ABC and VSR
Results 1 to 7 of 7

Thread: ABC and VSR

  1. #1
    Fallriver's Avatar
    Fallriver is offline Senior Member
    Join Date
    Feb 2009
    Posts
    1,233

    DefaultABC and VSR

    Tucker's pretty good on everything he needs to do, but is being 'lazy' (?) lately....not coming on the first call, not sitting on the first sit, and I don't want to keep repeating it....I'm assuming VSR will help that?
    No, VSR will not necessarily help with that.* VSR is for when you have the behaviour under stimulus control, which means that when you say Sit, there is at least an 85% chance Tucker will offer the sit, regardless of the situation.

    Tucker is not being lazy...there is a problem with his ABC's* * Whenever you don't get the expected response from a cue, you must look at the ABC's of the behaviour you are training.*

    A - antecedent (things that happen before the behaviour, or stimuli, or cues)
    B - behaviour
    C - consequence

    Focus mostly on the A's and the C's if you do not get the B.* First, the biggest problem with food based training is the antecedent.* When we train behaviours, we have bait bags waving in front of our dogs, we have their leash on, we are at training class, and we generally smell like a hot dog factory.* These are all antecedents to your dog's behaviour.* Your dog learns quickly that when you have your bait bag on and smell like hot dogs, the chances are very good that he will get a prize in exchange for a sit.* This is good and this is how we get the enthusiasm and attention in the initial stages of training.* This is how we prompt the sits so we can reward them.* Once we have a fairly reliable sit however, we need to quickly look at cleaning up all of our antecedents, except the primary stimulus (in this case, the verbal cue or signal to sit).* We need to systematically teach our dog that regardless of the antecedents surrounding our behaviour, the outcome will still be the same:* cookie for a sit.* Otherwise, when we ask for a sit, but don't smell like hot dogs and don't have a bait bag, our dog notices that a previously relevant antecedant is missing. They will understandably be confused about what we are asking them to do, because only part of the antecedant equation is presented to them.* We often misinterpret this as "my dog is stubborn", or "my dog will only do it for the cookies", when your dog is only confused and not sure of what you are asking.* Having too many antecedents is very much like bribing.* It is OK in the initial stages, but we want to drop it as quickly as possible.

    To clean up our antecedants, we need to get the food off our bodies and get rid of the bait bags (I could dedicate an entire post to my distaste of bait bags because they are a major antecedant problem).* Have the food on a table or in your mouth, or run to the cookie jar.* Ask for the sit in a variety of different places too, and with you in a variety of different positions.* Ask for a sit while you are laying down, hopping up and down on one foot, in different tones of voice.* If we do this, we will systematically teach our dog that the only antecedant for Sit is when we say Sit!* Anything else is irrelevant.
    OK, so we've cleaned up our antecedants, but we still don't have a lot of reliability in our sits.* So, now we must focus on the Consequences.* As we learned in this thread, there are only four consequences to your dog's behaviour:

    Good thing starts
    Good thing ends
    Bad thing starts
    Bad thing ends

    If you have a good, clear cue (you've cleaned up your antecedants), and you don't get behaviour, then you have to look at the consequence.* The behaviour must have a reinforcement history behind it to begin to be reliable.* Behaviour that is reinforced strengthens, so to strengthen the behaviour, we need to build a sufficient reinforcement history to make the behaviour reliable.* If we feed our dog every time he sits, we begin to build a reward history.* This is good:* our dog quickly learns that sitting pays off and not sitting does not.*
    The pay scale matters though.* If we give our dog one piece of kibble every time he sits, you will likely get a sit in low distraction environments.* If, one day you ask him to sit when a parade of cats is going down the street, you might have a bit more difficulty.* Your dog will quickly weigh the pay-off of chasing cats against the pay-off of one piece of kibble for the sit, and you will likely not be happy with his math.*
    Before we go to VSR, we need to be variable in the reward we offer for the behaviour.* Dogs like to gamble the same as we do and the first step to VSR is random reinforcement.* We will still feed every correct response, but the reward will not be the same.* If the reward is variable and unpredictable, your dog is less able to do the chasing cats vs. one piece of kibble math and is much more likely to hang around for the surpriseÖmaybe some hot dogs, maybe some cheese, maybe a game of tug, or a terrific chase game.* It might very well turn out to be one piece of kibble, but if historically he has received the above highly valued rewards for his sit, he will gamble that today might be his lucky day!*
    In addition to variable rewards, we need to introduce differential rewards and this is how we can help to fine-tune our dogís responses.* Before we consider going to a VSR, the dog should have zero latency for the sitÖmeaning, that no time elapses between your giving the sit cue and his butt plunking down on the floor.* Sometimes it is hard for novice trainers to increase the criteria of the sit (only rewarding the fastest half of the sits), so I sometimes find that differential rewards can make life easier.*
    To use differential rewards, we will give the single piece of kibble for OK sits and save the most terrific treats and the most enthusiastic praise for the fastest sits.* In addition, we will use the single piece of kibble for low distractions or familiar environments and the extra yummy stuff for novel environments or higher levels of distractions.* We can also use jackpots for their very best efforts (many pieces of food spaced out over a number of seconds).
    If we get no response, then what do we do?* Remember that consequences drive behaviour.* We have four options if we ask our dog to sit and he doesnít (actually we have five, but we wouldnít repeat our cues, would we?).

