I started training my first dog when I was 11 years old. Like most if not all people my age, the way I learned to train a dog was that you physically make the dog do something by pulling on the leash and/or manipulating the dog's body into the position that you want and attaching a word to it. So, you want the dog to lie down, you say "Down", a word the dog does not understand and has never heard, and you pull the leash down towards the ground. If the dog doesn't understand that, then you "help" him by pushing on his side, pulling his front legs out from underneath him, or, in extreme cases, digging your thumb in between his shoulder blades and pushing down. Eventually the dog would go down. I never questioned these training methods, and always had happy, obedient, well-trained dogs.
In 1999, after I graduated from college, I went to work as an instructor assistant, and later an apprentice instructor, at Guiding Eyes for the Blind in New York. We were still using traditional training methods then that really were no different from the methods I had grown up using.
By the way, when I say "traditional training", what I mean is compulsion-based training. As I said above, we physically forced the dog to do what we wanted to, and after hearing the words paired with the actions enough times, the dog realized that it could avoid the discomfort of a snap on the leash if it performed the action that went with the word. We never, ever used food rewards in those days. I can't remember if anyone explained to me when I was a new apprentice the reasons for not using food in guide dog training. (Probably not -- I don't think it was even discussed back then.) I think that the rationale would have been that we wanted a dog who would work out of a desire to please, and also that we didn't want a dog thinking too much about food because then it would be more likely to scavenge food on the ground, and also that we would think the dog wouldn't work if it wasn't hungry, et cetera. Of course I never questioned whether maybe we should use food; I just did as I was told.
When I left Guiding Eyes for The Seeing Eye, I continued to use compulsion-based training. In fact, we used more compulsion there than I had at Guiding Eyes. To make a huge generalization that I nevertheless stand by as being accurate, the dog population at Seeing Eye at the time I was there had much higher levels of initiative and drive, and the corresponding higher levels of energy and distraction that needed to be managed. We managed this by heavy use of compulsion. I was taught that every Seeing Eye dog that I trained had to be acquainted with the technique called "forced pull" or sometimes "forced lead". This was used when, at some point in the first few weeks of training, the dog chose not to respond to the Forward command for whatever reason. (He was distracted, he was hot, he didn't feel like it, whatever.) I would say "Forward", the dog would not respond, and I would then move the slip collar, aka choke chain, high and tight on his neck, behind his ears and behind his lower jaw bone, and pull the chain as tight as I could until the dog literally could not breathe. He would thrash around in a panic to try to escape the pressure until finally, maybe by accident, he would move forward. When that happened, I would release the pressure and praise. I think of forced pull as the classic guide dog example of negative reinforcement: the dog moves forward, and the pressure is released. The behavior that caused the unpleasantness of being unable to breathe to stop -- moving forward -- was reinforced. This was not a pleasant technique to apply. It could be damaging to the relationship between dog and handler, and could cause the dog's attitude toward work to become less happy. It was not pleasant to apply. Nevertheless, it was effective most of the time. And the beauty of it was that it gave you a tool in case the dog decided to quit pulling in class or even in the field. Those dogs knew what it meant if the trainer said Forward and then slid the collar up on their neck. A lot of the time just the action of moving the collar up would send the dog surging forward into the harness without having to tighten the collar at all.
That was basically the answer to any refusal of commands at Seeing Eye. I say it, you don't do it, you can breathe again when you decide to do it. Harsh? Yes. Effective? Yes, usually. Always effective? No. These techniques worked great on the confident, high-end dogs. The more sensitive dogs did not always react well to these techniques. There was a lot of inconsistent pace, "lack of willingness" (to engage with the person who made it impossible for you to breathe? can't imagine why that would be), angling in crossings, and other work problems of that nature. Those dogs were dropped from training most of the time, or else dropped back to the beginning for another cycle.
