Kayce Cover, M.S. Ed., Old Dominion University, Norfolk, VA
23529
Abstract: Presenting an innovation
in training technique leading to clearer, more precise, more efficient communication
of training requirements, which results in improved support of the animal’s
learning process. The technique requires no special equipment, adds no cost
to the training process, and yields immediate benefits. This innovation is called
the “intermediate bridge”.
What it is:
Bridge: In animal training, a bridge
is a signal that marks the instant, in time, wherein an animal meets success
(as defined by the trainer).
Terminal Bridge: A signal that marks
the instant an animal successfully completes a behavior requested by a trainer.
Intermediate Bridge: A series of
continuous and instantaneous signals marking a progression of successful instants
advancing toward a successfully completed behavior. A steady stream of articulated
syllables issued as an animal begins to cooperate with a trainer, and continued
until the animal begins to deviate from the requested behavior (at which point
they are stopped till the animal returns to compliance) or until the behavior
is successfully completed, (at which time the string of ibs is punctuated with
a terminal bridge).
The Intermediate Bridge (ib) is
a tertiary reinforcer (in that it is a reinforcer conditioned through its association
with a secondary reinforcer), that allows a trainer to give continuous and instantaneous
feedback to an animal working to complete a requested behavior. It can be given
verbally, allowing modulation of tone, emphasis, speed, and volume, which can
increase its effectiveness and utility. The appropriate use of the ib can dramatically
reduce training time, and the amount of error during learning. It can also be
an important tool in reclaiming behaviors in the process of disintegrating,
or in preventing deterioration of behaviors. It can be used to anchor an animal
meeting a new challenge, and unsure how to act, in appropriate behavior. It
is a very effective tool for rapidly increasing duration on target, or for increasing
the duration, intensity, speed, height, or other parameters of almost any behavior..
Origin:
I have been testing this tool for
over ten years, since I first realized that I, and other trainers trained by
me, were modifying bridging practices in this direction, but always with an
apology for “diluting” the bridge.
At an I.M.A.T.A* conference one
year, I presented a video of pigs from the University of Maryland, (Cover, et
al, 1990) standing for voluntary collection of blood via a 5 inch needle, thrust
into the vena cava, one inch from the heart. I had overseen the project and
trained all personnel, so any discrepancies in their technique might be laid
at my door. Reviewing the video, I watched the trainer’s practice of interrupting
a behavior midstream with a (terminal) bridge, followed by a reissue of a cue
– then terminal bridge (tb) – then cue – then tb – cue,
till finally the behavior was finished. The behavior was essentially started
and stopped and started again, over and over again in one behavior. Watching
it, made me consciously realize that I did this, and all the trainers I trained
did this – sometimes using other bridge choices (one student uses “Good!”
for example, a bridge I used at the National Zoo, but which I replaced with
“X” at around 1987.) This practice was very successful (five pigs
took an average of one hour per animal to complete this training). We felt obliged
to feed very frequently, as a result of all these bridges, so we set up the
reinforcer in a lab squeeze bottle (flat orange soda), and used the bottom of
the bottle as a target. Thus, we could target the pig and reinforce him by making
a squeeze on the bottle (target) without interrupting the flow of the behavior
(or the soda). However, when reviewing the tapes later, I questioned the need
for all that food reinforcement with, in our case, the soda. So I started developing
a specific cue to tell the animal it was on the right track, headed for sure
success, but not finished yet. This allowed us to give all the feedback and
support wanted, without stopping the progress of the behavior. It also allowed
us to increase the reinforcement of the animal for working hard, without satiating
the appetite or throwing the diet off. Most of all, it dramatically increased
my training speed while reducing the error rate equally dramatically. Later
experience showed that the ib can effectively save a behavior in progress, but
in danger of failure. We have also been able to routinely call animals away
from extremely attractive distractions (opportunities to chase other animals,
run away, fight, and/or bite) with the skillful use of the ib, in less than
fifteen minutes after initial introduction of the bridges and the target with
animals absolutely new to these techniques. (Not that this is recommended!)
Applications:
- The uses of the ib include:
- Extension of the duration of
behaviors - Acceleration of the animal’s
progress toward completion of cued request/behavior - Reduction of latency
- Intensification of focus –
despite severe distraction or challenge - Definition of the traits and
limits of behaviors - Error reduction in learning
- Smooth transitions between behaviors
- Apprehension of increasingly
subtle cues
Logic:
In training, once an animal has
learned to do something, it generally must complete a behavior before it gets
any feedback, reinforcer or punisher. Sometimes this leads to the destruction
of an entire behavior, as Keller Breland related (Breland and Breland, 1961)
(http://psychclassics.yorku.ca/Breland/misbehavior.htm).
Suddenly, some influence can become strong enough to disrupt the conditioned
response of the animal to the cue. So, a raccoon that was putting “money
in the bank” starts rubbing the money “in a most miserly fashion”
as if it were washing food. Breland’s conclusion: there is an inherent
tendency for behaviors that cross instinct to degrade. However, as an animal
trainer, I knew that my trained animals routinely resisted their “urges”
(competing instinctual drives), when executing cues. Often, these urges stem
from strong drives – drives toward attacking prey, engaging in sex, and eating.
