I am a clicker trainer. All of my own dogs are clicker trained and many of the classes that we teach at my training school, AutumnGold are “clicker-centric”. Clicker training is not only a scientifically sound approach to teaching dogs new things, but is also a kind, enjoyable, and bond-strengthening method of training – something that benefits both dogs and their people.
7-MONTH-OLD SIMON LEARNS EYE CONTACT AND SIT/STAY
For the uninitiated, clicker training is a relatively simple technique that involves pairing the click sound made by a small, handheld cricket with the delivery of a food treat. After several repetitions of this pairing (Click-Treat; hereafter CT), in which the click sound reliably predicts the treat, the sound comes to possess the same properties as the presentation of the treat itself – a pleasurable emotional response. Clicker training packs an enormously powerful positive punch for both the dog and the trainer because it allows the trainer to precisely target tiny bits of behavior at the exact moment they are occurring. The click sound becomes analogous to saying to your dog “That’s it!! That thing that you are doing right this instant is what will earn you the yummy treat that is coming shortly! You are SO very smart!”
CHIPPY LEARNS THE DOWN SIGNAL VIA CLICKER TRAINING
A second advantage of clicker training, a benefit that it shares with all training that emphasizes positive reinforcement, is that it shifts a substantial proportion of training control to the dog. This empowerment leads to a dog who loves to learn new things and is eager to “find out what’s clickin’ in each training session“. (Seriously there is nothing not to love about clicker training).
BABY COOPER HEELS FOR CLICKS
So let’s deconstruct the click.
When I was teaching companion animal science at the University of IL, I spent a fair amount of time lecturing about two principle types of learning – Classical conditioning and Operant conditioning. Clicker training provides a great example of a training method that involves both forms of learning:
- Classical conditioning occurs when a subject responds to relationships between two or more stimuli. The basic elements of this type of learning are a meaningless stimulus (initially called a “neutral” stimulus) that elicits no response and a meaningful (unconditioned) stimulus that elicits a response without any prior conditioning. The consistent pairing of the two stimuli, with the neutral stimulus always preceding the unconditioned stimulus, results in a change in the meaning of the neutral stimulus. Because the neutral stimulus consistently precedes and thus predicts the unconditioned stimulus, it begins to elicit the same response that is elicited by the unconditioned stimulus (think Pavlov’s dogs: A ringing bell, and then food). Generally speaking, many classically conditioned behaviors involve emotional responses such as pleasure/joy or fear, with the dog having little or no voluntary control over his/her response.
- Operant learning (also called instrumental learning) occurs as a result of the consequences of a (usually voluntary) behavior. This terminology originates from the concept that we are continually “operating on” our environment, and subsequently alter our behaviors in response to their good or bad consequences. Like other subjects, dogs tend to repeat behaviors that have desirable consequences. We say that these behaviors are positively reinforced (+R).
- What’s the difference? Classical conditioning is concerned with establishing relationships between stimuli and functions as a primary way in which animals learn about their environment. Trainers think a lot about “predictors” in a dog’s world and often will manage a dog’s environment to reduce or eliminate stimuli that predict unpleasant experiences and try to increase stimuli that consistently predict pleasant experiences for the dog. Conversely, operant conditioning involves primarily response-consequence relationships in which a dog learns to volunteer a behavior in anticipation of pleasurable consequences (+R).
Both classical conditioning and operant conditioning take place during clicker training:
Classical: Click (neutral stimulus) consistently precedes and predicts Treat (unconditioned stimulus). After several repetitions, the Click takes on the properties of the treat and is now said to be a conditioned stimulus. Trainers typically refer to it as “conditioned reinforcer” because the CT is used as a +R.
- Operant: Dog offers a behavior (sit), which results in the presentation of CT (positive reinforcement). Dog says “Yum! Sitting results in a treat! I like treats. I will increase the frequency that I offer a sit!”
- Classical: This last one is really cool, because it provides additional evidence for why our dogs SO enjoy clicker training. The voice cue “Sit” is added to the training process when the dog is reliably offering sit for CT. The trainer then begins to only CT when the dog offers the behavior in response to the voice cue (and no other time that the dog sits). Over time, as the dog attains proficiency (offers sit reliably in response to the cue), the cue “Sit” becomes a classically conditioned stimulus because it reliably precedes and predicts an opportunity for CT to the dog (with the operant sit behavior thrown in the middle). This means that the cues that the trainer uses with trained behaviors become imbued with the same characteristics as the click sound; the voice cues themselves become something that the dog enjoys and looks forward to, because they are always paired with an opportunity to earn a CT.
As I said, What’s not to like? Happy Training!
