sameness, as a channel
Two tuning forks, both made for the same note — A, 440 Hz. Strike the first and hold it near the second. The second fork is never touched, never struck, and nothing reaches it but the air. Yet after a moment it begins to sing on its own: faintly at first, then building, until it is sounding the note it was never given.
It is not being pushed. A push would work at any pitch. This works only insofar as the two forks agree — retune the second fork even a little and the answer fades, not gradually in spirit but by an exact curve. Below, the second fork's pitch and its patience are yours to change. The curve is its voice plotted against disagreement; the marker is where you have left it.
The second fork's voice: 0% of the fullest answer it can give.
At the default patience, the answer keeps at least 71% of its full voice only inside about ±1 Hz of agreement — a window 2 Hz wide in a 440 Hz note. Detune by 5 Hz and it is down to about a fifth. Nothing about the first fork changed; only the agreement did. The width of the window is set entirely by patience: a fork that forgets quickly (low Q) forgives a sloppy match, and a fork that remembers long (high Q) answers only to an almost exact twin.
Drawn slowed and widened so the eye can follow: a real 440 Hz fork rings thousands of times before fading (Q in the thousands), which narrows its window to a fraction of a hertz, and its answer builds in a fraction of a second. The arithmetic underneath the drawing is not widened — the voice at each detune is the steady-state amplitude of a driven, damped oscillator, computed exactly, and the numbers above are read off it.
A companion, in a way, to the harmonograph: there, a pattern lived in the small difference between two swings. Here difference is what closes the channel. The two pages make opposite claims, and both are true of their own apparatus.