|
The Hidden Value of
Rewards
Kay Laurence
© Teaching Dogs 2006. Vol 4 Iss 4&5
|
|
"Reward your Dog". We've heard
this many, many times in many various formats. It takes a lot of experience
to get the best from a reward - where the reward delivers everything the
dog needs to want to offer the behaviour again and again, with passion. Often
delivery of a reward is not enough, many other factors influence the effect
of the reward.
Back in the olden times, class instructors were perpetually yelling "praise
your dog". It seemed to be the hardest challenge for many folk to praise
the dog in a way that was actually REWARDING for that dog. We grow up with
the illusion that a "good dog" is sufficient, the dog will understand immediately
that the task was carried out correctly and that even when said through gritted
teeth, you really mean "good dog". Yeah.
A reward is only effective if the receiver of the reward finds it rewarding.
It sounds simple, but poor delivery can make the reward more trouble than
its value and have a backlash effect on the training and learning process.
Run down the Reward Check List and make sure that for the dog you are training,
this minute, that the reward is doing the appropriate job.
Type of Reward
Food
Not all dogs find all food rewarding. Dogs can be more activated by the anticipation
of the food, rather than the food itself. In fact I know a few dogs that
can successfully catch the food at the back of the mouth, store in their
for several minutes, and cough the lot up to enjoy at a later moment. Especially
when training has finished and all the other dogs are watching.
Wet food with a high scent stimulating the dog. Dogs do not need previous
knowledge to know that meat is what they want. They can learn that cheese
is rewarding, but it is not quite the effect of raw, fresh meat.
The factors influencing your food decision are:
1. Desirability
Does the dog want that reward. How can you rate desirability? offer two choices
of food in either hand, half close the fist and waft the scent of the food
under the dog’s nose, until you are able to see a preference in response
to one food more than the other. If no preference is shown, then there is
probably no preference. It is the smell of the food that will motivate the
dog, not the process of eating it. Great smelling food, quickly eaten, is
not better than poor smelling food that takes 3 minutes to chew. Think of
ice cream! Great food, great flavour, but not chewable.
2. Pocketability
Can you carry this food in training easily? Will it ruin your pocket, will
it go off quickly if left for a few hours? Can you quickly take one piece
out of your pocket at a time?
3. Stimulating Scent
Does the food give off enough scent to interest the dog? Remember your hands
will also become scented and act as a continuous reminder of the availability
of rewards.
The food maybe good for the dog, but a dried kibble is usually quite low
in scent until dampened with warm water. This certainly removes its pocketability
when it will turn to mush after 15 minutes.
4. Portion control
Can you deliver small enough chunks to be quickly eaten? Raw meat is an excellent
treat, but the very devil to flick off your fingers if in very small chunks.
For the toy dogs this may be the size of half your finger nail.
5. Price
Some great foods would be greatly appreciated by many dogs. As a special
treat, for special occasions your pocket may stretch that far, but for every
day training it may get too expensive, either to purchase or in time to prepare.
6. Preparation
Treat slicing becomes a daily task, and can contribute to warming up the
dogs for their training sessions. But nearly all food will require some preparation,
which is not always practical unless you have a good knife and chopping board.
Chicken pieces are great treats, but need cooking, peeling off the carcass,
and then dicing. If slightly over cooked they will disintegrate in your pocket.
7. Deliverability
Can you deliver with your hands quickly? Can the dog take the treat without
fingers? If the treat is delivered to the floor can the dog see it easily?
Liver cake is a great treat, good scent and nutritionally balanced, but a
disaster if it crumbles on contact with the floor. It encourages the dog
to spend energy finding every last crumb - or becomes a distraction at a
later time.
8. Thirst factors
Many foods that are high in flavour or scent are also high in salt (sodium)
content and will leave the dogs very thirsty for some hours afterwards.
Verbal Praise
This is not an easy skill to acquire.
Some dogs can become over excited with praise, others are quite immune to
lashings of flattery. For a dog living in a busy "verbal" household, praise
can just pass them by as so much white noise. Only the response of the dog
is important, and praise needs to be used appropriate to the dog and to the
behaviour.
