# Three original two-line PCM candidates · full event transcripts

**Candidate only; not cleared for children.** Before use a human educator must listen to every **whole file** on the intended device, check for clicks and comfortable room/volume conditions, record reviewer/date/device/setting/stop plan, offer exact no-sound route and respect pass. Current status: **HUMAN LISTENING NOT DONE / DEVICE NOT APPROVED / LOCAL VOLUME NOT APPROVED**. No compulsory headphones, amplification, looping, singing, tapping, movement or recording. A cue is never a body or safety command. Programmed amplitude is not measured ear-level decibels or a comfort guarantee. No child data.

All files are original **22,050 Hz, 16-bit mono PCM, 3.2 seconds**, no speech, background track or samples. The six shared place starts are **0.20, 0.70, 1.20, 1.70, 2.20, 2.70 seconds**, evenly separated by 0.50 s. Each active event lasts **0.20 s** with 0.012 s entry/exit fades. Line A is a smooth electronic 330 Hz sine at 0.016 of digital full scale. B is a second original electronic timbre with 330 Hz fundamental plus half-strength 660 Hz harmonic, multiplier 0.010. Source/timbre differences label the parts; pitch, level and timbre are not child variables in this pack. At a shared start the two synthesized parts are mixed, not two physical speakers. Peak is numerically checked but not a listening test.

| Candidate | Complete place-by-place onset / wait transcript | Exact no-sound score |
| --- | --- | --- |
| [Together](cue-together.wav) | Place 1 at 0.20 s A+B; 2 at 0.70 both wait; 3 at 1.20 A+B; 4 at 1.70 both wait; 5 at 2.20 A+B; 6 at 2.70 both wait. All six individual events: A 0.20/1.20/2.20 and B 0.20/1.20/2.20. | A circles 1,3,5; B squares 1,3,5; dash waits elsewhere, six shared columns. |
| [Turns](cue-turns.wav) | Place 1 at 0.20 s A; 2 at 0.70 B; 3 at 1.20 A; 4 at 1.70 B; 5 at 2.20 A; 6 at 2.70 B. All six events: A 0.20/1.20/2.20 and B 0.70/1.70/2.70. The other line waits at each place. | A circles 1,3,5; B squares 2,4,6; dash waits elsewhere, one line per column. |
| [Meet then separate](cue-mix.wav) | Place 1 at 0.20 s A+B; 2 at 0.70 both wait; 3 at 1.20 A; 4 at 1.70 B; 5 at 2.20 A; 6 at 2.70 B. All six individual events: A 0.20/1.20/2.20 and B 0.20/1.70/2.70. | A circles 1,3,5; B squares 1,4,6; dash waits elsewhere. Read both / neither / A / B / A / B. |

**Full silence transcript:** all files are silent 0.00–0.20. Together sounds 0.20–0.40, 1.20–1.40, 2.20–2.40; silence 0.40–1.20, 1.40–2.20, 2.40–3.20. Turns sounds 0.20–0.40, 0.70–0.90, 1.20–1.40, 1.70–1.90, 2.20–2.40, 2.70–2.90; silence 0.40–0.70, 0.90–1.20, 1.40–1.70, 1.90–2.20, 2.40–2.70, 2.90–3.20. Mix sounds 0.20–0.40, 1.20–1.40, 1.70–1.90, 2.20–2.40, 2.70–2.90; silence 0.40–1.20, 1.40–1.70, 1.90–2.20, 2.40–2.70, 2.90–3.20. Faded endpoints may have zero samples within active intervals.

No-sound rows can be raised round/square tokens with dash waits or exact neutral column descriptions. They fully meet composition-plan/texture reasoning; they do not prove auditory discrimination, performed coordination, actual comfortable volume, cultural importance or audience receipt. If reviewed playback/real sound-making occurs, record cue/device/setting and the actual learner response separately. No file provides either fresh public check's layout. [Owned synthesis code](../generate_assets.py) reproduces all bytes.