Standard Approach in Aquaria:
Positive-reinforcement: teaching a stimulus and rewarding
a desired response:
Human whistles
at the instant a cetacean does a
desired action (jump high, come to trainer, retrieve item, etc.),
Then human provides food and/or stroking
as a more tangible reward seconds
later when cetacean comes to human.
Human does
not whistle or reward when trainee does not meet firm criteria
Human raises
criteria gradually, always within natural ability and variation of the
trainee, so behavior becomes more advanced. (Shaping)
With some animals like dogs, clickers have been used in place of
whistles.
Standard References include:
Ken Ramirez,
Shedd Aquarium (Illinois), 1999. xiv+578 pages.
Animal Training: Successful Animal
Management Through Positive Reinforcement
ISBN 0961107499 prices,
libraries,
search
in book
This book has text by Ramirez, and reprints dozens of
articles by trainers throughout the United States. He thinks it is
distracting and unnecessary to know what
animals think, since it is hard even to know what humans think.
Both books quote the late philosopher Gregory Bateson as
saying that positive reinforcement "is a method
of communicating with alien species."
When trainers whistle to reinforce a cetacean's behavior,
and others are present, some articles discuss confusion over which cetacean
the reinforcement whistle applies to. Trainers generally decline to use a unique
whistle for each cetacean, even though some species seem to use distinct
whistles themselves. Trainers call individuals with visual shapes (hourglass,
circle) rather than sounds.
One article mentions adding mats of neoprene and Corprene (a composite of cork and rubber) to a tank for
research, to reduce
echoes which would otherwise distract cetaceans.
Karen Pryor, Ringpress (England), 2002, xvi+202 pages.
Don't Shoot the Dog: The New Art of
Teaching and Training
ISBN 1860542387 prices,
libraries,
search in 1999
edition
This book has many examples of humans teaching humans, and
humans teaching animals, with positive reinforcement. It also has comparisons
to other methods. She believes both animals and humans think about their
responses, and learn faster after grasping a concept. She notes that animals'
reactions to her training cause her to change what she does, so she recognizes
that they train her.
Humans expect to be trained at school, work, and some
volunteer programs. Elsewhere Pryor recommends training humans with positive
reinforcements, without telling them that the positive reinforcements are
aimed at change, because they will dislike being trained. For example a human
who is enthusiastic during some topics or activities and bland during others
will encourage the first topics more than directly requesting them. The
implication is that people respond well to pleasurable feedback from another
person, but don't want to admit how effective it is at making them change.
Pryor says:
I found myself fascinated... with what could
be communicated between us--from me to the animal and from the animal to me (p.
xii)
Killer whales are famous for anticipating
shaping. Their trainers all have the same joke: You don't have to train
killer whales, you just write the behavior on a blackboard and hang it in the
water, and the whales will follow the script. (p.46) [Morton '02)
says it another way "They can read your mind."]
I can sympathize with biologists who want to
observe the natural behavior of animals without disturbing or interfering
with that behavior in any way, and who thus reject gross interference such as
training... But I remain convinced that shaping sessions offer a fruitful way
to combine both approaches and that both field and laboratory workers who
can't or don't consider this tool may be missing a bet. (p.156)
Subjects like to learn through reinforcement
not for the obvious reason - to get food or other rewards - but because they
actually get some control over what is happening... animals brighten up
(p.160)
Dolphins in Aquariums Communicate with
Technology:
Dolphins used underwater buttons with 4 choices ('12)
Dolphins pointed narrow echolocation beams onto an array of
hydrophones which acted like a touchscreen
('08, video)
Dolphins used
touchscreen to show they recognized a musical note ('05)
Other species also communicate with technology
Reverse Training
As noted above, Karen Pryor recognizes that dolphins train
her, to adjust how she interacts with them. People willing to have less
control can give cetaceans more chances to train people:
A cetacean
can positively reinforce a human for desired actions: The cetacean
whistles at the instant a human does something desired (provide toy, do
gymnastic feat, play music, etc.), then pulls rope to dispense nut or grape
as tangible reward.
Sides of a tank can have rows of signs visible above and below water, symbolizing common
interactions and specific trainers. Going
to a sign is a request by a cetacean for a specific trainer or
interaction.
A cetacean can teach movements to a human, who is outside a glass wall of a
tank. The cetacean can whistle and reward a human when the human learns and
does what the cetacean chooses. Initially a human in the tank can signal to
the human outside, so cetacean sees initial skills of humans. Initial skills may
be gymnastics, yoga, pilates, dance, music, etc.
|
Cetacean-Human Vocabulary
Basics
Whistle Communications:
Human-Cetacean
Yes, No, Maybe
Keyboard
Communication
Body Language
One-way Sign
Language
Shared Music
Linguistics
Research
Brainstorming
- Possibilities in the Wild to Improve Cetaceans' Condition?
