>>Message Sent by: Gail Fisher <gail@alldogsgym.com> >>See http://www.click-l.com/ for faqs, rules, links and other info >>See http://www.escribe.com/animals/clicker/ for searchable archive >> Uh oh, again I need to clarify and perhaps disagree with Carol (Don't get a complex Carol -- it's just that you're . . . well . . . prolific <g>). What Carol described , i.e., 1:1 ratio of reinforcement done quickly, is not what we call a speed trial. There's certainly nothing wrong with what you're doing. but it's not a speed trial. A speed trial is not defined by you giving treats rapidly and cuing quickly (i.e., while the dog is eating). Speed trials are the dog's behavior coming rapidly -- without treats! What you're doing is fast repetitions with short latency, but when you interrupt the behavior with a treat each time, you're not able to get as fast as the dog can get. Speed trials require that the dog's behavior be on a variable ratio or reinforcement schedule. (Gary Wilkes, Morgan Specter and I had a series of articles on this subject in Clicker Journal 2 issues ago, I think). By the way, I'm speaking specifically about dogs here, not animals in general. Dogs can be on a much longer schedule of reinforcement than many, many other species. Further, dogs are reinforced by verbal praise. So when I'm talking about doing 10 or 15 or 20 behaviors in a row without c/t -- that doesn't mean you can't tell the dog he's doing GREAT! And most dogs respond well to that once you've gotten to this level of training. When doing speed trials, it's important to maintain precision while building speed. Building speed and accuracy go hand in hand in my opinion. It doesn't work to have lots of fast . . . somethings, unless they're all the *same* something. For example, you can count fast by 5's, or you can simply spew numbers out fast. If you're inaccurate, well it's gibberish, isn't it? If what you want is to speed up your 5 count, then it's important to be accurate. It doesn't matter how fast you recite numbers 5-10-12-15-25-22-30-35-40-43 -- it will never build the skill of counting by 5's! Kind of like "I speak English . . .Maunday, Tuesday, Thursday,Wednesday, Sattaday, Friday." (Any Godfather fans?). This doesn't represent fluency in English any more than doing behaviors quickly (but not accurately) represents fluency in dog behavior. As someone recently wrote on the list, practice doesn't make perfect unless the practice is perfect. Practicing a golf swing, a baseball pitch, a tennis serve, a musical instrument -- all skills that require maintaining accuracy and precision while building speed. You might be able to hit a lot of piano keys in a short period of time, but without practice and accuracy, it won't be "The Flight of the Bumble Bee". I also feel that it is critical to doing speed trials that the dog be on a random schedule of reinforcement. Carol wrote: >>>For a speed trial, I keep a handful of treats, so I can feed a treat quickly after the click. I put the treat in the dog's mouth - I don't toss it on the floor - having the dog take the treat from the floor would slow the repetitions. The point here is to get the repeats going just as fast as the dog and I can manage it.<<< There are lots of reasons that this is limiting. First of all, the trainer has to be right near the dog. Second, that means there's most likely body language involved -- and I want the dog concentrating on his behavior, not on where I am or what I'm holding in my hand or where I'm facing. A challenge: If you think your dog knows what "sit" means -- If your dog is facing away from you and you say "sit," will your dog sit facing away from you? If your back is to your dog and you say "sit," will your dog sit behind you? I think most people will likely find that the dog only knows "sit" in the context of as it relates to our body and body language. Facing us, or at our left side, for example. Carol also described two cues for one behavior -- something I also think is limiting. Not that this is a critical point, but if the idea is speed (which gives reliability, etc.), then having the dog become dependent on 2 cues is not the most reliable approach. In order to get the speed going as fast as the dog can manage it, you can't give a primary reinforcement for each behavioral repetition. Eating interrupts the trial. Each speed trial ends with the delivery of the primary reinforcement. The time frame for repetitions (and it's important to keep count if you're interested in improving) needs to increase -- that's how you know if you're improving. A speed trial is [blank] behaviors in [X] time frame, keeping records and improving the dog's performance. 5 sit/down/sit combinations in 20 seconds, for example, working up to being able to do 12 of the same combinations in 10 seconds. You need to be able to do the series without giving a treat after each behavior. If a dog can't go for 5 or 6 repetitions of a behavior without getting a click and a treat . . . well in my opinion, that dog isn't ready for speed trials. Further, if the dog requires the handler be standing so close, again it's limiting. What if the behavior you want to work on is "go out to a sit (or down if you're in England)". If you're inserting food into the dog's mouth for each repetition, how do you build speed with the dog sitting away from you? And what if you're doing agility, or any other sport that doesn't involve the dog being at heel position, or otherwise close to you? How would you do speed trials for that? As I wrote in my Clicker Training Volume II Manual, speed trials help the dog get to behavioral fluency. Much like fluency in a language, behavioral fluency means that the behavior is reliable, accurate, fast, stable and resistant to both distraction and extermination. We teach speed trials in our Level 2 class, after the dog has several behaviors on cue. They're an excellent way to build accuracy and true *understanding* of the cue(s). Here's an example: We like to start with a dog that knows 3 behaviors on cue -- for example, sit, stand, down (which is actually 6 behaviors -- sit from stand, sit up from down, stand from sit, stand from down, down from stand, down from sit). Any 3 behaviors will do -- at a seminar in Illinois, a wonderful Min Pin did a speed trial with random sit, down & rollovers. Sharp little dog -- first time he'd ever done a speed trial and it was an awesome demo. At my seminar with Corally, we had a super Beagle that demonstrated doing a speed trial. Anyway, start by building speed with 2 behaviors -- say sit and down -- slowly at first with frequent clicks and treats, maintaining accuracy/precision (nothing to do with competitive training -- a sit is a sit is a sit). So it might be Sit, down, sit up c/t. Sit, down, sit up, down, sit up. c/t. Sit, down c/t. (2 behaviors, 5 behaviors, 1, 10, 4, etc randomly reinforcing and building speed with those 2 behavior. Then add another behavior -- say stand from a sit. Start again slowly, such as stand, sit, stand c/t. Sit, stand, sit, down, sit-up, stand c/t. etc. Gradually building speed as the dog becomes more confident. The object is to be able to drill a series of cues with total accuracy (well, 80% or better, anyway) in a short time. Once the dog is doing it, you can maintain this level of accuracy (which translates to dependability) by practicing for just a couple of minutes once or twice a week. If you stop practicing it, however, you lose the edge. Speed trials are fun and shouldn't be difficult even for the novice clicker trainer -- it's just a question of having the dog be ready -- which means performing a behavior accurately on cue. Gail -- Gail Fisher All Dogs Gym (r) Manchester, NH 03103 http://www.alldogsgym.com (603) 669-4644 FAX (603) 641-6767

Home