Why We Click

There is no longer any doubt. Clicker training is here to stay.

More and more animal trainers are using it.

Although I work with dogs, not dinosaurs, I too am a dedicated clicker trainer, as are most of the instructors who teach for me at AutumnGold. However, while the theoretical underpinnings of clicker training are solid – and date back to Skinner’s original operant conditioning studies of the 1930’s – there is surprising little published research regarding its application to dog training. (There are even fewer studies of its effectiveness for training velociraptors. Huh. Who knew?). Even more surprising is the fact that the results of the dog studies that are available are not unequivocally in the “Yay, Clicker Training!” camp. Rather, their results have been lukewarm at best, with some showing only limited (or no) benefit.

Lucky for us, Lynna Feng, a graduate student at La Trobe University in Victoria, has taken on clicker training as her PhD research topic. I wrote about Lynna’s first paper in the Science Dog essay, “The Meaning of Click“. In her most recent publication, Lynna surveys the popularity of clicker training among dog trainers and asks the question “Why do we click“?

PUPPY COOPER HEELING FOR CLICKS

The Study: Lynna and her team used several approaches to this study. They directly interviewed a group of 13 dog trainers (8 clicker trainers, 3 non-clicker trainers, 2 uncommitted); reviewed a series of 7 best-selling dog books whose primary subject is clicker training; and examined five different click-dedicated websites. Data were coded and analyzed using a validated procedure called qualitative content analysis. This process describes the targeted phenomenon within a framework of predefined questions. In this case, the questions that the researchers were attempting to answer were:

  1. What is clicker training?
  2. Why do people use clicker training [with dogs]?
  3. What methods are generally considered to be “best practice” in relation to clicker training?

Here is what they found. Keep your clicker close by cuz this is really good stuff.

Results – What is Clicker Training? The study generated a lot of information, so I will attempt to distill this into results that are most relevant to dog trainers:

  • Philosophy or Technique? The majority of dog trainers view clicker training as both a training philosophy and a technique.  As a philosophy, the non-aversive and dog-centric nature of clicker training plus dogs’ positive experience of clicker training were emphasized. As a training technique, trainers espoused more practical views, such as “clicker training as a way to identify desirable behaviors in a dog with a clear signal (marker)” and “Using a clicker communicates to my dog that a primary reinforcer will be coming soon“.
  • The Dog’s Experience: Generally, trainers focused more on the communication properties of the clicker (i.e. marking a behavior and communicating to the dog that they are doing the ‘right thing”) rather than on its function as a secondary reinforcer. This is a rather important difference from the results of empirical research which suggests that the primary function of the click sound to an animal is as a secondary (conditioned) reinforcer. However, when applied to dog training, the clicker appears to be viewed first and foremost as an important communication tool that enhances the training experience and promotes learning. (More about this later).

Results: Why Click? The most consistently reported reason to clicker train was the perception that it helps dogs to learn more rapidly and effectively. Trainers also reported a number of additional reasons that they choose this training method:

  • Clicker training promotes active learning and encourages dogs to think for themselves. The end result is a dog who is eager to learn, more attentive, happy and confident. (This benefit was strongly contrasted with traditional training methods that focus more on instilling command compliance).
  • For the trainer, clicker training was reported to be easy to learn, encourages accurate timing and proper technique, and allows the trainer to better understand their dog’s learning process.
  • Many trainers mentioned their relationship with their dog, saying that clicker training strengthens their bond and improves communication (there is that communication thing again….).

CLICKER TRAINING IS REPORTED TO IMPROVE COMMUNICATION WITH OUR DOGS

Results: What are the “Best Practices”? Several broad categories of advice were reported:

  • Introducing the clicker: Both “charging” the clicker (repetitions of click-treat [CT] not contingent upon a targeted behavior) and introducing the clicker by using CT to reinforce a simple, known behavior were recommended. There was no consensus regarding which approach was preferred or superior. When asked about the number of repetitions of CT needed to establish the click as a secondary reinforcer, a wide range of estimates was provided – between “a few” to several hundred!
  • How to click: A wide range of opinions surfaced regarding how exactly to use the clicker. Almost all agreed that clear training criteria were essential and that the signal (click) should be applied at the exact moment that the dog is engaging in the targeted behavior. Conversely, there was not consensus about what the dog should be doing immediately after the click. This is the never-ending “click ends behavior” vs. “click during the behavior and dog maintains the behavior” argument. To date, there is no empirical evidence that supports or refutes either approach. (In fact there is no published research that even compares the two). Many (but not all) of the sources agreed that the trainer should maintain “one click/one treat” and “the treat should follow the click as closely as possible”. (Note, this second rule has some supportive evidence with dogs in the literature).
  • When to click: When asked if there were particular contexts in which clicker training was more appropriate than others, a number of sources stated that clicker training is useful for teaching almost any new behavior to dogs, and is especially effective for those that involve multiple steps or that occur at a distance from the trainer. Other sources stated that clicker training should only be used in a “controlled environment” (whatever that means) or only for certain types of training. (My particular favorite response in this section was the warning by some trainers that clicker training should be not be used “unless you are a professional“. Hmmm…Really?).

