Can clicker training facilitate conditioning in dogs?
Introduction
The dog-human bond is special because it represents a unique example of coevolution dating back at around 15,000–30,000 years ago (for the debate on wolf/dog's domestication see for instance Vilà et al., 1997, Thalmann et al., 2013). Dog companionship is primarily due to dogs’ communicative, relational and cooperative skills that made them able to engage successful interactions with humans (Bensky et al., 2013, Kaminski and Nitzschner, 2013). Very likely, these abilities made them particularly suitable, over the years of domestication, for numerous working roles.
Dogs can be trained with methods based on individual learning through which the dog creates an association between antecedents and their consequences. A novel behaviour can be elicited using different strategies summarized well by Kaplan et al. (2002). Briefly, by means of modelling and luring the dog is respectively coaxed or guided by a treat in the desired position and then it is rewarded with the aim of reinforcing that final position and hence the behaviour to reach it; by means of capturing, when the dog spontaneously manifests the desired behaviour it gets a reward. These methods, mostly known and applied by trainers and owners, suffer from intrinsic limitations, which are a repeated forced action from the outside, i.e. from the trainer in the modelling, and an unpredictable time from the dog showing the behaviour by chance in both luring and capturing. Alternatively, with the shaping technique it is possible to systematically reinforce increasingly closer approximations of the target behaviour. The final desired behaviour is reached by adjusting animal’s spontaneous responses, which are small achievable steps progressively rewarded towards the definitive behaviour (Skinner, 1951).
The use of positive reinforcement through shaping is a good candidate as a generic method and its successful use dates back to the work of Most (1910/1954) and Konorski and Miller (1937), who anticipated the task analysis of the behaviourism and its employment in operant conditioning (Skinner, 1938, Burch and Pickel, 1990, Mills, 2005). Operant conditioning is a process of learning whereby an association is formed between an arbitrary stimulus and an arbitrary response in virtue of a positive reinforcer. An instance is shown in Fig. 1a: a dog is rewarded to emit a response (push the handle) in the presence of a stimulus (a bread box). The effect of the reward is that of strengthening the association between the stimulus and the response (Skinner, 1951). Another form of conditioning exists and it is called classical conditioning. Differently from the operant conditioning, in classical conditioning an initially neutral stimulus (e.g. a bell) is repeatedly paired with a second unconditioned stimulus (e.g. food) that naturally evokes an unconditioned response (e.g. salivary reflex). After subsequent pairings of the two stimuli, the neutral stimulus becomes conditioned, as it is capable to evoke by itself the unconditioned response. In this form of conditioning, the dog learns that the conditioned stimulus is a predictor of the unconditioned one, which in case of food is a reward (Pavlov, 1927). The two forms of associative learning (operant and classical) can be used in combination. This is what happens in the so-called clicker training. In clicker training, the animal is reinforced to emit a response as it would happen in operant conditioning, but as soon as it displays the response and before the arrival of the reward, a conditioned stimulus (a distinct “click clack” sound emitted when a mechanical device is pressed and released) is provided (Fig. 1b). Because the clicker reliably anticipates the reward (i.e. the unconditioned stimulus) it then becomes a conditioned stimulus or secondary reinforcer.
The acoustic secondary reinforcer, in clicker training, is used also to mark the exact behaviour that results in food. While training a new and desirable behaviour, the clicker is retained to be effective in that the animal may instantaneously identify the precise behaviour that is rewarded and should be repeated in the immediate future. Also, the sound is meaning to fill in the temporal delay between the behavioural response and the primary reinforcer (Pryor, 1999).
An increasing number of professional dogs’ trainers and pet owners is using clicker training to shape dogs’ and other domestic animals’ behaviour (Lindsay, 2000, Tillman, 2000). Pioneered by Marian Breland and her husband for training animals at a distance (Gillaspy and Bihm, 2002), today growing popularity of the technique is accompanied by the idea that dogs trained with food alone are slower in learning and that clicker is the most effective stimulus to be used as a secondary reinforcer (Pryor, 2005). Indeed, the clicker sound has specific features that can be more effective than the use of a word or a whistle, commonly adopted by trainers and owners, as for instance consistency across persons and moments, and high detectability.
