I’m an auditory person. I grew up in a whole family of musicians. I love language and sound and music of all sorts. For the first half of my life I lived almost exclusively in the company of musicians, and in the second half I have few musician friends. This second half has made me conscious of the ways we musicians are different!
At work, I’m the one who gets asked to check the voice mail when we can’t understand the name. When there is a strange noise, eyes turn to me for identification of it. I can almost always tell if someone can’t hear me well, and I can immediately tell if someone is not listening to me (these are very different!). I can also ignore visual stimuli very well if I am listening to someone or something. (With regard to visual skills, I’ve been told someone could come in and rearrange the furniture in my house and I wouldn’t notice….)
So naturally I am interested in my dogs’ perceptions of sound and verbal cues. Frequent readers will know that I am honest about my limitations and frailties as a trainer, so I think you all will believe me when I say that ironically I seem to have three dogs in training who all have rather low aptitude for verbal cues. I.e., it’s probably not just my limitations in this case. So we all have to work extra hard on words.
Verbal cue discrimination training, where you teach a dog to respond only to the correct word, can be stressful for any dog. If not done with care, the dog can have a very high error rate, which is discouraging to many dogs. So I gave a lot of thought about how I could reduce that error rate.
In A Secret for Training Two Dogs I described my strategies for teaching one of my dogs to stay on a mat while I trained another. I briefly discussed my methods for releases. I have chosen to use each dog’s name, spoken in a certain tone, as an individual release. Dr. Patricia McConnell demonstrates this method in “Examples of Wait with Multiple Dogs.” This post covers how I went about teaching the discrimination of each dog releasing only on her own cue.
Does She Know the Cue At All?
Most of us at some point discover that our dogs don’t know their verbal cues nearly as well as we think they do. Here’s an experiment for those who have puppies or inexperienced dogs and haven’t worked on this before. Try this if your dog is familiar with Sit and Down, but not Stand.
Cue your dog to sit. Then look at her just like you are giving her a real cue (be as convincing as you can) and say “(Your dog’s name), Purple.” Or some other word that doesn’t sound a bit like “Down” or any other cue she knows. Most dogs will promptly lie down.
It usually turns out that your dog didn’t really know the word, “Down.” She didn’t need to, since whenever you have said something to her when she is sitting, you meant for her to down. So you can say anything and she will do it.
The transition from responding whenever “human-says-a-word” to learning to listen to the verbals can be difficult and stressful. That’s why I decided to apply the principles of reduced error learning.
Reduced Error Learning
I don’t use the term “errorless learning” because it is both impossible in real life situations and sets a depressingly high standard for most people (and also, I have to add, the most well known studies involved lab animals that were food deprived. I just don’t want to be associated with that). I wrote about this in Errorless Learning II. I have adopted Susan Friedman’s terminology of “reduced error learning” because I think it’s more realistic.
The original concept as promoted by Skinner is great. I do absolutely follow the practices of this kind of learning, which I would describe as “setting your dog up to succeed and to reduce stress in learning, including with creative manipulation of the training environment and props.”
An example of this is the process of making the right choice easier at first during an olfactory discrimination, such as the cups game. If you are teaching your dog to foot target the inverted cup that covers a smelly treat, first you start with only that one cup. Encourage her to use her nose to smell the cup and treat. Let her repeatedly practice touching the cup with the treat under it. Lift the cup and give her the treat each time.
To start the discrimination, you introduce a second cup without a treat, but you introduce it way over to the side where the dog can’t reach it. You gradually move it closer and closer while the dog is still touching the correct cup. In this way you have made the correct choice easy and the incorrect choice hard, and the dog is gaining a reinforcement history for touching the cup with the treat. Only after this process would you start mixing the cups up.
