Classical Conditioning (notes from Clicker Expo)
February 3, 2013 at 9:56 pm
There were some big lightbulb moments for me at the recent clicker expo, and these were related to the topic of Classical Conditioning.
Generally, the term “conditioning” can be a bit misleading because it tends to conjure up images (for me, anyway) from The Clockwork Orange or The Manchurian Candidate and suggests loss of free will, as if we are turning our dogs into robots. Technically speaking, the term “conditioning” simply means learning, and Classical and Operant Conditioning refer to the ways all living beings LEARN.
According to Dr. Susan Friedman, Classical & Operant Learning are always working together in real life. They always overlap. We artificially separate the concepts for teaching.
In Kathy Sdao’s seminar on “Classical Counterconditioning for Agression” she draws out the differences between Operant and Classical Conditioning:
Operant Conditioning/Learning happens in the realm of observable behaviors that we can mark and reward. These behaviors are freely chosen by the animal in order to earn reinforcement or escape punishment. Clicker training is Operant Learning. Golden rule: Behavior is driven by Consequences.
I did this drawing after my last Clicker Expo
Classical Conditioning/Learning on the other hand, is in the realm of reflexive or respondent behaviors – all the hardwired emotional, subconscious stuff that an animal has no choice over. These reflexive behaviors (eg, Flight or Fight) are learned through repetitive association and tied up with survival in some way. You pair something neutral with something that elicits “excitement” often enough, and the neutral stimulus will trigger off excited feelings. You repeatedly pair something neutral with something scary and the neutral thing will trigger fearful emotions. Antecedents lead to Behavior.
“Classical Conditioning is a powerful foundation for Operant Conditioning. Classical Conditioning will not get new behavior. It will put existing behavior under different antecedents.” – Kathy Sdao.
At Clicker Expo, in different presentations, this memo came up several times: We can’t start clicker training an animal who is fearful or anxious.
Classical Counter Conditioning is the first thing that should happen in order to calm the limbic brain, before the animal is able to “behave”. In the case of triggers that elicit fear responses, we pair these with very good things. This memo came up in Julie Shaw’s and Debbie Martin’s “Behavior Modification Clinic” Lab and also in Sarah Owing’s presentation about helping “WallFlower Dogs”.
Note: This example is based on real life. Whenever Boogie hears a “ding!” bell, he runs to the window and barks. Even if the “ding” is coming from the kitchen, the TV or if I accidentally touch a glass with a spoon. Maybe in his previous home, this “ding!” sound was the doorbell.
Here’s a much more detailed illustration on Counter Conditioing that I did for the Ahimsa Dog Training Manual:
Kathy Sdao talked about the ways in which Counter Conditioning can be ineffective due to these common mistakes:
- Weak Unconditional Stimulus. (the toy or treat is not valuable enough; the love of this is not stronger than the fear of the trigger)
- Trainer’s hand is in the treat bag and the dog is too focused on this
- Rhythmic trials. The ” trigger + treat” event happens at regular intervals to become predictable.
- Inadvertant Avoidance Conditioning. eg, if we keep treating before the dog sees the trigger, we might accidentally condition the food to become a “warning signal”
- We present the treat without the trigger (eg, dog gets the high value treat anyway, when nothing happens) – treat loses value
- Contingency issue. If we forget to treat when trigger appears
- NOT following up with Operant Conditioning.
Classical Conditioning is also not considered practical in the real world or in the long term because it is too easy to not do it correctly 100% of the time for it to be effective. This is why we need to follow up with Operant Conditioning of replacement behaviors, which made me think instantly of BAT…
Another example of Classical Conditioning was in Ken Ramirez’s Lab on creating value in “Non-Food Reinforcers”. He shared a story about a whale (or dolphin?) that wasn’t feeling well and wouldn’t eat and they needed him to take his antibiotics. As they were unable to reward with food, they used “reinforcement substitutes” like belly-tickling, clapping hands, praise etc. and these were just as reinforcing to the whale because they had been previously paired with food over a long time.
Similarly, we can train any novel stimulus - a toy or a human action (eg, clapping hands, thumbs up, “Good Boy!” etc) to be reinforcing if we pair this often enough with Primary Reinforcers (food, social interaction, play) during training. This pairing has to be maintained so that the non-food reinforcer stays emotionally meaningful to the animal. Some Ken Ramirez quotes:
“Yes, this is exactly like charging a clicker”
“A toy is not intrinsically reinforcing. It is reinforcing because it is paired with the Primary Reinforcer of PLAY”.
He also talked about learning how exactly your dog likes to play with a toy. Every dog is different. Example with tennis balls: Some dogs prefer chasing and fetching; some like to chew, or roll the ball around; some dogs like to peel the skin off. (Boogie is all of the above) Similarly with “touch” as a reinforcer. Each individual animal likes to be touched a certain way and only by certain people that he/she already has a relationship with.
“Value disappears from a conditioned reinforcer if you don’t know how to maintain it”
Susan Friedman in her closing speech at Clicker Expo also brought up the classical conditioning aspect of clicker training. A clicker is also a conditioned reinforcer… we infuse it with emotional value and meaning because it gets paired with food. The click not only marks behavior, it also elicits respondent behavior/happy emotions (“woohoo! I got it right!”) Not only do we have to be precise with our clicking, we also have to ALWAYS back up every click with a treat to ensure that this powerful training technology is effective.
She showed video examples of trainers not using a clicker correctly… eg, trainers who click several times before giving a treat OR trainers with animals who are responding to other signals and are not getting what the click means and are focusing on the food instead… and as a result, the animalsdon’t perform as requested or they get frustrated and walk away, or get cranky.
Susan Friedman: “If you click, dammit, TREAT!”
If you were at Clicker Expo and if I have misinterpreted any of the information in this blog post, please feel free to let me know!
Next blog post: When animals make mistakes – dealing with these in the least intrusive way.
Entry filed under: Articles, links, Events, Training. Tags: .