Originally published in December 2012; expanded and revised for 2019. The video in this post was featured at Tate Behavioral’s ABA Conference in October 2019 by Dr. Megan Miller.
A poisoned cue is a cue that is associated with both reinforcing and aversive consequences. Poisoned cues were probably the norm for a period in some types of training, and still are common. If you tell your dog to “sit” and he gets a cookie if he sits but gets a push on the butt or jerk on the leash if he doesn’t, then “sit” is a poisoned cue. The term was coined by Karen Pryor.
It’s only once you start training without pain and discomfort that you and your dog find out about “unpoisoned” cues. In positive reinforcement-based training, a cue is a signal that means reinforcement is available for a certain behavior. Cues become associated with very good things. There is far less stress attached for the dog, since a wrong guess doesn’t carry a painful, uncomfortable, or scary penalty. You can see the joy on the faces of dogs who are trained with positive reinforcement when they hear a familiar cue.
Getting Rid of Summer’s Poisoned Cue
I am proud of the training I share in this post and video. I made a plan and successfully replaced a poisoned cue for my dog Summer, which improved her quality of life.
Summer, who passed away in 2017, was my crossover dog, and unfortunately carried some baggage from that. We took three levels of obedience courses at a traditional dog club. Although I had searched for positive reinforcement training after reading about it on the Internet, after looking around locally I figured maybe it didn’t actually exist in the real world. In the dog club I found a place where people cared for their dogs very much and were interested in training them, just not completely how I wanted to train (and sometimes not at all how I wanted to train). I did not partake of the harsher methods at the club, but as most of us know, even the less physical practices such as stern voices and physically pushing them into positions can be hard on lots of dogs.
How the Cue Got Poisoned
Summer was quite sensitive to pressure on her personal space. (Even Zani knew this. She regularly got Summer to move away from a desired spot just by lying down right next to her.) So when I taught Summer to stay by issuing a stern STAY “command” and rushing my open palm to within an inch or two of her face, it was unpleasant for her. In addition, the stay exercise itself was probably scary. She was a 10-month-old stray when I got her and our bond was still tenuous. Hearing all the other trainers raising their voices at their dogs, and particularly my leaving her and walking away while she was surrounded by this activity, were also hard on her.
I’ve come to realize over the years just what a difficult and unnatural behavior “stay” is. I admire Summer for learning it in that environment.
Here is a short YouTube video showing an AKC competition group sit stay. Watching it will explain the cause of Summer’s stress better than I ever could in words. The stay cue happens at about the one-minute point, but it’s worth watching the rest. Imagine if this environment was your dogs’ introduction to the stay behavior! And all the problems are exacerbated if your dog is small, since she can be in physical danger during the stay exercise. (The offleash version of the group stay has since been eliminated from AKC competitive obedience because of safety concerns.)
So even after I crossed over to positive reinforcement-based training, Summer responded to the stay cue with a flurry of stress behaviors, even though she did stay. For years I tried half measures to rehabilitate the cue. I tried to counter condition the verbal cue with pairing with treats. I softened my voice practically to a whisper. I decreased the hand signal until it was barely a twitch. But I still got reactions to the cue and hand signal, and also she consistently showed stress when I turned my shoulder away and left. I tried to counter condition that too, but the cues kept bringing the stress back.
Pretty clearly this was a poisoned cue. In positive reinforcement training, a cue signals that reinforcement is available for a certain behavior. A poisoned cue is one that has been taught with a mixture of reinforcement and an aversive, either deliberately or accidentally. So even though I used treats to reinforce Summer’s stays, both the verbal cue and the hand signal were poisoned not only by their intrinsic unpleasantness but by their association with what was initially a frightening experience for Summer. The cues came to mean, “Brace yourself because bad stuff is happening, and by the way you get a treat at the end if you make it that far.” I have written about my own personal experience with a poisoned cue here.