PROFESSIONALS ONLY

Conclusions: The researchers’ goals with this survey study were to identify specific questions regarding clicker training with dogs that can be addressed through future empirical research. Here are several questions that they identify:

  1. Are the benefits associated with clicker training related to the level of experience that dogs have with this training approach? (It is hypothesized that some of the more pronounced benefits that many trainers noted, such as enhancing the bond that they have with their dog and improving communication, come about as a result of prolonged and consistent clicker training over a period of weeks and months).
  2. How long does it actually take to “charge” a clicker (i.e. to build the association that click always predicts treat)? Is it a few pairings, a few dozen, several hundred? Are multiple training sessions required or can this be accomplished in one session? Related to the previous question – What role does repeated and prolonged exposure to “charging” the clicker play in dogs’ ability to respond to clicker training?
  3. And my favorite: Does clicker training improve communication and enhance the relationship between the handler and her dog? If so, how does this develop and how is it expressed (or measured)?

Take Away for Dog Folks: The point made several times in this paper was that the benefits of clicker training to dogs and to their trainers appear to develop over time as the dog becomes more and more experienced with clicker training. This contrasts sharply with the methodology used in studies of clicker training with dogs, all of which tested the efficacy of clicker training with dogs who had little or no previous experience with a clicker and used a relatively small number of CT repetitions, short training sessions and simple target behaviors. This led me to wonder “Maybe the studies that have been published to date have studied something that is fundamentally different from what we, as trainers (and our dogs) are experiencing as clicker training in the real world“.

Here is what I mean: One possible reason that dog trainers, many who believe emphatically that clicker training is a highly effective tool, are at odds with the less than stellar results of the published studies is that perhaps we are not talking about the same things. In other words, the way in which clicker training has been studied with dogs (and, one could argue, with other species as well), is not the way in which clicker training is actually used in practice. Several important differences were identified in Lynna’s study. The two most important are: (1) In practice, clicker training takes place over extended periods of time; (2) It almost always includes an established and positive relationship between the trainee (the dog) and the trainer (usually the owner).

The primary point that I came away with from this paper was that despite some continued attempts  to make it so,  clicker training as applied with dogs is not a purely behavioristic methodology. Rather, if one considers all of the new information that we have regarding the dog’s cognitive abilities, including their well-documented ability to read and understand human communication signals, then it is likely that the actual practice of clicker training involves much more than a rigid application of CT without any personal (relationship), cognitive, or emotional component. Since the studies that are in existence have studied clicker training using highly controlled behavioristic methodologies, perhaps they did not effectively measure or capture the depth and complexity of the phenomenon that is taking place when we use clicker training with dogs.

The results of this paper will hopefully be the impetus for new studies of clicker training that include dogs who have an extensive clicker history and who have established relationships with their trainer. Other elements that require study include the variety of ways in which the clicker is applied during a training session (end of behavior vs. maintain/keep going), and the many areas of training in which it is currently used (complex chains, distance training, different sports and working contexts). Personally, I look forward to reading more about Lynna’s studies and thank her and her team for undertaking work that is of immense interest to dog folks, especially those of us who are dedicated clicker trainers.

So, why do I click? I click because it works, because my dogs love it, and because, like so many others report in this study, it contributes to my ability to communicate clearly to my dogs and enhances the loving bond that I have with them. Happy training, everyone!

Cited Reference: Feng LC, Howell TJ, Bennett PC. Comparing trainers’ reports of clicker use to the use of clickers in applied research studies: methodological differences may explain conflicting results. Pet Behavior Science 2017; 3:1-18.

 

 

 

New Book! “Only Have Eyes for You: Exploring Canine Research with The Science Dog”

Only Have Eyes for You: Exploring Canine Research with The Science Dog” (paperback version) is now available! Click on the image below for more information and to order. (Kindle version will be available soon!)

Book description:  In her second Science Dog book, Linda Case tackles commonly held beliefs about canine nutrition, pet foods, behavior, social cognition and training. Each of the book’s 32 chapters explores a current issue that is of interest to dog owners and pet professionals and presents the scientific evidence that supports or refutes commonly held claims and beliefs. Learn about pet food ingredients and research showing that what is on the label may not always be in the food, about measures of food quality (and what consumers may not know about the foods that they buy), and about the safety and digestibility of popular dog treats and chews. Other chapters review new information regarding how dogs communicate, factors that help or inhibit a dog’s ability to learn, and the effectiveness of different types of training. Find out if dogs are capable of “knowing what someone else knows”, if they feel empathy for their friends, if they bark for no reason, and if they are capable of feeling guilt following a misdeed. Learn more about breed stereotyping, factors that influence our perceptions of dogs, and which canine characteristics most influence our attraction to particular dogs. This newest Science Dog book has something for everyone who works with and trains dogs, as well as for those who simply love dogs and enjoy learning more about our canine best friends.

