poisoned cues
It’s been raining for the past 6 days and I am tired of training in the house and sitting at my desk. Thank goodness we have a rain break today and we can get outside. I couldn’t think of any excuses for not getting all my office work done that I have been putting off, so that is what I have been doing all week. Scoop has had to make do with one wet walk in the rain each day, and playing at games here in the house. Jim and I have also watched a few movies; District 9 (interesting) and something else that must not have been too relevant as I can’t for the life of me remember what it was.
I also re-read Steven Pressman’s The War of Art, and I watched Dr. Jose Rosales-Ruiz’s dvd The Poisoned Cue. As yet the after effects from “war..” haven’t much been felt, since I still have unfinished writing projects this week, but I have a start since I am sitting here updating Scooby’s blog. If you have not read The War of Art, you should. My friend Sandy Rogers gave it to me, and I find if I read it once a year (should be two or three times) then my resistance to finishing projects diminishes greatly! It is such an easy read, and a joy to find the inspiration in his pages. Resistance to writing, or painting or saving the universe is futile for weeks after you finish this easy read.
Another inspiration this week was in watching Dr. Jesus Rosales-Ruiz Poisoned Cue video. Apparently Karen Pryor first coined the term, and I located her technical description of the poisoned cue on her website, but Jesus’s video was a lot more explanatory. Of course it is 2 hours and 44 minutes long, hers is just one page:) The poisoned cue is pretty much a description of what the term means. You take a perfectly well trained cue, and poison it using not very nice training methods like coercion or intentional or unintentional punishment, instead of positive reinforcement. You can poison a cue, like your dog’s name, or your recall command using punishment, and voila, your cue creates stress and displacement behaviors instead of a joyful responses to your cue. One newish negative buzz word Jesus discusses is command. I think that Karen and henceforth others, are saying that a command now has the connotation of a behavior which was taught not using positive reinforcement. I might argue that I can teach a command using positive reinforcement without poisoning the word command🙂 but anyway, we all know that cue is the new command, just like brown became the new black in clothing choice a few years ago. Whatever the terminology, mixing up the use of positive reinforcement with corrections in training can get you into big trouble.
It is possible to poison your cue, poison your training aids like leashes, or poison an environment. In our sport it is probably possibly to poison the entire competition agility ring. I know I have seen the start line, the bottom of the contact obstacles or weave poles poisoned many times. All you have to do is punish the dog for not staying in a 2o2o position at the same time you are trying to clicker train him to nose touch in the same spot, and voila! you can unknowingly poison the cue touch. Just saying the word touch could make your dog slow down, leap off the end, lick his lips, or head out to sniff the nearest bush. Correct your dog for coming out at the 10th pole, and in short order you can have a big weave problem getting your dog to go beyond pole 10. Punish the entry and your dog will stay as far away from weave entries as he possibly can. You have all seen the ‘exhaust circle” that some dogs do to avoid making an entry error. Poisoning can happen after just one correction, and poisoning can happen by accident. The dog who injures himself or gets frightened on a piece of equipement, can effectively poison the obstacle, and subsequently the training or show environment. You don’t need to beat your dog while training to get stuck in a poisoned quagmire, but the effects are the same.
Jesus’ video had some great demonstration of how cues are poisoned, and good ideas for getting out of some of the situations, like reshaping the behavior and totally changing your cue . What I really liked in the video is the method they used to train some simple behaviors using cues which were taught using positive reinforcement, to reinforce and teach other behaviors. The example was done to show the difference between a positively reinforced and a poisoned cue. One of his students demonstrated using a recall command cue to teach a new behavior. The principle being that any cue you have taught using r+ you could use entirely to teach another behavior. So I did my own little experiment with Scoop.
I decided to use his “break” (release from position) behavior to teach him to walk over to a chair and lie his head on the chair. The behavior and cue of Break would be used instead of a click. Scoop is normally rewarded for almost every response to break, so that fits the criteria that you can use a positively trained behavior, and henceforth the cue for that behavior, to reinforce another behavior. Scoop’s criteria for break is that when he hears the word he stops any behavior he is doing (like sitting) and comes to me and is rewarded with a game of tug or a treat.
