by Karen Pryor
Click and wait?
Discussion from the Berkeley ClickerExpo from Karen Pryor’s Mail
Here’s an interesting clicker training issue that sparked a lot of discussion in Berkeley.
One of the great benefits of using a conditioned reinforcer, such as the clicker, is that it allows you to identify and strengthen a precise behaviour even though you cannot give food at that particular instant. The click functions as both a marker signal, identifying movement, and as a bridging stimulus, bridging the gap in time and space between action and reinforcer. That is the very basis of all the wonderful shaping we can do with the clicker. We can reinforce in the middle of a jump, or at a distance. We can catch the flick of an ear, a lifting paw, a tiny shift in weight, that will be past before food (or some other reinforcer) can be delivered. We can even click the moment an animal makes a good decision: turning away from a temptation, controlling an urge to jump up.
Then the food comes.
But what if the food, or other reinforcer-earned, expected, anticipated-doesn’t come right away? What happens when you click for a behaviour, touching a target, say, and then delay the presentation of food? What does the animal do? Does it matter? Is there a length of delay that is ’too long?’
In Chicago, Jesus Rosales-Ruiz and his students reported on a phenomenon they have been studying in sheep. What Jesus and his students did was to train sheep to touch a target for a click and a treat. Then they instituted a delay. After the click, the student waited five seconds, motionless, before presenting the food. What happened? The animals began exhibiting a new, untrained behaviour such as pawing the ground.
As a biologist I would call this ’displacement behaviour.’ Displacement behaviour is an activity that arises when an animal is in conflict: wanting to do something but unable to do it. Displacement activities are usually related to either feeding or grooming. Horses and cattle nose or paw the ground. (In humans, nail biting or scratching the head when perplexed-grooming-related activities-might be considered forms of displacement behaviour.)
Jesus and his students found that this behavioural response was reversible. Reduce the delay to less than two seconds, and the animal waited attentively. Put the delay back in, and the intervening behaviour arose again.
So that’s what happens in sheep. Would it be the same with our dogs? After the Chicago ClickerExpo, Dr. Rosales Ruiz undertook additional research with the help of two well-known clicker trainers, Sherri Lippmann and Virginia Broitman, authors of the Bow Wow series of videos. Sherri and Virginia felt that their highly experienced dogs would not be fazed by a mere five-second delay. After all, in real life delays occur occasionally. Sometimes we give a click when the animal is far away, and it has to come back to us for a treat. Sometimes we see a great behaviour and give a mouth click, but we’re not quite ready and have to rummage through the bait bag, or even jump up and go to the kitchen, to get the payoff.
What would happen, then, if in a series of trials after each click you just paused for five seconds? Broitman and Lippman did that. Wow! Even with experienced clicker dogs, displacement behaviour cropped up right away, and included barking, as in “What on earth is wrong with you, where’s my treat!” Delay, even for five seconds, had a huge effect on behaviour.
While it will of course be a while before the research is completed and the report written up and published, these initial findings have some useful implications for us. It’s true that the click is a bridge, a stimulus that bridges the gap in time and distance between behaviour and reinforcer. But, as KPCT Company president Aaron Clayton puts it, it is not the Golden Gate Bridge. Dr. Rosales-Ruiz points out that the tie between click and treat is classical conditioning, not operant conditioning. The actual pairing of the two, the conditioning, is only maintained by repetition. As clicker trainers, he says, we should respect that.
If you routinely click for a behaviour, and then fumble around in your bait bag making the dog wait while you find a suitable treat and hand it over, you are taking a chance. You risk seeing a rise in undesired or superstitious behaviour, such as wiggling, panting, getting up or lying down repeatedly, sniffing the ground or looking away. You take a risk that that ’hand moving toward baitbag’ becomes the reinforcing element, making the unwanted behaviour harder and harder to get rid of. You weaken the effect of the clicker as bridge (and while there’s no research evidence yet, personally I betcha you weaken the effect of the clicker as marker signal, too.)
Kathy Sdao, in her Clicking with Class sessions at ClickerExpo, sometimes makes everyone transfer a handful of beans from a paper cup to their treat hand, and then dispense them, one by one, back into the cup, as treats. Can you do that? Can you do it fast? Kathy’s aim is to see to it that you learn to keep your treats ready and in your hand before you start a training session. And that you learn to restock the supply in your hand between behaviours, NOT after you’ve clicked and while the dog is waiting for a reward.
It’s not that you can’t have a gap; it’s just that there are sound, demonstrable reasons why the careful trainer does everything possible to eliminate prolonged, unnecessary, and habitual gaps.
And what about those dogs and dolphins and other beasts that work at great distances and for long durations without tangible reinforcers? Yes, you can do that, but that capacity is trained; distance and duration are extended and strengthened by other conditioned reinforcers such as additional cues; not solely by the power of the click. A good topic for another day!
With Kind Permission from Karen Pryor: Many more wonderful articles can be found at www.clickertraining.com