                                                                             ♥♥♥♥♥♥♥♥♥♥♥♥♥♥

coversnip

Just Show Me A Sign

Like many dog trainers, I use both verbal and gestural (hand) signals as cues with my dogs. With our students at AutumnGold, we introduce both verbal and physical cues at the same time, but generally emphasize verbal signals because this is what most pet owners prefer to use with their dogs.

AG Down Stay

AUTUMNGOLD STUDENTS USE GESTURES AND VERBAL CUES WHILE PRACTICING DOWN STAYS

All of our classes include instructions for fading gestural cues in favor of  verbal cues for owners who wish to use primarily verbal signals. Students are taught to “lead with the verbal cue and follow with the gesture“, thus establishing a classical relationship (verbal signal predicts gesture signal). This connection allows the trainer to gradually fade the hand signal and eventually to rely primarily on the verbal command.

On the other hand (literally), hand signals are a lot of fun to teach and come in handy in a wide variety of exercises. For these, we offer a dedicated “hand signals” class, for students who are interested in teaching their dog distance signals and hand cues for direction or jumping. This is great fun for dogs and their people and is also helpful for students who are interested in competing in dog sports.

Chip Agility Jumping  Chip Down Signal                      CHIPPY SHOWS OFF HIS HAND SIGNALS FOR JUMPING AND DOWN

However, like many dog training practices, the use of verbal versus hand signals with dogs has not been formally studied. Until recently, that is.

Enter Biagio D’Aniello and his team of scientists at the University of Naples (among others) in Italy. I have written about this group’s research on previous occasions. They work with retrievers who are trained for water rescue work and are reporting new information regarding the dog’s communication skills and ability to learn through observation (see “With a Little Help from My Friends” and “Doggie See, Doggie Do“).

This time around, the researchers asked whether dogs who are trained to respond equally to verbal and gestural cues show a preference for one type over the other.

The Study: A group of 25 certified water rescue dogs were enrolled. The group included 10 Golden Retrievers and 15 Labrador Retrievers, composed of 12 males and 13 females. Per training protocols for water rescue, all of the dogs had been trained to respond to both verbal and gestural cues. The dogs were tested in four behaviors; sit, down, stay and come. The study was conducted in three phases. Phase 1: The four basic commands were given using gestures only. Phase 2: Commands were delivered using a verbal cue only. Phase 3: (Here is where things get tricky). Both forms of a command were given, but incongruently (i.e. they conflicted with each other). For example, the verbal command for “sit” was paired with the gesture for “down”, the verbal command for “come” was paired with the gesture for “stay”, etc. The frequencies of correct responses were recorded in the first two phases, and a “preference index” that indicated the percent of correct gestural responses was calculated for the third phase.

Results:

  1. Just a sign, please: When gestures alone were used, all of the dogs responded correctly to all four commands, with the exception of a single error (one dog missed a “down” signal). In contrast, when verbal cues were used, the dogs made a total of 18 errors. The most common mistake was failing to lie down in response to the verbal command “down”. These results suggest that dogs who were trained using both verbal and hand signal cues (and when no attempt was made to emphasize one type of signal over the other), the dogs responded more consistently to gestures than to verbal cues.
  2. Location, location, location: While dogs showed an overall preference for gestures over verbal commands, this preference was not found when the verbal command to “come” was paired with the hand signal “stay” and the owner was located a distance away from the dog. In this case, the majority of dogs (56 %) responded to the verbal command. This difference suggests that although the dogs tended to pay more attention to hand signals than to verbal commands, this preference may be overridden by the preference to stay in close proximity to the owner.
  3. Girls may be more visual: An interesting result of this paper was the sex difference that was found. Female dogs showed a strong preference for responding to hand gesture cues, while males were more likely to respond equally to both types of cue. (Note: Although there is a bit of previous research suggesting that female dogs concentrate more on visual cues than do males, the small numbers in this trial coupled with the method of scoring lead the researchers to interpret this result with caution – in other words, this may be a “statistical hiccup”).

This pilot study suggests that when dogs are trained to both hand signals and verbal commands, they will respond most consistently to hand signals. The study also suggests that context is an important factor, in that having a preference to be close to the trainer may override a preference for gestural signals, leading a dog to choose the signal (verbal or gestural) that leads to proximity.