This is how my Scoop experiment went. I have cookies and a clicker in my hand and am standing a few feet from a chair. Cookies and clicker in hand and me standing still are Scoop’s prompt for “game on let’s shape something”. He immediately offered to step towards the chair, I said break and when he turned to me I clicked and he got his cookie when he got to me. I was able to shape him going all the way to the chair one step at a time, and lying his head on the seat, in one short session. The skill of going to the chair was not difficult, but I really was surprised that rewarding him for break given at the same time that I would have clicked while shaping the behavior, would work that easy. Yeah, I saw it on the video, but till you do it yourself, it does not count:) The next day I tried using his name, another behavior which was taught with reward in a positive fashion, and I was able to quickly shape him to go to a wooden tub and put his head inside it. Again, it took just one session of a few minutes. I used his name at the same point that I would have clicked and it worked the same way. After I said his name as the replacement click, he came to me and I said yes and gave him a treat.
In the video they show a dog’s reponse to shaping behaviors using approximately the same methods as I described in my extremely shortened version, but at the same time they also showed how the dog responded to shaping behaviors with a recall cue which was taught with some coercion. The dog was recalled to the handler and if the response was not perfect, it was gently forced to come to the handler by pulling on it’s harness until the dog reached the handlers front, then the dog still received a click and treat after the punishment. This essentially “poisoned” that recall cue (turning it into a command). A totally different word was used for the cue for this forced recall to make it different from the positively trained recall cue, and make it obvious during the test when the command was used for shaping as opposed to the positively trained recall. Using the same methods described above using the cue to reinforce another behavior, they tried the same shaping experiment using the “poisoned cue behavior”. (no coercion was used while they shaped the new behavior) The results were obvious. Stress, some displacement behaviors while working, slow performance and an unhappy attitude were all easy to see in the dogs performance in the video. The dog did learn the skill through shaping using the negatively reinforced cue, but the difference in the performance of the behaviors was very evident.
I will definitely watch this video again, and I suspect I have not condensed and described it well enough to have you think you now know all there is to know about poisoned cues. The hostess on the video is Alexandra Kurland, and her commentary about horses while unnecessary for us dog trainers, does not interfere in any way with the interesting lecture and information that Dr. Rosales-Ruiz is sharing. You can find the video for sale on Kurland’s website, theclickercenter.com.
So you know this is not all I thought about or did today, here are some of the photos from our adventures on the ranch today. Every winter if we get enough rain (we did!) we get a big pond at the bottom of our property. The dogs love to play in the water. It is great exercise when they run through the 18 inch deep pond, I wish it would stay all year long. My photos aren’t as great at Marcy Mantell’s that I usually share with you, but they are the best I could do.
I hope it isn’t raining or snowing where you live but if it is, do some reading or watch the great Ruiz dvd to learn more about animal training.
Nancy
- let’s go, let’s go, let’s go!
- the pond
- cute Scoop!
- Panic in the front, then Ace and Scoop all stacked up
- Having a run
Jim pruning
Rammy and the ewes watching it all
Loading tweets...
Kory,
Since we don’t reward every single behavior or every part of a more complex behavior, I don’t believe not rewarding is true punishment as long as these behaviors or portions of behaviors have been positively trained, or are supported by other behaviors which are reinforced. The durability of a behavior however will depend on how carefully it was taught and reinforced, and if it will stand up to distractions and duration and delayed reinforcement at times.
The fluency of any behavior goes back to how it was taught and the history of correct reinforcement.
Thanks for writing,
NJG
I really enjoyed this post but still have one question — at what point does simply not giving a reward, become, in and of itself, a “negative reward marker”?
Some say, well, yes it is! but then how do the dogs know the difference between this and wnen we are just progressing further in training? Many obedience/agility students have wondered about this, especially when weaning off treats/toys before trialing (I see this especially if the student has been very verbal about “good try” then they don’t reward and now don’t even have a reliable verbal reward marker when the primary reinforcer is gone.
cool! I hope you enjoy it as much as I did.
NJG
Thanks so much for this post and the informative discussion afterwards. I have to admit to being guilty of poisoning cues. I ordered the dvd and am looking forward to being educated!
Great post. And follow-up discussion, too. Thanks.
