|
ANTICIPATION ..... You give one cue and the dog skips ahead to another behavior. We set up behavior chains without realizing it and then are perplexed that the first learned behavior is suddenly gone and is replaced with the new behavior.
Mark
it / feed it:
photo of Higgins courtesy of Steve Navratil > |
 |
As much as we stress the essential rule of training:
"consistency is important"
- in reality, it can be harder to be unpredictable.
We, as trainers, tend to do things in patterns. "Spot, Sit ... Down." The dog hears "sit" and is already thinking ahead to the "down" cue that his learning history tells him is sure to come. The sit response weakens as the trainer unknowingly strengthens the connection between the two cues. Within a small number of trials, the dog downs on the cue to sit.
The reason dogs make this connection is because we teach them to! We give a known cue to set them up to perform the new behavior. Sit - shake. The dog notes the sequence that "something they already know" is predictably followed by "something new." The connection is strengthened by pattern and repetition.
Dogs gravitate to the point of reinforcement. The new behavior is often accompanied by jackpots and smiles. It becomes more rewarding to execute the new novel behavior than the boring old set-up behavior that precedes it. The 'plain old sit' isn't getting big feedback and is often a utilitarian behavior. Sit before you go through the door. Sit before you get to shake! The dog is excited to get passed the sit and on to the more "fun" behavior!
| How
dogs learn. In the initial stages of learning a new behavior, the behavior of "laying on side, then rolling onto back with belly in air and flopping over to the other side" doesn't have a cue of its own yet. The "down" cue has preceded the new roll-over behavior each time you set him up to practice. Until the dog recognizes the new behavior has its own name and you separate the two behaviors, "down" is the cue to roll over. Down, roll-over, down, roll-over - pretty soon you say "down" and the dog rolls over. |
 |
When to add the cue. When the dog is at 80% reliability (you can predict in 8 out of 10 trials that the dog is going to reliably perform the new behavior) it's time to add the cue. The cue PRECEEDS the behavior.
"Down" - mark/feed - "roll over!" *prompt behavior* - mark/feed.
The trick for working through the normal phase of anticipation and getting to the point of *stimulus control is to sandwich the now less-frequently reinforced new behavior between heavily reinforced old behaviors so it is clearly its own link in the behavior chain.
* stimulus control: the dog performs the behavior reliably when he hears the cue, and does not perform the behavior unless and until he hears the cue.
Keep the old behaviors strong by rewarding the set-up behavior more frequently and the new behavior less often. Sit, wait (with-both-front-feet-firmly-planted-on-floor), click/treat, shake, wait click/treat.
Use a differential schedule of reinforcement: reward the best ones, don't reward the mediocre ones. You reward the shake only if the dog waited for the cue. False start? Give another cue to sit and wait, pause, give the cue to shake, if the dog waits and then responds to the cue, c/t.
You might explain it further by adding higher criteria to the foundation behavior: "sit-stay, circle c/t, sit-stay, back up a step c/t, shake, sit-stay, circle c/t." Or insert other known behaviors so the dog has to listen for which you want and reward randomly throughout: "sit, down, wait c/t, sit, stand c/t, down c/t, roll-over, wait, sit c/t."
Mix it up!
Once your dog has a rich repetoire of polished behaviors, give cues in random
order so the dog has to focus intently and go smoothly from one to the next.
Variable schedule of reward: reward slot machine style: build long strings without reinforcing each individual behavior. Release at unexpected points to celebrate a successful series of randomly cued behaviors.
See also: Tricks!
