…But he does it perfectly at home.

RETRIEVAL CUES AND CANINE LEARNING: PART 2

Debby Boehm

Cues Part I explored retrieval cues, what they are and how they work. Examples illustrated their impact on the learning/performance process. Part II continues to address their cue actions in learning, confusion, corrections, and proofing.

In Part I we learned that any additional information encoded with the main learning may become retrieval cues. They contain information about the context and mental and biological states present when learning occurred. They may include: handler presence and proximity, training collars or equipment, teaching location, verbiage (including the command words), attitude and expressions, corrections, presence of rewards, etc. They function reciprocally, accessing main learning which, in turn, accesses memories of the additional information. Retrieval or remembering of learned information is best when the performance and learning situations are the same.

During the initial teaching process, many encoded cues of additional information are needed to remember the desired learning.

Example: Food is used to guide (lure) the dog into the down position. Some of the cues encoded might be: food, hand motion, handler presence and proximity, training collar or equipment, teaching location, verbiage (including the command words), attitude and expressions.

Pet owners may be happy to continue to produce all these cues to elicit the correct response, but dog sport competitors drop out cues so a verbal and/or signal command will access down learning. Dog handlers also train in different locations to add cues so learning is accessed / remembered in a variety of situations.

Incidentally, the phenomenon of encoding the contextual cue of the dog’s proximity to the handler is profoundly illustrated when teaching the drop on recall. Dogs taught as described in the example above show a marked preference to drop in close proximity to the handler.

When cues access incorrect or stressful learning, it is efficient to retrain, changing as many cues as possible when teaching the desired new learning. Change commands, location, rewards, technique, moods, etc., so cues that access (remember) the new learning are separate and distinct from those that access the old learning. Understand that using cues that access old learning when teaching new will access the undesired learning along with the new.

If cues are missing, incomplete, competing or conflicting, access to the desired learning is disrupted resulting in partial, wrong or nonperformance. Performance is a function of access to the correct learning. The following examples illustrate some retrieval cue effects on remembering.

Dogs may not access correct main learning because of similar or confusing contextual cues.

Example: Signal and directed jumping exercises have very similar cues: ring setup, dog’s position in the ring, handler’s position relative to the dog, and the use of signals. These similarities cause confusion in inexperienced dogs’ because the main learning for both exercises is encoded with many of the same cues.

Dogs may not access correct main learning because of missing cues.

Example: A dog learned to go out to baby gates (directed jumping exercise), but is now asked to go out to a pole. The contextual cue of the baby gates accesses the go-out learning but that cue is not present. Since the dog has not learned to go to a pole, this cue is not encoded with the correct information. The dog may look confused, and may fail to go out.

Dogs may not access correct main learning because cues were unnoticed:

Example: Because of a loud commotion, your dog does not see or hear the dumbbell toss during the high jump exercise. The dumbbell lands behind the jump and out of view. When sent, he may fail to go because cues that access to information to retrieve the dumbbell (presence and toss of the object) are missing.

Dogs may not access correct main learning because of partial or inconsistent retrieval cues.

Example: The handler gives the recall command too quickly after the judge’s order. Instead of the usual distinct command from his handler, the dog hears two voices, “Call your dogfluffy come”, with no break between the judge’s order and the handler’s command. This cue is not consistent with cues that access correct learning. He may prick his ears, rock forward, or even start to get up, but his confusion is evident.

Often, corrections are used to (fix) training problems like those in the examples above. Aversive collar pops, shocks, and manhandling are some corrections used for wrong, partial or nonperformance, and to motivate the correct response. Some dogs are trained exclusively with correction. It is expected that by avoiding areas of correction the dog learns the desired response. Unfortunately, a response that avoided one correction may be completely different from the response necessary to avoid the next one.

Corrections are intended to mean many different things in many different contexts: too wide, too slow (lagging), too fast (forging), too close (bumping), stop/sit, start, go, retrieve, etc. In reality, correction indicates only that what the dog is doing results in discomfort, and should be avoided. Corrections are single aversive retrieval cues, accessing learning of what/who/where to avoid, and they occur when the dog’s response or non-response is incorrect. Corrections do not make missing cues appear, partial cues complete, or indicate which competing cues need attending to. Aversives do not convey any information regarding the desired response, unless that response is avoidance.

Dogs may not access desired learning because the cue presented was not encoded with the desired learning:

Example: A dog learns to sit in response to “sit”. He is now commanded “platz” (German “sit”). No surprise; the dog does not sit.

In the above example it is easy to predict that the dog will not respond correctly. We don’t expect that he will understand that “platz” means sit, --easily perceiving that “platz” does not access the desired learning because it was not present to be encoded with sit learning. Yet, corrections are used in the same way as the ‘platz”, word. Undefined and bewildering cues delivered with the intention of accessing desired learning when, in fact, they have no access.

Example of corrective heelwork training: With the dog sitting, the handler steps into position, commands, “heel”, and starts off. The dog fails to move, so he is corrected. The jerk on the collar only conveys information that sitting (next to the handler) is wrong. After a few steps, the handler decides to halt: the dog keeps going and is again corrected.

