Summary: A report on a research study conducted into the relative efficacy of using a clicker paired with a food reward, compared to using food alone. The hypothesis was that the clicker would lead to faster acquisition and that this would be more pronounced on a complex or potentially confusing task compared to a simple one. The study found that the clicker as a marker was no more effective than using food alone as a reinforcer for the simple task, but it was more effective for the complex task. 

Introduction: Why study clicker training?

Clicker training is a method of dog training that has increased in popularity over the past 20 years.¹ Clicker training was introduced to the general public in an article in Scientific American by Skinner² who originally proposed the use of a cricket (the small plastic children’s toy that is now known as a clicker) because it is high pitched and easily attended to by the animal. Clicker training was further popularized by Keller and Marian Breland, who studied under Skinner.³ Breland and Breland (and later Bob Bailey) are considered the pioneers of using a clicker, or clicker-like devices (whistle) to train a variety of species.⁴ Clicker training’s explosion in the dog training field is primarily attributed to Karen Pryor (, who has published multiple books on animal training and articles on training dolphins and porpoises to offer creative and innovative behaviors.^5-10

Currently, clicker training (also called marker training in the training industry to incorporate all clicker-like devices  including voice, whistle, and vibration) is a supported method of training by independent animal behavior and training certification bodies such as the Certification Council for Professional Dog Trainers, and the International Association of Animal Behavior Consultants.^11,12 However, even with its long history, very little research exists to support these claims in pet dogs. In addition, the function of the clicker is also debated.

Function of the clicker

While competing theories exist to explain the function of the clicker, very little research has been done to investigate the underlying mechanisms. Feng et al.¹ and Dorey and Cox¹³ conducted systematic reviews of proposed theories as to why clicker training is effective and suggested that it appears to work as a conditioned reinforcer (pairing a neutral stimulus with an effective reinforcer), though more research is needed to confirm this assertion. Others have suggested that it may work as a bridging stimulus.⁵ However, given the click’s short duration (known as trace conditioning), this is contradictory to robust research that suggests bridges are more effective when the sound is played from desired behavior to delivery of the primary reinforcer (delay conditioning).¹⁴ Finally, the clicker may function as a marking stimulus, an unexpected and novel event that causes the animal to attend to the behavior immediately preceding it.¹⁵ However, as Dorey and Cox highlighted, a clicker is not unexpected or novel (as it is used repeatedly to train an animal), so its function as a marking stimulus is questionable.¹³

Assuming that the clicker functions as a conditioned reinforcer, it is important to understand the research regarding a clicker’s efficacy in the learner’s acquisition of new behaviors when compared to using food only. Otherwise, introducing a clicker to pet owners may not be necessary. Forgoing it would avoid the challenges of using a clicker (timing, coordination) as discussed by Feng et al. (2017).¹ However, because there are only five studies that investigate the use of clickers with domestic dogs, it is important to also consider research that was done with other domesticated animals to get the full breadth of research conducted.

Clicker training efficacy

Systematic reviews of existing research on the effectiveness of clicker training with domesticated animals (e.g., goats, horses, dogs) have mostly found the clicker to be effective at training new behaviors, but no more effective than using food alone.^13,1

In the first study of its kind, Feng et al. (2018) examined the use of clickers with pet dog owners, as opposed to professional researchers or trainers.¹⁶ Because the reported use of the efficacy of clickers comes from dog trainers who train dog owners,¹ this was a significant step at trying to understand the anecdotal evidence of clickers being more efficient when teaching new behaviors. The study evaluated 45 dog-owner pairs divided across three groups: two groups that took part in a six-week basic obedience course with blinded instructors for the clicker-plus-food and food-only conditions, and a third group that was a waitlist control that did not take part in the six-week course. The researchers used survey data and behavioral measures to assess impulsivity and dog-owner relationship and found no significant difference between the clicker-plus-food and food-only groups. The researchers also collected data on owner-reported training experiences. The clicker-plus-food owners found teaching their dog to nose-target an object easier, but no differences were found for any other behaviors. The study concluded that no significant difference between clicker-plus-food and food-only groups was observed.