The opposite of this process would be to put out multiple cups with only one with a treat under it, and mix them up each time the dog gives a try. Even if your dog knew a foot target, there would be no clue as to which cup to touch. And even if it seems like it would be obvious for them to touch only the smelly cup, well, I’m here to tell you that my hound couldn’t do it, even when I made it much easier than a bunch of mixed up cups. With several cups, the failure rate is apt to be so high that many dogs will quit after a few attempts. This is the difficulty with trial and error learning.
Applying Reduced Error Learning to Cue Discrimination
Verbal cue discrimination means you teach your dog to respond only to the correct verbal cue and not other words. The way this is generally done is to repeat the cue for one behavior several times, and reinforce correct responses. (If you are not getting correct responses, you aren’t ready to work on cue discrimination.) After about four of these, say a completely different word instead. If the dog doesn’t do the behavior (yay!), or hesitates, quickly mark and reinforce.
Note that this is harder for the dog than firing off a bunch of different cues the dog knows. Because in this exercise the dog must be discriminating enough and confident enough to do nothing if the word is not a real cue. Plus, in so many situations we reinforce clicker dogs for guessing. The first time you practice this it can be like pulling the rug out from under the dog’s feet.
So how can we reduce stress and errors? In addition to choosing words to begin with that were very different from the correct cue word, I also chose to use at first a different tone of voice and/or volume for the cue. I took pains to make the non-cues as far away in the auditory sense as they could be from real cues.
Since the whole point of individual releases is that one dog comes and the other/s stay put, I practiced with each dog by herself, going through the following steps to insure that she learned to respond to her own release cue and not the other dogs’.
Special note: I heavily reinforce my dogs for being on their mats, and I don’t require them to move when I give the general release cue, “OK.” Because of this I incorporated a hand target and/or other body language at first to encourage them to move, then faded it. Others would probably not need to do this.
- With the dog on her mat, I called her with her release word followed by invitation to hand target or other body language that invited her to move.
- Then I called her with her release word without a hand target.
- Then I said a word that was very different from the dog’s release word and in a different tone (I blurted it out, high and squeaky). I reinforced her for not moving. If she got up, I quietly escorted her back to the mat, walking side by side with her to avoid using body pressure. (This hardly happened at all, which was one of my goals.) If the dog did get up, I made the non-cue word even more nonsensical. Quieter. Or perhaps I turned away. Anything I could think of to make it less cue-like. Once she started getting it: lather, rinse, repeat.
- I started interspersing the dog’s release word. I reinforced when she came, and for the other words, I reinforced when she stayed. If she stayed for her own release word, I beckoned her a little. If she came for another word, I quietly escorted her back to the mat as described above.
- I gradually worked into using a normal tone of voice for the non release words. I continued to reinforce for correct behavior/s, staying or releasing appropriately.
- The final step was to work in the other dogs’ release words to make sure the subject dog wouldn’t release on them. At this point I was saying all the words exactly the same way without helping the dog. The goal was that she released for her own and was steady for the others.
This method worked very well for Clara and Zani. Clara in particular got it very fast, and I loved how she lay there very relaxed on the mat while I said the other dogs’ release words.
Summer had the hardest time. She alone started offering other behaviors for the non-cue words. That meant that the first few times I used a non-cue word, I had to withhold reinforcement or else reinforce a random behavior performed on the mat. In most cases she tried her “rewind” trick, a backwards inchworm move. I figured out to reinforce very fast, before she was able to move, and we got through it.
But then after I got her to stay still on the mat through the non-cue words, she lost confidence about coming on her own release word. She was not getting the difference.
I did some extra sessions with Summer. After I reviewed the video I realized what the problem was. My squeaky cues were actually prodding her to action. She is a bit sound sensitive and I think they stressed her out a tiny bit. In any case she responded by trying something, anything. So I did the obvious, and instead of squeaky blurty non-cues, I said very quiet ones. That did the trick. I was able to raise the volume almost immediately, and she is catching up to the other dogs.
I’m getting close to my goal of having all my dogs present and unfettered while I train one, with the others reinforced for their self control on their mats!
Thanks for reading! Coming up: