What CAN’T a Medical Alert Service Dog Do plus new notes on science plus draft

By January 4, 2019Blog

Now that so many people are closely observing their pup or dog for signs that they are acting unusual, as well as trying to train and reward for various alerts, it makes sense that there is more talent out there than we previously thought.

Some dogs just do not have enough social interactivity with people to care, though. Independent, or bonded to other dogs. They might notice things but have no motivation or training to communicate.

Maybe the pre-alert is easier than we think. Once a dog knows a cue or odor is important, they will anticipate, and eventually associate an odor (or whatever), with the oncoming symptom. Dogs are such good anticipators that most obedience classes make us teach the dog NOT to anticipate, but to wait for the next cue. Not to break a stay, even though they know their owner has a bad habit of always saying “Come” after a stay.

They anticipate so much in our lives with them:Our tone of voice when a visitor has stayed too long and we are going to start hinting about leving or doing some chore, or when we are thinking about getting up to go out.

Medical alerting by dogs is something of a mystery that is still being deciphered. Does a task have to be formally trained, or is it equally valid if a dog seems to teach itself the task? Does it matter how a task was trained, as long as it is useful?

When this field first arose, it was hard enough to understand how (and why!) a dog would alert an owner to their medical condition such as low blood glucose (LBG). It was harder still to understand why a dog would alert to a psychiatric condition like an anxiety attack. It became doubly hard when so many owners self-reported that dogs were alerting BEFORE any detectable symptom was present, giving advance warning to impending seizures, emotional symptoms, LBG, and other events where there are no medically detectable precursors.

The controversy about medical alerts and self-trained dogs, and owner-trained dogs, will become more and more intense as more and more people train their own dogs.

Let’s look at this topic from the simplest aspect first: how the medical alert SD industry teaches an alert to an existing condition. I will then describe how I believe an untrained dog “teaches itself” to alert to a condition, how owners train their dog unconsciously, and how all this creates a pathway for a dog to teach itself to alerts before a symptom can be detected by us.

When teaching an alert to any symptom, many scientists and trainers work from the viewpoint of traditional science-based training.

Here is the sequence as often done by professionals:

  1. A known Symptom Cue (odor, behavior, or other cue) must be identified. For Diabetic Alert Dog (DAD) training, the Symptom Cue might be perspiration, saliva, or breath condensates. When a patient is having a proven meter-detected low, these odors are collected and preserved by various methods .

    A behavioral Symptom Cue might be fast breathing or crying. Some behaviors might also possibly associated with odors, but for the purposes of training, the trainer will usually just mimic the behavior. Some owner-trainers are also collecting body odors during a panic attack or migraine, and experimenting with training alerts to these odors.
  2. The trainer does not need to know the chemical composition of a Symptom Cue.
  3. The dog is exposed to the Symptom Cue in a systematic way. Usually, food and social reinforcement is given at the time the Symptom Cue appears, or when it is artificially produced. This could be an odor, or a behavior like crying. The dog learns a positive excited associaition with the Symptom Cue.
  4. A physical alerting response behavior is also trained and reinforced. Often a nose-nudge is taught as an Alert Cue.
  5. The alerting behavior is then cued whenever the Symptom Cue is present.
  6. The dog begins giving spontaneous alert behavior whenever the Symptom Cue is present. 
  7. Now there is an Alert Cue, which is phased out as the dog begins to alert independently on its own initiative.
  8. The Alert must be accurate and consistent.
  9. Good trainers and programs track and document accuracy of the Alert.

However, many medical and psychiatric behavioral alert behaviors did NOT start out this way.

Most of these alerts have been discovered by dog owners noticing that their dog that seems to alert to a Symptom Cue spontaneously. Analyzing this mystery, we find that although the dog was not formally trained, it somehow alerts to the cue.

Two possibilities arise:

  • The dog might have an instinctive response to the cue
  • It might have learned a response through some unknown reinforcement or association.

How does this happen?

We need to look at this process from different viewpoint. Learning theory of course still applies here; it is just that the sequences are different. This type of alerting has two unique features:

  • The learning process happens differently from the expected order.
  • The learning process  has an organic growth that is triggered by events in real life.

If we theorize that the alert to the cue might be instinctive, then we can look at how dogs are purposely trained using known instinctive behavior. This type of training proceeds differently than shaping a new behavior or correcting undesirable behaviors.

By exposing a Border Collie to sheep, or a Pointer to a bird:

  1. We first trigger the instinct to herd or to point.
  2. Once the instinct is triggered, we have a behavior to work with. It can be rewarded in various ways.
  3. We then may isolate and put cues on various parts of it.
  4. The behavior is then useful to us and can be controlled.