Distractions were handled as follows: we taught the dogs the meaning of "Pfui" during the first week. I did this by scenting a stuffed animal with fox urine or something like that from a hunting supply store, and then placing the stuffed animal in the middle of an otherwise empty stretch of sidewalk. As I approached the stuffed animal with a dog, I would watch the dog closely to see the moment when it noticed the stuffed animal. At that moment, I would say "Leave it", a command the dog most likely had never heard. 95 percent of the time the dog would ignore me and lunge for the stuffed animal. When it did, I would calmly say "Pfui" and, with two hands on the leash, give the strongest correction I was capable of. This was a correction the dog could feel -- enough to jerk the dog off its feet. I knew the correction was effective if the dog dropped its tail, came slinking back to me with its ears pinned back trying to appease me, yelped, or otherwise lost all interest in the stuffed animal. If it remained interested in the stuffed animal, it got Round Two -- the same type of correction, but harder. Usually by this point the dog was afraid to approach the stuffed animal at all, and would stop at a distance and look at me uncertainly and wag its tail low and submissively and wait for direction, which I would then be happy to give it because I knew that now the dog got it. But you better believe that the next time I said "Pfui", it had an effect on the dog.
When I was working a dog and knew it was thinking about going for a distraction, I would first say "Leave it", to give it a chance to avoid trouble. If the dog then returned its attention to me, it got praise. If it didn't, it got a verbal correction, "Pfui", followed by the repetition of "Leave it". If that wasn't effective, it got a "Pfui", then a leash correction, then a "Forward" command, then a "Leave it". This worked fairly well with the huge caveat that we had to understand both exactly what level of correction the dog would require AND what level of correction a client was capable of giving. If we got either of these things wrong, the dog would "come up", that is, transform from a responsible, well-behaved dog to a distracted. out-of-control dog that had minimal respect for its handler. When this happened, our options were down to dog switch or management tools (Gentle Leader or possibly even pinch collar if it was happening with a grad at home). This type of screw-up was easy to do because often a dog that had needed firm leash correction when it first came in was totally willing to give up its agenda to the trainer once it knew what the trainer was capable of doing to it. The dog would behave perfectly all through training with minimal or no leash correction needed and then turn into a monster with a client. Or we would get a new client in class who showed us that he or she was perfectly capable of giving a strong leash correction physically but was perfectly incapable of giving that same leash correction when a dog was on the other end of the leash because he or she just didn't like to correct. This happened semi-frequently, and it was common for instructors to say, "If she would just CORRECT THE DOG…!"
I first learned about positive reinforcement training for dogs towards the end of my time at Seeing Eye, when I was getting ready to leave for grad school. It should be said that traditional training was not all compulsion- and punishment-based. It always included an element of reinforcement -- praise. We were taught, and taught clients, that dogs worked because they loved our praise. A dog that didn't "love praise" was unwilling. We didn't really talk about reinforcement in scientific terms (reinforcement = anything that causes a behavior to be repeated). We learned and taught that dogs loved praise just because. It's true that some dogs do. Some dogs are naturally so enthusiastic about people, because of breeding and temperament, that we don't have to pair praise with anything in order for it to be reinforcing. But for an equal number of dogs, praise is a low-level reinforcer or not a reinforcer at all. and if you really want to make things happen, some other method must be used to get through to the dog. I also believe that a lot of times during my early years training guide dogs, praise WAS reinforcing -- but not because humans were great and the words that came from our mouths made dogs happy. I think praise was reinforcing because what it said to the dog was "You're not in danger of being busted right now", and this caused the dog to wag its tail out of relief. Of course I have no way of proving this. But I am pretty sure I'm right. Anyway, I read as many clicker training and positive reinforcement training books as I could, read the Clicker Solutions email list religiously, and then bought an Aussie puppy while I was in grad school to try it out on.
Positive reinforcement-based training is a term that describes training based on reinforcing the dog for desirable actions as opposed to punishing undesirable actions. And let's be clear before we go any further that a leash correction, in behavioral terms, is designed to be a punishment -- something that decreases the frequency of a behavior. A leash correction that does not decrease the frequency of a behavior is an ineffective punisher. So if you have a dog that's sniffing the ground and you say "No" and give a leash correction and the dog keeps sniffing the ground, you haven't punished it at all. In behavioral terms, you haven't done anything at all. More on that later. Anyway, when you want to train a dog using primarily positive reinforcement, you also have to be able to manage the dog's environment to a pretty good extent. There are lots of undesirable behaviors -- digging, barking, chewing inappropriate things, jumping on people, really almost all the natural dog behaviors -- that are self-reinforcing to a dog, meaning that if you put a dog in an environment where those things are able to be practiced, the dog will do them more frequently. So you have to be mindful as a trainer and make sure you don't put your dog in a situation for which you haven't given him the tools needed for "good" -- desirable to humans -- behavior.
