Re: [PyrNet-L] Obedience school?

["Tracy Bassett" <espinay@dynamite.com.au>]

<<Treats are a wonderful motivation, and should be used as a reward. BUT...
you don't want your pup to react to a command only when they want a treat.
>>

Positive reinforcement methods (operant conditioning) are really gaining
popularity now as a great way to train dogs - I believe wholeheartedly in
using these methods.  Problems can develop with food training though when
the training is not carried through the correct progression to its ultimate
end.

You start off teaching a dog an exercise using treats every time as a
reward.  But this is not where it ends. Once the dog is responding to the
command every time when it is being rewarded every time (ALWAYS accompany
the treat with praise), you then begin to randomise the reward.  For
example, do the exercise twice (eg 'Sit) and give the treat/praise both
times.  Then do it again and just use praise.  Give a treat the next time,
then twice again without treats etc etc.  Gradually you can increase the
number of times you do the exercise without giving a treat. But make sure
you keep the reward schedule random.

Randomising the reward actually works to increase the likelihood of the
behaviour occurring rather than decrease it.  It works something like a
poker/slot machine.  If you put a dollar in a poker machine, pull the
handle, and get a dollar out every time, you would soon get tired of it and
stop playing.  Because the poker machine pays out on a random reward system,
you keep playing - next time you may hit the jackpot!  Same thing goes for
the dog - the dog keeps 'playing' (eg sitting on command) because next time
he may just hit the jackpot and get the treat!  If he gets a treat every
time he does it, the chance of him performing the behaviour actually
decreases.

So, basically, you start teaching an exercise with food every time, and then
gradually decrease the frequency of the food once the dog 'understands' what
is required of that command.  (make sure he does truly understand first, if
you go too fast and the dog is not doing the exercise every time, go back a
step in the training progression chain) You do this with each exercise
taught, so a dog who is in the process of being taught a number of exercises
in a row may, say,  be sitting  on command with very infrequent use of a
food reward, dropping on command on a close randomised schedule, and
standing on command using food every time.  It also may be learning at the
same time to sit straighter beside its handler (if training for competition
for example) using frequent food rewards in a training session.  The rewards
are used here (and later phased out to a long randomised shedule)  to 'tidy
up' an exercise.

The thing to remember though is that the reward should not be removed
completely (and by this I mean no reward of any kind ever) as by doing this
you will extinguish the behaviour (eg sitting on command).  you still need
to reward occasionally (even if it is once a week - not hard really - say
sit, and give your dog that bikkie he would have got anyway!)

There is a lot more I could say about training with food rewards, but it
would probably take pages! :-) Instead I will just suggest some books.

Karen Prior's 'Don't Shoot the Dog' is excellent, or take a look at her book
"Lads Before the Wind".  This second one is about training dolphins, but
gives some insight into the use and methods of operant conditioning using
positive reinforcement. And if you want to read more on the
scientific/theoretical principals of operant conditioning, here is an
article http://www.valdosta.peachnet.edu/~whuitt/psy702/behsys/operant.html


Tracy Bassett
Canberra, Australia
espinay@dynamite.com.au
http://members.dynamite.com.au/espinay/index.htm