It is certainly a good stand by in the absence of other rewards. Association
needs to be well conditioned to make praise effective, and the link between
the emotional state and the verbal praise must continue for the rest of the
life of that praise. The chosen words or tone, must be said consistently
in conjunction with activities the dog finds pleasurable. This is easy to
incorporate in every day interactions with the dog.
You can relax a dog with verbal praise that has been associated with naturally
relaxed behaviour. This praise needs to reflect the mood in cadence, tone
and pitch.
Equally, you can excite a dog with praise that has been associated with exciting
activities, such as hunting for food or a toy or greeting. Again the praise
needs to reflect the mood.
Physical reward
This can vary from physically stepping away from the dog, to a deep massage
- again depending on what that dog finds rewarding, NOT what the person finds
rewarding. Often a trainer or owner is shaped by the first dog they train.
This may be a dog that found physical contact very rewarding and the owner
was equally comfortable touching and stroking the dog. Another dog they train
may find this quite irritating - physical praise needs to be good for the
dog and appropriate to the behaviour.
My dogs very much like physical contact, it is usually conditioned as very
young puppies and in association with attention. When living in a multi-dog
household, one-to-one attention is highly rewarding and many behaviours can
be rewarded in that spotlight, that would otherwise be objectionable - such
as grooming.
But the dogs all like different physical contact, from the major hug Arnold
enjoys to the scritch on the sternum that Kent enjoys. The collies love attention
and fussing, but generally detest the idea of a cuddle. They like to lean
on you for contact, but would not miss grooming if it never happened again
in their lives. I have met collies that find physical proximity uncomfortable,
probably an unplanned outcome of generations of farm breeding (most farm
dogs on wet days are not popular neighbours when stinky from varieties of
sheep and cow dung). These dogs will often exhibit stress when taught close
heel positions or even in freestyle cued to weave through the handlers legs.
They are far happier at a distance, where they can also use their eye to
monitor cues and activities.
We've turned around a few recall dogs with an excessive amount of physical
flattery and verbal nonsense. Especially effective for the dogs that would
return for a slice of liver, but eat and be off. I hold the collar with one
hand and physically praise the dogs with as much hand contact, quite vigourous
rubbing and verbal singing for AT LEAST 60 SECONDS. This is a considerably
long time, especially if you actually keep an eye on your watch. If done
effectively you should be puffing a bit at the end as well. The dog will
certainly want to have a good shake to settle the fur back into order, and
then resume hunting. Ten or more sessions of this on a successful recall
and you can see the dogs become addicted to it. There is a warning also:
beware transferring every ounce of dirt from the dog to your hands and clothing,
and if carried out effectively you will have a dog that sticks to you more
than is healthy ….
Go on, off you go!
A Toy
Toys in themselves can be intrinsically rewarding. A dog may simply enjoy
the texture and mouthing on an object - particularly if the object has a
pleasant association from puppy times. The dog may enjoy the noise production,
or tossing the toy around, playing with their feet and generally enjoying
the interaction.
Toys are excellent rewards for behaviours where the dog needs to relieve
some stress, or entertain themselves alone for a while.
Games
These can be particularly beneficial when played in partnership with people.
The game itself is rewarding and the dog will associate the reward with an
object, a cue word, if used with consideration, and pleasure in our company.
Additionally during play the dog learns a lot about the way we move, play,
change balance, give up, persist and our body cues.
Games need to have rules and can vary in intensity, dangerousness, points
scored and duration. They can be very exciting or simply a mutual tug to
relieve stress. Games can arouse the dog, teach the dog self control and
transfer emotional context to the proceeding behaviour. Games with other
dogs can also be used as a reward, when appropriate to a specific situation,
such as self control around play mates, followed by the playtime.
With all rewards the value, duration, placement and delivery must be pre-planned
and thought out, as each can affect the effectiveness of the reward.
