Why does Reward Based Training Work?

Welcome Back to Science Sunday

Using food to lure your dog into a sit position is one of the first ​training "tricks" we learn, but how does rewarding a dog actually elicit the desired response?
Humans often use a sort of experimental approach. The sequence might go like this - enter a dark room, flick light switch, room lights. The action of flicking the switch had the desired effect and we usually use the same approach when presented with the same situation, had it not worked we would try another solution. Psychologists, of course have a term for this; it's called "win-stay-lose-shift strategy".  If it worked keep doing it , if it didn't, try something else. Is this the strategy that dogs are using when responding to reward based training?
To find out, Molly Byrne, and a team of researchers at the Department of Psychology, Boston College looked at this very question. A group of adult dogs (323) were taught that if they touched an inverted cup they would receive the treat hidden underneath. Next they were presented with two cups located equal distance from the dog and the dog was restricted to selecting only one of the cups per trial. The paper cites previous work by the same team to eliminate olfactory cues. On subsequent runs it was recorded whether they chose the same cup or the other cup and whether the previous run was successful. To eliminate the possibility of previous conditioning, the experiment was repeated with 334 puppies.
Quoting from the study:

• We find initial evidence that the majority of puppies and adult dogs are using a win-stay-lose-shift strategy informed by their previous choice in contexts where they lack information about the location of a hidden treat. Specifically, both adult dogs and puppies were more likely to shift to a new location when they were not successful on their previous trial.

 Specifically 66% used the win-stay strategy  when there was a reward on the previous trial and 45% used the lose-shift strategy if there was not one in both puppies and adults.
Our takeaway from this should be the importance of treating every time when trying to establish a behaviour to avoid the dreaded "lose-shift" side of the equation.
"But Pat, wait" I hear you saying  "you always say you are a Balanced Trainer,  have you suddenly joined the pure positive crowd."
Not at all. Rewards are useful for establishing a behaviour, but once the desired command is learned and established its time to start phasing out the treats in lieu of praise, occasional treats and discipline. Its like gambling, the gambler is not rewarded every time, but maybe next time.  Knowing when and how is where the professional part of professional dog training comes in. What do you think?

---

References:	Evidence for Win-Stay-Lose-Shift in Puppies and Adult Dogs , Molly Byrne, ​Department of Psychology, Boston College
	Why Does a Reward During Training Change a Dog's Behaviour?, Dr. Stanley Coren, Psychology Today, August 06, 2020