I wondered why some very good trainers were stumped by this problem while I,
and others, found a way around this problem.
I subsequently came to believe that
a conditioned animal may have an increased tendency to produce a behavior in
response to a conditioned cue, without being consciously aware of what it is
doing. Hence, when the behavior breaks down, the animal may be equally as unconscious
of the difference in his actions leading to failure, as he was to the actions
which previously led to success. However, if, when the behavior begins to break
down, the trainer supports the animal (and the behavior), using ib’s,
the animal will often continue to complete the behavior correctly. The challenge
can often be withstood, and the behavior remains intact. For example: the animal
has been called; he starts to come but a ball bounces across his path; he looks
startled for a second, and has ceased his forward motion. At this instant, the
trainer can interrupt the dog’s preoccupation (e.g., with a loud noise
or command), and as the dog looks toward the trainer (and therefore away from
the distracting object) the trainer can issue intermediate bridges, to influence
the animal to continue in this direction. The intermediate bridges will draw
the animal back into the interrupted behavior, almost as if he had a line tied
to his nose. The efficiency of this technique can be dramatic. Another example:
an animal which is running away in panic, can be stopped by a “interrupt”
such as “NO!”. The instant the animal “startles” in
response to the “NO!” it is modifying its behavior back toward the
originally cued behavior, and the trainer can reinsert an intermediate bridge,
and begin shaping the animals behavior to reclaim a trial that otherwise would
have ended in failure.
It goes something like this (examples
of applications):
(From an explanation to another
trainer)
“Imagine if you were my animal. The first thing I do is teach you, using
10 trials, the meaning of the terminal and intermediate bridges, and the target.
You now have three keys to communication within five minutes. You know when
you have been successful (terminal bridge). You know when you are sure to be
successful (intermediate bridge), and you can clearly see the requirement for
your success (the target). Initially, I place a target so close to your finger
that just by breathing, you move enough to (almost accidentally,) touch it.
(You, being a primate, investigate your environment hands first, not muzzle
first – I hope- so I present information so that your hand can interact first.)
Next, I present the target twice as far away, but it is still very close. You
start toward the target. I start a steady stream of intermediate bridges (ib’s).
As you approach closer they get a bit faster and louder – this intensifies your
effort, perhaps because it confirms that you have this system figured out. Suddenly,
you are distracted – as suddenly the stream of intermediate bridges stops. This
is enough to pull your attention back to my request – and you make eye contact
with me, or the target – and I immediately start up the stream of intermediate
bridges. You make contact and get a terminal bridge (tb).
Without any negation, punishment
or loss of fluidity, I have you back on task and focused on my feedback.
Let’s say that the distraction that
initially pulled you away was another dog that you want to go bite – I stop
the intermediate bridge, but it is not enough to refocus your attention on my
request. I immediately interject a negation in a sharp loud tone – a pattern
interrupt. This causes you momentary pause. I start the intermediate bridge
stream again (at your pause, not before), guiding you away from failure and
back to success. It is compelling, if you try it.
As you master a behavior, the intermediate
bridges (ib’s) are faded away, leaving you with a clean cue, which elicits a
subtle, immediate response. You are still experiencing ib’s – but on new material,
not for everything that you have mastered. Because the ib allows the trainer
to closely define exactly what is wanted, for the animal, you end up with really
clean, prompt behaviors. If my horse is grazing, a football field away from
me, I can get up on a mounting block, say in a conversational tone “ready?”
and my horse immediately leaves her food and comes over to pick me up. This
is just a typical result of working this way. I have side stepped many testing
issues and diversionary tactics that I would have experienced had I not used
the ib.
However, let’s say I call my horse,
and she is on her way to me, but a snake cuts across her path. I see the situation
developing, but am too far away to directly intercede. However, I have this
communication base. I immediately start the stream of ib’s – essentially telling
the horse that at this instant – while she is still calm and has not yet reacted
to the snake in some excited way – she is headed toward success. I now issue
a cue – “back”. The horse pulls her head up a bit – she is still thinking
about that snake, but I have a share in her attention, and her pulling her head
up is an increment in the direction I want. So, I intensify the ib’s. She gains
confidence and starts backing up, to my rallying bridging. Then I say “left,
walk” – and at this point she gives a last piercing look at the snake (which
I also was naming, by the way) and commits to moving to her left, etc, etc.”
…
Why this isn’t featured in
the OC literature if it is so effective:
Unlike an experimental psychologist,
who may be interested in exactly how a behavior breaks down, and how resistant
it is to breakdown, etc, most trainers are interested only in success. And,
it is doubtful that science can ever keep up with the advancements we make in
the field, simply because the rigor of science makes scientific progress a slow
and ponderous process. I might be able to train 100 behaviors before I can prove
anything about one behavior. So, one reason that the use of the ib is not prominent
in operant conditioning research literature and practice may simply be that
none have gotten around to exploring it yet.