“Scoop’s criteria for break is that when he hears the word he stops any behavior he is doing (like sitting) and comes to me and is rewarded with a game of tug or a treat. “
Is “break” actually a behavior, or is it another conditioned reinforcer, essentially identical to the clicker or the verbal CR “yes”? Any time we say “yes” or click, the dog is free to CEASE behaving (is ceasing the task actually a behavior?) and come get it’s primary reinforcer. So “break” sounds like it could be used interchangeably with “yes” or click, and maybe is not a cue for a new behavior, but more of a conditioned reinforcer that marks the END of behavior and the availability of a PR. Perhaps a better test of the theory would be to use a positively-trained behavior like sit, down, back up, rollover. Using a behavior that POSTPONED his return to you and thus postponed the primary reinforcer would be a more interesting test, since both “break” and his name signal him to come running straight to the PR.
“they also showed how the dog responded to shaping behaviors with a recall cue which was taught with some coercion. The dog was recalled to the handler and if the response was not perfect, it was gently forced to come to the handler by pulling on it’s harness until the dog reached the handlers front, then the dog still received a click and treat after the punishment.”
It’s been widely debated that this actually isn’t an appropriate use of punishment because in the actual experiment, the occurrence of the targeted (“punished”) behavior did not decrease, thus dragging the dog in like a fish on a line was not punishment in the scientific sense; it was simply very aversive to the dog. He never learned how to prevent it by coming faster, nor did he learn how to “turn it off” by offering to come after they had begun dragging him in, which would have been an example of –R. Also, getting a click and treat that is NOT contingent upon the dog performing the target behavior (he got the c/t even after being dragged in by the handler against his will) also doesn’t fall into the category of +R because it’s noncontingent.
A variation on this experiment that is very interesting is to teach a new behavior using only positive reinforcement until the dog demonstrates that he understands the target behavior and has a strong positive association with the cue. Then use positive punishment and/or negative reinforcement to fine tune it – for example, to eliminate looking away at distractions while heeling. A dog that is taught using +R until he has a strong positive association with the cue and behavior will not show the same signs of stress or ‘poisoning of the cue’ if aversive methods are later employed.
Charlene
Charlene,
Thanks for writing. Scoop’s “break” IS a very specific behavior. It is a recall to me. I taught it the same way I teach a recall. If I don’t want Scoop to come to me, I use a secondary release which is more conversational, both “Ok”, and “off ya go” allow my dogs to nick off and go play or do as they please. My marker word and click of course are used as event markers but may or may not end the behavior. At times my dogs are expected to continue a behavior through a click and the reinforcement comes to them instead of vice versa.
I used the break and recall cues to train the other behaviors as that was most similar to the behavior training which appears in Jesus’s video. YOu asked, “Is “break” actually a behavior, or is it another conditioned reinforcer, essentially identical to the clicker or the verbal CR “yes”?” After this training experiment I would have to happily say that yes, my recall and break commands definitely can operate as a conditioned reinforcer. And I would assume as demonstrated in the video, if I had taught these cues with coercion or some amount of aversion,they would not have operated as well as cr’s.
I am going to play with some other trained behaviors this week, like left, right and walk back to train another behavior. What I won’t be experimenting with is poisoning a cue so I can see if it has the opposite effect:)
I agree with your observation on the use of the term punishment. I assume you are speaking hypothetically and not advocating (in your last paragraph) that I use negative reinforcement or positive punishemnt to “fine tune” cues which were taught using positive reinforcement?
Nancy
Very interesting reading, going to order the video. Thank ewe for for writing. I have been thinking about you in all your rain. It’s raining here too!
love this post. i have never understood why some people use a negative marker. i had an instructor who did that and she would be frustrated that i would not mark unsuccessful attempts. it did not make sense to me, and when i saw karen pryor at a dog show i was lucky enough to have a few minutes to ask her about negative markers, and that is when i first heard the term “poisoning the cue”. no matter how nicely said, its still a nm and does more damage than good, in my opinion.
on a similar vein, i am driven crazy when i see people continually go back and try to get their dogs to finish weave poles when they’ve popped out. i truly believe that it doesn’t teach them anything….it just helps them get their Q. any thoughts on this?
I think every dog is different and each situation is different when a dog does not complete weaves. When training I tell the handler to ignore the error and have the handler continue forward motion past the end of the weaves. If they want they can turn around and weave in the opposite direction to try and get the dog to forget the error going the other way. I never “correct” the incomplete poles. In the ring it is hard to pattern yourself to ignore a weave error and go on. Only if you are suspecting that the dog may come out will you be prepared to comletely ignore the error. Usually we just react, by stopping, and that can pattern the error to be repeated.
Great post Nancy!