Despite the alleged efficacy of clicker training, scientific evidence in its support is still lacking (Miklósi, 2015). In order to address the efficacy of learning through clicker training, one possibility is that of systematically comparing the training time needed to learn a new behaviour when clicker training is used as compared to other training methods. Two studies investigated this issue using clicker with horses (McCall and Burgin, 2002, Williams et al., 2004) and in both cases no reduction in training time was recorded when learning was assisted with clicker rather than without clicker. Similarly, results obtained with dogs showed no advantage related to clicker use: dogs trained with the clicker learned the new behaviour in the same amount of time as dogs trained with food alone (Smith and Davies, 2008).
Another possibility to verify training efficacy is that of evaluating the strength of conditioning, which can be assessed throughout extinction. Extinction occurs when the secondary reinforcer is presented alone and, since it no longer predicts the arrival of the primary reinforcer, conditioned responding gradually stops (Gleitman et al., 1954). If clicker training is more effective in shaping a new behaviour, this should resist longer to extinction than when it has been shaped with other training methods. One study on horses showed that no difference in extinction emerged between the groups trained with clicker and with food alone (McCall and Burgin, 2002). Conversely, a study on dogs showed that dogs trained with clicker required more trials to extinct the behaviour than dogs trained with food alone (Smith and Davies, 2008). However, a difference of treatment existed in this experiment between the two groups of dogs (i.e. those trained with clicker and those trained without secondary reinforcer): during extinction trials, the reward was withheld from both conditions, but dog in the clicker condition continued to get the secondary reinforcer whenever the target behaviour was displayed. The fact that dogs trained with the clicker continued to hear the predictor of reward during the extinction phase could have possibly delayed extinction because dogs could have been waiting longer for a reward in virtue of the association between primary and secondary reinforcers.
Yet another alternative to test clicker training efficacy is that of focusing on the comprehension of the association between the antecedents and their consequences. An advantage resulting specifically from clicker training may be displayed when the newly acquired behaviour is applied efficaciously to different novel situations, an ability known as generalization (Pearce, 1987). No study has been conducted on generalization ability after clicker training.
With the present study, we aimed at determining whether the use of a clicker to train domestic dogs to perform a novel behaviour decreases training time or increases ability to generalize the learned behaviour when compared to the shaping obtained with other reinforcers. Also, for the first time, we investigated whether the use of clicker significantly differs from the use of a spoken word, as would be expected if the clicker provides a sharper feedback. To this goal, we compared the performance of dogs either trained with clicker and a treat, or trained with a spoken “Bravo” and a treat, or trained with treat alone, and then we tested them in a generalization task. Dogs had to learn a novel motor action, which was pushing up a handle with the nose and the muzzle in order to open completely a bread box. Then, we presented dogs with two different boxes: a slightly modified version of the bread box and a completely different box in order to test dogs’ generalization ability.
Section snippets
Materials and methods
The experiments comply with European Community (2010/63/UE) and Italian laws on animal experiments by the Ministry of Health (decreto legislativo n.26 del 4 marzo 2014). All dogs were recruited by direct contact with private voluntary owners and tested after authorization of dogs’ owners.
Discussion
Clicker training might rely on the combination of operant and classical conditioning. While the animal learns to provide a response in the presence of a given stimulus thanks to the reward, this association would also be strengthened by the presence of an acoustic secondary reinforcer that, presented immediately after the response and just before the reward, predicts the arrival of the reward. Indeed, what we expected, in line with what referred by dog trainers and owners, is that the standard
Conclusions
Although we should be cautious in drawing any strong conclusion from statistically non-significant results, our study is consistent with previous works conducted in different laboratories with both dogs and horses (Smith and Davies, 2008, McCall and Burgin, 2002, Williams et al., 2004), which, taken together, point toward no advantage in favor of the shaping method using one acoustic signal over another. Moreover, at best the allegedly beneficial effect of clicker would seem to be rather small
Acknowledgments
Sincere gratitude goes to all dogs and their owners. We thank also Michele Grassi for valuable comments on the statistical analysis, Cristian Grigoletti for the rendering of the images of the apparati and two anonymous reviewers who helped improving a previous version with their valuable comments. The work was partially supported by a UniTs-FRA2015 grant to CC.
References (42)
- et al.
The world from a dog’s point of view: a review and synthesis of dog cognition research
Adv. Study Behav.