Duration
The length of time a reward takes will punctuate the rhythm of the behaviour
being taught. If a dog has run away from another dog on a recall, come across
a field, the reward duration should equal the length of time the dog has
had to maintain concentration on the behaviour and the amount of self control
the dog needed to leave a rewarding situation.
If you are teaching the dog to paw tap an object, the reward should be in
equal duration to the behaviour. Tap, bite, swallow.
To keep short punctuation with physical praise I would not exceed 3 seconds,
and to extend the delivery of a treat I would stretch the process up to 10
seconds. This can begin with the click, followed by an 8 second mini drama
that ends in the food delivery. Remember anticipation of a reward is as,
if not more, rewarding than the acquisition itself. As soon as the dog hears
the click and looks for the reward, you can begin by moving towards the food
container, I have even asked the dog to join me to hunt for the container,
open up the pot, ensure a good waft in the direction of the nose, then shuffled
the food around the pot for a few seconds trying to decide which piece looks
the absolute best one for that absolutely superb behaviour. All of that interactive
process IS the reward.
A long reward delivery carries a danger that the behaviour can be forgotten
in the all consuming reward moment. This is particularly true when teaching
new behaviours, especially puzzle solving. The reward needs to be simple,
non-distracting, consistent and quick.
But if you are practising memory skills then using long duration rewards
as a distraction can improve concentration. Food delivery may be quick, but
if the process of eating the reward, begins with a mini search, followed
by decision making sniff, then a quality assurance tasting session the duration
of the process can become unrewarding. Using familiar foods will reduce this
with food that is easily picked up and swallowed.
Mabel demonstrated that although a branded treat may have everything required
in flavour and size, it easily got stuck between teeth and required heavy
duty tongue gymnastics to complete the process, by which time she had completely
forgotten what she was doing and fluency never ever peeked over the horizon.
Other dogs will collect the food, and still be eating during the next behaviour
- multi taskers!
Games can vary greatly in duration, a three second tug time is perfect. Long
games will not only interfere with memory skills, but extract energy from
the reserve. Develop a variety of games of different energies, a gently tug
that can last longer or the vigorous tug workout that cannot be repeated
too many times.
Placement
This can make or break a behaviour. Although the click promises a reward,
if the reward takes too long to occur, or is preceded by a lengthy search
then it loses value and can become punishing. I have seen dogs flinch when
they hear the click, because the opposite happened and the food was delivered
to the dog at such an alarming rate and with such vigour that the dog anticipated
an unpleasant process. The behaviour rate definitely slowed down in both
cases.
When learning new behaviours the placement of the reward should set the dog
up at the optimum place to begin the next behaviour. If the dog gets the
treat at the physical point of completion, then it will need to initiate
one or more movements to find the starting place of the behaviour. If you
are practising the drop down position from standing then click for the correct
movement but reward the dog in the standing position. This sets the dog up
to finish the reward and without delay begin the behaviour again. This is
valuable when teaching fluency, especially in connection with repetitive
muscle movements.
If the food is delivered to the place of completion two outcomes will affect
the behaviour. Firstly, the reinforcement balance will be heavily loaded
to the end result, the outcome, with both the click and the reward at the
same point. This leaves the opening behaviour less reinforced, and very often
it is the opening of the behaviour that ensures the correct completion. Secondly
the dog can relax the muscles on completion during the reward process. This
is particularly useful if you want to teach duration of a behaviour, but
a nuisance if you want to teach quick responses, or fast repetitions of a
behaviour.
When teaching the dog to drop on cue you are looking for:
1. A quick relation between the cue and the behaviour
2. A contraction of the muscles that will lead to an accurate
dropping movements (as opposed to a sit down, or bow down)
3. A poised position on the floor.
If the poised drop position is to be used as a control mechanism or demonstration
of control, it is highly likely that the dog will need to leave that position
very quickly for another movement. The muscles need to be held poised and
ready for action. By rewarding in a standing position the dog will learn
to drop, hold the muscles and then rise quickly for the reward. If you are
looking for duration in the poised drop position then the click can be delayed
in small increments.