Reference: Clicker training diagram adapted from Canine and Feline Behavior and Training: A Complete Guide to Understanding Our Two Best Friends, LP Case, page 105, Cengage, 2010.
23 thoughts on “Deconstructing the Click”
Thanks foor sharing this
My Rhodesian Ridgeback was clicker trained for an advert he was in. He was 4 months old at the time. I went onto do standard obedience training with him without the clicker. He is 6 nearly 7 years old. The other day a lady in our local dog park was clicker training her dog, at the sound of the click my ridgeback hurtled across the park, sliding to a halt as he came to a sit in front of her, just as he had been trained as a pup. Much amusement all round but it demonstates how powerful a training tool it is.
Hi Carolyn – Wonderful story! 🙂 Thanks so much for sharing!
Here is a different paper, published later than Ms. Wood’s, with different results regarding click vs verbal. I can’t put a direct link since it goes straight to a download, but if you Google the following, you’ll get it. “To click or not to click: Positive reinforcement methods on the acquisition of behavior” by Alexander Blandina at the U. of Florida. His comparison of the different results is quite interesting.
Really good points, Gerry.
The click is technically a bridge but signals “correct and reward is coming” so the “automatic” positive neurological response of anticipation is the same as the reward would elicit, just much milder. This secondarily reinforcing response to the click would fade if not intermittently maintained with a primary reinforcer and it would no longer be a bridge between response and reward.
Agreed! (And Janice explains it better than I did). 🙂
Thanks so much Janice, really really helpful to hear your take on it!
I’m just working on project that uses the click as a bridge stimulus vs. a secondary reinforcer. After some review of the literature, it seems the terms bridge stimulus and secondary reinforcer are often used interchangeably, when (to me) they are very different. One paper was pointing out that Pavlovian conditioning must be performed when pairing the clicker as a bridge stimulus, but Pavlovian conditioned behaviours result in involuntary responses. I don’t see or am curious as to how one would measure the involuntary response that occurs when using the clicker as a bridge (I feel there isn’t an involuntary response, and that the click noise means “move towards the person/device with the treats”, I don’t think this is involuntary).
Which brings me to my next point, using the clicker as a secondary reinforcer (aka conditioned reinforcer) although it’s often said that trainers use the clicker as a secondary reinforcer, I don’t think they really do (I’ve watched a lot of Youtube videos!). A definition of a secondary reinforcer is a stimulus that has been paired with the primary reinforcer and becomes reinforcing. Which technically would mean one wouldn’t need treats anymore as the click is reinforcing. However, I believe when we use classical conditioning to pair “good dog” with petting/scratching, the good dog becomes a secondary reinforcer because it in itself is reinforcing, and continues to be.
I hope this makes sense, not sure anymore! Would love any thoughts of advice.
Hi Erin – Thanks for your note and thought-provoking comments. I will address these sequentially to the best of my ability (others, feel free to jump in!).
1. My understanding of the use of the word “bridge” is that it was originally introduced as a metaphor to convey the connection between the primary reinforcer (food treat) and the conditioned reinforce (click). And, as such, classical conditioning (involuntary response) does appear to take place in that if you are standing close to a dog and have repeatedly paired the click-treat, the dog’s eventual response is to anticipate the consumption of food (salivation, pleasure – involuntary response) when she hears the click, in the same manner in which she would anticipate eating when actually seeing/smelling the food treat. For this reason, the connection between the click and the treat is Pavlovian (classical), rather than operant (instrumental). That said, if one consistently asked for an operant behavior in response to the click and only +R if that behavior was offered, then that portion of the sequence would be operantly learned. So, in essence, you would be inadvertently “throwing in” an operant response when you really do not need to (and in fact, may “muddy the training waters” in terms of the connection between the click and treat). This is why, imho, when we initially “charge the clicker” we really should not ask for any behavior at all (although many training instructions say to ask for “sit” or “eye contact”), but rather try to stay purely within the classically conditioned realm of having the click consistently predict the treat (with no voluntary behavior involved).
2. I think that the issue here is that petting is not truly a secondary (conditioned) reinforcer because social interactions such as petting, praise, even friendly eye contact are not initially meaningless to the dog, but rather they too are inherently reinforcing, albeit possibly with lower reliability/intensity than food treats. The clicker, on the other hand, is expected to have no inherent properties as a pleasurable stimulus at all, so of the two, it alone is a “pure” conditioned reinforcer when used. When we use “good!” then treat, we are really pairing a weak primary reinforcer with a stronger one (food), but it can still be defined as a classically conditioned relationship if the pairing is consistent. Last, essentially, we should be able to not use the food treat with the clicker, but if one does this, over time, we lose the classical connection and the click “loses its power”. This is consistent with Pavlovian conditioning, since the dogs would stop salivating over time if the bell stopped predicting the food over multiple trials.
