Like kids with their Halloween candy, do dogs rank the treats that we provide to them? Many trainers selectively use what we call “high-value treats” for some behaviors and “low-value treats” for others. However, other than subjectively observing the level of our dog’s pleasure at receiving different types of treats, do we have actual evidence that treats vary in their influence upon learning? For example, does my Chippy learn faster or more efficiently when I use homemade tuna treats versus small, dry biscuits? Intuitively, I think that he does, since he almost climbs into the oven when he smells his treats baking (the only thing I am even remotely capable of producing in the kitchen, by the way).
Lucky for us, some scientists have already asked this question and are providing some answers.
The Study: Mariana Bentosela and her co-investigators trained a group of 13 adult (pet) dogs to offer eye contact using either a high incentive treat (dried liver pieces), or a low incentive treat (dry kibble pieces) as the positive reinforcer. The methods used to train this behavior are the same used in the study that I reviewed in the “Trick or Treat” blog of October 31. All of the dogs rapidly learned “gazing behavior” using both types of reinforcer and all consumed the treats that were provided, regardless of the type. Following the initial acquisition phase, dogs were rested for one hour and a second training session was conducted. In the second session, training was exactly the same, except that all of the dogs now received dry dog food pieces as reinforcers. The investigators refer to this as “downshifting” the dogs who had been previously trained using liver treats. In this study, the dogs who were trained in both sessions with dry kibble are the control group. Behaviors that were measured included the number of trials to acquisition, gaze duration, food rejection, and general behavior and communication signals.
DRIED LIVER (HIGH INCENTIVE) DRY DOG FOOD (LOW INCENTIVE)
Study Results: The researchers were interested in finding out if dogs, like several other mammals, would react to a change from a high-value to a low-value reinforce with a behavior that was initially taught using the high-value treat. Here is what they found:
- During the initial training phase, although all of the dogs learned to offer gaze, dogs trained using the high-value treat maintained eye contact for a significantly longer duration than did dogs trained using the low-value treat (23.1 vs. 15.7 seconds, respectively).
- During the second phase, dogs who were downshifted from liver to kibble rapidly decreased their gaze duration, while the control dogs who had been offered kibble in both trials showed an increase in gaze duration.
Dogs who had been trained with liver and then switched to kibble rejected the food on significantly more occasions when offered kibble than did dogs who were trained consistently with the kibble pieces. Most commonly, the dog would approach the treat being offered, sniff the treat in the trainer’s hand, and then walk away without eating it.
Take Away for Dog Folks: First, we must note that the study groups used in this experiment were quite small (6 dogs in the control group and 7 dogs in the experimental group), and that the behavior that was trained is very simple and is easily acquired by most dogs. However, even with these limitations, the results suggest that learning is influenced by the value of the positive reinforcer that is used and that dogs are affected by changes in anticipated reinforcer value. Dogs trained with high-value treats showed a stronger response (longer duration of gaze), but also demonstrated signs of extinction when suddenly switched to a low-value treat and even rejected the reinforcer!
Most trainers who pay attention to different value reinforcers use high-value treats when training a difficult or complex behavior. Typically, we pair high-value treats with the behavior that is most difficult or the part of a sequence that the dog is least motivated to offer. (If back-chaining, the last behavior gets the highest value treat). The results of this study tell us that, yes, using a variety of treat types can be helpful, and that yes, high value treats may enhance learning. However, the results also suggest that we might want to pay close attention to downshifting from high to lower value treats during training sessions. Dogs may not only notice the switch, but it can affect the stability of a previously trained behavior, at least when that behavior has been only recently learned and is not yet reliable.
So trainers – Continue to rank your treats and pay careful attention to the value that they have to your dog, both when you are using the high-value stuff and when you throw in that handful of dried kibble or biscuits now and then. Happy Training!
WILL WORK FOR HOT DOG.
Reference: Bentosela, M, Jakovcevic A, Elgier AM, Mustaca AE, Papini MR. Incentive contrast in domestic dogs (Canis familiaris). Journal of Comparative Psychology 2009; 123:125-130.