Some SD training might happen this way, although accidentally at first. Some dogs appear to have an instinct to respond to some biological cues. Once a dog reacts to a diabetic low, or to the odor of fear or other body chemicals in a panic attack,  the owner or trainer has something to reward. A dog that responds to human crying as if to a panicked puppy, is also giving the trainer an instinctive behavior that can be rewarded.

Whether by way of an instinctive response, or a curious response to odor, or an association between an odor and a human behavior, alert training can still happen accidentally. With an owner with an untrained pet dog, here is what I believe happens:

  1. The untrained dog responds more or less randomly to a biological or behavioral symptom. Some possibilities are: chemical odors excreted in sweat or breath, blood pressure changes, pain chemicals in the body*, or heat such as inflammation. Behavioral changes could be seizure behavior, trembling, crying, self-harm, or any “surprising” behavior that does not happen frequently enough for the dog to become habituated to it.
  2. The untrained dog must already have a natural tendency to notice and react to any of those.
  3. It must notice them with a safe reaction; Some dogs react with interest and affection, some by being playful, some by lying close, some by fearful avoidance and hiding, some do not react at all. Some reactions are dangerous: some dogs will react to a seizure in a person or another dog with a violent attack, so there is a wide range of possible behaviors among different dogs.
  4. Some accidental training may occur when a person who is feeling sick or upset calls the dog over to comfort them, or simply allows it to stay next to them or on them. The dog learns that the symptom cues always result in social affection and contact.
  5. It then starts going to the person when it notices the symptoms.
  6. The owner or a family member eventually makes an association between the symptom cue and the dogs behavior.
  7. The person is thrilled by this, and rewards the dog in some way, usually by increased social interaction. The person might interpret the dogs behavior as wanting to comfort, wanting to warn, or even that it understands the owners condition. It is not really important WHY anyone thinks the dog is displaying the behavior. All that matters is that some reinforcement is taking place.
  8. Next time the symptom appears, the dog shows even more awareness of the symptom.
  9. The person now realizes that something important and useful is happening. He or she starts consciously training, by rewarding the dog more intensely for its behavior.
  10. The owner might then put a formal Alert Cue on the natural alert, or teach a more definite cue behavior such as a nose-nudge, and asks for it when the dog is showing its natural alert. Or, the owner might still encourage the dog without consciously knowing that they are ‘training” the dog. Either way, the dog now displays a formal Alert Cue.

Now we have a dog is starting to perform the technical definition of a trained Service Dog task. The owner achieves increasingly reliable communication with their dog.

As trainers, we can speed up this process and train in a different order to get reliable alerts:

  • We can teach the dog using odor samples in containers, not with an actual client.
  • We can train the dog by mimicking behavioral cues like crying or self-harm, not with an actual client.
  • We do NOT have to know exactly what odor we are collecting. 
  • It does not matter what the dog is sensing.
  • We experiment. Some trainers report success by collecting odors from a seizing or migraining person, preserving them, and using those odors to train an alert behavior in the dog. If the dog then starts alerting during the symptom on a real client, then we know we are on the right path.  If not, we can keep trying, or try with a different dog.
  • If odor cannot be collected, we can still  try to make sure that every time the Symptom Cue is present, the dog gets rewarded. If the owner is too ill, another person can reward the dog every time.
  • We keep an open mind. Are pain alerts due to increased heat from inflammation, causing a dog to cuddle closer to the heat source? Maybe so, but there is another possibility. Certain chemicals are released in painful parts of the body, and fluid can be withdrawn from a painful place in a syringe, and injected to other parts of the body. Pain will then be felt wherever the “pain chemical” is injected. (Sorry, I do not have the reference to this “pain-chemical” research.) So, here is  another clue to a possible alert to a chemical odor. 

For instance, self-harm is a behavior that many owners are training their dogs to interrupt:

  1. During the actual self-harm, the dog might be frightened, so the owner trains when in a relaxed and cheerful mood, and mimics self-harm.
  2. By calling the dog over for praise and treats and toys, the dog can have a positive and excited attitude rather than a worried one.
  3. The dog starts to respond to real self-harm, and the owner now has a dog that interrupts it by intruding into the owners consciousness and demanding attention, thus refocusing the owner away from self-harm.

This is a simple association, but has dramatic effects. Many people are not fully aware of starting to do unwanted behaviors, because the behaviors may induce a trance-like or dissociating (and very rewarding) mood. So, the dog can be a valuable interruptor. It is also a positive interruptor, since the owner feels rewarded by the dogs presence and intrusive behavior. Diverting their attention to the dog is a mentally healthier behavior.

