What if Classical Conditioning Does Not Work
Science has at times come under attack as being too cerebral, and failing in the “real world” with “real world” problems. At times, it is a valid concern. Many products and services claim to be “science based” or “founded on science,” but for whatever reason, the application fails. To achieve dog training success, it helps to know potential pitfalls – so they can be avoided!
We know that Classical Conditioning is a learning process that evolved early. It’s shared by nearly all, if not all, species of animals can do it. Slugs can do it. Classical Conditioning is used to train bees in research!
Recall that classical conditioning is very simple, but not simplistic. It’s the way cats learn to love the sound of can openers. For a refresher: CLICK HERE. If the animal in question reacts to sound of a cheese slice being opened, from two floors away, out of a dead sleep…they clearly can make robust associations. The problem is not that these dogs do not have the capacity to learn an association. Something else is going on.
There are a five basic reasons why Classical Conditioning seems to fail. All are rather easy to fix.
In part one, we learned that cats learn to like can openers because can openers predict food.
Suppose that the steps are reversed. The cat is given food, and then the scary devil of a can opener sounds.
Practically speaking, it’s a bit absurd. Which is in part the point. It makes little sense to reverse the order. If the order of steps are reversed, then food predicts can openers.
By changing the order of BUZZ – FEED to FEED – BUZZ, Classical Conditioning still works as expected. The problem is that food predicts scary devil sounds. With repetition, we can expect that the cat will become more worried, more apprehensive at the sight of food. It’s a bit like realizing that gifts from a jerk usually come with nasty strings attached. Gift or no gift, it’s not positive.
When people switch to animal training, presenting the information happens backwards rather frequently. Usually, people use treats so they can distract, survive, and perhaps even bribe to get through a bad situation. It happens more often than one would expect. The training set up becomes:
- “I have special treats!…..I’m gong to cut your nails.”
- “Ask strangers to offer food and maybe the dog will go to them and make friends.” (Invariably this leads to “He was brave enough to take the food, try to pet him.” Which is really: “I have food, I’m scary, if you come for it, I will touch you.”)
- “I see a scary stranger coming. Quick get food ready before the dog notices the stranger!….and then the dog notices the stranger.”
In all of these examples FOOD appears first. Food is followed by an “oh crap” realization by the dog. They were trapped and tricked with the promise of good, only to find out that food led to a bad outcome. It’s literally how animal traps work. “Put cheese in the (humane) mouse trap…GOT YA.” Backwards is a trap. It causes animals to be wary of the bribe. Good dog training isn’t a bribe, it’s a thoughtful process.
Classical conditioning still “works” CORRECTLY. Food is warning the animal that bad things are imminently likely. Food becomes the devil.
In the real world, this process serves a purpose. This is how individuals learn to avoid things that are noxious and do harm. Over time, the warning alone is sufficient to trigger fear. The sight of treats causes anxiety, fear and apprehension.
Once the order is correct, consistency is the next issue. Imagine that our feral kitty from part one, lives with a family that eats a significant amount of canned food. Throughout the day, the cat hears BUZZ a number of times. Sometimes the cat experiences BUZZ followed by food. Other times the cat hears BUZZ and is not fed (Because the family is opening soup). Can openers are unreliable in predicting cat food. Some cats might wort out a weak association. (Instead of woo hoo can opener!, the cat might saunter in, mildly optimistic.)
Many people have read that skipping treats is “better”, and like a slot machine. They have been told that the dog will work longer and harder for less food. There is a place for variable reinforcement in skills training. However, in classical conditioning, few repetitions, with long breaks between obtain faster overall results. Don’t work harder. Work more consistently!
Go Big if you’re trying to make a straight Classically Conditioned Association. The brain is wired to notice things that are strange, novel, or surprising. Most people can’t remember when or where they found loose change. However, details about a valuable item such as a wallet or jewelry are easily recalled for years or even decades. Use the good stuff – meat or cheese. Special food “wakes” the brain just like a found wallet, prompting it to search and remember the reason for the amazing food.
If using food rewards for skills training, choose something distinct for rehab work. Use medium value pea sized training treats for skills. Reserve the meat or cheese for addressing fear.
All dogs have past learning that they draw on. It might be obedience. It could just also be things the dog happens to notice. For example, most dogs love bowls. Bring an empty bowl into a dog class and almost all the dogs start to happily, excitedly try to get to it. Most dogs have an association between bowls and food because that is how they are fed.
Imagine, out of convenience, you place some cheese in a bowl. It’s at the ready so that you can easily reach the food while teaching the dog that something is good. For simplicity, let’s assume the dog is learning to like nail clippers. The dog sees the clippers and then the person reaches into the bowl and gives the dog a piece of cheese. Since the cheese is “hidden” in a bowl, assume that the dog is unaware if there is food in the bowl or not.
The bowl already has a past association. Bowls rocks. While nail clippers are being paired in the correct order, it’s possible that the flow of information is “blocked” by the bowl.
Human handlers assume the dog is learning to like the clippers..
The dog knows it has been given food. But what for?
The brain has no reason to look for a new explanation (Clippers led to food), when an old explanation suffices (Bowls lead to food). Bowls do often lead to food. It’s a good mental short cut that “blocks” the new association to clippers.
Fixing this problem is rather simple. Weaken old associations that get in the way. The canned soup eating family demonstrated can be weakened through inconsistency. Hide food in multiple locations so it’s not about the bowl. If a dog thinks that bait bags rock, start wearing an empty one to bore them with it. If a dog starts to worry when the leash is short, gently, shorten the least 3 or so times per walk when there is no dog around. Clear out the old baggage for a fresh start!
This idea is very similar to blocking. While blocking involves interference from former learning, overshadowing involves interference from something happening at the same time. Two pieces of information are simultaneously presented to the animal. One is noticed at the expense of the other.
People do something similar when they are reading subtitles. It’s damn hard to watch the action and read the words. It is also very hard to read the words and listen to them at the same time. One piece of information overshadows the other. It possible to miss large chunks of information.
In addition to these “science-y” issues, a few more can crop up for more serious or complex problems. We’ll focus on them shortly, after looking at how to adjust for these challenging issues.
The thing to take away is that most processes in science have rules. There are times rules can be broken and there are times they should not. As the old adage goes, you have to know the rules before you break them.
Next up: What do you do if the problem is so serious, it’s traumatic for the dog?
Side note: I’ve taken liberties by using many terms that are not terms of science. In the blogs covering fundamentals, I’ve done this intentionally for ease of readability (and because jargon is so damn confusing for those new to these concepts!)