Behavioral contrast for key pecking as a function of component duration when only one component varies.

McSweeney et al. (1993) worked with pigeons in a two-part schedule. One part always gave food for key pecks. The other part never gave food. They changed how long the no-food part lasted.

The birds pecked a key. The team watched response rates. They wanted to see how the length of the bad part changed the good part.

The biggest jump in pecking happened when the no-food part lasted 30 to 60 seconds. Shorter or longer times gave smaller jumps.

Only the bad part's length mattered. Changing the good part's length did not change the jump size.

de Rose (1986) ran almost the same test and got the same curve. Both labs show 30-60 s is the sweet spot for boosting rates.

Rogers-Warren et al. (1976) saw zero contrast when both parts gave equal food. K et al. still saw contrast because they held food steady and only changed time. The studies differ in what they controlled, not in the basic rule.

Ginsburg et al. (1971) stretched pre-food extinction up to 40 min and saw bigger contrast. K et al. kept the blackout inside the session and found a narrow peak. Together they show timing matters both before and within schedules.

When you set up a mixed schedule, keep the lean part near half a minute. Too short gives little gain; too long wastes time. This 30-60 s window is easy to watch and adjust in any session. Try it next time you run reinforcement thinning or error-correction loops.