The controversial clicker
Should we stop using a clicker? Is it really needed? A recent study by Chiandetti et al. published in November of 2016 in Applied Animal Behavior Science has raised many questions about the efficacy of the clicker. The study, titled “Can clicker training facilitate conditioning in dogs?” looks at the differences among the use of a clicker, the use of a word as a marker, and the absence of a deliberate marker altogether. Under the specific set-up used in the study, the researchers found no pronounced difference among the three conditions (view abstract here). The authors did not suggest people stop using a clicker in training. They were careful to point out the limitations in their study, and encouraged readers not to jump to conclusions. Yet, that’s exactly what many people have done. Those who don’t like to use a clicker point to the study as proof that a clicker is irrelevant, while those who use clickers take offense, cry foul, and claim research bias. However, most of the questions that I have heard are raised by individuals who have not read the original research paper, and only heard about the study through third-party descriptions.
My thoughts
I think it is important to recognize that studies of this type ask very specific questions that determine how the study is conducted, and the results need to be interpreted with great care. After reading the article, I am not surprised by the authors’ findings. The dogs in the study were trained to push open a breadbox. After training the initial behavior through shaping, the researchers attempted to look at whether the clicker followed by food, a verbal marker followed by food, or a food reinforcer alone affected the dog’s ability to generalize the learned behavior under two new conditions. One condition was labeled simple, a slightly modified breadbox; the second condition, which they labeled complex, required the same motor skills from the dog, but the apparatus was completely different.
The researchers concluded that there were no important differences in how well the dogs generalized to the new apparatus, whether the trainer used a clicker, a verbal marker, or the food-only reinforcer. However, I think it is important to point out that the task was reasonably straightforward, and the trainers were able to deliver the reinforcer immediately after the task was completed. Often, the value of the marker is more pronounced with complex behaviors where precision is a key. The real purpose of the marker is to provide contiguity, an immediate connection between the behavior and the food reinforcer. Because the researchers were careful to make sure reinforcement was available immediately, which is good training, the need for a marker was less critical.
Most important, this study was not focused on the speed of acquisition of the initial task, opening the breadbox, just on the speed of generalization after the initial behavior was learned. The study’s data did show that the dogs trained with the clicker acquired the initial task in fewer attempts and in a shorter time frame than the other dogs. Additionally, the dogs trained with the verbal marker learned the initial task faster than those trained with just the reinforcer. The data indicates an average of approximately 80 attempts (12 minutes) for the clicker trained dogs, 110 attempts (21 minutes) for those trained with a verbal marker, and 125 attempts (27 minutes) with food reinforcement only. However, once the study moved to its primary focus, generalization, the difference between conditions was less pronounced.
Different study = different results
One of my colleagues, Lindsay Wood, did a study for her master’s thesis work in 2006 titled “Clicker bridging stimulus efficacy.” In that study she asked a very different question, and concluded that a clicker was more effective than a verbal marker (view study here). Lindsay’s study involved shaping dogs to touch a target in the middle of the room. She showed that dogs trained with a clicker learned more quickly than those trained with a verbal marker.
Why such different results regarding the efficacy of the clicker? The results aren’t really that different, because the studies asked different questions, used different testing methodologies, and defined efficacy differently. Both studies, from my perspective, were well-designed and carefully conducted. Neither makes sweeping generalizations about what the study proves, and both were careful not to overstate the importance of their results. The Italian study compared the efficacy of the clicker in teaching generalization, while Lindsay’s study looked at the speed of initial behavior acquisition.
But I also believe studies like these highlight the difficulty of conducting research in a controlled setting that accurately reflects real-world training; therefore, it is difficult to draw concrete conclusions from a single study. It is often those with strong opinions about the use of a marker who are quick to want to glorify one study or attack another. Personally, I think both studies have merit. They provide food for thought and allow me to ask further questions about what I can do to make my training better. I have not looked at or questioned whether a clicker is faster for teaching generalization. When teaching generalization, I tend to focus on other things such as fluency, gradual exposure to the new stimuli, and tools beyond shaping (the only teaching technique used in the two studies).
What neither study intended to clarify was the skill level or other tools needed to be a successful trainer. After all, the clicker is just one of many tools that can aid in shaping behavior. There is no magic in the clicker; it is just a toy noisemaker. The effectiveness comes from the way we use it combined with other tools. The click facilitates consistent, immediate communication. The accuracy of a clicker is based on our ability to observe behavior and react appropriately. Like most training tools, it requires skill to use effectively. I know Lindsay to be a skilled trainer and I presume the Italian trainers were also skilled, but it was not a factor that was accounted for in their study, so it is impossible to evaluate that key factor based on the reported results.
Using a marker
So, back to my original question, is a clicker necessary? While it may not be necessary for learning to occur, a clicker is an excellent tool that, when used properly, can facilitate better training. Animals can and do learn without deliberate human-created markers; a marker may not be needed when reinforcement can be delivered immediately. Below are some general guidelines and thoughts about marker signals that I have shared in response to the many recent questions I have received.
Why I use a marker:
- A marker signal is useful in communicating the exact moment that a behavior is performed correctly.
- Animals find predictors of reinforcement in their environment; if you do not teach a marker, they will find a naturally occurring marker. Because of that fact, I like to be more deliberate and teach a specific marker that I can control.
- A marker is helpful for complex behaviors, when the trainer desires to reinforce a precise movement or action, or when the behavior is performed at a distance and immediate delivery of reinforcement is not possible.
- I like the precision and consistency of a mechanical marker, but verbal, visual, tactile, and other types of markers can be effective if used with care and precision under the right conditions.
- Markers assist me in the evaluation and coaching of young trainers, as a marker allows me to more clearly evaluate their timing—it is a useful teaching tool.
My guidelines for marker use:
- A marker is most effective when used while the correct behavior is still taking place, not after the behavior is complete.
- The best marker is one that the animal can perceive easily, is unique to the environment, can be replicated by all trainers easily, and has no previous negative association. For these reasons, the most effective marker will vary depending on the animal and the environment.
- The marker alone will not create stronger behavior; it must be timed right and paired with effective reinforcers.
- Good use of a marker requires practice. I prefer that trainers play training games and develop the motor skills to use a marker in practice before using it with an animal.
Continue asking questions, keep improving
Professional trainers need to be aware of all the tools that are available. Few tools are needed in every training situation, but good trainers understand the relative strengths and drawbacks of each tool. The best trainers will keep asking questions to better understand the techniques we use and to understand the science underlying each procedure. We do not need to have a knee-jerk reaction to studies where results differ from our own experience. It should simply prompt us to ask more questions. Are the study’s findings scientifically sound? If so, why do they differ from my experience? Should I change my practices, adjust my understanding of the principles at work, or simply wait for more data and more information?
Whatever our response, we should remember that we are on a constant journey of learning and discovery. If we keep an open mind, our end goal should be to make our training and communication with animals clear and consistent. As practitioners, we should keep our eye on the science and keep up with new information, but we should never lose sight of our primary objective, the best animal care possible.
Happy Training,
Ken
Post new comment