Notably, owners using clickers perceived teaching their dog to nose-target an object as easier than the group that used food-only, but there was no evidence to suggest this was due to clicker efficiency rather than preferred training method. This study did not measure efficacy of the acquisition of new behaviors. However, we know that owners perceived an object-targeting behavior easier to teach with a clicker based on survey data, so this behavior is a good candidate to use when working with pet owners to explore acquisition of new behaviors.¹ We currently do not have any studies that measure the efficacy of clicker-plus-food and food-only groups when pet owners are teaching their family pets, even though efficacy is supposed to be one of the major benefits of clicker training hypothesized by Pryor.⁵

Goal of this study

The goal of my study was to further investigate the use of clicker-plus-food compared to food-only with pet dog owners. More specifically, we aimed to assess the efficacy of pet owners teaching their dogs two moderately challenging skills: a cone-targeting behavior and a 360-degree spin, using these two reinforcing methods. The cone-targeting behavior was chosen because it has been reported that owners find it easier to teach that behavior with a clicker.¹⁷In contrast to the cone-targeting behavior, a second behavior was chosen that requires the owner to move in a way that may communicate multiple social cues and could potentially confuse the dog. Rossi et al. (2014) discussed how dogs can be sensitive to head movements, body posture, and pointing,¹⁸ and Edwards et al. (2019) discussed how interactions with dogs in a veterinary setting, such as restraint, leaning over, and staring, can cause distress and raise welfare concerns.¹⁹ A spin behavior, taught with a hand target, was chosen because it requires the owner to be close to their dog, and involves leaning over and other potentially confusing body positioning. This behavior presents an opportunity to investigate whether the clicker can be effective in training a behavior and allow the dog to clearly understand what is being reinforced in the presence of a human who may be moving in a manner that could be confusing in terms of what is being asked.

Hypothesis of this study

Therefore, our hypotheses for this thesis are:

1. As owners found teaching the cone-targeting behavior with a clicker easier than without,¹⁶ we expect to see more dogs acquire the behavior in the clicker-plus-food group than in the food-only group.
2. Given the added body language that may confuse the dog while learning the spin behavior, we expect to see more dogs acquire the behavior in the clicker-plus-food group due to the anecdotal evidence of a clicker giving the owners more precise timing.⁵
3. In addition, because we expect more dogs to acquire both the cone-targeting and spin behaviors in the clicker-plus-food group, there would be a higher percentage of goal behaviors (touching the cone target or performing the full spin) when divided by the number of trials due to the faster/higher occurrence of acquisition.
4. And finally, we expect to see more goal behaviors (touching the cone target or performing a full spin) performed in the clicker-plus-food group due to the higher occurrence of acquisition.

Materials and Methods

Participants and facilities

Fifteen participants were recruited via a social media platform to complete the research by attending a one-hour group training session (without their dog), followed by a 90-minute training session with their dog that was scheduled one to three weeks following the group session.

Seven adults, six women and one man between the ages of 29 and 60, were randomly assigned to the clicker-and-food group. The seven dogs in this group were five females and two males between the ages of 1 and 10. The dogs in this group were owned for one to nine years. All dogs were mixed breeds, with the exception of one Bull Terrier.

Eight adults, seven women and one man between the ages of 26 and 69, were randomly assigned to the food-only group. The eight dogs in this group were three females and five males between the ages of 2 and 10. The dogs in this group were owned from three months to ten years. All dogs were mixed breed, with the exception of one Australian Shepherd.

Materials

Owners were provided with the choice of using one of three food treat choices: Happy Howie’s beef roll, Happy Howie’s lamb roll, and Happy Howie’s turkey roll. Multiple treat options ensured that all treats had the same texture and consistency but provided flexibility in case of diet restrictions for the dog. Treats were cut into small cubes, approximately the size of green peas. At least 100 treats were available for each training session. Small pea-sized turkey and hot dog treats were also available for dogs who were not interested in the Happy Howie treats.

The treats in a bowl

A small dog bowl was provided to hold the treats during the session.  Owners were instructed to keep the bowl of treats on a 3-foot rolling table, or to carry it while holding it behind their back. This provided easy access to treats while training and ensured the dog could not eat directly out of the bowl.

A target stick from the Treat and Train device was used for the targeting training sessions.

Target stick

A Nest Cam was used to record the training sessions on video. Walkie-talkies were used to communicate the start and end of the training and rest sessions to the owner, without disturbing the dog. The stopwatch app on the iPhone was used to time the training and rest sessions.

The filming setup

Procedures

First, owners were invited to a group session, without their dogs, to learn about the two behaviors that they would be training and how those behaviors would be trained. Owners were randomly assigned to either the clicker-and-food group (owner reinforces behavior by clicking then providing a treat) or the food-only group (owner would reinforce behavior by providing a treat only).