The “Holy Grail” of Medical Alerting is alerting BEFORE the symptom

Alerting BEFORE a symptom cue appears….wow! This is  the amazing thing about medical alerts. They are controversial, but my experiences helping out at two of the very best and rigorously scientific DAD training programs have convinced me.

Dogs, being a species with a talent for making associations and anticipating “what comes next”, can learn to recognize behaviors/odors the owner is unaware of, and start an untrained alert BEFORE the owner is aware of the condition or symptom.

Perhaps the person is unaware of doing anxious behaviors in early stages of a panic attack, or perhaps they have never been aware of any precursors for the oncoming symptom (such as a seizure). Yet, dogs sometimes alert prior to panic or seizures.

Many DADs eventually start  to alert to an oncoming low up to 20 minutes prior to a detectable low, while a meter still reads normal levels.

A dog purposefully trained to interrupt a behavior such as self-harm may then anticipate by alerting to the precursor signs the owner might give before their unwanted behavior starts.

This kind of alert is the result of a very simple process that dogs as a species are good at: association, also known as cause and effect.

Here is an example: A dog learns that every time an owner cues “sit”, that the next cue is usually “down”. The owner might be frustrated when the dog downs when it is cued to sit, but every experienced trainer knows the dog is just anticipating a learned sequence. For obedience trials, anticipation is often unwanted behavior, because the dog must learn to wait for the next cue instead of anticipating it.

But for medical alerts, we want to encourage this association and anticipation. If we train the dog to alert to a Symptom Cue, many dogs begin to anticipate environmental cues that happen before the event. The dog begins to do what I call “Precursor Alerts”.


The major breakthrough in Diabetic Alert training came when the dogs spontaneously began alerting BEFORE meter readings showed any abnormal glucose readings. Most dogs began alerting fifteen to twenty minutes before any low could be detected with a meter.

Several reputable programs and owner-trainers have theorized that there must be a separate odor the dog associates with a FUTURE low: an odor that is present when the blood sugar is lowering, but before the low is detectable in the blood by meter testing.

With a Precursor Alert, an owner can use medication or change behavior so that the symptom never goes full-scale. Preventing a blood sugar low can avoid the small but cumulative damage each minimal low causes. Early medicating for an oncoming panic attack or migraine can avoid hours of misery and sometimes reduce the total amount of medication needed.

Importantly, a dogs’ Precursor Alert can break through normal human denial. None of us wants to “give in” to the feeling of an impending fibromyalgia flare, migraine or panic attack, and there is a temptation to “stay strong,” tough it out, and not take medication early. People might not want to retreat from a stressful or triggering environment if they will risk criticism from friends or family. A dog can interrupt that mood of denial, helping their person to realize that they must deal with the situation NOW.

So, what is the procedure for training a dog to do an alert to an unknown Precursor Cue that precedes  the actual Symptom Cue? After all, we cannot collect an odor, if we do not know whether a symptom cue is going to happen in the future.

The initial process is the same as the professional trainer sequence given previously: the dog is trained to alert to the Symptom Cue.

  • The owner watches carefully for false-positive alerts. This is a tricky stage. Reinforcement for false-positive alerts will encourage false alerts.
  • However, not-reinforcing false alerts  may extinguish the dogs’ initial attempts at alerts to the unknown precursor cue.
  • Most trainers follow a middle path: they mildly acknowledge the possible false alert, but do not do a full reinforcment. But they observe carefully,  in order to begin to notice the dogs true alerts.

The best results I know of have been gained when the owner keeps detailed records of dog alerts and symptoms. Diabetics can keep notes on all of the dogs alerts, and also on their blood glucose (BG) readings when they test twenty minutes later.. In that way, a pattern can be detected. With behavioral symptoms, constant self-checking on mood can give clues about rising anxiety.

Even if the owner is not watching the alerting process carefully, the dog may simply start pre-alerts on its own, and eventually even the most oblivious owner may notice a connection: that an alert is followed later by the symptom.

Now that people in the mobility-(and medically)-challenged community are becoming aware that dogs can alert this way, they know to watch their dogs behavior. Does this lead to the self-delusion that the dog is alerting? Yes. I’ve observed people announcing that their new 8-week-old puppy is alerting to precursors of complicated medical and behavioral conditions. I distrust these claims, because I believe that most “accidentally” trained alerts require many months of close relationship to develop. But previously, the opposite happened: nobody believed this kind of alerting could happen, so their lack of knowledge led people to either be in denial, or not want to risk ridicule about proclaiming their belief that their dog was alerting.