Take Away for Dog Folks

The finding that dogs (usually) respond better to hand signals than they do to verbal cues is probably not surprising to most trainers. This certainly supports our understanding of dogs as being highly responsive to body language and non-verbal cues. Still, it is always gratifying to find scientific data that supports one’s (previously unsupported) suppositions.

This is Data

Do hand signals have enhanced saliency? However, is it possible that there is more to the differences found in this study than is explained by the dog’s proclivity for reading body language? This paper lead me to think more deeply about these two types of signals; specifically about the type of hand signals that we choose to use.  The majority of hand signals that we use in dog training are far from being  arbitrary signals. Rather they are structured in both form and function to direct the dog’s attention or body to part or all of the targeted behavior. For example, a commonly used hand signal for “down” is  a sweeping motion from the dog’s “nose to his toes”. During training, this gesture easily doubles as both a lure when food is held in the hand and as a vehicle to deliver positive reinforcement when the hand delivers a food treat once the dog attains the down position. A reliable response to the hand signal alone is achieved by gradually removing the lure from the signaling hand and switching to +R from the opposite hand. We are then left with a hand signal that has, well, enhanced saliency for the dog, if you will. A second example is the use of body language and hand signals to inform a dog about the direction to run or jump in agility training. The physical signal itself has inherent meaning to the dog (we all get this). This signal is then enhanced by pairing it with food or an opportunity to tug. Contrast these gesture examples to the variety of verbal cues that we use with dogs (sit, down, come, etc). All of these, of course, are completely arbitrary from the dog’s point of view. We could just as easily use the word “down” to train a down command as the word “pumpkin” or “fluffy butt”. While we do enhance saliency by pairing these terms with reinforcers, they cannot be structured in the same way that gestures can to be naturally obvious (salient) to the dog.

So, in addition to dogs being highly attentive to body language (I think we all agree on that), it also seems that the hand signals that we select function to naturally attract our dog’s attention and direct behavior. The trainer “beefs up” this attraction by pairing the signal with positive reinforcement. Therefore, gestural cues may always have one step up over verbal cues when comparing the two (when the owner is in close proximity). Here is an idea – try training a sit using a down hand signal or teaching an agility dog to jump in the opposite direction from which you are pointing. In addition to this being a bit of a training challenge (more than a bit, I suspect), I would hypothesise that when arbitrary gestural signals are compared with verbal cues, we might see a leveling out of the preferences for gesture versus verbal signals. Just an idea……any researchers biting?

A role for individual preference and reinforcement history? I also pondered what the influence of an individual’s preference might be in this type of testing. All dogs tend to have certain exercises that they enjoy more than others. Some of these exercises may be inherently reinforcing for the dog while others may simply be preferred because they have a strong reinforcement history with the trainer (i.e. the exercise has been practiced and reinforced more frequently). In the case of this study, we might expect that dogs trained for water rescue work would be highly bonded to their owners and would also have a very strong reinforcement history for the “come” command. It would be interesting to explore verbal versus gesture preferences in dogs who are trained for different types of work, who may have different behavior preferences and reinforcement histories. Such a test would be analogous to the study that this same group did with dog’s looking back for help, in which they found some very interesting differences.

In practice: From a practical viewpoint, as a trainer, these results suggest to me that we should be doubly careful when fading hand signals in favor of verbal cues, especially when training a dog’s less preferred behaviors. While this research suggests that dogs are asking us to “just show me a sign”, it also seems that their responses will be influenced by a number of factors, including looking for the cue that tells them what they want to hear!

Dottie Come when Called

COMING WHEN CALLED IS DOTTI’S PREFERRED BEHAVIOR!

Cited Study: D’Aniello B, Scandurra A, Alterisio A, Valsecchi P, Prato-Previde E. The importance of gestural communication: A study of human-dog communication using incongruent information. Animal Cognition 2016; DOI: 10.1007/s10071-016-1010-5.

The Meaning of Click

Hi. My name is Linda and I am a clicker trainer. In the spirit of full disclosure, I admit that I have been using a clicker for many years. My use began with the common gateway secondary reinforcer, the verbal cue (“Yes!”). While that worked well for a while, I eventually found that I needed more. I wanted a marker that was accurate and clear to my dog and something that could provide that immediate “ah ha!” moment in dog training that we all crave.

Cooper Clicker Training Heel

BABY COOPER HEELS FOR CLICKS

Recently, my husband suggested that perhaps I am too dependent upon my clicker. It is possible that finding them all over the house, in the pockets of my jackets and jeans, in the car, and oh yeah, one in the refrigerator, had something to do with his concern. I emphatically denied this and insisted that I could quit clicker training any time that I wanted to.

