BONGO BIRD
Alex the Talking Bird
By Irene M. Pepperberg, Ph.D., and Robyn J. Bright
Birds can learn to talk... and to understand
Training a bird to talk can be either terribly exciting or extremely
frustrating. Why do some birds learn to talk and others not learn? The
answer depends on many factors, including the type of bird you're
working with, its personality and age, your relationship with the bird
and the techniques you use to train it.
BIRDS DIFFER IN LEARNING ABILITY
Certain species are considered better talkers than others. Among the
larger parrots, African greys are considered one of the best, followed
by Amazons, especially yellow napes, blue fronts, red loreds, and double
yellowheads. Macaws, such as scarlets and the blue and golds, can learn
to talk, but they are usually loud and rough voiced. Cockatoos can also
learn to talk, but unlike macaws, they tend to have sweet voices.
Neither cockatoos nor macaws are as easily trained to talk as African
greys and Amazons.
Mynahs, especially greater Indian hill mynahs, are also well known for
their speaking abilities. One mynah, for example, would respond well to
children, saying "Hello" when a child approached, then
"How are you?" if the child said hello in return, and finally,
"What's your name?" much to the child's delight. Smaller
species of parrots can also learn to talk very well, especially
budgerigars and members of the Psittacula genus, including ringnecked,
Alexandrine, and plumheaded parakeets. Cockatiels, lovebirds, lorikeets
and conures may also learn to talk, but they are usually more difficult
species to teach. There are exceptions to every guideline about species'
capabilities, however, one reason being the personality of the bird
being trained to talk.
Personality plays a large role in whether any bird learns to speak, and
some birds never learn, no matter how long or intense their training.
Others pick up everything with apparent ease. In fact, you must be
careful about what you say and do around a bird that learns easily. Many
times swear words and annoying noises like a baby crying have been
picked up by talkative birds!
If you want to teach a bird to speak, choose a bird that appears alert
and interested in what's going on around it, one that seems to listen
when you speak to it and is vocal in general. The best-talking budgies
are usually those that constantly babble and chirp before being trained
to talk. Males also tend to be better talkers than females, especially
in budgerigar and cockatiel species. It is important to spend time
observing a bird's behavior before making a purchase.
The age of your "student" and the number of other birds in the
same household also affect the likelihood that a bird will learn to
speak. Young birds especially hand- raised ones, are much more likely to
talk than older birds. The best time to begin training is around 4 to 6
months of age with smaller birds like budgies, and between 6 and 12
months of age for larger birds like African greys and Amazons, although
younger parrots can begin picking up words.
One bird is easier to teach than two birds: Two birds are more likely to
speak to each other in their own language than to speak English to you.
Birds are, however, social creatures and do not thrive in solitude. A
bird who is left alone should have a mirror and other toys. Such
playthings will not affect the likelihood that the bird will speak or
become tame. Although some bird owners believe that a mirror negatively
affects their pets' ability to be tamed, we have not seen evidence of
this in our 50 years of combined experience with birds.
The bird's living conditions, including its social and physical
environment, will greatly affect its willingness to talk. Treat the bird
as you would a human. Be a friend, not an owner, and place the bird
where it will receive plenty of attention and social interaction. If the
bird is frightened, work slowly at first; a fearful bird will not attend
to training. If you put yourself in your bird's position, you can
realize how large you are in your bird's eyes. A bird that feels
comfortable around humans will be more inclined to talk than one that
runs away, growls or hisses. Patience is vital, not only to gaining the
bird's trust, but to teaching the bird to talk.
REPRODUCING HUMAN SPEECH
Once the bird is relatively calm, an owner can choose among many
different techniques for speech training. Two methods are most common.
In one system, you simply repeat the desired utterance to the bird as
clearly and as precisely as possible. The other technique involves
playing a record or tape that consistently repeats a single word or
phrase to the bird. Each system has advantages and disadvantages. A
recording provides consistent input, while a live trainer may begin to
slur and change input during the training period, perhaps even making it
difficult for the bird to understand what it's supposed to learn.
Recordings can be used unwisely, however. Birds will tire of and may
eventually stop listening to a recording played too often or for periods
greater than 15 minutes. A bird will not learn from a recording that it
has 'tuned out:' and a live trainer is more likely to keep the bird's
attention. Reinforce any recordings you use by speaking the same word or
phrase that is played, pronouncing each word carefully and keeping
training periods short. Birds can certainly learn to talk without
recordings and in fact, most do; a record or tape simply provides more
input than owners may be able to furnish on their own.