The process does not even have to be conscious on the part of the owner or trainer. Here is an example from my personal experience, that resulted from non-intentional behaviors of the owner.

A 12 pound self-trained “Guide Dog” who had a legally blind and hearing-challenged owner? Yes. I would have said this was nonsense, and not possible, but I observed the process of the unconscious training by the owner.

Frosty, her Hearing Dog was  trained for her ONLY for on-leash obedience and HD sound alerts, by the Hearing Dog Program. No Guide training was done, and the partner was informed that tiny Frosty was not trained to help her with mobility issues.

She had many physical issues as well as low vision.

Frosty learned that she stumbled at curbs, and began to stop before he reached them, anticipating her falling (or more likely, feared himself being fallen or stepped on!!!!). Even if her falling did not hurt him personally, he certainly was emotionally affected by her sudden fall, and unusual behavior and voice when hurt. She also stumbled on any rough pavement surface. Frosty also made this association, and would slow down when he noticed any surface changes on the ground ahead. His caution gave his partner advance warning to be careful and go slowly.

Part of their communication was that Frosty’s partner kept some tension on the leash, and was able to feel the dog stopping. We had attempted to teach the owner loose-leash heeling, but we failed, and Frosty was walked on a harness with a tight leash. Probably his partner was unconsciously realizing that she was more aware of  the environment when she held the leash with tension.

To the owner, it was a miracle that her tiny Frosty stopped her before she stumbled. To Frosty, a dog of only normal intelligence, it was simple cause and effect. Association, then anticipation.

Since the owner had some vision, she did not need more than this amount of guiding, and the pair functioned very well at this level.

Professional Guide Dog Programs use much more complex Guide training than this. They purposefully teach a dog to enlarge its perception of its personal space. The dog can learn to avoid an overhead item at the owners head level, because it has learned that such things affect its own body or emotions. It therefore learns to add a “human-shaped” space to its body image, and negotiate as if it was taking up the same space as both human and dog. Many other tasks are also trained. But Frosty’s example is simple association and anticipation.

To me, the ultimate working Service Dog  team is a human and dog who predictably understand each others signals and cues, together with some kind of pre-symptom alert.

What can dogs learn to alert to? Nobody knows. We will only find out if we experiment, and “Go Where No Trainer Has Ventured Before”. It is Owner-Trainers who have come up with every SD function yet discovered. They have unknowingly broken ancient barriers that scientificially educated trainers KNEW could not be breached. Because these trainers knew it could not be done, they never tried. But if a naive and “uneducated” person is NOT told that something is “impossible”, they often achieve it, or come very close.

Every new alert that someone discovers has potential for future scientific research. For instance, nobody knows what the precursor is for a seizure. Electrical activity in the brain, odors, or blood pressure have been theorized as precursors. However, one of our Hearing Dog partners had a HD that seizured. She told me that the dog had an “acid-like” smell just before it seizured. I asked if she (the human) had a good sense of smell, and she said yes, she did, that she probably could have had a career as a “Nose” employed by the perfume or wine industry. Perhaps research should focus on odors as the clue that seizure-alert dogs notice 3o minutes prior to a seizure. That story is hearsay, but possibly could be one more tiny piece of the puzzle. Clues like this can start the formal scientific research process.

We who work in SD Programs are in awe of  the many important discoveries that the medically and mobility-challenged community and their dogs have made. With Internet communication, owners and trainers are sharing experiences and discoveries daily, and training innovations are growing logarithmically.

Should the question we now ask ourselves  be, “What can a dog do? “
No, the better question is, “What can a dog NOT do??”
Let’s find out!

 


new methods and scientific analysis


Now that we have Panksepps’ research and theories to study, I find that looking at both his “Emotional Circuits” AND learning theory is the way to go. Can’t do it myself yet, since I do not have the Emotional Cicuits stuff learned well enough. But Panksepp fills in a lot of the blanks that Learning Theory analysis is insufficient for.

Like · Reply · 6 · 1 hr

Martha Hoffman Panksepp also has made a bridge possible between the IPO and Ringsports people who have their own jargon and theories about emotions and drives/instincts, that turn out to be more connected to what Panksepp researches. Giving those trainers and breeeders more respect, I hope, since so many of them have spent their lives studying dog behavior and training. And especially, behavior and temperament inheritance.

Karen Pryor wrote a wonderful article (it’s in one of her books, the collections of different papers and articles) about an orchestra conductor rehearsing an orchestra. She described how he used every quadrant, training all the different imstrument sections almost simultaneously, creating cues and fading them to near invisibility, using his facial expressions, hand and baton, and body movements, to communicate with several sections at once, and ending up with the orchestra one step further toward operating as he wanted.