If I am looking to settle a dog down for a relaxed duration, then I will
teach the dog to change onto one hip, on the cue “settle”, and in this case
the reward will come to the dog after the (soft) click. (I discriminate with
different clicks to indicate a yeah, go to it - chase the food with the ordinary
click, and a muffled click for the relax, you done good, I am returning with
your food, click.)
Placement needs to be variable, appropriate to the behaviour you are trying
to teach and the dog that is trying to learn it. I am teaching Dottie to
turn her head to her right. I sit in a chair, with food at nose height. To
prevent the movement being over strongly linked to one position only I vary
the base position with her in the stand, sit or down. As she moves her head,
I click and offer the food on her left. She does not need to move her feet
to collect the food, I can maintain the base position, and by turning her
head to the other direction, I am setting up the opening for the muscles
on the right side of her neck to want to move. Once she has acquired the
movement, to extend duration of the pose I would swap the delivery point
to the outcome of the movement - holding the head to her right.
The placement changes as the outcome of the learning changes. Once the behaviour
is established I will vary the placement.
Checking the function of the ears: I will often test a verbal cue
by tossing a treat away, and giving the cue whilst the dog is picking up
the treat. At that moment they are not looking at me and can only listen
to the cue, but, (and there is always a dog called But) some dogs concentrate
so much on the eating process they do not hear ANY cues at that time.
Make reward collection simple: Try not to use placement as a competition.
Imagine completing your week's work, and going to the pay office to collect
the brown envelope. "Ah, this week your pay is somewhere in the third office
on the right …… you'll need to go and find it".
Nice. Thank you. That process could easily add distaste to the reward and
reduce its value.
If you are a Treat Tosser, become a Treat Placer and do not rely on your
tossing accuracy, especially for the visually challenged dogs. Remember to
wait for the dog to look at your hand before you throw, so that they have
an outside chance of seeing when the food has gone. If there is a danger
of the food disappearing into a confusing background, then use a tray, sheet
or bowl where the dog can be assured of acquisition. Set this up at the point
to commence the behaviour again.
If you want a variable and accurate placement as you extend the opening behaviour
then begin the reward drama as soon as the dog looks at you after the click,
and take the treat to the exact spot.
Placement can vary from low movement, unobtrusive placement to running with
you to collect a toy from another room, the car, or the kitchen. The reward
begins as soon as you start the sequence that ends in a reward. Do not neglect
to teach this sequence so that they become familiar with it.
Delivery
The style of delivery plays a large part. When concentrating hard on click-listening
the delivery needs to be consistent, fast and with quiet body language. The
movement to deliver does not want to intrude on the dog's concentration or
detract from the click. When acquiring new behaviours the click will be more
important to the learner than the reward. They need to pin point exactly
what was happening when the click happened, compare it to what happened on
the previous clicks, begin to see the pattern forming and experiment with
the direction the learning is taking. For many dogs this puzzle solving process
is engaging and rewarding. It allows us to use a lower value reward than
perhaps the finished behaviour because of the self rewarding nature of learning,
as opposed to practice.
At the other extreme we teach the recall, through a nose touch to hand, with
an exciting chase-the-sausage game. On the click, I begin the body language
of a ten pin bowler, and give the sausage chunk as much acceleration as possible
down the room. To make it more exciting I vary the direction, often feigning
one way but calling the dog quickly to look at me, and see I've thrown the
other way. The game is chasing the moving sausage, not hunting for the unseen
sausage. Mini hot dogs, 1" or 2cm long are perfect for this. They bounce
along with a "can't catch me" cheeky little swagger. Most dogs will not resist
the challenge for long. The effect it has on the recall is immediate - the
faster the chase for the sausage, the faster the recall to touch. It will
depend on whether you time the cue to touch whilst the sausage is being killed,
or after swallow - success for a running recall is healthier AFTER swallow,
not during eating.
By making a game out of the style of delivery we begin to double the value
of the reward, and induce an emotional colour to the behaviour we are trying
to teach. The fun of sausage chasing becomes inextricably linked to recall,
making recall almost as much fun.
