Can a decay process explain the timing of conditioned responses?

The author built math models that use decay to explain when animals respond. He asked if these models can explain the scatter we see in real response times.

He used computer runs and equations, not live subjects. The goal was to test if a fading memory trace could create the typical spread in latencies.

Pure decay models failed. They could not copy the scalar property, the fact that spread grows with the target time.

The math showed that decay adds the wrong kind of noise. It makes timing either too exact or too wild, never the natural proportion we see in data.

Rapport et al. (1996) showed pigeons remember several recent intervals, not just the last one. Their memory-weighting idea fits the new view that read-out, not fading traces, sets variability.

Capaldi (1992) pushed nonlinear dynamics for behavior. Parmenter (1999) agrees that nonlinear tools help, but warns that simple decay is the wrong kind of nonlinearity.

Delano (2007) found hyperbolic decay in reinforcer value. This looks like a clash, yet E tested choice, not latency spread. Different measures let both papers be right.

When you graph latency data, expect the width to scale with the mean. Do not blame a fading cue or memory trace. Instead, look at how the animal samples its memory of the target time. Build teaching programs that give clear, stable reference times and check for scalar spread as a healthy sign, not error.