 

She pointed out that this man, after a lifetime of genius-level conducting, would never be able to analyze his process in learning theory terms, or in any scientific terms at all.

Like · Reply · 3 · 37 mins

Martha Hoffman Daily, I eagerly read the posts of the owner-trainer Service Dog groups.They are mostly amateur trainers or pet owners. But, they are constantly discovering new ways to teach alerts to things that medical science says are impossible. Or finding creative training techniques.

 

The owner-trainers on FB are training dogs to react and alert to seizures, migraines, low blood pressure, anxiety attacks, anger, and pain. They are also collecting odor samples from when they have these symptoms, and using those to train. They do not care that they have been told that these things are undetectable by odor. They are unhappy about being differently abled or sick, they are determined to fight their conditions, and many are often home with time to train their dog. Many have great success, and are happy with their dogs.

 

Their dogs often are THEN proceeding to alerting on unknown precursors to their symptoms, about 15 minutes or more before symptoms arrive. (Dogs are great anticipators and can work backwards through many associations).

 

None of this is “scientific” None of this can be explained YET by science, or by the owners and trainers themselves.

 

Diabetic alert dog training was completely ridiculed by the medical and dog training professsions when I was volunteering to help set up the first DAD program in the US.

 

The trainee dogs accompanied a trainer to a diabetic kids summer camp where kids are monitored every few hours, 24/7.

 

After two weeks, when the dogs had alerted and usually been proven right by blood tests, selected the clothing of a child who had had a low from a big pile of other kids clothes, led a doctor into a shower room where a child was having a low…one of the Diabetic-specialist doctors who had been skeptical of this “woowoo craziness”, told the program that this was the greatest medical advance he had ever seen in his lifetime, and that he was proud to have witnessed it.

 

FIRST comes a discovery. THEN comes research and analysis, by others. The discoverer has no obligation to provide proof or explanations. The world of Service Dog Alerting is self-motivated, and is rushing along at a fast pace, each persons discovery building upon the next.

 

The US Dept of Defense funded two million dollars toward one formal research study of medical detection. Secretly. Their plan: Identify the chemical the dog were alerting to, and build a machine to identify it.

 

That’s the eventual result of amateur citizen-scientists bold and un-scientific attempts. They are the pioneers, and science follows behind to unravel the mystery.

Should the question we now ask ourselves  be, “What can a dog do? “
No, the better question is, “What can a dog NOT do??”
Let’s find out!

 

Commando Brando checking out the telephone.

Commando Brando checking out the telephone.

mhhd site photo pilar look play

Pilar uses a signal shown in both wolves and dogs: looking back over the shoulder to signal others to follow them somewhere.

Medical alerting by dogs is something of a mystery that is still being deciphered. The field was begun by owners of dogs who were reacting in unusual ways to medical symptoms. Trainers and owners are currently finding ways to formalize the training process in order to get consistent results with a wider variety of dogs and owners. But dogs are a step ahead of us, and also sometimes alert before any detectable symptom is present. Many questions arise.

  • Does a task have to be formally trained, or is it equally valid if a dog seems to teach itself the task?
  • Does it matter how a task was trained, as long as it is useful?
  • Should medical alerting tasks be trained the same way as other SD tasks?
  • Does the ADA definition of a task encompass a task that is done by a dog of its own voliton, with no formal “task training” for that behavior?

When this field first arose, it was hard enough to understand how (and why!) a dog would alert an owner to their medical condition such as low blood glucose (LBG). It was harder still to understand why a dog would alert to a psychiatric condition like an anxiety attack. It became doubly hard when so many owners self-reported that dogs were alerting BEFORE any detectable symptom was present, giving advance warning to impending seizures, emotional symptoms, LBG, and other events where there are no medically detectable precursors.

The controversy about medical alerts, self-trained dogs, and owner-trained dogs, will become more and more intense, as more and more people train their own dogs. and here is the strikethrough

Let’s look at this topic from the simplest aspect first: how the medical alert SD industry teaches an alert to an existing condition. I will then describe how I believe an untrained dog “teaches itself” to alert to a condition, how owners train their dog unconsciously, and how all this creates a pathway for a dog to teach itself to do advance alerts.