A bird learning to talk must hear input that is clear and unchanging,
but just as important, it must attend to the input. To attract and
maintain your bird's attention, hold it in front of your mouth while you
speak so it knows the sounds are directed toward it. Two training
periods per day, each from 10 to 15 minutes long, will suit the
attention spans of small birds, although larger birds might train for
somewhat longer periods. Simply talking with a bird as you would another
person between training periods may also help teach the bird to speak.
For example, tell the bird what you are putting into or taking out of
its cage each time you feed it or clean up.
Talking about objects and actions is a training method in itself and is
referred to as "training by association." Most owners come
upon this method by accident, and it is a particularly powerful
technique. One Amazon, for example, had never talked, even though its
owner had been trying for two months to teach it to say
"Hello." The owner always greeted her bird with "Good
morning" when she removed the cover from the bird cage, and one
day, before she said anything, her bird said "Good morning" as
she removed the cover. Obviously, the Amazon had associated the cover
being pulled off with the phrase "Good morning."
Many bird owners will tell you their birds say "Hello" when
the telephone rings or "Come in" when the doorbell buzzes,
even though these birds have not been trained to talk. The technique
simply involves associating actions or objects with speech. For example,
say a word, such as "Hello," every time you walk into the
bird's view, or use a specific word or phrase during feeding. The bird
may learn to produce these words by association.
A bird that is good at association can also acquire a word or noise that
you may wish it never learned. Screen what you let such a bird hear
because it is likely to learn words said loudly in anger or agitation!
Noises are also easily learned. Our lab mascot, a parakeet named Tribble,
faithfully mimics the sound of a computer printer, and it's a wonder
he's lived to his first birthday because of it!
It is difficult, but not impossible, to stop birds from saying unwanted
words or noises. If the bird is out of its cage and/or receiving
attention when it makes the unwanted word or noise, put the bird back in
its cage and ignore it. Birds that do not like to be sprayed with water
can be taught a negative association by spraying them when they make
unwanted sounds. Another technique to discourage annoying noises is
simply to cover the bird's cage for 5 to 10 minutes each time the bird
produces these sounds.
If you want your bird to talk, do not first teach it to whistle.
Although birds enjoy learning to whistle, many people have had
difficulty teaching their bird to talk if it first learns to whistle.
Don't teach whistling until after the bird has learned its first words.
Learning its first word or phrase is always the bird's most difficult
task. A bird may need as little as one week or as long as six months of
training before it produces its first sounds, and some birds never learn
to talk. Many birds mumble or babble words while learning. This behavior
is normal and means that the bird is very likely to learn if it
continues to hear the word or phrase it is babbling. Once a bird has
spoken its first word or phrase clearly, it will learn new words more
easily.
All the above methods are designed to teach a bird to talk, but are not
designed specifically to help the bird understand what it is saying. A
bird that says "Hello" may not realize that word should be
used only to greet someone. A bird that can learn to reproduce sounds,
however, can learn the meaning of what it says. The next section
illustrates how to accomplish this.
TEACHING COMMUNICATION, NOT JUST TALK
Only a few techniques can be used to teach your bird to do more than
mimic human speech-that is, to use words meaningfully and in appropriate
context. Prior to the studies in our laboratory, for example, several
American researchers (Joseph H. Grosslight with Wesley C. Zaynor, and
O.H. Mowrer) tried to use procedures of the psychological laboratory to
teach birds to communicate meaningfully. These researchers put their
birds in social isolation and played tapes of sounds they wanted them to
mimic, or they repeated sounds in the birds' presence without any
reference to the meaning of the words.
The birds were generally rewarded with food if they responded properly,
for example receiving a nut in exchange for the word "Hello."
The utterances then actually became requests for food, rather than
meaningful communication. Not surprisingly, the view of parrots as mind
less mimics became entrenched both in scientific literature and in the
minds of the public.
During the 1940s and 1950s, however, European researchers, such as Otto
Koehler, showed that African grey parrots could learn nonvocal tasks
like matching quantities that are generally thought to require complex
intelligence. Other researchers, working both in the field and in
aviaries, subsequently observed behavior that indicated parrots use
their natural vocalizations meaning fully with each other, which they
probably learned from other flock members. These two lines of research
suggested that the American psychologists failures to achieve meaningful
communication with their birds might be due to inappropriate training
practices, rather than to any inherent lack of ability in their
subjects.
I decided to test this premise. I started my research with an
approximately 1 year-old African grey parrot named Alex in June 1977.
Combining the experimental rigor of the laboratory with what little was
known about parrot communication in nature, and borrowing ideas from
projects that examined the bases for human social learning, I devised
techniques to teach Alex to communicate meaning fully. These techniques
have been successful, as you will see.
