Timing and latency and short sessions

I have been pretty successful with my clicker training so far. Noch has been a brilliant student.
But I realise that I can working on getting better, as I said in my last post, and so I am formalising my training and keeping a log.

But geeze, it sure as hell is hard to keep a log and stick to the set goal for that training session! I realise I train intuitively a lot -- which can actually be good and bad. Bad because it probably makes Noch's job confusing if the actual goal changes. I do see the advantages to keeping to one training session=one goal. I also see the advantages of forcing oneself to write down that one goal and how to reach it on paper.

I have also realised that my training sessions may be a bit on the long side(!)

Apparently a good rule of thumb is 5 repetitions to one training session.
Then break for a short while (go write in you log).
Then do another training session.
In total a training period should last about 5 mins maximum. This means that you have time for approximately 2-4 training sessions in one training period. Wow! So short eh? I can see that clicker training really doesn't have to take a lot of time at all.

These incredibly short sessions put a lot of pressure on the handler to set up good criteria -- which is great! It also keeps the dogs perky and bright eyed and loving it (and ending when they still want more). Of course it is painful for the handler to stop when the going is soooo good! But I think it is one of the keys to getting the dog to lvoe cliker training.

The repetitions themselves have some timing critera too: latency (the time between signal and offered behaviour eg. handler standing up and waiting, handler presenting target, handler giving command) should be between 0,2-3 seconds depending on the nature of the excercise. 0,2 seconds being for commands/signals the dog already knows, 3 seconds being for when the dog is starting to learn a new behaviour. When you are free-shaping this latency can be increased to 5-10 seconds.

Again this strikes me as really fast! But it does make sense -- in practice this means for example presenting a target and after 3 seconds of the dog failing to touch it the target gets removed and hidden behind my back for 5-10 seconds. It is then presented again for 3 seconds. These 3 seconds in other words is the window in which the dog has to present his behaviour.

Additionally you want to avoid letting the dog do two incorrect repetitions in a row, and really really avoid three incorrect repetitions in a row. If this is happening you may want to consider that your criteria are too high and you may need take them down a notch so you are setting your dog up to suceed.

My book advises 5-30 min training passes per "go".
This means 3-6 training periods each containing 2-5 training sessions containing in turn 5 success/fail repetitions.

So much to think about! Geez, I am now SURE that Noch has the easy end of this bargain!

Lastly -- I found a brillant way of teaching loose-leash walking that I am dying to try! It echoes my thoughts of pondering if you could use horse rein signals with a pulling dog. I mean we humans instinctively pull on our dogs to communicate with them through the leash. Actually no different than using a bitless bridle on a horse. In the same way that pulling on a horse that hasn't been taught to move away from pressure is meaningless, so is pulling on a dog -- it will only be met with resistance. Sometimes I even ponder if it is possible to teach a dog to canter and trot and do groundwork like horses -- it would be fascinating to try! Anyhow Shirley Chong has this great way of building up to loose leash walking that is based on her experince with teaching race horses to have soft mouths again:

http://www.shirleychong.com/keepers/LLW/LLW%20Step%201.html

Setting critera when clicker training

Well I am pondering how to go about the incredibly challanging task of teaching Noch how to pass other dogs without going beserk (he looooves other dogs and will bark and jump around in excited frustration and forgets I exist).

I am thinking it comes down to me improving my training skills. Things like timing and setting him up to suceed seem incredibly important. But I am a bit lost as to how to go about it without getting frustrated and (being the crossover trainer that I am) resorting to negative reinforcement in sheer frustration...

So I've ended up reading my clicker book again "Klickerträning för din hund" (in swedish) by Morten Egtvedt and Cecilie Köste. A brilliant book, but unfortunately not-so-great layout. The font they've used for the body text is sans serif making it hard to read. An eyesore for someone like me who works with layout on a day to day basis. But yeah, apart from that it's full of brilliant advice. I think I am going to start a training log to help me hone my criterion-setting skills (and have a clear training plan).

But yeah the one thing I wanted to mention was the following (summarised and partly quoted from the book):

Dog trainers using positive reinforcement set their dogs up to suceed, so that they may have lots of opportunities to reward their dogs for good behaviour.
Dog trainers using negative reinforcement set their dogs up to fail, so that they get lots of oppotunities to correct their dogs.

That thought I found really really interesting!! And made me go "ahhh.... oh yeah".

Lastly a You Tube video, some really good criterion-setting-up going on here:

Snowball the dancing cockatoo

Just some links about the now world-famous dancing cockatoo(!)

Article:

"Snowball's chance, Cockatoo's extremely rare sense of rhythm may help explain how the brain relates to music. By Adam Loberstein, August 14, 2008."

http://www.signonsandiego.com/uniontrib/20080814/news_1c14bird.html

Video that goes with the article:

http://video.signonsandiego.com/vmix_hosted_apps/p/media?id=2073032&item_index=2&all=1&sort=NULL