The first correction (starting off) was intended to indicate that the dog should go forward, but the correction actually encodes information that sitting next to the handler has unpleasant consequences and should be avoided. It does not convey what the correct response should be. At the halt, the intended meaning of the second correction is to stop and sit. In reality, it conveys only information indicating that continuing forward is wrong. The dog has received only corrective input accessing information about where/what to avoid, not what response is correct.

Furthermore, if this dog tries to avoid the next correction by responding in a way that avoided the previous one, he will not be successful. Usually, his efforts are not understood or rewarded. In fact, they result in another correction. As the cycle continues, the dog is subjected to cues (close proximity to the handler, arm and leg movements, collar and leash, verbiage, training context, and correction) that access learning of who/what/where to avoid.

Correction for being wrong is not the same as showing the dog how to be right. The dog may react to correction in a way that mimics the desired response. Reacting to “heel”, the dog moves forward to avoid correction (the desired response), but learning encoded by the aversive is to avoid sitting next to the handler when certain cues (like movement of the left leg and corrective arm) are present. Heelwork taught by correction encodes learning to avoid all areas outside of heel position. Unfortunately, corrections given when the dog fails to move on “heel” or stop and sit on “halt” encode avoidance of heel position. This is quite different from reinforcing and teaching directly the desired response in heelwork: synchronized movements with the handler, holding position to the left leg regardless of speed or direction. In that scenario, cues presented access the desired learning.

Training and learning through a corrective/aversive process is complex. Since corrections only indicate information about what is wrong, dogs must develop strategies that result in behaviors that avoid correction. Like us, they guess. They sort through cues to determine which are predictors of correction, and then guess at responses which may or may not result in correction. Guessing and performing behaviors that may result in correction are risky, and stressful. Learning is not efficient and may encode unpleasant / stressful-state cues.

Some dogs guess well, but others guess badly and are frequently wrong. When repeated wrong responses are corrected, information that responding is wrong and should be avoided is learned. Responding may diminish or cease entirely. Confused and frustrated, dogs quit trying.

Example of guessing strategy we humans can relate to: A youngster is given information indicating a wrong answer, such as 5 times 7 is NOT equal to 12, and asked to respond with the correct answer. .He doesn’t know, so he guesses. He may guess (incorrectly) many times in an attempt to come up with the right answer. So many times, in fact, that he may get frustrated and give up. Incidentally, even if he does guess the correct answer, he still doesn’t know how to solve the problem.

Some dogs are able to work their way through the maze of cues and responses to “master” this convoluted task, yet others’ dismal performance and dejected attitude indicate this is well beyond their grasp. Corrections demoralize and de-motivate a dog who is trying his best to figure out what is right when he is only receiving information about what is wrong.

Even during shaping, where correct responses are rewarded and incorrect responses are not rewarded, repeated wrong guesses can result in frustration, confusion and inhibited or diminished responding. Frustration and diminished responding can occur even when no aversives are applied.

Corrections draw the dog’s focus to the who/what/where that produce/cause correction. This includes information about training context and equipment, presence and proximity of the handler, and the mental and biological state of the dog and handler during learning. At first, it would seem that more intense focus would be desirable. However, this type of attention is directed to predict and avoid correction. The problem (avoiding correction) requires an immediate solution, (the dog wants the corrections to stop). Some dogs even guess that biting the handler might be a response to achieve this goal.

Correction for wrong or nonperformance from a positively reinforced dog has the same action and result. Correction: collar pop, physical correction, or verbal “no”, indicates only that what the dog is doing/about to do is incorrect. Because correction was not part of the teaching process, it does not convey any information about how to be correct. The dog must guess (correctly) or be redirected verbally or physically. In this case, avoidance learning with cues of context and state are encoded.

Proofing as defined for use in this article is the process of altering the teaching context with the addition of distractions (competing cues) to achieve reliable performance. The goal is to teach the dog that the requested behavior must maintain priority and be performed even though there is a very interesting distraction, one that may even act as a retrieval cue for a different behavior.

At first, it may appear that the easiest solution is to correct the dog for wrongly going to the distraction, to cause the dog to avoid it. Correction is the method of choice for aversion training. But closer analysis indicates correction may not provide the best result. Ignoring and avoiding are not the same as focus on the exercise. In fact, these processes compete for limited cognitive resources. Active avoidance (ignoring) and attention both use cognitive resources. When used to actively avoid, fewer cognitive resources are available to focus attention on the exercise.

Example: A training situation is “set up” with a distractor (empty bag, toy, whatever) placed in proximity to the thrown dumbbell. When sent for the dumbbell, the dog goes, but notices the distraction and investigates, resulting in correction and avoidance. On the next retrieve, the dog makes a wide arc, avoiding the distraction. By using so many resources to avoid the distraction, he may be unable to access desired learning necessary to complete the retrieve.

Proofing (like initial teaching) is the process of sorting through cues to determine which require response and which should be disregarded.

Retrieval cues that access desired learning make correct performance possible. Missing, partial, and conflicting cues result in disrupted access and performance. Indicating wrongness is not the same as reinforcing the dog for being right. Avoiding distraction is not the same as focus on the exercise. And presenting cues that access only information about who/what/where to avoid is not the same as presenting cues that access desired learning. Careful teaching so desired learning is clearly accessed will help your dog perform perfectly beyond the confines of your own home.

Copyright Debby Boehm. DO NOT REPRINT WITHOUT PERMISSION.