Category: Getting Started ¤
  Author: Karen Pryor ¤
   Title: Stimulus Control ¤

A new clicker trainer posted about her very understandable confusion. Her
dog now throws dozens of behaviors at her, but won't give his old, fairly
reliable "stay." What to do? 

The dog is "telling" us that he knows some behaviors, but he doesn't know
any cues. Even the old "stay" command doesn't mean to him what it means to
the owner. To him it means lie down, but if that doesn't work, jump up and
try something else.

This dog is in the first stages of learning clicker training (owner, too)
and coming along great! but now it's time to establish a few cues. Then the
owner can sort out what she's asking the dog for. I would start with his
very favorite behavior, whatever that is, and begin by clicking the
behavior a few times, (and treating) and then by offering the cue; clicking
the behavior. treat. Silence. Let the behavior happen. Do nothing. Offer
the cue INSTANTLY (before the behavior happens again) and let the behavior
happen and click and treat. Alternate between cue/B/CT and no cue?B/ no C/T
four or five times. Quit. 

Try again in a couple of hours. Learning the first cue is Pavlovian
conditioning, so it's slow and unconscious (see Furbish video) but after a
while it will get easier. Now on maybe the third session, you hope the
animal will pause, after offering an off-cue behavior. Give the CUE as a
reinforcement for pausing. I.e. shape the behavior of waiting for the cue.
(See the whippet/foxhound puppy on Clicker Magic.) Now you are getting
somewhere, and pretty soon you'll have your first cue.

Now shape longer downstays, without a cue, just by starting with tiny
increments of 1 sec or so, building to 15 or 20 secs before c/t. Then add a
cue, NOT the old Stay cue, a new one, and you should have a pretty good
stay come back fairly quickly. SHAPE the cues for your other behaviors, and
you'll be miles ahead of where you started.=20

Some people use the word wrong when the dog offers a behavior in the
absence of the cue. I believe this should be saved for later, when the dog
already understands the concept of cues, rather than used as a training aid
during these very early stages, when it can become discouraging to the dog
(and frustrating to the owner) because there is still at least 50% off-cue
behavior.

I am adding below the new handout on stimulus control that people will be
getting at the seminar this weekend. The four criteria are the same, from
DSTD, but there's a new paragraph of explanation in front. To me, the whole
business of Stim. control is exactly as important as the shaping. it is
half of training. But it is not as easily come by, or as instantly
reinforcing, as shaping. Hang in there, when you have built a dog that is
truly cue-wise, and learns a new cue in a minute or two, it's just magic. KP

"Reliability," or Perfect Stimulus Control

A conditioned stimulus, or cue, is different from a command. A command is,
essentially, a threat: do this, or else.  If the dog does the behavior, we
may reward, but if the dog fails to obey the command, we are obliged to
correct or coerce or lure the behavior into happening, or face more
failures or avoidance in the future.

A conditioned stimulus, or cue, on the other hand, is a sort of switch: it
turns on a specific behavior. The dog knows that it can voluntarily give
behaviors that will lead to reinforcement. The cue simply identifies which
behavior will work at this given moment. If the dog fails to do the
behavior, then we know some aspect of that cue has not yet been fully
conditioned. We will then need to set up a situation in which we can
reinforce successful responses to that particular criterion.

To get good stimulus control, pick one or more simple behaviors, and
methodically establish the criteria for each cue, treating each aspect as a
separate shaping program. You may also need to take time to reinforce
responses to each aspect of the cue, under many different circumstances and
in new and distracting environments. However: Once your subject has learned
to respond swiftly and correctly to three or more cues, conditioned in this
careful manner, it will learn to generalize: a cue is a cue. You will then
be able to teach new cues, and to get instant, correct, reliable responses,
much faster: often in minutes. The "rules" for cues ("identify the cue, do
the behavior now, do it right, don’t do something else, don’t do this new
behavior on some old cue") will transfer rapidly to new cues. Compared to
the traditional method of teaching and enforcing each command and its
attached behavior separately, establishing an understanding of stimulus
control is a far more efficient way of building a large repertoire of
reliable behavior.

The Four Aspects of Perfect Stimulus Control

1.  The behavior always occurs immediately  upon presentation of the
conditioned stimulus. When you say "Sit!" the dog sits immediately. 

2.  The behavior never occurs in the absence of the stimulus. While working
or training,  the dog never sits if you didn’t say or signal "Sit!."

3. The Behavior never occurs in response to some other stimulus. When you
say "Down!" the dog never sits, instead.
 
4.  No other behavior occurs in response to this stimulus. When you say
"Sit!" the dog does not offer another behavior such as lying down, or
standing there staring, or licking you.


Karen Pryor   May, 1997

Niki