When teaching an alert to some symptom, many scientists and trainers work from the viewpoint of traditional science-based training. Here is the sequence often done by professionals:

A known symptom cue (odor, behavior, or other cue) must be identified. For Diabetic Alert Dog (DAD) training, the symptom cue might be odors that occur when a patient having low blood glucose (LBG). Perspiration, saliva, or breath condensate carrying the odor of the symptom cue could be collected for later use in training. A behavioral symptom cue of a panic attack might be fast breathing, or crying. A behavioral symptom cue can be mimicked in training situations.

If the syptom cue is an odorThe trainer does not need to know the chemical composition of the symptom cue , only how to collect it and preserve it. 

The dog is exposed to the symptom cue in a systematic way. Usually, food and social reinforcement is given at the time the symptom cue appears, whether spontaneously, or by some method of preserving odor or mimicking behavior. 

An alerting response behavior is also trained and reinforced. This might be a nose-nudge or any other behavior from the dog. 

The alerting behavior is cued whenever the symptom cue is present.

The dog begins showing the alertinwhen it detects the symptom cue.

The alert must be accurate and consistent.

However, many medical and psychiatric alert behaviors did NOT start out this way. Most of these alerts have been discovered by dog owners noticing that their dog that seems to alert to a cue spontaneously. Analyzing this mystery, we find that although the dog was not formally trained, it somehow alerts to the cue. Two possibilities arise:

The dog might have an instinctive response to the cue.

The dog might have learned a response through some unknown reinforcement or association.

How does this happen?

If we theorize that the alert to the cue might be instinctive, then we can look at how dogs are trained using known instinctive behavior. This type of training does not proceed in the same way as shaping a new behavior or correcting non-desired behaviors. By exposing a Border Collie to sheep, or a GSD to an agitator with a tug toy, we first trigger the instinct to herd or to bite. Once the instinct is triggered, we have a behavior to work with. We then may isolate and put cues on various parts of it, until the behavior is useful to us and can be controlled.


Some SD training might happen this way, accidentally at first. Some dogs appear to have an instinct to respond to some biological cues. Once a dog reacts to a seizure happening, or to the odor of adrenalin [?] in a panic attack,  the owner or trainer has something to work with.

With an owner with an untrained pet dog, here is what I believe happens:

The untrained dog responds more or less randomly to a biological or behavioral symptom. Some possibilities are: chemical odors excreted in sweat or breath, blood pressure changes, pain chemicals in the body*, or heat such as inflammation. Behavioral changes could be seizure behavior, trembling, crying, self-harm, or any behavior that does not happen frequently enough for the dog to become habituated to it. This only happens if it is a dog that already has a natural tendency to notice and react to any of those. It must notice them with a safe reaction; Some dogs react with interest and affection, some by being playful, some by lying close, some by fearful avoidance and hiding, some do not react at all. Some reactions are dangerous: some dogs will react to a seizure in a person or another dog with a violent attack, so there is a wide range of possible behaviors among different dogs.

What about dogs who seem to train themselves? I have already observed this process many times in Hearing Dogs. I believe I can show how the unique relationship between humans and certain temperament types of dogs can produce a seemingly impossible result. Dogs can self-train Hearing Dog Alerts whether the owner is Deaf or Hearing. Here is the process:

An important sound such as a door knock or a phone ring happens, and the owner or family notices it.

The owner gets excited and reacts to the sound.

The dog notices the excitement. People might run to the phone, or open the door to visitors.

The next time the sound happens, the dog again notices the excitement.

The dog makes an association, and begins to react to the sound with some behavior such as barking or running.

The dog joins in with the human activity, running as a group to the sound (or to the door). Some dogs may feel that they are joining their family group to defend territory, some may join in to greet visitors. Some may join in simply through social facilitation, following the mood of the group.

At some point, the dog anticipate the coming human response, and responds first, before any person notices the sound.

If the owner is Deaf, they may notice the dogs unique new behavior, and respond with happiness and praise. Everyone may be interested and intrigued that the dog is running to the phone or to get the kids when dinner is announced. This is reinforcing to the dog.  If the owenr is not Deaf, they may be annoyed, or interested. But the human response will not be as intense as that of a Deaf owner who just realized the dog is informing about phones, doorknocks, or someone calling them to dinner.

Now we have a positive feedback loop. The owner watches the dog, encourages the dog, shows off the “trick” to friends and family, and the dog receives more and more attention. Possibly, treats are introduced; but the HD temperament of dog is reinforced by any social contact, as well as by any activity. The dog begins to anticipate more and more, and becomes active without the faciliation of the owner moving or helping. the dog is driving the interaction now, and using it to get the owners attention and

Alternatively, a person who is feeling sick or upset might call the dog over to comfort them, or simply allow it to stay next to them or on them. Once the dog learns that the symptoms result in social affection and contact, it may start going to the person when it notices the symptoms.  