INTRINSIC REWARDS
One feature common to all our training procedures is the consistent,
exclusive use of intrinsic rewards. The reward for using a word to label
an object is the specific object to which the label refers, not a
general item like food. So if Alex correctly identifies a cork, we give
him the cork. This procedure consistently ensures the closest possible
connection between the objects and their labels, and prevents Alex from
using his labels as a means to get food, unless, of course, they are
food labels.
Because Alex is a laboratory subject rather than a pet, he is often
asked about objects and concepts that he may not find particularly
interesting. To keep him attentive and motivated, we reward him for a
right answer with the right to request vocally ("I want X") a
more desirable item than the one he has identified. This system provides
some flexibility, but it keeps his reward vocalizations meaningful: Alex
will never, for example, automatically receive a slice of banana after
identifying a cork. He must specifically request banana ("I want
banana"), and we will not respond to such a request until he
completes the appropriate prior task.
THE MODEL/RIVAL (M/R) TECHNIQUE
Our primary training system, called the model/rival, or M/R,
technique, is based on a protocol developed by Dietmar Todt, a German
ethologist interested in social learning in parrots, and on the work of
Albert Bandura, an American psychologist studying how social modeling
affects learning. M/R training involves three-way interactions between
two humans and the avian student. We use the M/R technique primarily to
intro duce new labels and concepts, but also to shape pronunciation.
During M/R training, humans demonstrate how the label is to be used. In
a typical interaction, Alex sits on his gym, on his cage or on the back
of a chair, observing two humans handling an object in which he has
already shown some interest. He watches one human act as a trainer of
the second human; this second person acts as a model for Alex's
responses and as his rival for the trainer's attention. The trainer
presents an object, asks questions such as "What's here?"
"What color?" "What shape?" and gives praise and the
object itself as a reward for a correct answer.
Disapproval for incorrect responses (errors similar to those Alex may be
making at the time, such as unclear utterances or partial
identifications) is demonstrated by scolding and temporarily removing
the object from sight. Alex therefore observes the effects of an error.
The model is asked to try again if the identification was wrong or to
speak more clearly if the response was (deliberately) incorrect or
garbled. Alex is often included in this procedure, and although his
first - often unclear - attempts at the new vocalization are rewarded,
he is shown by the model how he must improve.
Unlike the modeling procedure developed by Todt and other researchers,
this protocol involves repeating the interaction while reversing the
roles of the human trainer and model, and occasion ally includes Alex in
the interactions. Our protocol shows Alex that the interaction is indeed
a "two-way street": Since the same person does not always ask
or respond to the questions, Alex sees that the procedure can be used to
make changes in his environment.
Inclusion of role reversal in M/R training appears to counteract what
would be, for our project, the drawbacks associated with Todt's method:
Todt's birds were exposed only to pairs of individuals who maintained
their respective roles, and his birds did not transfer their responses
to anyone other than the human who posed the questions. In contrast,
Alex responds to, interacts with and learns from all the trainers with
whom he comes in contact.
SENTENCE FRAMES
After Alex begins producing a new label in the process of
(appropriately naming) a new object, we use an additional procedure to
clarify his pronunciation. We present this object to him along with
several "sentence frames" phrases like "Here's your
paper!" In this manner we can present a target word, such as
"paper" frequently and with consistent stress, without
presenting it as a single, repetitive sound that Alex might find boring
and might therefore ignore. This combination of vocally repeating the
label and physically presenting the object resembles the behavior
parents sometimes use when introducing words for new items to very young
children, which appears to have two effects. Alex hears the label the
way it is to be used and learns that context, and he learns to reproduce
the label without associating simple word-for-word imitation with
reward.
REFERENTIAL MAPPING
We use another set of procedures, called referential mapping, to
assign meaning to new utterances that Alex occasionally produces
spontaneously. The vocalizations are generally combinations and
variations of the English labels he has already learned. For example,
after learning the word "Gray," he produced "Grape,"
"Chain" and "Cane." Unlike the vocalizations that we
train by using the M/R procedure, he does not necessarily use these
labels to describe or request new objects or circumstances. Our
procedures for dealing with these spontaneous words, however, neither
attempt to nor need to evaluate the intentionality of Alex's behavior
Rather, we merely respond to the novel speech acts as though he were
intentionally commenting about or requesting objects, actions or
information.
Referential mapping includes:
Showing how the use of the word affects others.
Because other studies on both humans and birds suggest that experiencing
the appropriate consequences of an utterance may assist learning, we
respond to Alex's vocalizations with an appropriate object or action.
For example, we treat him as if he does indeed understand the
significance of what he is saying. Whether he intends to produce the
combination is unimportant; we simply demonstrate that these phrases can
be meaningful and can be used to control, or at least influence, his
environment and the actions of his caretakers. Some animals may have
similar experiences in the wild: Young birds may, through their
interactions with adults, learn not only what to sing but also how song
is to be used.