Basics. Human groups who don't know each other's
language have usually developed contact
pidgins: "minimal vocabulary and grammar," which both groups
can use. Some pidgins are derived from both languages, some from just
one.
Linguistics generally uses "Language" to mean
human words or signs (as in sign languages Lieberman
2015, Chomsky
2005, Chomsky
2015). It probably can include human whistled languages.
This page will temporarily call a joint communication between species a
contact pidgin.
DareWin.org
offers over 200 hours of video and
sound recordings of cetaceans, so anyone can try to correlate behavior
with vocalizations. They have humpbacks, sperm or cachalot whales, pilot
whales, melon-headed whales, blue whales, fin whales, Fraser's dolphins,
spinners, bottlenose, orcas, common dolphins. Woods Hole
offers 2,000 sound recordings of 60 species of
marine mammals; total time may also be over 200 hours, though there is no
ability to correlate sounds with behavior. Project CETI tries to correlate
sounds, actions and meaning.
Whistle Communications: Human-Cetaceans
Herman in Hawaii created a contact pidgin of
at least 34 whistles with female bottlenosed
dolphin Phoenix.
34
are listed in Herman, "Cognition and Language Competencies of Bottlenosed Dolphins", in Schusterman et al, Dolphin Cognition and Behavior,1986,
p.230.
33
are listed in Herman et al, "Comprehension of Sentences by Bottlenosed Dolphins," Cognition, 1984, p.144.
23 are
listed in Herman, Cetacean Behavior, 1980, p.415)
Trainers elsewhere use gestures which
cetaceans lack hands to reproduce. Trainers elsewhere also merge object and
action in one command (take ball to hoop).
By using whistles, Herman created a system
capable of 2-way communication. By splitting commands into object and action,
he could reorder commands without fresh teaching (take hoop to ball).
Another step would be to create a few standard
whistles or pulsed sounds (like an alphabet) and combine them to create as
many words as needed.
Herman's goal was to prove Phoenix understood
communication and sentence order, not maximum vocabulary. He defined whistles
for:
13 mobile objects: Ball, Pipe, Hoop, Person,
Fish, Frisbee, Surfboard, Basket, Net, Loudspeaker, Water jet, Dolphins Akeakamai and Phoenix
4 fixed objects: Gate, Window, Panel to choose
tape recordings, Channel between 2 tanks
10 transitive verbs: Mouth-grasp, Tail-touch,
Fin-touch, Go over, Go under, Go through, Toss, Spit, Fetch, Place in (last 2
also take indirect object)
4 modifiers: Right, Left, Surface, Bottom
3 feedback: Correct (yes), Wrong (no), Erase
previous command
Herman knew the dolphin understood these
whistles when he put them together in new combinations, and the dolphin did
what he asked without further training 80-90% of the time.
Richards et al, working with Herman, trained a
dolphin to make distinct whistles for objects, "so that, in effect, the
dolphin gave unique vocal labels to those objects" ("Vocal mimicry of computer-generated sounds
and vocal labeling of objects by a bottlenosed
dolphin," 1984)
Herman had earlier taught 6
whistles to a female bottle-nosed dolphin,
Kea, "for three objects (ball, ring, cylinder) and three
actions (touch, fetch, mouth)", and the dolphin followed his
instructions:
"immediate generalization of response
that occurred. In addition to mouthing the three familiar training objects in
the presence of the mouth name, Kea correctly mouthed on their first
appearance a plastic water pipe, a wooden disc, and the experimenter's open
hand. The same type of immediate response generalization occurred for touch
and fetch... Kea understood the concepts of 'touchingness,' 'mouthingness,' and 'fetchingness.'
" (Cetacean Behavior, 1980,
pp. 413-418)
Kea was taken by others and released at that point in the
research. Herman does not say what happened to Phoenix, and his later
publications do not discuss the whistle communication, so further work was
probably classified. Herman did get funding
from the US Navy, starting in 1985. These whistles can develop into a
contact pidgin, since both humans and dolphins can create the whistles.
He says it was hard for humans or machines in 1980 to hear
whether the dolphins created the same whistles, to make requests of the
humans. In fact understanding whistles does not seem that hard: many humans
recognize dozens of bird whistles. Or a click system like Morse code can be
used instead of whistles. Morse operators have learned to transcribe clicks
fluently.