He called my bluff and suggested that I try using food alone, no clicker. Admittedly, I did not react well.

cold dead modified

PERHAPS I AM A BIT DEPENDENT

Hyperbole aside, why is it that many trainers, myself included, are so completely sold on clicker training? While the short answer is a forehead thumping “Duh…..because it works so well“, a longer exploration into clicker training, plus a bit of science, is needed to fully understand this phenomenon.

Operant learning: There is a large body of  scientific evidence supporting the effectiveness of using consequences to teach new behaviors, a type of associative learning called operant learning or conditioning.  Although the consequences that are used can be either aversive or pleasurable, most trainers focus on pleasurable consequences, or positive reinforcers. For dogs, a universal primary positive reinforcer is food, though verbal praise, petting, and play are also important. (Note: A primary reinforce is a stimulus that is inherently rewarding to the animal, with no need for prior conditioning). Animals learn most efficiently when the targeted behavior is immediately followed by delivery of the positive reinforcer. Even brief delays between the behavior and the reinforcer can slow or prevent learning.

The timing issue: Herein lies the problem. In the practical context of animal training, there are numerous  situations in which it is impossible for a trainer to deliver a primary reinforcer at the exact time that the desired behavior is being offered. Examples with dogs include when teaching retrieving, targeting distant objects, or moving a paw or other body part in a very precise manner. Secondary reinforcers help to solve this problem. These are signals that are clear to the animal, such as a sound or light flash, and which are purposefully paired with a primary reinforcer. For marine mammal trainers, a whistle is used. For dog trainers, it is the click.

clicker-training_gif

Click-Treat: The sound of the clicker is transformed from a neutral (meaningless) stimulus to a conditioned (secondary) stimulus by repeatedly pairing the click sound with the delivery of a food treat (the primary reinforcer). After multiple repetitions of Click-Treat (hereafter CT), in which the click sound reliably precedes and predicts the treat, the click begins to possess the same properties as  the treat itself. Clicker training allows the trainer to precisely target (mark) tiny bits of behavior at the exact moment they are occurring. The click sound becomes analogous to a bridge in time – saying to the 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!”

Well, at least that is what we think the click means to our dogs………

The meaning of click: Recently, a team of Australian researchers reviewed clicker training and examined the mechanisms through which clicker training might enhance learning (1). They looked at each of the three functions that dog trainers typically attribute to the click –  a secondary reinforcer,  a marker of behavior, and as a bridging stimulus. Although we typically give equal weight to all three of these functions, the current evidence, collected primarily in laboratory animals and pigeons, is telling us differently:

Secondary reinforcer? As described earlier, once a clicker is “charged” as a secondary reinforcer, it should possess the same reinforcing properties as the primary reinforcer (treat). This means that the click sound alone, without being followed by a treat, is expected to cause an increase in the targeted behavior and help learned behaviors to be resistant to extinction. An unpairing of the connection between secondary and primary reinforce should also lead to a lessening of these effects. All of these outcomes have been tested in rats and pigeons and the evidence overwhelmingly suggests that a conditioned signal (click), when consistently paired with a primary reinforce (treat) does indeed take on the properties of the primary reinforcer. The researchers also provide  evidence (in rats) of a neuropsychological nature – dopamine release has been shown to occur at times that would be expected if a secondary reinforcer was the driving mechanism for learning.

Event marker? Almost all clicker trainers, when asked to explain why clicker training works so well, include some version of “it precisely marks the behavior that I wish to reinforce, at the exact moment that it is happening“.  I agree with this account, given my own practical training experiences. But, of course, belief is not the same as evidence.  What does the current science say about using an auditory signal to mark behavior? As a marker, the signal (click) must draw the animal’s attention to the event. So, if a signal functions to mark behavior, we would expect to see an effect of the signal, though at a lower intensity, when it is not paired with a primary reinforcer. For dogs, this means that hearing the “click” sound, regardless of its pairing with food, should emphasize that moment and thus enhance learning whatever behavior is occurring. Again, though not tested with dogs (yet), this hypothesis has been tested with laboratory animals. The evidence suggests that learning is somewhat enhanced by a marker alone but that the pairing of the marker with a primary reinforcer is decidedly more potent. While “click” may indeed be a marker for behaviors, this function is intricately related to its role as a secondary reinforce rather than marking an event simply by bringing the animal’s attention to it.

Bridging stimulus? The bridging stimulus hypothesis focuses on the “a treat will be coming to you soon” portion of clicker training and applies when the dog is a distance away or there is a temporal (time) delay between the behavior and delivery of the food treat. According to the bridging hypothesis, rather than simply marking the behavior, the signal communicates to the animal that reinforcement will be delayed (but is still promised). A limited number of published studies have examined this function, but the evidence that is available suggests that an auditory signal (such as a click) may bridge the temporal gap between behavior and food. However, all of the studies used a type of training process called “autoshaping” which is a highly controlled and contrived experimental process. Whether or not a click acts as a bridge in the practical and varied setting of dog training remains to be studied.

