine fields, 250 Hz to 6 kHz. Leave any frequency blank if it was not measured; blanks fall back to the population distribution for that band rather than to a guess.
Units must match how the audiogram was recorded, dB HL or dB SPL.
Transducer selects the conversion used for dB HL.

Units and transducer both matter, and neither fails visibly. On the same audiogram the prior at 250 Hz differs by about 10 dB between the default table and ER-3A insert earphones. If you are unsure which applies to your setup, leave the transducer on unspecified (legacy), or leave the audiogram blank, rather than guessing.

Only tables measured at BTNRH are offered. Published figures for other transducers were removed after the ER-3A row from that source proved up to 15.5 dB low against BTNRH's own measurements. Measure a transducer before adding it.

4. Running the test

Click Run Test. The listener sees the response scale and the instructions available from View Instructions.

Phase 1	Brackets the listener's range. Descends in 10 dB steps until "Can't Hear", then ascends from the start level until "Too Loud" or the ceiling.
Phase 2	About 100 trials. Each one is chosen where the current estimate is least certain, so the listener will sometimes hear very similar sounds in succession. This is expected, and the instructions tell them it is fine to repeat an answer.

5. Reading the result

The plot shows equal-loudness contours: one curve per category boundary, giving the level that elicits CU 5, 10, … 50 against frequency.

The shaded band around each curve is ±1 standard deviation. A wide band means that boundary was not pinned down; treat it with caution rather than reading the curve alone.
If an audiogram was entered it is drawn as a dashed line, converted to dB SPL. It should sit near the lowest contour.

Loudness Growth	CU per dB between the CU 5 and CU 50 contours, averaged over bands. Higher means a narrower dynamic range.
False-Alarm Rate	Responses above "Can't Hear" to sounds well below the listener's threshold, with the number of such trials in brackets. A screening indicator of attention, not a precise rate. Shown as `--` when too few trials qualify.
Posterior SD	Median uncertainty across all boundaries. The single best summary of how much to trust the run.

6. Saving the results

Two downloads appear when the test completes.

Download trials (TBT)	The raw record: phase, trial, frequency, level, response.
Download contours (ELC)	The result: band, frequency, CU, level and SD, 90 rows.

Completed sessions also appear under Previous Sessions and can be reloaded.

Session storage is not a record system. It lives in this browser on this machine, holds the 50 most recent sessions, and is lost if the browser data is cleared. Download the CSV files for anything you need to keep.