(2013) - et al.
Should old dog trainers learn new tricks? The efficiency of the Do as I do method and shaping/clicker training method to train dogs
Appl. Anim. Behav. Sci.
(2014) - et al.
Social learning in dog training: the effectiveness of the Do as I do method compared to shaping/clicker training
Appl. Anim. Behav. Sci.
(2015) - et al.
Comparison of 3 different reinforcements of learning in dogs (Canis familiaris)
J. Vet. Behav.
(2013) - et al.
Do dogs get the point? A review of dog–human communication ability
Learn. Motiv.
(2013) - et al.
Robotic clicker training
Rob. Auton. Syst.
(2002) - et al.
Does training make you smarter? The effects of training on dogs’ performance (Canis familiaris) in a problem solving task
Behav. Proc.
(2008) - et al.
Equine utilization of secondary reinforcement during response extinction and acquisition
Appl. Anim. Behav. Sci.
(2002) Adapting to the human world: dogs’ responsiveness to our social cues
Behav. Proc.
(2009)- et al.
Training methods and owner-dog interactions: links with dog behavior and learning ability
Appl. Anim. Behav. Sci.
(2011)
Clicker increases resistance to extinction but does not decrease training time of a simple operant task in domestic dogs (Canis familiaris)
Appl. Anim. Behav. Sci.
Personality traits in the domestic dog (Canis familiaris)
Appl. Anim. Behav. Sci.
The efficacy of a secondary reinforcer (clicker) during acquisition and extinction of an operant task in horses
Appl. Anim. Behav. Sci.
A toast to Most: Konrad Most, a 1910 pioneer in animal training
J. Appl. Behav. Anal.
Relative efficacy of human social interaction and food as reinforcers for domestic dogs and hand-reared wolves
J. Exp. Anal. Behav.
Most domestic dogs (Canis lupus familiaris) prefer food to petting: population, context, and schedule effects in concurrent choice
J. Exp. Anal. Behav.
Obituary: Marian Breland Bailey (1920–2001)
Am. Psychol.
The S-R reinforcement theory of extinction
Psychol. Rev.
Analysing breed and gender differences in behaviour
The Behavioural Biology of Dogs
On two types of conditioned reflex
J. Gen. Psychol.
Cited by (34)
The effectiveness of visual and auditory elements of a compound stimulus in controlling behavior in the domestic dog (Canis familiaris)
2021, Journal of Veterinary BehaviorVariation-set structure: Extraction of important segments in pigeon learning
2021, Behavioural ProcessesCitation Excerpt :Establishing a direct connection between the segment and the reward contributes to reinforcement learning (Mendel and McLaren, 1970; Sutton et al., 1992; Kolodny et al., 2014); while establishing connections among segments that are indirectly connected to the reward (the reward is at the end of the segment chain that leads to it) can take place by means of local reinforcement (Kolodny et al., 2014). In the latter case, the connection reinforcement occurs when a novel segment leads to another segment that, in turn, leads, eventually, to the reward; and, therefore, the first segment can be perceived as a secondary reinforcer, similar to the use of a clicker in training dogs (Chiandetti et al., 2016). In language acquisition variation-set structure facilitates the ability to extract important segments from a data-stream by finding a repeated segment within different but related utterances over a short time period (Küntay and Slobin, 1996; Goldstein et al., 2010).
The effects of a choice test between food rewards and human interaction in a herd of domestic horses of varying breeds and experiences
2020, Applied Animal Behaviour ScienceIs clicker training (Clicker + food) better than food-only training for novice companion dogs and their owners?
2018, Applied Animal Behaviour ScienceHelping Pet Owners Change Pet Behaviors: An Overview of the Science
2018, Veterinary Clinics of North America - Small Animal PracticeClicker training does not enhance learning in mixed-breed shelter puppies (Canis familiaris)
2020, Journal of Veterinary BehaviorCitation Excerpt :Furthermore, to our knowledge, no one has compared positive reinforcement training methods with puppies who were new to training. All previous research used mature dogs (McCall and Burgin, 2002; Williams et al., 2004; Smith and Davis, 2008; Chiandetti et al., 2016). Therefore, this study aimed to fill a gap in knowledge by evaluating clicker efficacy for a developmental demographic, puppies, frequently targeted for clicker training programs.