My Aussie pup was soft, sweet, bright, people-oriented, and very willing to please. I don't think I ever said "No" to him and I certainly never gave him a leash correction. He was beyond easy to train. Positive training methods and the clicker were awesome! We began competing in rally obedience and easily earned our Rally Excellent title -- not a difficult title like Utility Dog, but still, a title that required a high level of communication between handler and dog. We accomplished this in nine trials -- the minimum needed to earn the title -- and won first place eight times out of the nine, with a couple of perfect scores and all the rest only a couple of points off, like 197 or 198 out of 200. He was very enthusiastic about training and made me look like a great dog trainer. I tried using clicker training with Hilda, my early-retired Seeing Eye shepherd, and she did all right with some things, like the retrieve, and terrible with obedience, so I just gave up, figuring maybe she had been ruined for the clicker by her previous old-school training.
The first time I heard of clicker training for guide dogs was probably 2008 or 2009. I was still living in Arizona and was raising a puppy for Guide Dogs for the Blind. At that time I knew GDB was using the clicker to some extent but had no idea for what guide work tasks. The puppy raisers were given a new assignment that we understood to be a trial for incorporating marker-based training into puppy raising. (The clicker is a type of marker, and a marker is just something -- usually a sound -- that captures the exact second the dog performs the behavior that you want. Once the behavior has been marked, you have a few seconds to get the treat to the dog. This is helpful because it is nearly impossible to time the delivery of the treat in a way that perfectly communicates to the dog just what it is getting reinforced for. Clickers are the most popular marker because they are 1) unique, 2) distinctive, 3) cheap, 4) portable, but you can use anything, including your voice.) GDB had the puppy raisers introduce "Nice" as the marker word. They gave us instructions on making the association between the marker word "Nice" and the the delivery of the food. They only wanted us to use the marker word for one behavior - teaching the dog to go to its mat and lie down. The manual covered the steps in making this happen. I don't remember the exact steps, but I do remember being impressed by the almost-idiot-proof presentation of the information. There was almost no way we could have screwed it up.
When I went to Leader Dog in 2013, I knew they used a positive-reinforcement-based training program, and was very excited to learn it based on the success I had had with my Aussie and in teaching my GDB puppy to go to her mat. The one piece of the puzzle I had never understood was how these methods worked for instinctive distraction, the kind where the dog was more interested in the squirrel or the other dog than in any kind of treat you could provide. I was very excited to learn how this method worked in practice for guide dogs with higher levels of instinctive distraction. I have now been using this type of training method for the past two years, and below are my thoughts on it as of right now.
1) It is much more humane to the dog. There is absolutely no reason in the world not to lay a foundation first -- reward the dog for correct behaviors, attach a word to them, build up the skill as gradually as the training period allows -- before using punishment and compulsion when the dog has no way of understanding what is expected of it.
2) Use of food reward makes the formation of a trainer-dog bond (and a client-dog bond in class) much faster. We use mostly Labs, and for most Labs, he who controls the food controls the dog. A dog that is interested in the trainer is more engaged in the learning process.
3) The majority of clients are happy to not have to give hard leash corrections. It is understandable that the longtime grads will have difficulty transitioning to a method so different from the methods they have had success with in the past. It is our job as instructors to point out the advantages of the new method -- and to make sure that the overall result is a better working dog. It is my experience that if you can show a client a better working dog, he or she will accept a new training method. If the dog is not a better working dog, you can hardly blame the client for questioning the new method.
4) The public is far more accepting of positive handling methods than of traditional ones, including hard leash corrections. While I myself could care less what the public thinks, I DO care about our clients and the amount of public interference they have to deal with. Accusations of animal abuse, founded or unfounded, are very common today -- look at the New York City carriage horses as just one example -- and a large segment of the public is not interested in listening to explanations about keeping the dog's attention on the work and keeping the blind handler safe. They are more worried about the dog's welfare.
5) There is no quicker method for teaching the very valuable guide dog skill of targeting. Whether we're talking about a seat in a classroom, a push button at a crossing, or a building entrance, I was never able to teach targeting with anywhere near as much precision when I relied on verbal praise and leash correction. The increase in efficiency when using food reward has been very impressive.