Take Away for Dog Folks

The bulk of the current evidence coming from other species, primarily lab animals who are tested in highly controlled conditions, tells us that the major way in which clicker training enhances learning is through the click’s function as a secondary reinforcer. As far as event marking and acting as a bridging stimulus, these may be in effect, but if so, they are in a supporting role rather than being the star players. So what might this information mean for we who love to click?

  1. In its role as a secondary reinforcer, the click takes on the pleasurable properties of the primary reinforcer, food treats. Pairing of the click with the treat (charging the clicker) is essential to both establish and maintain these properties.
  2. While clicking without treating will work for a short period of time, repeated uncoupling of the click from the treat will extinguish the connection and the click will stop being effective as it gradually reverts to a neutral stimulus.
  3.  Although most of us refer to the click as “marking” behaviors, the actual marking properties of the click appear to be intricately linked to its function as a secondary reinforcer, rather than having any stand-alone strength in this capacity. Ditto for bridging stimulus.

Bottom line? Given  these three suppositions, if you are a trainer and are in the habit of clicking without treating, you may want to stop doing that (2). The power of the click lies principally in its strength as a secondary (conditioned) reinforce, so maintaining that connection appears to be key.

As for me, this evidence provides further support for the strength of clicker training with dogs. Don’t think I will be going through any 12-step program to reduce my dependency anytime soon.

12 Step Click

Happy Training!

Cited Papers:

  1. Feng LC, Howell TJ, Bennett PC. How clicker training works: Comparing reinforcing, marking, and bridging hypotheses. Applied Animal Behaviour Science 2016; Accepted paper, in press.
  2. Martin S, Friedman SG. Blazing clickers. Paper presented at Animal Behavior Management Alliance Conference, Denver, CO, 2011.

 

The Consequences of Consequences

Operant learning is all about consequences. Most trainers and behaviorists are well-versed in the uses of pleasant and aversive stimuli as dog training consequences. These can be constructed into a 2 x 2 matrix that includes the type of stimulus (desirable/pleasant or aversive/unpleasant) as one factor and the intended behavioral change (increase or decrease response frequency) as the second factor (1).

Quadrants

Learning occurs when one of these four consequences lead to a change in the dog’s behavior:

  • Positive reinforcement (+R): Delivery (acquisition) of a desirable stimulus (food treat, praise, petting, play) results in an increase in the response. Example: Dog increases “sit” response to acquire petting and a food treat.
  • Negative Reinforcement (-R): Disappearance (avoidance) of an aversive stimulus (head collar pressure, collar jerk, harsh voice) results in an increase in the response that allows the dog to avoid the unpleasant stimulus. Example: Dog increases “sit” response to avoid pressure applied to head collar or collar jerk (or to avoid a horrible “ehhhh” sound emanating from the trainer. How do people produce that sound, anyway?)
  • Positive Punishment (+P): Delivery of an aversive stimulus results in a decrease in response. Example: Dog decreases standing/lying down/moving away from a sit position to avoid pressure applied to head collar or collar jerk (or the horrible “ehhhh” sound).
  • Negative Punishment (-P):  Removal of a desirable stimulus results in a decrease in the response. Example: Dog decreases standing/lying down to avoid losing access to food treats, petting praise attained whilst sitting quietly.

Petting for sitting 2         angry yelling              Petting (Desirable)                                       The dreaded “Ehhhh” sound (Aversive)

The Hypothesis: There is a general (but certainly not universal) consensus among trainers and behaviorists that training methods that emphasize positive reinforcement are more effective and more humane than those that emphasize the use of aversive stimuli. To date, there is some evidence in  the scientific literature that supports +R methods as more effective than -R (see Yogi Bear Dogs ).  And now, there is a study that examines the effects of these two different training approaches upon dogs’ levels of stress and their relationships with their owners (2).

The Study: Stephanie Deldalle and Florence Gaunet of the University of Paris-Nord and the Laboratoire de Psychologie Cognitive conducted an exploratory study that observed two dog training schools and their students during a series of advanced training classes. One school used primarily +R methods in the form of food treats, praise and petting to increase desired behaviors in the dogs. The second school used primarily -R methods in the form of collar corrections (pressure/jerks) and physical manipulation (pushing the dog into a sit). Neither the schools’ instructors nor their students were aware of the study’s objectives or that their school had been selected because of the type of methods that were used.