Great points, Erin. Not sure that I answered what you are asking, or that you will agree, but enjoy the chance to chat about this! Linda
Thank you so much for your in-depth reply. And I think I’ve been struggling with the fact of comprehending “Click equals INSTANT good feelings” (secondary reinforcer) vs. “Click equals someone/something over there gives me treats” (bridge). But as Janice pointed out below, the positive neurological response of anticipation is there in both cases, and when I think of the click technically as a bridge it makes some sense. Sooooooooo interesting. Thanks so much, this is really helping to get through the literature and prepare for my project!
This is straight from Karen Pryors website http://www.clickertraining.com/node/3273
“Click and treat”
“After assembling the basics, pair the secondary conditioner, the clicker, with the primary motivator, the treat. It’s easy! Simply click and then toss your cat a small treat. Wait until she eats the treat and makes eye contact before you click and treat her again. It can take between 5 and 20 repetitions until the cat has a strong positive association with the clicker. The clicker, once paired with the treat, becomes a powerful communication tool, communicating to the cat when she is doing a desired behavior.”
So I do wonder why she waits for the cat to make eye contact, as maybe that’s intended to be a “restart” point, but this operant learning.
Hi Katherine – Thanks for sharing this! Do you know if this study has been published in a journal? I did a search and could not find it anywhere. Regardless – interesting information – thanks so much for reading and for sending this link along! Best, Linda
I just looked, too, and could not find that it had been published in a journal. It looks like it was her graduate degree thesis. Glad I was able to pass it along, though!
I am an enthusiastic but so-so clicker trainer. While clicker training is easy to understand, it is not necessarily easy to master–it is a fairly high level technical skill. Most people can learn to do it, but it requires practice. Precise timing and swiftness are essential, and breaking complex behaviors into trainable pieces is an art. Dr. Sophia Yin is beginning to publish work that precisely measures both performance of trainer and response of dog–pretty fascinating.
Hi. Just a question that I have wondered about for some time. In group classes, I used to use clickers. However with many people I found, holding the leash, giving hand signals and treats was about all they could handle. When the clicker was added, it was more difficult for the pet parent. I switched to the old “good dog” instead and achieved the same results. Is there some benefit to clicking over marking with ‘good dog’?
Hi Charles, Thanks for your comment. We actually face the same issue in our beginner (manners) classes, so make the use of a clicker completely optional. Most of our advanced classes are taught as clicker classes, but again, the clicker is not required and we encourage students to use a clear, one-syllable word (good, yes) as a conditioned reinforce if they decide not to use a clicker. Even that can be challenging if the student does not recognize the difference between giving lots of praise (which of course is fine) and the use of a short, consistent, conditioned reinforce. So, my personal opinion is that using a word is fine, but that it is not quite as precise as is a clicker (see my response to Janice above). Thanks for reading and discussing! Linda
Reblogged this on Dogs On The Ball! and commented:
Clicker science explained- a great read!
Thanks for the reblog!
Has there been direct comparison of the results of using a cue word such as YES! instead of the clicker? If so, was there a difference in acquisition rate and retention? Many people find it difficult to always have a clicker in hand especially dealing with shaping puppy behavior, but the voice is always there and can be used just as quickly.
Hi Janice – Thanks for your comment. I am not sure if there are any published studies that compare the use of a mechanical sound (click) with the use of a word cue (yes, or good) as a conditioned reinforce, but think it would be a great study. I agree that a clicker is not always feasible, and you are right – we always do have our voice with us! I used “yes” for years before using a clicker, simply because we competed in obedience trials with our dogs and my reasoning was that I always had my voice with me, but could not take a clicker into the ring. That said, once I switched (and I do think that both work, BTW), I did find (this is personal opinion, mind you), that the clicker provided a much more precise and clear signal to the dog and was remarkably better at targeting very small increments of behavior. Perhaps this occurs because it is such a unique and binary sound, and we are always talking with our dogs, so the “yes” sound may be less distinct, perhaps? Also, we have had a fair number of students whose dogs were frightened by the sound of the clicker, and this often persisted even when they tried using a softer clicker. In those cases, using “yes” was the alternative and it worked well for them. I would love to see a study with puppies and adult dogs comparing the two – both for acquisition rate and retention. Another call for behavior grad students out there! 🙂
Reblogged this on peakmemory and commented:
Believe it or not, clicker training also has applications for human learning. See http://tagteach.com/ for details.
Thanks for the reblog! Yes, I have heard about Tag Teaching – a very interesting concept! Thanks for the link!