The owner or a family member eventually makes an association between the symptom and the dogs behavior.

That person is thrilled by this, and rewards the dog in some way, usually by increased social interaction. The person might interpret the dogs behavior as wanting to comfort, wanting to warn, or even that it understands the owners condition. It is not really important WHY anyone thinks the dog is displaying the behavior. All that matters is that some reinforcement is taking place.

Next time the symptom appears, the dog shows even more awareness of the symptom.

The owner now realizes that something important and useful is happening. He or she starts consciously training, rewarding the dog more intensely for its behavior.

The owner might then put a formal cue on the natural alert, or teach a more definite cue behavior such as a nose-nudge, and asks for it when the dog is showing its natural alert.

The dog now starts doing the trained behavior as well as its natural behavior. Now we have a dog is starting to fulfil the technical definition of a trained Service Dog task. The owner achieves increasingly reliable communication with their dog.

As trainers, we can speed up this process. We can collect odor samples, mimic physical symptoms, teach alert behaviors, and get reliable alerts. We do NOT have to know exactly what odor we are collecting.  For a condition like seizures, it does not matter to us what the dog is sensing. We experiment. Trainers report success by collecting odors from a seizing or migraining person, preserving them, and using those odors to train an alert behavior in the dog. If the dog then starts alerting during the symptom, then we know we did it right.  If not, we can keep trying, or try with a different dog.

With more mysterious conditions, for instance electrical disturbances in the brain, or blood pressure changes, we can try to make sure that every time the symptom is present, the dog gets rewarded. If the owner is too ill, someone else could reward the dog every time, as long as the owner is being properly cared for.  

*Research has documented that certain chemicals are released in painful parts of the body, and these molecules can be transferred by researchers and injected to other parts of the body. Pain will be felt wherever the “pain chemical “ is injected. (reference:______________)

For instance, self-harm is a behavior that many owners are training their dogs to interrupt. During the actual self-harm, the dog might be frightened, so the owner mimics self-harm, but with a relaxed and cheerful attitude. By calling the dog over for praise and treats and toys, the dog can have a positive and excited attitude rather than a worried one.  The dog starts to respond, the owner now has a dog that interrupts self-harm by intruding into the owners consciousness and demanding attention…refocusing the owner away from self-harm.

Alerting BEFORE the symptom:

Alerting BEFORE a symptom or condition appears….wow! This is the amazing thing about some medical alerts. It may be controversial, but my experiences helping out at two of the very best and rigorously scientific DAD training programs have convinced me.

This kind of alert is the result of a very simple process that dogs as a species are good at: association, also known as cause and effect. Here is an example: A dog learns that every time an owner cues “sit”, that the next cue is often “down”. The owner might be frustrated when the dog downs when it is cued to sit, but every experienced trainer knows the dog is just anticipating a learned sequence. For obedience trials, the dog must learn to wait for the next cue instead of anticipating it. But for medical alerts, we want to encourage this association and anticipation. If we train them to respond to a symptom as a cue, then the dog begins to anticipate the cue, based on environmental cues that happen before the event. In formal obedience-style training, this is unwanted, but in medical alert training, it is our Holy Grail.

So, what is the procedure for training a dog to do an alert to some unknown cue that precedes the trainer cue?

The initial process is the same:

The dog is trained to respond to the cue.

The owner watches carefully for false-positive alerts. This is a tricky stage. Reinforcement for false-positive alerts may encourage false alerts. However, not reinforcing them may extinguish the dogs attempts at alerts to a precursor cue. The best results I know of have been gained when the owner keeps detailed records of dog alerts and symptoms. Diabetics can keep notes on all of the dogs alerts, and also on their blood glucose (BG) readings when they test. In that way, a pattern can be detected. Constant self-analysis can give clues if anxiety is rising, for instance.

Even if the owner is not watching the alerting process carefully, the dog may simply start pre-alerts on its own, and eventually the owner notices a connection: that an alert is followed some time later by the symptom.

The process does not have to be conscious on the part of the owner or trainer. Here is an example from my personal experience:

I know of a 12 pound self-trained “Guide Dog” who had a legally blind and hearing-challenged owner. I would have said this was nonsense, and not possible, but I observed the unconscious training by the owner. She had many physical issues as well as low vision, and often stumbled on rough surfaces. Frosty, her Hearing Dog we trained for her (but only for on-leash obedience and HD tasks!) learned the association, and would slow down at surface changes on the ground. The dog also learned that she stumbled at curbs, and stopped at them, anticipating her falling (or more likely, feared himself being fallen on!!!!). Even if her falling did not hurt him personally, he certainly was affected by her sudden fall, and unusual behavior and voice when hurt. Part of their communication was that the owner kept some tension on the leash, and was able to feel the dog stopping. Since the owner had some vision, she did not need more than this amount of guiding, and the pair functioned very well at this level.