Methods: A group of 24 owner-dog pairs training at the +R school (hereafter +RS) and a group of 26 owner-dog pairs at the -R school (hereafter -RS) were studied. The dogs represented a variety of breeds and ranged in age from 8 months to 7 years. One researcher attended two sessions of a one-hour advanced class at each school and collected data for 50 minutes during each visit. Data collected included the owner’s behavior and the dog’s response and body postures when walking on a loose lead and when responding to the “sit” command. Within each session the frequencies of +R and -R stimuli used by the observed owners were also recorded. Owners were asked to complete a questionnaire at the end of the session.

Results: No differences in owner demographics or dog characteristics were found between the +RS and the -RS groups. Training results were classified by activity:

  • Response to sit command: Dogs trained using primarily -R showed significantly more mouth licking (38 % vs. 8 % of dogs), yawning (12 % vs. 0 % of dogs), and lowered body posture (46 % vs. 8 %) when compared with dogs trained with +R, behaviors that are all associated with stress. Altogether, 65 percent of dogs in the -RS group demonstrated at least one stress-related behavior, compared with only 8 percent of dogs in the +RS group. Conversely, significantly more dogs in the +RS group offered spontaneous gaze to their owners during the sit command when compared with dogs from the -RS group (88 % vs. 33 %), a behavior that is interpreted as an invitation to visually interact and a positive relationship.
  • Walking on a loose lead: Although not statistically significant, more dogs in the -RS demonstrated a lowered body posture while walking when compared with dogs enrolled in the +RS (15 % vs. 4 %). However, reduced body posture while walking was relatively uncommon in both schools. Similar to their response during the sit command, significantly more dogs in the +RS group offered spontaneous gaze to their owner during heeling compared with dogs in the -RS group (63 % vs. 4 %).

Yawning Dog 4 tongue flicks stress               Stress Yawn                        Mouth Licking                           Avoiding Eye Contact  

Butkus Eye ContactOFFERING GAZE

Take away for dog folks: It is important to note this was a preliminary and exploratory study that compared students who were training their dogs at one of two possible schools. This methodology can allow comparison of the behavior of  dogs trained using two sets of training instructions (which is exactly what the researchers did), but  cannot be used to make general conclusions about training schools that use different methods because only one school of each type was studied. Although this may seem to be a minor point, it is an important one that cautions us to take care when interpreting the results of this study. (For a refresher on the need to study groups rather than single entities, see The Steve Series of this blog).  The results of this study suggest that:

  • The emphasis upon negative reinforcement when training dogs to perform basic manners exercises (sit, walk on lead) can cause stress, demonstrated as reduced body posture, tongue flicks, yawning, and avoiding eye contact.
  • Conversely, the emphasis upon positive reinforcement may improve a dog’s confidence and relationship with her owner, as evidenced by offering voluntary eye contact (and by the absence of stress-related behaviors).
  • Perhaps most importantly (and unique to this study design): All of the dogs in this study had been in training for at least several months and were enrolled in an advanced class. Those dogs who were trained using -R were most likely to respond with stress to the owner’s verbal command “sit”. The researchers suggested that the training itself and its attendant commands/cues had become a conditioned aversive stimulus for these dogs. In other words, there were consequences to the type of consequences that were used during training.

 

Juno and Carrie Default Eye Contact       Jack Russell Terrier lying beside it's accidentCONSIDER THE CONSEQUENCES OF CONSEQUENCES  

References cited:

  1. Canine and Feline Behavior and Training: A Complete Guide to Understanding Our Two Best Friends, LP Case, page 89, Cengage, 2010.
  2. Deldalle S, Gaunet F. Effects of two training methods on stress-related behaviors of the dog (Canis familiaris) and on the dog-owner relationship. Journal of Veterinary Behavior: Clinical Applications and Research2014; In press. Abstract

Yogi Bear Dogs

Do you live with a Yogi Bear dog? You know what I mean – one of those smarter than the average bear dogs? I am quite certain that I live with several.

Smarter than Average Bear

For example, Chippy my Toller, excels at retrieving rings and carefully placing them over a pylon, riding a skateboard (sometimes recklessly, in my opinion), and playing a game of his own invention called Agility Ring-Toss. Our ever resourceful Brittany, Vinny, demonstrates his version of canine genius by finding a hidden object after seeing it for less than a second, and our youngster, Cooper, shows himself to be a Puppy Einstein each time that he learns a new trick, typically before  I have a chance to complete my prescribed schedule of shaping. And Cadie, now 13-years-young, demonstrated her canine intellect by learning not only to fake a hurt paw, but to limp pathetically on three legs and fall to the ground in a full-blown melodramatic swoon when she decided to ad lib her own version of this popular trick.

All of their antics – Clicker trained.

Cadie Chip Vinny Cooper May 2013CHIP, VINNY, COOPER AND CADIE CASE

Clicker training is a popular and by all accounts highly successful training technique that requires the use of positive reinforcement, the “treat” part of click-treat. Many trainers, myself included, are convinced that focusing on positive reinforcement during training promotes a  “love of learning” in our dogs. And, it follows, these dogs develop into individuals who are “smarter than the average bear”.

