Shaping Canine Potential: Differential Reinforcement’s Power to Create Complex Behaviors
Anyone who’s trained a dog knows about Pawsitive Reinforcement.
Positive = adding something (a treat, scritchies).
Reinforcement = anything that increases the likelihood that a behavior will increase.
When we first start training our pups, we’re typically using positive reinforcement on a 1:1 basis—sit-yes-treat; look-yes-treat; recall-yes-treat.
But when it comes to shaping more complex “behavior chains” (a series of behaviors that must be performed in precise order to receive reinforcement), trainers can use Differential Reinforcement.
Differential reinforcement = selectively reinforcing desired behaviors, while withholding or reducing reinforcement for undesired behaviors.
Differential Reinfocement has many applications:
DRI = Differential Reinforcement of Incompatible Behaviors. Reinforcing a desired behavior (let’s say sitting) that fundamentally prevents an undesired behavior from happening at the exact same time (let’s say jumping). Or playing “find it” instead of barking at a dog across the street.
DRA / DRO = Differential Reinforcement of Alternate/Other Behaviors. Reinforcing a desired behavior that’s not necessarily incompatible, just different than the undesirable behavior.
DRZ = Differential Reinforcement of Zero Behavior. Rewarding that nice, calm, relaxed dog who is doing nothing at all (instead of doing something they shouldn’t).
But the one I use the most is DRE.
DRE = Differential Reinforcement of Excellent Behavior. Rewarding the “perfect” sit over the sloppy side-saddle one. Rewarding the immediate recall over the take-your-time version.
DRE focuses on rewarding a dog for making choices that are closer to the desired behavior, and not reinforcing lesser versions. When training complex behaviors such as service tasks, agility sports, or working roles, DRE can make all the difference. It not only encourages dogs to learn more complex behaviors but also strengthens their ability to perform these behaviors reliably over time.
There are a few different ways to do it:
Quantity—one can “jackpot” a big handful of treats, whether higher or lower value, as a way to differentiate from a single treat. This works very well when “capturing” new behaviors. The sudden, unexpected jackpot creates immediate learning. Especially when coupled with outsized facial expressions and happy-talk, a jackpot unleashes maximum dopamine (just like a slot machine).
High value vs. low value—one can continue giving treats on a 1:1 basis, but the exceptional behavior gets a high-value treat (chicken), while the mediocre version of the same behavior gets a lower-value treat (kibble). This can be helpful when a behavior is beginning to develop. The pup still gets a reward every time, so they learn that even a mediocre performance is still a way to get a paycheck. But Rover also begins to learn that certain subtleties can lead to a bigger bonus.
Reward vs. no reward—this is best used when a behavior is already well-developed, because it requires precise timing. When a behavior is just being learned, missing the moment to give a reward can slow-down the overall learning of the new behavior. So it’s better to use one of the other techniques above early-on. Later, when the pretty-good version of the behavior is rock solid 100%, we can start to use reward vs. no reward to reinforce only the excellent version of that behavior.
Case Study: my talking dog Dude.
He’s a Husky-Shepherd, who speaks with the cutest low growl/howl (cute to me, scary to some). I wanted him to say “Ah-roo” with two syllables and a distinct high-low in the melody (more cute, less scary).
First I captured any performance of the low growl-howl “roo” by jackpotting it. A big handful of treats and an exaggerated “I love you” face one would use with a human baby. Compared to the normal one-treat delivery for other behaviors like sitting, leash-walking, etc…very quickly he learned that the treat vending machine pretty much explodes when he speaks “roo.”
Then, I started shaping. When he started speaking on cue (and even off-cue, which is OK with me), I started shaping the two syllable “Ah-roo” using high value-low value DRE. If he said “roo” he’d get a regular treat. If he said “Ah-roo,” he’d get a milkbone. To be fair, this is both quantity and high-low at the same time, since the milkbone is much larger than the normal treat.
Finally we moved to reward/no-reward. A regular “roo” just didn’t cut it anymore…sorry bud, no treat. Only “Ah-roo” with two distinct syllables gets a treat. He learned “Ah-roo” is the exact phrase that gets him a paycheck, and he really loves payday.
This same approach can work to teach a dog to retrieve a specific item. Through DRE, one would use quantity, high-low, and reward/no-reward to teach the dog (over time) that it only gets paid when it picks up and brings the item, rather than just touching it or holding it briefly. This method teaches dogs to aim for the top-tier performance by recognizing that the better their behavior, the better their reward.
For dog owners, employing DRE isn’t just about getting results; it’s about having a conversation with their dogs. Especially for pups who are motivating with other rewards beyond just treats (pets, scritchies, play). How much fun is it using different levels and quantities of these rewards to teach Fido new things. It’s possible to have 50 training interactions in just a few minutes—and it truly feels like having a conversation.
Whether you’re a seasoned trainer or a first-time dog owner, understanding and utilizing DRE can significantly enhance your training approach. It’s not just about teaching tricks; it’s about cultivating an eager, educated, and enthusiastic canine companion. By focusing on excellence, we can help our dogs reach their full potential in whatever tasks they tackle, making the training process both rewarding and enjoyable for everyone involved.
