I have just spent 3 intense but incredibly rewarding days learning from the incomparable Bob Bailey. It was a full-immersion R+ training geek out and I thoroughly enjoyed it!
If you don't know who Bob Bailey is, then, in short, you should! Go here to read more about his fascinating role in the history of positive reinforcement training and the roots of modern-day "clicker training". (When I say "fascinating" I mean it - I'm talking about pigeons trained to guide missiles by sitting inside and pecking at a screen, ravens conducting surveillance by flying to upper-story windows and taking photos with cameras attached to their heads, Navy dolphins trained to swim in the open ocean on circuitous routes of up to 12 hours and return to the trainer, and more).
It will take me some time to process all the learning and reflect on how I will integrate it into my own practices as well as my teaching, but here I will share some highlights and brief "take aways" from the seminar.
Keys to good training:
- Precise timing
- Observing behaviour
- Quick decisions
- Fast trials
- More trials
- More reinforcement
- Higher expectations (of yourself too!)
- Changing your behaviour
You can understand the principles of good training, you can read about it in books and on the internet and write academic papers on it, but if it doesn't go from your head to your hands (i.e. put it into practice) then it means nothing.
Hearing about good training, thinking about good training, believing in good training, planning to do good training, is NOT the same as DOING good training!
Good training is not about hierarchies and relationships - relationship with the animal is nice, but we don't NEED it in order to train.
(My interpretation: Good training deepens relationships, but the one is not dependent upon the other).
1. "ANY trainer, using ANY method, can train ANY animal, to do ANY behaviour, given enough time"
Don't just ask "did the animal learn the behaviour?" instead ask yourself "how could I have trained this in less time?" and "Could the behaviour be better?"
(Plus I would personally add, "was the horse a joyful and eager volunteer in this learning experience?")
Every trainer can point to some behaviour or other they have trained. Success is not just getting behaviour, it's getting behaviour quickly and accurately.
(My interpretation: Good training is both ethical AND efficient)
2. If you are making poor decisions and training badly, and the animal manages to figure it out despite you, the ANIMAL should take the credit, not you.
3. "Animals are built to learn. They are learning all the time, not just when we want them to."
(Training is learning that happens in the presence of a human)
"Evolution prepared animals to learn well, and learn quickly. If learning occurs slowly there is a REASON!" (Spoiler alert: it's probably you).
If the animal is not learning, most likely the trainer has done something wrong - e.g. not defining behaviour, not making it worthwhile for the animal to do it, not controlling distractions.
This is the best time for novelty and creativity. Use your imagination, visualise what you want.
Define, describe, simplify, make mistakes (rehearse without the animal).
First (before we start training) we must ask ourselves "what do I have?" then, "what do I want?". The training plan describes the path from what you have, to what you want. Train to fill in the gaps.
"Split, but do not reject opportunistic lumping!"
Don't get in the way of learning - have a plan, but if the animal is ready to leap ahead, don't hold them back
E.g. We might have a plan to get from A-B-C-D-E-F but the animal may leap straight from B to E. (Get out of their way, and for Dog's sake keep up!)
Execute! Don't waste your most precious resource - TIME
Over-analysis (thinking too much) can get you into trouble.
"Don't allow INDECISION to paralyse your ACTION! " You must be able to make fast, good decisions in the moment. Having a clear plan is key to this.
Training period (e.g. 10 minutes)
Training session (e.g. 1 minute block or 10 trials)
Trial (one attempt)
Train ONE behaviour per session; ONE criterion per session.
Approx 8/10 trials should be successful before you move on / add criteria (that doesn't mean it will always be 80% successful - it may be a much lower ratio early on).
Define new behaviour for any training session in ONE SENTENCE! If you can't define it in one sentence then it is too complicated for one session.
"If you can't precisely define the response you are training, STOP TRAINING"
Every session should have an objective. If you have 3 consecutive session failures then go and bang your head against a wall 3 times and then review your plan.
PAVLOV & SKINNER - Pavlov's respondent (classical) conditioning works on reflexes, and Skinner's operant conditioning influences voluntary behaviour.
Don't rush to start training. Hang out with Pavlov for a bit - i.e. just feed the animal. How fast does he eat? What is his behaviour like? Is he anxious? Is he ready to start training?
When you're training, Skinner is on one shoulder, Pavlov is always on the other -
"Operant and respondent behaviours lie along a continuum. All operant behaviours are accompanied by respondent behaviours."
"An animal's behaviour may appear totally under operant control, yet there is always an underlying respondent component"
Don't try and train an anxious animal!
(My notes: the emotion that the animal is feeling while learning becomes tied in to the specific behaviour they are learning. This issue is commonly overlooked by horse trainers. Look into 'poisoned cues' for more on this.)
If you're training your animal, you're utilising operant conditioning. You may not be doing it systematically or mindfully or even consciously, but believe me its happening anyway.
GET behaviour with reinforcement.
REDUCE or ELIMINATE behaviour with extinction.
SUPPRESS behaviour with punishment.
Both primary reinforcers (e.g. food) and conditioned reinforcers (e.g. a click, whistle, word etc) strengthen behaviour. Conditioned reinforcers lose strength with improper use, i.e. always pair them with a primary reinforcer! (In simple language - click AND treat, not just click).
The delivery of the conditioned and primary reinforcements is two separate acts.
"Click for ACTION, feed for POSITION"
Generally speaking, the LONGER THE DELAY in delivery of the reinforcer the LESS EFFECTIVE the reinforcement process.
"You get what you REINFORCE, not what you WANT"
Use measurable behaviour. Don't try to work out what you think the animal is thinking. Just look at the behaviour.
If you're getting more of a behaviour, it is being reinforced somehow. This applies whether or not YOU perceive the behaviour to be "good" or "bad".
"Reinforcement of unwanted behaviour is worse than failure to reinforce wanted behaviour" - i.e. if there is junk behaviour in amongst wanted behaviour, you're better off NOT reinforcing it.
80% of training problems stem from issues with A) timing B) rate of reinforcement C) criteria.
Timing must be PRECISE. RoR must make it worthwhile for the animal, otherwise they will find other behaviours more worthwhile. Criteria - if its too high, RoR will be low. If criteria is too low, you are likely feeding the animal for non-contingent behaviours and therefore may be training unwanted behaviours.