Nonetheless, the ib can be related
back to descriptions of complex and chained (schedules of ) reinforcement, as
described in OC literature. However, the significance of this tool seems to
be totally overlooked, in that subsequent experimentation apparently (as far
as I have been able to find) fails to explore these tools. Further, researchers
do not appear to exploit the power of the ib’s (or targets – beyond
getting lever pushes or animals to station) much, if at all, in the creation
of concepts and behaviors in research.
However, once aware of the importance
of this tool, one sees it, undefined, everywhere. The basic concept of the ib
underlies the child’s game of “Hot and Cold,” and the development
and use of homing devices and retrieval pingers. Further, at an I.M.A.T.A.*
conference in the late 70’s or so, Richard Humphries presented work he
conducted from his time with ABE, where cats were trained to be guided by acoustic
beeps (transmitted to receivers in their ears). The cats could be guided directionally,
and led to stay in one spot for a period of time, using these signals. The ib
was foreshadowed, but yet to be named, defined and fully exploited.
Comparison between a Keep Going
Signal (KGS) and an IB
The question has sometimes arisen,
“How is the intermediate bridge different than a Keep Going Signal?”
From reading explanations of the application of the KGS, and watching demonstrations
by practitioners, I offer these comments, acknowledging that I am not an expert
on the use and application of the KGS:
The ib is a form of intricate coaching, issued during the defining of a behavior,
or to prevent disintegration of a behavior, or to repair a disintegrating behavior,
or to give an animal feedback on how it is responding to a distraction or challenge.
The KGS is apparently most often used as a reminder that there is a trainer
on the scene, where a cue was given some time past and success has not yet been
met. For example, in guiding a student in the use of the KGS, one Clicker Trainer,
an instructor and clinician, said that there was no reason to give a KGS if
the animal had not been working on the behavior for a half hour or more. In
one case (ib), the trainer coaches the most subtle nuance of the behavior and
gives the animal constant feedback on his progress toward success, in the other
(KGS), the trainer acknowledges that progress is being attempted if enough time
goes by before success is met. So, while both the ib and the KGS are directed
toward the same need, the application is significantly different.
With the advent of the ib, endless
debates about whether to click a less-than-desirable behavior (because at least
the animal is trying) or leave a void, or about whether it okay to let three
seconds pass between a negation and a punishment, disappear. With an ib, an
animal knows, much earlier, whether he is going to be successful, and can correct
in progress.
How to condition the IB:
5-minute directions for conditioning the ib, can be provided upon request. Please
see below.
An aside about the role of research
in training and training in research – a hope for further research and
greater collaboration:
Just as trainers benefit from learning
from the vast body of knowledge and experience contained in the Operant Conditioning
literature, researchers might really benefit from using communication and motivational
tools from the vanguard of animal training for better support of their research
projects. Rather than just study these issues, scientists can use them for producing
better and more valid research. (I was hired as a technical resource at the
University of Maryland so that their researchers could have these types of tools
at their fingertips.) Just as trainers will benefit from studying the contributions
of research behaviorists, researchers could produce less confounded experiments
if they used better methods of communicating the ideas being tested.
In summary:
When an animal is learning, the
trainer can support it continuously with an intermediate bridge. The instant
the ib ceases, the animal starts to question its performance and will usually
move to correct the behavior. It knows the instant it veers from perfect performance.
He does not get all the way to the end of the behavior before being told (by
negation, or lack of reinforcer) that his behavior will not result in success
(and pay!). Since the animal now knows the instant he deviates from perfection
– and he can correct immediately – the behaviors can develop very cleanly,
and the cues can quickly fade to subtle echoes of their initial forms. Any trainer
will probably agree that it is sloppy and counterproductive to have an animal
invest rehearsal time in incorrect behavior. Most trainers want each and every
transaction to advance the behavior. This is most likely to happen when the
animal gets immediate, precise, clear information on what is wanted and how
he is progressing to meet those wants. The animal can move toward success with
confidence because he has correctly assessed how to create success. And, success
in work tends to lead to enjoyment of work.
The trainer can use this same ib
signal to correct the breakdown of behaviors, prevent the breakdown of behaviors,
increase the efficiency of proofing behaviors, and support the animal’s
response to novel challenges. This is an extremely versatile and powerful training
tool. It is effective. It is efficient. Many professionals have reported an
increase in training efficiency of 25-75% (less training time), with superior
results.
Hopefully, people will actively
explore this tool. Having been involved in a wide range training tasks, and
having seen the power of the appropriate use of ibs under every circumstance,
I would not be without this versatile tool. However, try as one might to describe
the ib, it is best appreciated when seen in action. For examples of bridge use, and animals trained via the use of the Intermediate Bridge, please see synalia.com/press and YouTube, search Kayce Cover.
References:
Breland, K., and Breland, M., 1961. The misbehavior of organisms. American Psychologist,
16, 681-684.
Cover, K., Erez, B., Hartsock, T., 1990. Voluntary blood sampling of pigs. Video,
K. Cover and U. MD.
*International Marine Animal Trainer’s
Association
Copyright 2002 Kayce Cover; First
published in the October 2002 issue of American Animal Trainer Magazine. All
rights reserved.