"When can I stop using food?" is a question "treat-slinging weenie" trainers get asked a lot. So much so that I blogged about it here a while ago: Clicker Training Works Okay, But When Can I Stop Using Food?
Recently, I was asked to clarify when and how to phase out food for basic training accomplishments. Hence this blog post.
Some people don't bother to phase out the food because they don't see any reason to. There's really nothing wrong with carrying treats around and "paying" your dog for a job well done as it happens. This is especially true when working with dogs with reactivity or aggression issues. These dogs may need the extra motivation food gives to be able to maintain control in the presence of their "triggers" (other dogs, strangers, etc.). Then there's the added bonuses: food works to countercondition the dog's emotional response (changing "Oh no! another Dog!" to "Another dog! Food opportunities!") and eating often has natural calming effects.
For basic training, I use food because it's a great reward (reinforcer) with a lot of benefits, including:
- Dogs need food to survive so they have a built-in desire to work for food.
- It's fast so you can do a lot of repetitions in a little bit of time (a pea-sized treat is gone in less than a second, whereas even a quick game of tug takes longer).
- You can use a food treat as a lure -- the dog follows the food (e.g. into a "sit" position if you use the food like a magnet to bring your dog's nose up and back -- the back end of the dog usually responds by going down) and then gets the treat when they're in the right position. (Note: you should quickly fade the lure so the dog's doing the behavior without the treat lure, but that deserves a discussion all its own.)
- You can easily "feed for position" -- if you're doing a lot of quick repetitions of "Sit," for example, you mark the sit (with a click or a verbal "Yes!" or "Yip!") and then toss a pea-sized treat onto the floor to encourage the dog to get up to get the food. Then you can cue the "Sit" again right away. Result: 15 repetitions of "sit" in a minute, and you've just done a LOT of training in a tiny amount of time with a tiny amount of food. If you're working on the duration of "Sit," you "feed for position" by delivering the treat directly to the dog while they're still sitting.
- You can use a variety of foods for different results. High-value food (stuff your dog goes ga-ga over) can cut through a lot of distractions, making it possible for your dog to perform in distracting or stressful environments. It can also encourage top-notch performance -- really fast recalls, prompt sits, etc. Low-value food (often including your dog's kibble if you feed dry food) is great for rewarding easy behaviors, especially when you don't want the dog so focused on the food that he's not really learning what you're trying to teach. Plus using a variety of different food treats can help keep the dog from getting bored with "the same-old same-old."
For those who don't want to rely on food (for whatever reason), or who need to phase out food because it's not used in competition or their dog's on a special diet that makes using food more difficult (people feeding their dogs on the BARF diet don't want to carry "raw & meaty bones" around with them everywhere!), food is best used when first teaching a behavior. Once the behavior is "fluent" (it happens promptly on cue and the dog does it well each time), you can phase out the food in a number of ways.
- For "duration" behaviors (like sit-stays or down-stays), there will be a natural phasing out as you're asking the dog for longer durations. So when first teaching a down-stay, you'll "pay" the dog for every 2-3 seconds of down-stay. That quickly gets stretched into 10-second down-stays, then 20-second down-stays, and so on, until your dog's staying for several minutes (or however long a down-stay you need) in return for a single treat.
- Once a behavior is fluent, you can ask for 2-3 behaviors in a row before "paying" your dog with a treat. You can also "mix and match" behaviors, for example asking for "Sit" and then "Touch" (dog touching nose to your hand or to another target), then "Sit" again before saying "Yes!" and feeding a treat. Likewise, when you cue a complex behavior "chain" (several behaviors in a row, that are done in continuous motion) such as "Recycle!" (take this junk mail from my hand, trot over to the paper recycling bin, drop the junk mail in) the treat comes at the end. Of course you have to train each behavior separately and also teach the dog the proper sequence in the chain... but then you only reinforce after the dog's performed the entire chain.
- For many trained behaviors, use "life rewards" instead of food. These include "sit before we go through the door" (being released to go through the door is the reward for the sit), "walk on a loose leash before you go sniff that really interesting tree" (the "go sniff!" cue gives the dog permission to stop and sniff, reinforcing the loose leash walking by your side), and the very important "come to me even though you're playing with another dog and I'll let you go back to play with that other dog" ("Come!" --> "Good Dog! Go Play!"). You use what the dog really wants to be doing anyway as a reward for doing what you ask.
- For teaching impulse control and "on/off switches" (Go Wild/Freeze game), play is a great reinforcer! A game of Tug of War is the reinforcement for the "Take!" cue. Say "Take!" and offer the tug toy, followed by 10-12 seconds of tugging before you either let the dog win -- who wants to play if they never win?!? -- or asking the dog to "Give!" before saying "Take!" and getting back to the game.
- You can also switch to a "variable reinforcement schedule." In simple terms, this means you only reinforce the behavior some of the time rather than all of the time. This can actually strengthen behavior, as Casino owners know very well: video slot machines variably reinforce you when you push the Play button. You can further improve behavior by reinforcing the best responses to your cues with food or life rewards while still marking the average behavior with a "good dog."
- And then here's the mind-blowing part: when you've used a lot of reinforcement (food and play and other "life rewards") to train a behavior until it's fluent and reliable (the dog does it well and every time you ask), you can then use that behavior as a reward for other, less well trained behaviors! This is due to the "Premack Principle." Simply stated, the Premack Principle says that more probable behaviors will reinforce less probable behaviors. With a strong "reinforcement history" (the dog has been reinforced a lot for that particular behavior, in many different circumstances) the behavior becomes intrinsically rewarding for the dog. Start with the dog's "default" behavior -- what the dog will do when he's not sure what else he should do. The "default" behavior will often be the first behavior you teach, such as "Sit." Or it will be the one you practiced the most (assuming you used lots of high-value reinforcers). The more behaviors you build to fluency & reliability, the more behaviors you can cue to reinforce the behaviors you're still working on!
It's also important to remember that even when fading food, you shouldn't fade rewards! After all, how would you feel if your boss cut your pay every time you got good at a specific aspect of your job?
The good news is that your dog will find your praise and your look of pride and happiness rewarding, especially after you've built strong reinforcement histories through the use of food and play and other life rewards. Your dog really will begin working for you -- but only if you're generous with your feedback that lets your dog know "Well done!" Never take your dog's behavior for granted, and always show that you appreciate his cooperation!