Sessions for the groups were separate as not to allow the participants to know the  experimental conditions. The meeting started with a group session designed to teach the concepts of reinforcement, treat delivery, and how to train the desiredbehaviors (cone target and spin). This group session ensured that all owners came into their training sessions with equal knowledge. Next, groups were split into clicker-plus-food and food-only, and did an exercise designed to teach reinforcement timing. Participants were then shown a two-minute video of the spin behavior being taught to a dog and given a protocol on how they were to do this, starting with a hand target. Participants were told that on their training day, they would perform three three-minute training sessions, with a five- to ten-minute break between each session.  They were instructed to train for the entire three minutes, and repeat the behavior if their dog acquires the behavior in the session.

Next, a video was shown of the cone targeting behavior. The participants were told that they would be shaping successive approximations of their dog touching the cone and given a four-step protocol of how to do this.

After the group session was complete, owners scheduled their training session with their dog within one to three weeks. These sessions required that only one owner and their dog were in the training room at a time to avoid distractions. Owners were advised to feed their dog a light meal before the visit (light dinner being defined as 33% of their normal meal amount).

Before starting the first training session, participants were instructed to pair the clicker with food (clicker-and-food group) 20 times by clicking and feeding the dog one to three seconds after the click. For the food-only group, owners were instructed to practice delivering food 20 times.

Once completed, owners were instructed to use the walkie-talkie to inform the researcher. Next, the researcher explained that the remainder of the session will be six three-minute training sessions, followed by a five- to ten-minute rest period in between training sessions. The first three sessions would be used to train the first behavior, and the last three sessions would be used to train the second behavior (half the owners taught spin first, while the other half taught target first). The participants were also shown a quick recap video of the two behaviors that they were taught in the group session.

The researcher would then use the walkie-talkie to ask the owner if they were ready. If they confirmed, they were asked to put the walkie-talkie down and told that the training session would start in 30 seconds, with the researcher counting down the last ten seconds followed by “Go” to start the session.

The owner would then train their dog on the specified behavior for three minutes, with the researcher using the walkie-talkie at the three-minute mark to tell the owner to stop. The owner would then stop training and use the walkie-talkie to confirm that they stopped the training session. The three-minute training session was followed by a five- to ten-minute break for the next five sessions (during which time they could use the water available for drinking). For session four, the owner was instructed to start training the second behavior.

Once all sessions were complete, the owner and dog were debriefed on the purpose of the experiment. The owner was presented with a certificate recognizing their dog for helping with research, and the owner’s address was also confirmed to receive a gift-box that includes dog toys and treats.

Certificate of completion

Results

Coders were asked to watch the videos and score each reinforcement trial with an R, N, or S. R was scored when an owner reinforced successive approximations to the cone-targeting and spin behaviors. For example, during the training sessions for the cone-targeting behavior, an R would be scored if a dog was reinforced for approaching the cone, looking at the cone, or approaching and looking at the cone. For the spin behavior, an R was scored if the dog successfully touched the hand target of the owner. N was scored when the owner appeared to be resetting the dog. For example, in the cone-targeting sessions, if the dog neither looked at nor approached the cone for more than five seconds, owners were instructed that they could toss food to reset them and to help avoid the dog feeling frustrated. Similarly, for the spin behavior, if the dog seemed confused by the hand target and did not target the hand after 5 seconds, they could toss a treat to reset the dog. While this may potentially be confusing for the dog, this step was added to prevent the dogs from feeling frustrated when working with novice trainers who were teaching complex skills through shaping. Finally, an S was scored for a trial if the dog touched the target with their nose in the cone-targeting videos, or completed a three-hundred and sixty degree spin for the spin videos. A behavior was considered acquired when four S’s were scored, or three out of four trials were scored with an S.

Table 1 shows the total number of times the behavior was acquired in each condition along with the means for total trials (R + N + S), the number of times the desired behavior was performed (S), percentage of trials that were scored S (S divided by R+N+S), and total trials until the behavior was acquired (the fourth S scored or when three out of four trails were scored an S).

To understand how the clicker-plus-food and food-only groups compared, Fisher’s exact test was used to compare these two conditions for both the spin and cone behavior using the acquired data (Y and N) from Table 1. For the cone behavior, three dogs in the food-only group and one dog in the clicker-plus-food group acquired the behavior. However, there was no significant difference between the number of dogs that acquired the behavior between the groups (p = .28, Fisher’s exact test). For the spin behavior, all three dogs that acquired the skill were in the clicker-plus-food group; no dogs in the food-only group acquired the behavior. For this behavior, the results are potentially trending significant (p = .077, Fisher’s exact test).