To the owner it was a miracle that her tiny Frosty stopped her before she stumbled. To Frosty, a dog of only normal intelligence, it was simple cause and effect. Association, then anticipation.

Professional Guide Dog Programs use much more complex Guide training than this. They purposefully teach a dog to enlarge its perception of its personal space. The dog can learn to avoid an overhead item at the owners head level, because it has learned that such things affect its own body or emotions. It therefore learns to add a “human-shaped” space to its body image, and negotiate as if it was taking up the same space as both human and dog. Many other tasks are also trained. But Frosty’s example is simple association and anticipation.

Dogs, being a species with a talent for making associations and anticipating “what comes next”, can learn to recognize behaviors/odors the owner is unaware of, and start an untrained alert BEFORE the owner is aware of the condition or symptom. Perhaps the person is unaware of doing anxious behaviors in early stages of a panic attack, or perhaps they have never been aware of any precursors for the oncoming symptom (such as a seizure). Yet, dogs sometimes alert prior to panic or seizures. Many DADs eventually start  to alert to an oncoming low up to 20 minutes prior to a detectable low, while a meter still reads normal levels. Several reputable programs and owner-trainers have theorized that there must be a separate odor the dog associates with a FUTURE low: an odor that is present when the blood sugar is lowering, but before the low is detectable in the blood by meter testing.

With a pre-symptom alert, an owner can use medication or change behavior so that the symptom never goes full-scale. Preventing a blood sugar low can avoid the small but cumulative damage even a minimal low causes. Early medicating for an oncoming panic attack or migraine can avoid hours of misery and sometimes reduce the total amount of medication needed.

A dog purposefully trained to interrupt a behavior such as self-harm may then make the mental leap to start interrupting at the precursor signs the owner might give when their anxiety is rising, but the actual self-harm behavior has not yet started. Many patients find that they need less medications for various emotional conditions when they are alerted prior to emotions getting out of control.

In addition, a dogs alert can break through human denial. None of us wants to “give in” to the feeling of an impending migraine or panic attack, and there is a temptation to “stay strong” and not take medication early, or risk censure from friends or family by retreating from a stressful or triggering environment. A dog can  interrupt that mood of denial, helping people to realize that they must deal with the situation NOW.

To me, the ultimate working team is a human and dog who predictably understand each others signals and cues, together with some kind of pre-symptom alert.

What can dogs learn to alert to? Nobody knows. We will only find out if we experiment and “Go where no program has ventured before”. It is Owner-Trainers who have come up with every SD function yet discovered. They have unknowingly broken ancient barriers that scientificially educated trainers KNEW could not be breached. So, it never occurred to those trainers to even try. If a person does not know what is “impossible”, they often achieve it, or come very close.

Every new alert that someone discovers has potential for future scientific research. For instance, nobody knows what the precursor is for a seizure. Electrical activity in the brain, odors, or blood pressure have been theorized as precursors. However, one of our Hearing Dog partners had a HD that seizured. She told me that the dog had an “acid-like” smell just before it seizured. I asked if she had a good sense of smell (the human) and she said yes, she did, that she probably could have had a career as a “nose” employed by the perfume or wine industry. Perhaps research should focus on odors as what a dog that alerts 3o minutes prior to a seizure is sensing. That story is hearsay, but possibly could be one more tiny piece of the puzzle.

Humans will always imagine things, make up stories, or outright lie about their dogs. That’s unavoidable. But gradually we are sifting out what dogs can actually do in reality.

We who work in SD Programs are in awe of them and their dogs, and the discoveries they have made. With Internet communication, owners and trainers are sharing experiences and discoveries, and training innovations are growing logarithmically. What can a dog do? The real question is, “What can a dog NOT do??” Let’s find out!

[Or link to How do Hearing Dogs “train themselves”???] (NOTE: with dogs that “train themselves” to alert to sounds and function as Hearing Dogs, the process is similar, starting with a naturally talented and sound-reactive dog that notices human responses to sounds, and finds reinforcement both from joining in with the human activity).

Is one way more accurate? Well, program and pro-trained dogs also have a chance of failing to be consistent, due to the distractions of everyday life and not living with a pro dog trainer. They key is the communication and relationship between dog and person.

In the end it comes down to, dogs help humans and humans help dogs. 

Leave a Reply