6) The dog's drive to curbs, both up and down, is significantly better than I remember from when I was training without food. Back then, if the dog failed to stop at a curb, it was leash corrected and forced to do it over again. This often led to the dog developing some ambivalence to curbs, especially up curbs. I remember having to discuss alignment at down curbs a lot more in the old days than I do now. This is not to say that we never deal with dogs veering in crossings now -- we certainly do -- but for the most part, if you point them at a curb and say "Forward", they will go there with purpose and enthusiasm. Why wouldn't they? The curb has only ever been a good thing for them, never a bad thing.
The Concerns I Have About Using Primarily Positive Methods:
1) Punishment is the quickest way to decrease undesirable behavior, and I strongly believe that total removal of punishment from the trainer's tool box would be a mistake. I don't have any problem at all with teaching a dog what "No" means and using it, followed by a leash correction of a strength that is appropriate for the individual dog if necessary, for stopping undesirable behavior such as jumping up, trying to scavenge, or persistently sniffing. The emphasis here should not be on avoiding punishment at all costs, but rather at developing an effective understanding of the level of punishment that dog is likely going to require and taking that into consideration when matching a dog with a client.
2) If you are going to bring a dog all the way through puppyhood and formal guide work training without using punishment (because you want the dog to be happy, enthusiastic, and low stress), you had better have a plan for incorporating stress into your proofing process. Because, guess what, guide work is stressful. Period. There is no getting around it. When you can't see, you rely on the dog to make decisions in high-pressure situations. A dog that has never been pressured is more likely to shut down in real-life situations, especially once it is outside the "class bubble". The good thing about the old training methods was that once a dog made it through, you could be pretty sure that dog had a pretty good backbone, because those training methods WERE stressful, and dogs that couldn't handle it were dropped. (Though I must admit, so were some other dogs who probably would have been useable in a low-pressure situation.)
3) Positive training methods and clicker training in general rely on the handler being willing not to exceed the dog's current threshold -- whether it is for distraction or something the dog finds stressful. It is unreasonable to ask clients to do this. It is unreasonable to ask clients to alter their routes or lifestyle because their dog has difficulty dealing with some aspect of their daily life. I never want to be in the position of telling a client, "I know your last three dogs have done this route with no problem, but this dog isn't comfortable with -- " (insert distraction or environmental stressor here) "-- so you're not going to be able to do this route anymore." That is unacceptable. A dog issued to a client must be able to cope with the environment the client works in. That is the responsibility of the trainer, both to know as much as possible about the incoming client's home environment, and to do everything possible to expose the dog to that environment. Every working dog issued should, in my opinion, be known to be capable of functioning around large crowds, areas with many other dogs and/or other animals, working on busy streets with traffic coming from either in front or behind, crossing large or complex streets, and working in areas with no sidewalks. No client should be expected to have to baby the dog through these environments or, worse, avoid them completely.
4) The dog should have demonstrated ability to work without the presence of food rewards over a sustained period of time by the end of the training cycle. I do believe that random reinforcement throughout the dog's working career will keep work enthusiasm high, and I think most clients will see this too and will continue to use random reinforcement just because the benefit of superior work exceeds the cost of carrying treats, but I also don't believe that clients should be issued a dog that will not work without treats. Sometimes it isn't convenient to carry treats (job interview, first date, rain that makes treats soggy or frigid weather that requires heavy gloves), and the dog should still be able to function in these situations without the treats. Additionally, we know there are some clients who simply will not use the treats once they return home. I want to know prior to clients' arrival that all of my dogs will in fact work without food rewards for extended periods of time.
5) If we are not going to teach clients to apply proper leash corrections (i.e., leash corrections that function as punishments and decrease unwanted behaviors), we MUST be able to give them alternative methods to use that are just as effective in regaining the dog's attention.
6) We must make sure that our dogs are polite about taking treats, and are not overly fixated on the treat hand or treat bag to the point that they are stopping mid-block in hopes of getting a treat, or are getting out of position at curbs because they are curling around the handler's body to try to reach the treat bag.
Overall I see far more positives in the evolution of training methods for guide dogs over the years than negatives. I also acknowledge that in order to provide the best service possible to our clients, we must consider the negatives too and put effort into finding the best ways around them. I myself prefer to have as many tools as possible, and be able to use them as I see fit based on the individual dog, the individual client, and, of course, the fact that I work for an IGDF-accredited school and must follow IGDF guidelines. I love dogs, but the fact is that I do this job to provide enhanced mobility for people. not to make dogs' lives better, and I want to use the methods that produce the best working dog that is the easiest for the client to handle, whatever those methods may be.