Finally, because our data was non-parametric, Mann-Whitney tests were run to analyze the total number of trials, the total number of times the goal behavior was performed (S), percentage S (S divided by the number of total trials), and number of trials required until the behavior was acquired as reported in Table 1. When these variables were analyzed for the cone behavior, total trials (U = 21.00, p = .42), total S (U = 12.50, p = .070), S percentage (U = 21.00, p = .16), and trials until acquired (U = 1.00, p = .48), there was no significant difference found between the clicker-plus-food and food-only conditions.

Similarly, for the spin behavior, total trials (U = 13.50, p = .09), total S (U = 19, p = .18), and S percentage (U = 16.00, p = .16) were not significantly different between the clicker-plus-food and food-only conditions. Number of trials until acquired could not be compared for the spin behavior because no dogs in the food-only group acquired that behavior.

Discussion

For the cone-targeting behavior, there was no significant difference in the number of dogs that acquired the behavior across the two training methods. This finding aligns with previous research by Williams et al. (2004) and Smith & Davis (2008) which found no significant difference between acquisition rates when domesticated animals were trained with a clicker-plus-food or food-only.^.20,21 However, in the case of the spin behavior, there was a difference in the number of dogs that acquired the behavior in the two training groups. All three of the three dogs that acquired the spin behavior were in the clicker-plus-food group, and none of the dogs in the food-only group acquired this skill. This seems to suggest that a clicker may be beneficial for dogs in situations where conflicting information is present, such as added body language from the handler.^18,19 However, it is important to note that these results were based on three dogs and only show a trend, thus caution must be taken when interpreting this finding.

When examining the total number of trials, the number of trials the goal behavior  was performed, the percentage of trials in which the goal behavior was performed, and trials required to acquire the behavior, there was no significant difference in either group. This suggests that using a clicker-plus-food does not result in any efficacy benefits compared to food-only when measuring how often the desired behavior was performed, either as a total or as a percentage of trials. In addition, in the case of cone-targeting, there was no difference in the number of trials to reach the behavior acquisition criteria. A striking difference was found in the case of the spin behavior; only three dogs in the clicker-plus-food group reached criterion of acquisition, and no dogs in the food-only group acquired that skill.

The results of this study support what previous studies have suggested, that there is no significant difference in the acquisition of new behavior with a clicker-plus-food versus food-only, but there may be a difference depending on the behavior being taught. Therefore, claims that clicker training leads to faster acquisition of behavior compared to using food only need to be treated with caution. Additional investigation is required to determine the types of behaviors or environments in which a clicker may be more effective than food only in teaching new behaviors.

In this study, for example, it appears that pet owners benefited from using a clicker when teaching a skill that required lots of extra body movements and leaning over their dog; this can lead to confusing or distracting information from the handler, something the clicker may be able to overcome. These types of behaviors from the handlers can also cause fear in a dog, so using a clicker in husbandry training (cooperative veterinary and grooming care) should be investigated further based on the findings of this study.¹⁹

In conclusion, statements about clicker training being more effective than using food only may not be accurate. The efficacy of clicker training in certain situations, including those that include distracting environments, or when training behaviors that require handlers to use body language that could be perceived as confusing or threatening, needs further investigation.

In particular, more research is needed to investigate if and when clicker training is likely to be more effective for the pet-owning public, professional trainers, and researchers. Furthermore, given that the evidence suggests that clicker training is effective, but no more effective than using praise paired with food or using food only, it is important to give pet owners control over what reinforcement method they feel works best for both them and their pet dog.

The relationship between the trainer and the animal is especially important in the training of a highly social species like dogs. Thus, selecting a reinforcement method that works best for the individual relationship may be more important than the general effectiveness of reinforcement methods. Breland and Breland (1961) discussed the limitations of behaviorism, and the need to use the concepts of behaviorism after first understanding the species and the individual being trained from an evolutionary and ethological perspective. Because dogs have lived within our societies for 20,000 to 40,000 years,²¹the function of the relationship between the trainer and dog needs to be strongly considered when making training choices for dog owners and their pets.