    Good thing starts (positive reinforcement)
    We could reward the no sit (not a good option but one that we often inadvertently use).* There is something called repertoire reinforcement, in which a chain of single behaviours gets rewarded at the end.* An example of a repertoire is the formal recall where the dog must remain seated while we leave, come on command, sit at front, and return to heel on our signal.* If we reward the return to heel, we have rewarded that whole repertoire of behaviours.* Repertoire reinforcement is important to remember when we are tempted to repeat ourselves.* If we say Sit and we get the poop face instead, and then repeat ourselves and reward the subsequent sit, we have actually reinforced the first no sit simply by letting our dog proceed toward the reward.* Not a good path to take if you want your dog to sit on the first cue!

    Bad thing ends (negative reinforcement)
    We could conceivably tighten the leash after giving the sit cue then release it only when the dog sits.* As we discussed at length in another post, this may not be a good option.

    Bad thing starts (positive punishment)
    We could use an e-collar or prong collar as a correction.* Many researchers believe that introducing punishment poisons the cue, but that is for another time.

    Good thing ends (negative punishment)
    If we have a reinforcement history for the sit, then our dog will be aware of the fact that we have given the sit cue and if he doesnít sit, an expected reward will not come.* We can short circuit this a bit by marking the withdrawal of the reward with a non-reward marker (NRM).* This is the opposite of a click as it means what you just offered gets you nothing.* You can say No or Wrong as a NRM, but if you are fairly consistent with your criteria, you shouldnít need to.*

    Negative punishment is a built-in correction.* If we do not get the expected response, we donít reward.* The dog hears a cue that was previously linked with a cookie, and suddenly no cookie comes.* It will help him to learn that the behaviour has to be in the middle of the ABC sandwich in order to get the reward.
    I like to short circuit the NRM by turning my back on my dog.* So, the sequence would be the cue to sit, failure to sit, then I say Pfffft and turn my back on my dog for 30 seconds.* Then I turn back to him like nothing happened and start another trial.* The reason I like this approach is that it puts my dog into drive.* Novice trainers can be boring and inhibited when training their dogs and I find they just ask for the same thing in the same way over and over and over again and fail to meet the dog half way.* Turning your back is a strong statement and your dog will invariably run to the front of you, as if to say ďOK, Iím on board now, letís try again!Ē* Try it, it works well.
    If you ask your dog to sit and he doesnít however, donít blame the dog.* Chances are, the cue is weak or the reinforcement is insufficient.* Remember your ABC's and take a closer look at your cue or your reinforcers.* It will either be one or the other.* Most people get frustrated with loose leash walking and recalls because they are in too much of a hurry to fade the food and donít build sufficient reward history to make the behaviour reliable.* The have a C problem*

    Alrighty, so now we have a game plan for building reliability and zero latency.* Once these two aspects are in place, then we can move on to a VSR.* Up until this point, we have been rewarding every correct response (continuous schedule of reinforcement, CSR).* This is akin to putting money into a Coke machine.* Every time you put quarters in, you get a pop.* You are happy and you walk away from the machine.*
    As I mentioned previously, dogs are gamblers, so we like to put them on a VSR (variable schedule of reinforcement).* This is akin to a slot machine:* when you put quarters in, you may get nothing or you may get something.* You may get something small or you may get something big.
    Both the pop machine and the slot machine are quarter inserting behaviours.* But you never see somebody pumping quarters into a pop machine for 4 hours at a stretch, but you do with a slot machine.* The difference is the schedule of reinforcement.* VSR strengthens behaviour, so quarter inserting into slot machines is a stronger behaviour than inserting quarters into pop machines.* The only difference is the consequence, so once more, consequences drive behaviours, and a VSR creates a stronger behaviour than a CSR..

    So obviously, in order to make our dogís behaviours stronger, we need to put them on a VSR.* Before you consider doing this, make sure that you get a sit at least 90% of the time the cue is given.* If you are not getting a fast and correct response 90% of the time, then your behaviour is not under stimulus control (you have an ABC problem) and it will break down if you move to a VSR.
    There are many convoluted methods of going onto a VSR and there are also twofers and threefers to confuse us even more.* Luke from Georgia was kind enough to send along some very nice articles on VSR, one written by the Baileys.
    The problem I have with a VSR is that most people start them too early, or feel pressured to go to twofers and threefers too early and there isnít sufficient reward history or the antecedents are not cleaned up enough for it to be successful.
    Instead of going to a VSR, I would prefer to continue working on reliability and generalization first.* If we progress naturally from differential rewards (the value of the reward is contingent on the difficulty of the sit), then what we can do instead of going to a VSR is increase our criteria and only feed the more difficult efforts.* So, to move on to more reliable and generalized behaviour, we will no longer feed piece of cake.

