On the development of stimulus control.

Wildemann et al. (1973) watched pigeons learn to tell loud from soft white-noise bursts. The birds pecked one key if the noise was softer than a taught reference, another key if it was louder.

Sessions ran day after day. The team plotted how often the birds chose the "louder" key at each noise level. These plots are called stimulus-control curves.

Early curves were almost flat. Birds pecked at random no matter how loud the sound was.

With more trials the line bent into a clear S-shape. Birds now rarely pecked the wrong key near the reference point, showing sharp control by the noise difference.

Neuringer (1973), also with pigeons, seems to disagree. That study used click-rate instead of loudness and got flat curves even after training. The key difference is procedure: A used presence-absence training (clicks vs no clicks), while G used intradimensional training (softer vs louder clicks). Flat curves appear when the task is "sound or no sound," but sigmoid curves appear when the task is "more or less of the same thing."

Brown et al. (1972) had already shown that birds can match continuous tones to continuous responses, yet generalization to new tones stayed rough. The 1973 paper adds the next step: repeated practice along the same dimension smooths and sharpens that rough control into a tidy S-curve.

SLOANE (1964) plotted flicker-rate gradients and found each added S- pushed responding toward the safe end. G’s sigmoid curves are the end state of that push: many S- values along the continuum finally carve a steep, orderly slope.

When you teach a client to judge speed, pitch, or brightness, don’t stop at "stimulus or no stimulus." Build several comparison levels along the same dimension and give practice across them. Expect flat data at first; keep going until the curve bends. That extra training is what turns fuzzy responding into crisp, reliable discrimination.