Stimulus generalization, discrimination learning, and peak shift in horses.

Researchers taught three horses to press a panel for grain. They used a light cue to signal when presses paid off. Next they added a second light that looked slightly different. Only one light still led to grain. The horses had to learn which was which.

After the horses chose the correct light every time, the team tested new lights. These lights varied in tiny steps between the trained ones. They wanted to see if the horses would still press when the light was close, but not exact.

The horses picked the correct light almost every time. When tested with the in-between lights, they pressed most for a light that was even further from the non-paying cue. This 'peak shift' matches what pigeons and rats do.

The data show horses build stimulus control the same way smaller lab animals do. Size and species did not change the basic rules.

Hendry et al. (1969) showed that old cues can block new ones from gaining control. They first trained rats with a noise cue, then added a light. The light never gained power because the noise had already 'claimed' the job. Davison et al. (1991) did the opposite: they let the new cue win. The difference is timing. Blocking happens when the first cue is already strong. Peak shift happens after the animal has learned both cues and you test in-between versions.

Burgess et al. (1971) mixed two cues that told rats to press at different speeds. The rats settled on a middle speed. M et al. found a different kind of middle: the horses shifted their peak away from the bad cue. Both studies show animals do not average cues blindly; the payoff history tells them how to respond.

Baer (1974) tracked how stimulus control flips during fixed-interval schedules. Early in the interval the same light slowed pressing; late in the interval it sped pressing. That paper shows context can flip control. The horse study adds that even simple discrimination can push the peak of responding away from the non-reinforced cue.

If you work with larger animals or in equine-assisted programs, you can trust basic ABA teaching rules. Use clear SDs, reinforce consistently, and expect peak shift when you probe generalization. The horse data say you do not need rat or pigeon labs to check your procedures; the laws hold across barn and clinic alike.