Limitations

When interpreting the results from this thesis, it is important to consider key limitations that can be addressed in future research. While the sample size was sufficient for seeing trends, a larger sample size is a requisite to draw further conclusions. In addition, focusing on behaviors that are more varied (tricks versus husbandry and cooperative care behaviors), and potentially done in more distracting environments (rather than a neutral, low-distraction room), would help to determine behaviors and environments that may be more conducive to clicker training. In addition, while no participants had already taught their dogs these skills in the past, previous training experience was varied; controlling for training experience would be beneficial for future research. Finally, focusing on behaviors that require more awkward body language and inadvertent communication (leaning over, performing grooming or veterinary tasks) should absolutely be investigated based on these results. Given the significant welfare implications of cooperative care for veterinary and grooming procedures, this could potentially be a very high-impact area of study.

References

1. Feng, L., Howell, T. & Bennett, P. (2017). Comparing trainers’ reports of clicker use to the use of clickers in applied research studies: methodological differences may explain conflicting results. Pet Behaviour Science 3, 1-18.
2. Skinner B.F. (1951). How to teach animals. Scientific American 185:26:29.
3. Breland, K., & Breland, M. (1951). A field of applied animal psychology. American Psychologist, 6:6, 202–204.
4. Morris, E. K. (2003). Comments on the 1950s Applications and Extensions of Skinner’s Operant Psychology. The Behavior Analyst, 26:2, 281–295.
5. Pryor K. (1999). Don’t Shoot the Dog. NY: Random House Publishing Group
6. Pryor, K. (1975). Lads Before the Wind: Adventures in Porpoise Training. Oxford: Harper & Row.
7. Pryor, K. (2009). Reaching the Animal Mind: Clicker Training and What it Teaches us About all Animals. NY: Scribner.
8. Pryor, K., & Ramirez, K. (2014). Modern animal training: A transformative technology. In: F. K. McSweeney & E. S. Murphy (Eds.), The Wiley Blackwell Handbook of Operant and Classical Conditioning. (p. 455–482). Wiley-Blackwell.
9. Pryor, K. W., Haag, R., & O’Reilly, J. (1969). The creative porpoise: Training for novel behavior. Journal of the Experimental Analysis of Behavior, 12:4, 653–661.
10.  Pryor, K., & Chase, S. (2014). Training for variable and innovative behavior. International Journal of Comparative Psychology, 27:2, 218–225
11. Certification Council for Professional Dog Trainers. (2019). How to become a certified dog trainer. Retrieved October 27, 2019
12.  International Association of Animal Behavior Consultants. (2019). IAABC Core Competencies. Retrieved October 27, 2019.
13. Dorey, N. & Cox, D. (2018). Function matters: A review of terminological differences in applied and basic clicker training research. PeerJ. 6. 10.7717/peerj.5621.
14. Williams, B. A. (1991). Marking and bridging versus conditioned reinforcement. Animal Learning & Behavior, 19:3, 264–269.
15. Lieberman, D. A., et al (1985). Marking in pigeons: the role of memory in delayed reinforcement. Journal of Experimental Psychology: Animal Behavior Processes, 11, 611–624.
16. Feng, L. et al (2018). Is clicker training (Clicker + food) better than food-only training for novice companion dogs and their owners?. Applied Animal Behaviour Science. 204.
17. Feng, L. C., Howell, T. J., & Bennett, P. C. (2018). Practices and perceptions of clicker use in dog training: A survey-based investigation of dog owners and industry professionals. Journal of Veterinary Behavior: Clinical Applications and Research, 23, 1
18. Rossi, A., et al (2014). Visual attention in dogs and the evolution of non-verbal communication. In A. Horowitz (Ed.), Domestic dog cognition and behavior: The scientific study of Canis familiaris. (pp. 133–154). NY: Springer-Verlag Publishing.
19. Edwards, P. T., et al (2019). Fearful Fido: Investigating dog experience in the veterinary context in an effort to reduce distress. Applied Animal Behaviour Science, 213, 14–25.
20. Williams, J. L., et al (2004). The efficacy of a secondary reinforcer (clicker) during acquisition and extinction of an operant task in horses. Applied Animal Behaviour Science, 88:3–4, 331–341.
21. Smith, S. M., & Davis, E. S. (2008). Clicker increases resistance to extinction but does not decrease training time of a simple operant task in domestic dogs (Canis familiaris). Applied Animal Behaviour Science, 110:3–4, 318–329.
22. Breland, K., & Breland, M. (1961). The misbehavior of organisms. American Psychologist, 16:11, 681–684.
23. Botigué, L. R., et al (2017). Ancient European dog genomes reveal continuity since the Early Neolithic. Nature Communications, 8, 16082.