Duane W Batteau (1964 at 12:10)
developed machines for the US Navy, which translated human voices to higher
frequencies for dolphins to hear and translated dolphin voices to lower
frequencies for humans to hear. The work continued at least until 1967 when
the Navy classified its dolphin research. Batteau died, also in 1967, before
he published
results.
Herzing (2014)
created 3 whistles for "play
objects (Sargassum... scarf, and rope)", and found that wild dolphins
understand them, but has not found if dolphins produce the whistles.
US
and Russian
navies assign dolphins to detect mines and humans, and recover lost objects
underwater. There are no published reports on the level of detail used in the
instructions, but numerous videos
show their abilities.
Yes, No, Maybe.
Ramirez (1999)
says that trainers make a signal for a wrong
action which he calls No.
Similarly he defines Yes as praise from the trainer for a desired
action. However this Yes and No are much more value-laden than factual Yes
and No.
Herman and Forestell
said they taught dolphins to signal whether an object is present in or absent
from their pool, which they called Yes
and No. (1985).
If a more general
Yes/No can be taught, humans could
pose a variety of questions to cetaceans, about the fish they like, noise
levels, boat speed and preferred locations, etc. Presumably one can teach a
general Yes and No by varied questions: Ball present? Ball large? Sound loud?
Sound high-pitched? No one has published on the subject.
Smith, Shields and Washburn (2003)
report that a dolphin distinguished high frequencies, low frequencies and Uncertain. They taught Uncertain by
providing a response which gave a reward sooner than the wrong answer, but
not as fast as the right answer, and the dolphin used it for frequencies
close to the cutoff between high and low, in the same ways as humans did in a
parallel test (Smith
2010).
Toki
Pona is an invented pidgin with 120 words, capable of wide expression
(such as the Universal
Declaration of Human Rights, the Iliad,
El Amor Brujo,
simple phrases),
which would be a broader
start than the toy-based and military-based vocabularies used with cetaceans
so far. Humans can learn it in 30 hours.
Keyboard Communication. Several researchers
have used underwater keyboards,
which humans and dolphins can touch
to signal an action (Reiss
or here, Herzing,
Xitco). A dolphin defined a key as "I want a
small fish" and researcher Reiss (2011)
ignored it. Otherwise humans choose the few activities symbolized and seem to
provide the keyboard less than an hour per day, which limits it.
Body Language. Herman and Tavolga
(1980)
report Bastian's experiments having a female dolphin communicate to a male
dolphin, telling him to push the right or left paddle. They were in the same
tank, and a barrier blocked sight of each other but not sound. The female's
body seemed to send the signal, so body
language may be part of how cetaceans communicate. This would explain how
they can read humans as well as each other. Evans and Bastian (1969,
pp.432-433) could not find evidence whether the female intended to signal the
male, but did conclude the male read some signal from the female. They never
published the detailed results for others to evaluate. They thought it was
significant the female made the same sounds whether the male was in the tank
or not, but do not discuss whether she was vocalizing to the humans at that
point, and dismiss the significance of communicating with body language. The
authors say that among humans most evidence about intentions is
"linguistic in nature" (p.429), so dolphins' lack of words means
humans cannot know dolphins' intentions. By contrast, others believe humans
read body language (consciously or not) to judge intentions, more than words.
One-way Sign Language. Herman summarizes
other communication research with animals (accessed 2016, undated, latest
reference 1999). For 30 years, from the 70s to the 90s he created a gestural
vocabulary of "30 to 40 words"
for one dolphin, Akeakamai. These were "gestures of a person’s arms and
hands", so this was not a contact pidgin, since it was not designed so that dolphins could to
produce it too. As of 1980
(p.416), the first 22 gestures covered the same words as Phoenix' whistles
above, except Gate, Window, Panel, plus 2 words Phoenix did not have: Go left
and Go right. He lists 4 of the words Akeakamai
learned later: basket, question, present, absent. The same gestures could be
given on a TV screen, and were
immediately understood and followed.
Shared Music. Philosophy
professor Rothenberg tried to create human-humpback communication through
music, "the main thing is to listen to what's going on. If you want
to interact with an animal musician, you want to play something that leaves
space for them" (0:59). He says the humpback responded with musical
variations on his playing (3:52). His recording ("Never Satisfied") and unclear audiograms
are published (pp.236-239).
Linguistics Research.
Noam Chomsky, an expert in linguistics,
researches how human language works for humans. He says studying
communication with other species does not help us understand humans, but may
help us understand other species,
"we learn nothing about humans from the
facts that apes can be trained to mimic humans in some respects."
"it’s possible that training apes to do
things with signs can teach us something about the cognitive capacities of apes."
