A good study, and not surprising
Let me begin by summarizing: the main goal of the study was to compare the speed of new behavior acquisition in naïve dogs under three conditions: immediate delivery of a primary reinforcer, delivery of a verbal marker followed by a primary reinforcer, and delivery of a click followed by a primary reinforcer. The authors asked three separate questions under each of those conditions, all focused on speed of learning by young dogs with no previous training experience. The study concluded that, under the conditions of the study, a skilled trainer can teach a behavior just as quickly (if not more so) with just a primary reinforcer than with a clicker followed by a primary reinforcer.
I find the study to be well-designed, and I don’t find it controversial or surprising. As trainers, we know that under the right conditions the immediate delivery of primary reinforcement can be more effective than using a secondary reinforcer. However, it is also notable that the study, under those conditions, found no statistically significant difference in the speed of new behavior acquisition between clicker use and the use of primary reinforcement only.
Hyperbole can get us in trouble
When trainers are passionate about a topic, there can be a tendency to use hyperbole to make claims that are not necessarily valid under all conditions. For example, to make a point in some contexts, trainers might oversimplify and abbreviate the negative impact of punishment on learning by claiming that “punishment does not work.” But that statement alone would not be accurate. Sweeping generalizations can harm our credibility, despite best intentions. The Gilchrist et al. study examined specific claims such as, “the clicker speeds up training” or “new behaviors can be learned faster when a clicker is used.” The authors set up a study to test the accuracy of those types of statements, given the procedures and conditions specified in the study. The title of the study is provocative, but that’s just a title. My goal is to look at any study both critically and with an open mind.
A good study prompts reflection and evaluation
This study does not necessarily negate what we have experienced. It simply encourages us to look at the procedures carefully and consider the generalizability of the study’s finding to the situations in which we use similar procedures. In this study, under the parameters they set up, the authors demonstrated that the clicker did not increase learning speed for the targeted behaviors. In the study, the authors were careful to point out several rival hypotheses that might explain their conclusions. They stated a variety of reasons that might influence the study’s results, which is a standard research practice. In evaluating research, I always caution readers not to interpret more from the data than is actually presented; it is important to look critically at the parameters of a project while keeping an open mind.
Uses of the clicker
There are, in fact, training situations when I might choose not to use a clicker. Broadly speaking, if I can deliver primary reinforcement quickly and accurately, and I can meet my other goals, I might choose to use primary reinforcement without incorporating a clicker. The study doesn’t change those practices. It has been my experience that under most training conditions it is quite difficult to deliver the primary reinforcer directly and achieve the needed level of precision.
I am a strong proponent of using some type of signal to mark behavior in training. I find the use of a clicker (or other distinct stimulus) valuable for many reasons; the speed of behavioral acquisition, the focus of the Gilchrist et al. study, has never been a primary reason. In my personal experience, it seems to be true that animals learn faster when a clicker is used—but studies such as this one will help me examine that perceived benefit more critically.
Under the study conditions, the clicker does not appear to slow down or speed up behavior acquisition. However, there may be other conditions (more complex behavior, older or more experienced dogs, different species, and other reasons discussed below) where the clicker may be helpful in speeding up learning. These are interesting considerations that will require more investigation.
Training conditions (outside the scope of the study)
There are many conditions under which I find a clicker helpful, and I would welcome research on them. Here are a few:
- Working at a distance. The Gilchrist et al. study attempted to look at the effects of working at a distance, but the researchers worked only within a single room, and the greatest distance achieved in the study was about 20 feet. For me, the advantage of the clicker working at a distance is when the animal is out of sight and cannot see the trainer. The click (or other marker signal) allows the trainer to communicate accurate completion of the goal behavior and may serve as both a secondary reinforcer for the behavior completed and as a cue to come back for primary reinforcement.
- Behaviors that need to be repeated or need duration. I often train animals to offer behavior continuously, and the clicker lets them know when to stop and come back for primary reinforcement. Examples include vocalization (the dog barks when cued and continues barking until he hears the click), medical behaviors (the animal moves into position for a medical exam where duration may vary and he may not be in a position where he can see me, but he continues to offer the behavior until he hears the click), or dolphin jumps (when I worked with dolphins they were taught to leap into the air over and over again, until cued to stop and come to me for reinforcement). Although the Gilchrist et al. study looked at the duration of a simple behavior, the trainer in the study was able to deliver the reinforcer immediately at the end of the required duration, which is not the case in many of the practical scenarios I have described.
- Precision for complex behaviors. The study looked at a particular type of precision but did not investigate some of the more intricate ways that a clicker is useful in certain complex contexts. In film work, for example, we often found the need to change details of the behavior after the director looked at the behavior through the camera. We were often challenged with shaping the behavior on set, without having the time to go away and retrain. I worked with dogs that learned direction cues like right, left, up, and down, and the clicker was used to pinpoint exactly where their heads should be pointed. This specificity was critical to the camera shot and I found the clicker to be a very efficient way of getting that specificity quickly.
- Consistency. When multiple trainers are training the same animal, in an organization or a family, it is helpful to the animal to provide a consistent, coherent system of communication. I worked for many years in a zoological setting and I saw real advantages to having a consistent way of communicating to the animals—a system that would not be susceptible to the variability of the voice and the awkwardness of body movements. The click or the whistle created consistency across the team.
- Contiguity between behavior and reinforcement. Many trainers struggle with timely delivery of the reinforcer. In my experience, a clicker helps those trainers keep the reinforcer connected to the behavior while they improve their skills.
- Teaching tool. The clicker is a useful tool in teaching new trainers about the importance of timing and criteria. As a teacher of trainers, I have found that the use of the clicker helps them focus on criteria and attend to the animal’s movement to determine the right timing of reinforcement.
- Evaluation of training. As a consultant and teacher of trainers, I have found that when trainers use a clicker, I am able to better evaluate their timing and criteria.
Other elements of the study
There are many other aspects of the study that are worthy of mention. The authors accurately point out that the use of terms like “marker” and “bridging stimulus” can be confusing. While there are scientific references that use very specific meanings of those terms that are not synonymous with “secondary reinforcer,” in our community we tend to use all three terms interchangeably. That is a subject worth exploring at another time. The authors also used only one trainer for all of the trials, which helped control for skill differences that may have existed if they had used multiple trainers. But using just one trainer limits the generalizability of the results to other trainers. Also, the trials were not blind (the trainer knew what she was looking for) and that may have affected results. As with all research, additional replications would be helpful.
So, what does it all mean?
It is important that we welcome the critical scientific evaluation of what we put into practice every day as trainers. We must also continually adapt how we train as we discover more effective and ethical practices that achieve desired training goals. Training is such a wonderful process, a process that we know helps the animals in our care. It is important that we continue to evolve together and learn from both the science and the practice.