    So, if you ask your dog to sit when you are in the kitchen preparing dinner, I would call that piece of cake and I would no longer feed that.* If you ask for a sit on your front porch, that might be a more difficult sit, so you might want to feed that.* If you ask for a sit at the dog park, that is even further up the scale of difficulty, so I would feed that with something very yummy and maybe even jackpot it.* After a few trials, sitting on your porch will become piece of cake, so you no longer have to feed that.* Sitting at the dog park will be a bit easier, so instead of jackpotting, you might just want to flip him a piece or two.* After you have built a stronger reward history, sitting at the dog park will become a piece of cake, and then you can put the clicker away, but not the cookies yet.* We still need to go to a VSR to keep the behaviour strong.

    One disclaimer on this:* most of the behaviours that we ask our dogs to do are not innately reinforcing, so we give them a paycheque.* Once you have weaned your dog off the clicker, you will still need to randomly reward every now and again to keep it sharp (keep your dog on a VSR).* This is especially important for recalls.* I am completely happy with my dogsí recalls, but every month or so I secretly load my pockets up with cookies and take them off to the fields.* Iíll call them the same as always, but when they get to me cookies will suddenly fly from my pockets and they are just overjoyed!* They know it wonít happen every time they come, but the reward history for coming when called is well established, and they still randomly get rewards for coming when called, so coming when called remains strong because historically they have been payed to do so, and currently they get jackpotted every once in a while.
    By not feeding piece of cake, you donít have to worry about schedules or breakdown of behaviour.* When you are trying to decide whether a sit is a piece of cake or not, ask yourself if there is a 90% probability that the dog will sit given the distractions and environment.* If there is, it is piece of cake, if there isnít, feed it.
    Not really VSR, but I think that criteria shifting is a much safer and clearer way of fading food.*
    Dana


    To err is human:To forgive, canine."
    - Anonymous

  2. Remove Advertisements
    JustLabradors.com
    Advertisements
     

  3. #2
    Join Date
    Mar 2006
    Posts
    3,110

    DefaultRe: ABC and VSR

    How DID you know that he sometimes gives me a poop face instead of sit? ;D

    Yes, some places are definitely piece of cake (in the kitchen, of course, and usually in the field and at the pond before throwing the bumper...his sits are pretty firm in those environments).

    I've read variations of what you've described, and will print your advice out and heed it.

    I hope my printer ink doesn't run out before I print out your other training tips!

    Thanks again, Dana....I'll be checking back later to read more from you. (I never post, but look, at conformation/showing a lot. It's interesting to me, learning things there too...that's where I usually find you posting...but thanks for coming over here too.)


  4. #3
    luke from georgia is offline Senior Member
    Join Date
    Mar 2009
    Posts
    5,915

    DefaultRe: ABC and VSR

    Having too many antecedents is very much like bribing. It is OK in the initial stages, but we want to drop it as quickly as possible.

    To clean up our antecedants, we need to get the food off our bodies and get rid of the bait bags.
    amen!
    bribery has its value and is useful in certain situations, particularly in the beginning stages when a specific behavior is not something that comes naturally for the dog.
    however, we can and should train our dogs that responding to cues without knowing if and where the treats are going to be available is just as rewarding, if not more so, than offering behavior when they can see/smell the treat right in front of them.

    The problem I have with a VSR is that most people start them too early, or feel pressured to go to twofers and threefers too early and there isnít sufficient reward history or the antecedents are not cleaned up enough for it to be successful.
    Instead of going to a VSR, I would prefer to continue working on reliability and generalization first.
    great advise!!!
    often times, we do not establish the reward history sufficiently and then blame the dog for being a brat, and then there are many of us who are treat shy, as if using treats weakens the training and hinders the learning process. without even knowing what VSR means, a lot of people move to VSR way too quickly, in an effort to hurry up and wean their dog off of food/toy/life rewards, and so the training breaks down before it can get build a strong foundation.

    even after a dog's response is no longer dependent on getting a reward, it is such a pleasant suprise for the dog to get a treat when they least expect it. ;D

    "Properly trained, a man can be dog's best friend." ~ Corey Ford

  5. Remove Advertisements
    JustLabradors.com
    Advertisements
     

  6. #4
    CoCosMom's Avatar
    CoCosMom is offline Senior Member
    Join Date
    Feb 2009
    Posts
    8,070

    DefaultRe: ABC and VSR

    Thanks again Dana, I'm really enjoying these training posts and am printing them off for review!!
    ~Jo & CoCo


  7. #5
    Join Date
    Aug 2006
    Posts
    128

    DefaultRe: ABC and VSR

    You should put these posts in a book and sell it. This is America after all and theres some $$ in those posts!

  8. #6
    Join Date
    Aug 2006
    Posts
    128

    DefaultRe: ABC and VSR

    Oh I just read you were in Canada. Capitalist rules still apply

  9. #7
    Fallriver's Avatar
    Fallriver is offline Senior Member
    Join Date
    Feb 2009
    Posts
    1,233

    DefaultRe: ABC and VSR

    Hehehe...capitalist rules always apply 8) ;D
    Dana


    To err is human:To forgive, canine."
    - Anonymous

Bookmarks

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25