Demonstrating varied meanings of the word.
We also use a variation of the M/R technique to further demonstrate the
possible relevance of Alex's spontaneous recombinations. In this case,
two trainers model an interaction corresponding to the now-targeted
vocalization: One human says the new word, while the other produces an
object or an example of the term, or demonstrates the action to which it
refers. We then reverse roles so Alex sees that the exchange is neither
specific to nor controls only one person's actions. If Alex says the
word during the demonstration, he is shown and occasionally receives the
object or action - a "chain" of paper clips, for example. The
human trainers not only act out identification of the object or action
by responding to each other's queries of "What's this?" but,
when possible, use objects that demonstrate how the vocalization can
have varied applications. One example of this is producing
"boxes" of different shapes or sizes.
Elaborating on the word in sentence contexts.
Finally, we use another set of sentence frames to provide additional
cues about the appropriate contextual use of the object or action label.
While the targeted object is being manipulated or the action
demonstrated, either by the human or by the parrot, humans say such
phrases as "You're eating a green nut. Do you want another green
nut?" in which only the label for the action or object remains
fixed and stressed. As before, Alex hears numerous repetitions of the
label but in ways that demonstrate the connection between this label and
the object to which it refers.
ALEX'S ACCOMPLISHMENTS
Using these techniques, Alex's trainers have, over a period of 12
years, taught him tasks that were once thought beyond the capability of
all beings except humans or possibly certain nonhuman primates. Alex has
learned to name about 40 different objects, including paper, key, wood,
hide (rawhide chips), grain, banana, box, gym, peg wood (clothespins),
cork, corn, nut, walnut, showah (shower), wheat, pasta, cracker, scraper
(a nail file), chain, shoulder, block, rock (lava stones, hardened
Playdoh forms), carrot, gravel, back, chair, chalk, water, nail, grape,
cup, grate, treat, cherry, wool, popcorn, citrus, green bean and banerry
(apple). We have tentative evidence for labels such as bread and jacks.
He functionally uses the word "no," as well as the phrases
"Come here," "I want X," and "Wanna go Y,"
where X and Y are appropriate labels for objects or locations. If we
respond to his requests incorrectly, for example, substituting some
thing other than what he requested, he generally (75 percent of the
time) says "No" and repeats his initial request.
Alex has labels for seven colors: rose (red), blue, green, yellow,
orange, gray, and purple. He identifies varied shapes by labeling them
as two-, three-, four-, five or six-cornered objects. He uses the labels
"two," "three," "four," "five,"
and "six" (six) to distinguish quantities of objects up to
six, including collections made up of new objects, sets of different
objects and sets in which the objects are placed in random
configurations. He combines all the vocal labels to identify
proficiently, request, refuse, categorize, and quantify more than 100
different objects, including those that vary some what from training
exemplars. His accuracy has averaged 80 percent when tested on these
abilities.
We have also examined Alex's comprehension of "category." Not
only can he identify any one of a number of different hues or shapes,
but he also knows that "green," for example, is a particular
instance of the category "color." Thus, if we show him objects
having both color and shape, he can use this knowledge to respond to
questions of either "What color?" or "What shape?"
Alex has also learned the abstract concepts of "same" and
"different," and he responds using the word "None"
to the absence of information about these concepts. Thus, if we show him
two objects that are identical or vary with respect to some or all of
the attributes of color, shape and material, he can use the appropriate
category label to tell us which attribute is the "same" or
"different" for any combination. If, however, nothing is the
same or different, he says "None."
He can also respond to questions involving objects, colors, shapes and
materials not used in training, including some objects whose names he
does not know. Alex is indeed responding to the specific questions, not
merely on thc basis of his training and the physical attributes of the
objects: His responses were above chance levels when, for example, we
showed him a green wooden triangle and a blue wooden triangle and asked
"What's same?" If he were ignoring the question and responding
on the basis of his prior training, he would have determined and
responded with the label for the one different attribute, in this case,
"color." Instead, he responded with one of the two appropriate
answers "shape" or "man-man" (matter). Such mental
faculties were once thought beyond the capacity of any bird.
Despite our success in teaching Alex to communicate with us at an
apparently advanced level, he cannot talk to us the same way that we
talk to one another. For example, he does not tell us what he did
yesterday, nor can we ask him what he would like to do tomorrow. But our
work does suggest that these birds are intelligent creatures that can
interact with humans in some very interesting ways. Not all birds will
necessarily learn as much as Alex has. Some birds may even learn more.
The essential point an owner of one of these creatures must realize is
that birds are intelligent beings that deserve companionship and
intellectual stimulation.