Smarty PantsI’M A SMARTY PANTS

However, as we all know, anecdotal experiences (even lots of them) do not add up to scientific evidence. So, is there evidence that supports the belief that training, and specifically training that is reward-based, leads to smarter dogs?  A bit, it seems. Let’s look at two recent studies:

Study 1:  The first was conducted by Sarah Marshall-Pescini and her colleagues at the University of Milan (1). They examined the success of two groups of dogs presented with a novel problem-solving task. The first group consisted of pet dogs who had either no previous training at all or had completed a single basic obedience class early in life (untrained dogs; n = 54). The second group included dog who were extensively trained in a variety of dog sports such as agility, search and rescue, retrieving sports or schutzhund (trained dogs; n = 56). All of the dogs in the trained group had been trained using positive reinforcement (food treats and/or a toy as the reinforcer). During the study, each dog was individually tested for his/her ability to solve a box-opening task. A plastic container holding a dog treat could be opened either by hitting a paw pad or using a nose to lift the lid. Other than first watching their owner open the box to show them the treat inside, the dogs received no aid during the problem-solving session. The experimenters measured several variables including each dog’s success/failure, time needed to open the box, and behaviors while working out the problem.

Study 1 Results: Significantly more dogs in the trained group were successful at opening the box than dogs in the untrained group (34 vs 16, P = 0.0002). Trained dogs also spent more time interacting with the apparatus and less time orienting to their owners when attempting to solve the problem. Interestingly, the type of training that the dogs had experienced had no effect upon their ability to solve the task. In other words, agility, schutzhund, and SAR dogs were all equally proficient at solving a unique problem.

Study 2: The second study was conducted by Nicola Rooney and her colleagues at the University of  Bristol in the UK (2). Fifty-three dog owners first completed a detailed survey that asked a series of questions regarding their preferred method of dog training.  Following completion of the survey, each owner was videotaped in their home while they interacted with their dog and while training their dog to learn a new command. The experimenters measured degree of positive reinforcement (treats, praise, petting) versus negative reinforcement (harsh voice, collar corrections, swatting/hitting) that owners preferred and analyzed the dogs’ videotaped behavior.

Study 2 Results: This study found that dogs owned by people who reported using a higher proportion of punishment during training were less likely to interact with unfamiliar people visiting the home and were significantly less playful when compared with the dogs of owners who reported to train using primarily positive methods. In addition, the dogs of owners who stated that they used reward-based training and who were classified via survey results as being highly patient with their dogs tended to perform better in a novel training task. However, this difference was not statistically significant.

 Take away for dog folks: Each of these two studies have strengths and weaknesses, and together provide a nice bit of helpful information to trainers and dog folks. The first study is a rock star in terms of dog numbers; over 100 dogs studied in an applied dog study is a difficult and time-consuming feat – Bravo to the researchers for this!  And, their results tell us that indeed, dogs who are regularly trained are more likely to engage in a novel problem-solving task, to work independently of their owner, and to be successful at solving the task. In other words, dogs who are trained regularly seem to have “learned to learn”! What this study’s results cannot tell us, however, is whether or not there is any difference in the performance of dogs who are trained using primarily positive reinforcement  and those who are trained using more coercive methods. This limitation occurs because the latter type of training methods were not examined in this study. 

The second study, on the other hand, did compare dogs who were trained using primarily positive methods with those who were trained using more aversive (correction-based) methods. However, limitations of this study are that it utilized a self-reporting survey and did not measure dogs’ problem-solving abilities.

         Problem solving                         Survey

Direct Measurement of Problem Solving         VS.               Survey/observation Study

Results of the second study suggests that the use of coercive, punishment-based training methods with dogs may negatively influence a dog’s behavior with other people and may inhibit rather than support learning. It also suggests that using methods that emphasize positive reinforcement may lead to more confident and playful dogs – something that most of us certainly desire. IMHO, it would be nice to see a study of the first type that includes a group of dogs who were trained using more correction-based methods so that problem-solving ability can be directly measured in dogs that are trained using the two different philosophies. Any behavior students out there who want to give this one a go?

‘Til Next Time  – Happy Training & Enjoy Your Own Yogi Bear Dog!

References:

1. Pescini-Marshall S, Valsecchi P, Petak I, et al. Does training make you smarter? The effect of training on dogs’ performance (Canis familiaris) in a problem solving task. Behavioural Processes 2008; 78:449-454

2. Rooney NJ, Cowan S. Training methods and owner–dog interactions: Links with dog behaviour and learning ability. Applied Animal Behaviour Science 2011; 132:169-177.