Slot machines pay out on a variable schedule. People play them for hours. If every pull paid out a dollar, they'd pull a few times and walk away — where's the excitement? But when the payout is unpredictable, the behavior becomes almost compulsive.
Dogs aren't gamblers, but the same principle applies to reinforcement.
The Two Basic Schedules
Continuous reinforcement: Every correct response gets a reward. This is how you teach new behaviors — the dog needs consistent feedback to understand what's being asked. Think of it as the learning phase.
Variable reinforcement: Rewards come unpredictably. Sometimes after one correct response, sometimes after several. The dog doesn't know when the next reward is coming, so they keep offering the behavior to find out. This is the maintenance phase.
The mistake most owners make: they stay in continuous reinforcement mode long after the dog has learned the behavior. Then they wonder why the dog stops performing as soon as treats disappear. The answer is that they've never transitioned to a schedule that builds persistence.
Why Variable Works Better for Maintenance
When reward is continuous, its absence is immediately noticed. The dog does the behavior, no treat appears, and the dog's internal model says "this doesn't work anymore" — behavior starts fading within a few unrewarded trials.
When reward is variable, no-reward trials are just part of the normal pattern. The dog's internal model says "sometimes I have to do this several times before the reward comes" — so they keep offering the behavior.
This is why advanced obedience trainers can maintain incredible behavior with relatively infrequent food rewards. The behavior has been put on a variable schedule, and the occasional high-value reward is enough to keep it sharp.
How to Transition to Variable Reinforcement
The transition should be gradual. Don't go from rewarding every sit to rewarding every fifth sit overnight — the dog experiences that as the behavior stopping to work, and it can damage the training relationship.
Instead:
- Start rewarding every other correct response (VR-2). The dog notices nothing different.
- Expand to random 2-4 responses between rewards. Average of 3 correct responses per reward.
- Gradually extend the average to 5, 7, or even 10 responses per reward for well-established behaviors.
- Vary the size of the reward, not just the frequency. Sometimes a single treat. Sometimes a jackpot — five treats delivered one at a time. The unpredictability of how much is as reinforcing as the unpredictability of whether.
The key with fading treats in dog training is keeping the behavior strong throughout — if errors increase significantly, you've moved too fast.
Practical Application
For a dog with a solid sit: instead of treating every sit, treat randomly. Sometimes after one sit, sometimes after three. The dog stays sharp because they're invested — the treat might come this time.
For a recall: in daily life, only occasionally treat the recall. But when you do treat, make it a big deal — cheerful voice, multiple treats, genuine celebration. The variability keeps the recall reliable; the occasional jackpot keeps it enthusiastic.
One important note: don't use variable reinforcement during the learning phase of a new behavior. That's where continuous reinforcement belongs. Variable schedules are for maintaining what's already learned. (See also: positive reinforcement dog training.)
The goal isn't to stop rewarding your dog. It's to make rewards unpredictable enough that your dog keeps working for them reliably. Random is more compelling than certain.