For the dog whose world ticks quite slowly, a slow deliberate delivery will
match their learning style. You give the cue to wave a paw, they blink twice
(hard drive accessing database for data match), send message to rear end
to park, move body mass backwards into sit, shuffle upper weight to one side
to allow considered elevation of front limb. Click. Trainer, lifts hand in
slowly elevated movement towards box of food, reaches in and produces fresh
kill, rolls kill back into the centre of hand and beginning at shoulder height
slowly lowers the opening hand for easy delivery to mouth region.
This strategy will also work for the dogs that are chivviers - who, after
the click telepathically transmit messages of hurry, hurry, hurry to the
deliverer. The frenzy and frustration can become an increasing spiral, over
influencing the success of each behaviour. This may be a blessing for a dog
that begins a behaviour with a manana attitude, but an opposing style of
delivery can change the frenzied dog to being more relaxed. The highly animated
style of delivery can light the fire of the slow-to-warm beings.
But some dogs only work at one speed, and trying to change this can add frustration
and become punishing.
Access to another behaviour
If a behaviour is highly rewarding, then that behaviour can be used as a
reward for another behaviour. The cue for the rewarding behaviour will act
as the click. This is a great technique to use the emotion of the rewarding
behaviour to influence the new behaviour.
It is particularly obvious that if you use games to reward behaviours, that
the cues for the game can replace the click. But equally you can use a movement,
such as spin, jump or roll over, if the dog finds it rewarding, as a reward
for another behaviour. This is apparent in agility where the cue to hit a
contact point or turn on cue, is rewarded by the cue (click) for the next
apparatus which is rewarding.
The gundog or sheepdog will hold the steady position by the handler with
rock-like steadiness, because the behaviour is rewarded by their highest
reward "get out there and do the job" activity.
This technique is particularly useful when teaching chains of behaviours,
but handle the strategy with caution as the opportunity for the dog to self
reward must be protected.
Test the reward:
You can experiment with quite interesting and important results with a range
of rewards.
Begin with making a list of events that you consider rewarding for the dog.
Such as:
lick of paté
chunk of cheese
piece of raw heart
a "good boy"
pat on the head
a hug
jump in the air
licking your face
tummy rub
Focusing on one event, set the dog up for a simple free shaping session.
Perhaps choose an established behaviour such as paw tap, and transfer it
to a new object. After the click, give the reward, perhaps the pat on the
head. Take note of how many repetitions of the behaviour the dog will offer
in the next minute. If the rate of reinforcement, ie. the number of clicks
stays constant or increases, then the pat on the head IS rewarding. If it
decreases then remove it from your list of rewards. It ain't.
I would not advise you make a day of testing rewards, otherwise the value
of the click may become discharged, but it is worth making a check on that
"verbal praise", or that new super-treat, and checking how rewarding it is.
The maintenance or increase in rate of reinforcement will give a fairly true
indication of the value of the reward.
The cessation of the behaviour after one click and test reward is unusual
because the dog trusts the click. But cessation can begin after the second
delivery of the test reward because the dog couldn't quite believe you forgot
to reward them. This will be a sure indication that in that situation THAT
is NOT rewarding.
Not only will the mechanical measurement of the rate of reinforcement give
you direct results, but also the way the dog returns to have another go,
try again. Their passion to complete another behaviour will be related to
the passion for the reward, their confidence in the behaviour and their enjoyment
of the learning process. Dogs are nearly perfect learners and they will excel
in the last two factors, that can be disturbed by a low passion, poorly delivered,
badly placed reward.
Comeback keenness is a measure of teaching skills - where the increments
suit the learner, and motivation for a repeat of the reward. If you are learning
how to teach in appropriate increments, make sure your reward is motivating
- in every aspect.
reprinted with kind permission from Amy Shaler
Learning About Dogs Ltd
PO Box 13, Chipping Campden, GL55 6WX
+44(0)1386 430189
www.learningaboutdogs.com