"ape-language
experiments... might teach us about ape intelligence by training apes to deal
with problems that are outside their normal cognitive range."
Similarly we learn little or nothing about Tibetans when
Bolivians learn Tibetan or Tibetans learn Spanish. However either lets
Bolivians and Tibetans communicate ideas about living on cold plateaus.
Shared whistles or other symbols between humans and cetaceans would let us
communicate about common issues in local waters.
Published communication experiments with
cetaceans have had an entertainment focus, like "Fetch Ball," so
the value of the communication seems limited. The same effort teaching
"Flounder unhealthy," and "Loud noise here at dawn" can
warn cetaceans to avoid polluted fish and pile driving. Cetaceans can also
identify pollution to us.
Navies have taught dolphins and belugas
to retrieve objects and identify enemy intruders. Whether the instructions
are "communication" or "stimulus," they act as contact
pidgins. It is likely navies can tell dolphins what to search for, in what
direction, and how far, though nothing is published.
Herman mentions humans' unsuccessful efforts
to decipher cetacean sounds. None
of it was done by experts in linguistics. He ignores how nearly impossible it
is (without translators) to decipher the flow of any foreign language, even
to break it into words. Babies learning English take up to a year to
recognize that louder syllables often start a new word. Then they are guided
by adults until they recognize enough words to catch these when spoken, which
breaks up the flow into identifiable portions. The great difficulty of
understanding without mutual guidance is the reason for contact pidgins.
Lieberman and Chomsky disagree a lot on
linguistics, but both say language developed 40,000-100,000 years ago;
Lieberman says mental skills developed for millions of year, and language
flowered when our necks
got long enough to make certain vowels clearly. Chomsky says our physical
nature was long good enough, so the reason language flowered must have been a
mutation which
added to our mental skills. Chomsky also says human language is a
"completely unique capacity, there's nothing analogous to it anywhere in
the animal world. There are animal signaling systems, but they're completely
different in design and use, and just every dimension" (2:05). He says
"there is every reason to believe that this [human language] developed
pretty suddenly as an optimal communication system" (4:50). If so, we'd
expect other species to have optimal systems for communicating, just as eyes
have evolved in vertebrates, octopuses and insects. The difficulty is
creating contact pidgins between these different systems.
Brainstorming -
Possibilities in the Wild to Improve Cetaceans' Condition?
Teach "Loud noise here at dawn" to warn cetaceans to avoid sonic
tests, pile driving, and other loud human activities. Cheaper and better than
canceling tests when cetaceans are in the area.
Easily shared concepts include:
Counting (each small number can be
represented by an ideophone, such "one" represented by 1 click or
whistle, "two" represented by a pair of clicks or whistles,
"six" as two groups of 3 clicks each: ||| |||,
etc.); Killian et al. (2003)
found that a dolphin understood numbers of items presented visually, and
cetaceans are likely to understand numbers presented in sound.
Common items (fish species, objects,
individuals, up, down, north, south, east, west, styles or pieces of music,
etc.)
Enthusiastic (good, happy)
Time of day (distinctive whistles at
the same times every day)
Ask cetaceans to report harassment by humans?
Learn
an association between whistles and arabic numbers, to read and report boat
license numbers (chimpanzees learned arabic numbers), or
Identify
size of boat, location, time of day
Warn cetaceans to avoid polluted fish
Test
prey fish to determine pollution levels
Create
agreed symbols (visual or sound) for each species that cetaceans eat, and for
healthy/unhealthy.
Present symbols like "Flounder
unhealthy," to warn cetaceans to avoid polluted fish. Cetaceans could
also identify pollution to us.
Encourage cetaceans to let researchers approach closer to photograph and monitor health?
Minke
whales in Australia approach humans spaced along a rope. The reinforcement
seems to be that Minkes enjoy watching humans.
When
water is too cold for people, will whales voluntarily approach a boat
offering attractions?
Drones
can sample whale blow less intrusively than boats, to monitor
health
Cetaceans have been attracted by various
stimuli,
Belugas
in Churchill have been attracted by the Beatles (2015)
Dolphins at Herman's research center in Hawaii (1980)
chose music with a lever,
"including dolphin sounds, whale sounds, human voices, music, etc."
Takaki
(1998)
also mentions "levers that control lights
or bubbles"
Pryor
(1980)
mentions "providing of means for control of the environment by the
animals themselves."
Trail
distinct buoys (different shapes,
materials, sizes); play different music
when whales approach different buoys
Provide audio connection (VOIP, Skype, etc.) among
whales in different parts of the world (e.g. belugas St Lawrence + Churchill)
|