Good job! That’s great!
Yes, good boy.
Do you know these sounds? They’re the sounds of secondary reinforcement. They’re what we often use before—or in place of—giving a nice scratch on the withers or a piece of carrot. And they’re meant to show our horses we’re pleased with them.
The question is: Do they work?
Results from a new study by French researchers suggests they might not be. In a neutral field setting with researchers following a set protocol, secondary reinforcement just wasn’t effective. While results might differ between horses and trainers, the fact is, science can’t currently confirm that horses learn from audio rewards.
“In theory, it should work,” said Léa Lansade, PhD, of the French Horse and Riding Institute and the National Institute for Agricultural Research’s behavior science department, in Tours. “But for the moment, no scientific study has been able to prove that secondary reinforcement—such as saying, ‘good’ to the horse when he’s done the right exercise—can substitute a food reward to motivate him to do the requested exercise.”
In their study, Lansade and Ludovic Calandreau, PhD, tested the effects of secondary reinforcement on a group of 14 yearlings in a three-part study. In the first part, they taught all the horses to associate the word “good” with receiving food. This is known as Pavlovian conditioning—teaching the animal that a particular sound (originally, for the scientist Anton Pavlov, it was a bell) is associated with true positive reinforcement (such as a food reward). Each horse received 288 associations of the word “good” with a handful of grain pellets.
In the second part, they taught all 14 horses to recognize a hand gesture to receive a food reward. The horses had to touch one of two traffic cones according to the trainer’s gesture indicating which cone to touch. Each time the horse chose the indicated cone, the trainer gave the horse handful of grain pellets.
In the third part, the researchers randomly divided the horses into two groups. They repeated the same cone-touching session—but this time, they did not give food as a reward. Instead, half the horses heard the word “good” when they touched the correct cone. The other half heard nothing, and the researcher only stood motionless with a fixed gaze after the horse touched the correct cone.
The secondary reinforcement group—those that heard “good”—did no better in choosing the right cone than the control group that heard nothing at all, Lansade said.
“But that doesn’t mean that’s the way it would be for all horses in all situations,” she said. “Riders should try repeating our study with their own horses. It’s possible that in a different environment and with different horse-human relationships, the results would vary dramatically.”
Lansade said her study results would probably have been similar if they had used clickers instead of the word “good,” because both are auditory signals representing a food reward. But again, clicker-training’s efficacy might differ under circumstances outside of the “sterility” of the scientific study setting.
The current study doesn’t mean people should stop saying “good boy” or using verbal praise with their horses.
“They might as well continue, because it could still be useful with their own horses,” Lansade said. “If we want to be sure that we’re motivating our horses enough to learn, we should probably be adding the food reward, because our study indicates that the secondary reinforcement alone doesn’t seem to be sufficient.” (Editor’s note: Learn more about treat-training while riding at TheHorse.com/18144.)
She admitted that the study results surprised even the researchers themselves.
“I teach about how secondary reinforcement works and that it can be effective,” she said. “So I really hoped this research would confirm that. But in the end, we couldn’t show anything positive about it. But these are the results of a neutral scientific study. I would love to see the kinds of results people might get in their home farms with their own horses, because it could be quite different.”
The study, “A conditioned reinforcer did not help to maintain an operant conditioning in the absence of a primary reinforcer in horses,” was published in Behavioral Processes.