In addition I think Summer’s physical reaction, which usually included an abrupt drop of her head, was not only born of stress but had become a superstitious behavior as well. A tough combination to try to fix.
I finally understood, with some discussion and encouragement from my teacher, that I would have to re-teach the behavior with a new cue. Changing cues is one of my unfavorite things. Who wants to change their habits? But that is exactly what we are asking our dogs to do every time we train them. After watching Kathy Sdao’s wonderful “Improve Your i-Cue” DVD, I realized just how unfair I was being by not wanting to change my own behavior. Cues are comparatively easy for humans. We have language, and we usually choose a word related to the behavior (although we don’t have to). For dogs, learning verbal cues is an exercise in pure memory, using a sense (hearing) that they don’t lead with. They take in information more readily with smell and sight. So Kathy convinced me that not wanting to change my own habit was pretty selfish.
I picked a new verbal cue: “Hang out.” I wanted something that had relaxed connotations for me so I could always say it very pleasantly. I said it in a high, singsong voice.
Retraining to Replace the Poisoned Cue
When retraining, I couldn’t use the “New cue/old cue” technique of teaching a new cue. This is an application of classical conditioning in which you repeatedly give the new cue just before the old cue. (Then reinforce the dog for responding with the right behavior.) The dog starts to make the association and “anticipates” the old cue by performing the behavior. Pretty soon you can pause after the new cue and the dog may respond! You can fade out the old cue as the dog responds earlier and earlier, finally responding to the new cue by itself.
But if I had done that, I would have brought the bad feelings attached to the old cue right along for the transfer. Instead, I had to start from scratch. I had to change or rehabilitate three things: the hand signal (I dropped it completely), the verbal cue, and the motion of my turning away from her. Since Summer already had a very nice stay, with distance, duration, and distractions, I wouldn’t need to retrain the advanced stuff extensively. I just had to build good associations with the cue itself and my initial leaving.
I applied classical conditioning in a different way. I associated her new cue with a great treat. I did cue/treat/cue/treat/cue/treat while she was in a sit or down. I did this for several sessions and a few hundred reps. I was serious about this rehabilitation. I wanted the cue itself to have great associations. This wasn’t pure classical conditioning since I did have the one contingency on her behavior: she must sit still. But realistically, what dog is going to move away when you are shoveling treats at them like I was?
Then I applied desensitization. My turning away from her was a big stressor, so I started with the tiniest piece of a turn away without moving my feet (then treat!) and ever so gradually working up to taking a step away. If I got a stress reaction, I knew I had gone too fast and would go back to an earlier increment that didn’t bother her. Only after I had done these things for many sessions over several days did I start re-teaching stay in the usual way including adding the new cue.
We had a huge history to overcome. Her old reaction was such an entrenched behavior that pieces of it still crept in, even when “Hanging out.” I continued to do conditioning on the verbal cue and especially the turning away, which seemed to be the strongest trigger. This was a particularly difficult situation to rehabilitate, since there were so many longstanding triggers for her stress.
Changing a cue, even a poisoned one, is not always this extensive a project. I also had to replace her “Leave it” cue and that was much easier. I just stopped saying those words, trained the behavior per the Training Levels, and started attaching our new cue. The whole experience was so dissimilar to what we had done in obedience class (“You better believe it!” said Summer) that her old stress didn’t get triggered. I offer that as a ray of hope to the folks who watch the video and think, “Hundreds of repetitions? OMG!”
The video compares Summer’s “stay” behavior from some older clips from 2009, complete with stress reactions, and her new “hang out” behavior in 2012. In the second half of the video I show some 0f the training.
By the way, I trained Zani and Clara without using an explicit cue for stay. Sit, down, and stand all mean to stay in that position until released. But since I trained Summer with a stay cue initially, I think we were both more comfortable using one.
Anyone else have to replace a poisoned cue or have to rehabilitate certain aspects of a behavior?
Copyright Eileen Anderson 2012