Subscribe to the weekly “audio edition” via iTunes
y daughter will often practice the piano by herself, but lately, I’ve been sitting down with her and trying to be her practice buddy. Essentially, to make sure she doesn’t make the same mistakes repeatedly and develop bad habits.
Of course, despite my best intentions, this basically means that I can’t help but pounce on every missed note or rhythmic detail that seems to be going past her radar unnoticed. I might start with a gentle “nope” at first, but it quickly escalating to a firm “no!”, followed by a “hold it!!”, “stop!!!”, and eventually a “HEY! LOOK AT THE SCORE!”.
All of which she finds rather upsetting, and makes her want to avoid practicing. And I can’t blame her, because who wants to hear “no!” all of the time?
Meanwhile, our dog “DJ” seems to looove “practicing.” He can barely contain himself when my daughter works with him on a new trick (his latest one, “sit pretty” is the one you see above).
Why is there such a difference?
(When I asked, my daughter retorted “treats.”)
And sure, that might be part of it, but when you step back and take a look at the learning process I use with my daughter and that which she uses with our dog, there’s actually quite a big difference between the two approaches. While I seem to be focused on all the times she gets something wrong, she is focused on all the times DJ gets something right.
So…err…is it possible that she is using a more effective learning strategy with our dog, than I am with her?
Some dog trainers swear by a technique known as “clicker” training, which is essentially a form of operant conditioning – a learning method that goes all the way back to the 1900’s.
The gist of operant learning is that we are more likely to repeat a behavior if something good happens after we do it.
So when our dog hears us say “sit pretty,” and then jumps up, lays on the ground, barks, runs to grab a toy – and suddenly gets a treat when he accidentally sits back on his hind legs with his paws up like a chipmunk, his little doggy brain starts to learn that there might be a connection between “sit pretty,” sitting with his paws up, and his yummy treat.
It’d be easier if we could simply explain to him how to sit back, balance his body, and lift his paws up, but since we can’t, he just explores a range of behaviors, until he finds the one that leads to a treat. And over time, he starts to learn that when we say “sit pretty” it means we want him to sit up like a chipmunk.
This is essentially how clicker training works, except that you use a clicking sound instead of a treat, so that you can more precisely “highlight” the desired behavior at the exact moment the dog does it.
So in other words, instead of highlighting all of the mistakes a learner makes with a emotionally-charged “no!”, this method allows the learner to explore a little more freely, until you catch the learner in the act of performing the desired behavior.
Interesting…but does this work for humans?
This was recently tested out in a medical setting, where 23 med students and surgical residents were recruited to see how this learning method might compare with a more conventional approach.
Half of the participants were assigned to the clicker training group, while the other half comprised the control group.
Each learner was given the same task – learn how to tie a variant of the ''Tennessee Slider'' knot.
The control group
Participants in the control group began by watching the instructor demonstrate the skill twice, along with some verbal instructions.
Then, they were given a diagram illustrating how to tie the knot, and 15 minutes to learn the new skill.
After completing the 15-minute learning phase, each learner was given 15 minutes of practice time (the idea being, to get in ~100 repetitions before the final test).
The clicker group
The clicker group also watched a demonstration to start with – but then the instructor walked the learner through a script, in which the skill was broken down into 6 steps.
For each step, the learner would receive
- instructions (“The goal is to wrap the long end of the rope around your right palm and then hold it with a “fakey” pinch)
- a demonstration of that step
- and a “tag” point (a phrase, like “fakey pinch”, that served as a reminder of the essential part of the skill that would be marked by the clicker sound if they succeeded in executing it correctly)
Then, the learner would try to perform that subskill until they heard a click (or a short, emotionally-neutral “good”), signifying that they just did it correctly. Once they achieved a click, they would do 5 more repetitions to reinforce the skill before moving on to the next step in the sequence.
And when each of the 6 subskills had been successful learned, they were linked together to ensure that the learner could successfully perform the knot-tying procedure from beginning to end.
All followed by 15 minutes of practice, just like the control group.
To determine how effectively the skill was learned, each learner finished with a test of their knot-tying skills, where they were asked to tie the knot 10 times, while their instructor observed and evaluated the precision of their technique.
All 12 participants in the clicker group passed the test with flying colors, successfully tying 10 knots with proper technique.
The learners in the control group on the other hand, struggled. Only 4 of the 11 participants managed to tie 10 knots with proper technique.
Different rates of learning
Although the clicker group seemed to learn and perform the skill with greater precision, there was a downside of sorts – time.
It may not be a hugely significant downside (and might actually be indicative of a deeper learning process), but it took participants in the clicker group 1-2 minutes longer on average to progress from their first instruction to successfully tying their first knot (342 seconds vs. 279 seconds).
One last interesting tidbit
One of the participants in the control group failed to figure out how to tie the knot in the allotted 15 minutes. So after the test, the learner sat with the instructor, and using the operant learning method, successfully tied the knot in 201 seconds. And yes, it’s not fair to compare this time with the others, since this learner already spent 15 minutes trying to figure this out on their own, but still…makes you think.
Operant conditioning is not a new concept, but the way in which these principles are applied to learning, is not something I had ever given much thought to.
Specifically, the focus on letting the learner explore and feel their way towards the goal, and marking the correct movement when it happens. As opposed to focusing on all of the imperfections and creating a learning climate centered around simply avoiding mistakes. Somewhat akin to the difference perhaps, between playing a game to win vs. playing the game so as not to lose.
There’s more study needed of course, but in the meantime, I’m tempted to experiment with this the next time my daughter runs into a gnarly technical passage that needs a bit of problem-solving…after all, we do have a few clickers lying around somewhere (although just to be clear, there’s nothing magic about the clicky sound – a simple verbal cue like “good” is just fine too).
Have you ever tried something like this in your teaching?
To learn more…
I first learned of this study on the NPR podcast Hidden Brain. Check it out below, as the episode explores the concept and nuances in greater depth, and also paints a more vivid picture of what this kind of training looks like in practice.
When Everything Clicks: The Power Of Judgment-Free Learning @Hidden Brain podcast