Chimpanzee

1
Chimpanzee Language
Psych 1090 Lecture 13
2
The issue of whether animals can truly
communicate with humans is incredibly thorny.
The fights pro and con were brutal and, in a
sense, doomed the field
However, before it collapsed, the work had a
stunning impact not only on how we view animals,
but also on our understanding of language
3
And Im going to try to give you just a taste of
what happened
The work started with the Kelloggs
who brought up a chimp, Gua, along with their
son, Donald.
and who showed that simple exposure to language
wasnt enough to engender it in a nonhuman primate
4
Hayes and Nissen came next they trained a
chimpanzee named Viki
Although Viki succeeded on a large number of
cognitive tasks,
she acquired only a very few English labels
Was her failure due to her primitive vocal
tract or her cognitive skills?
5
The Gardners, Allan and Trixie, watched tapes of
Gua and Viki
and realized that they could understand
everything without sound
i.e., that the apes used a lot of gestures
They reasoned that maybe one could use ASL, the
language of the deaf, to work with the apes
6
They reasoned that ASL would allow them to
separate out issues of
the inability to produce labels vocally
from the possibility of learning a human language
Note that, at the time, not everyone agreed that
ASL was a true language
7
Some folks looked at ASL as a proto-language
others thought about it in terms of a rationale
for a gestural theory for the origins of human
language
and others began studying ASL for the first time
in great detail as a consequence of the ape
language controversy
8
Were what seemed to be simple signals like
you-me-eat?
representative of a real language, whether
produced by humans or apes
or were such utterances as primitive as they at
first appeared?
Until the ape studies, no one seemed to care very
much
9
But those issues were to comeFirst the Gardners
started their pioneering work with Washoe
Not too long afterwards, David Premack decided to
work with a more restrictive symbol system of
plastic chips
He trained Sarah to use these chips to
communicate with her trainers
10
Sarah with something like apple same as apple
11
And fairly soon after that, Duane Rumbaugh,
working with von Glasersfeld and others
developed a computer-based system of lexigrams
that allowed their ape, Lana
To function without any human intervention at
all, to avoid cuing
12
Original lexigrams were color-coded.
13
                                            
One of the later keyboards
14
And, in another lab at Columbia University, Herb
Terrace started a second ASL project with Nim
Chimpsky
Terrace had been trained by Skinner and felt that
language could be taught through operant
conditioning
15
Nim did not learn as much as Washoe,
And Terraces attacks on the field pretty much
brought it down
But, again, that was in the future and an
incredible amount of information was gathered in
the interim
16
The basic issuehow did animal language differ
from human language
could only be resolved if we could figure out the
limits of the animals abilities
AND insure that any animal failure was not a
consequence of poor training or vocal anatomy
17
So, that is why Im going to start with Premacks
1990 paper even though its not the oldest
Because he brings up fundamental issues of
similarities and differences
Between animals and humans use of symbols
Are they equivalently words?
18
Now, Premack was not the first to worry about
concept word
In 1960, Quine wrote a paper on the topiche
proposed the following
A linguist visits a primitive tribe, and while he
is talking to a tribal member, a fuzzy creature
runs across their path
19
The tribal member states Gavagai.
Our inclination is to assume that gavagai is
the label for that creature in that culture
But, in reality, gavagai could mean any furry
critter, anything that is running fast, a generic
term for a mammal
Or even lunch!
20
Or, of course, any number of different things.
The point was to make it clear that establishing
the referent of a sound pattern isnt necessarily
simple
And that we do it by context, by exclusion, by
categoryetc.
21
Now, if we have so much confusion about what a
word is or means for humans,
there is even more confusion about what it might
mean for an ape
and whether an ape would be mapping human terms
onto something already existent in ape terms
22
Specifically, when we train an animal to label
something,
are we first having to train it to understand the
concept of labeling?
Or are we just setting up some kind of paired
association that has no linguistic function?
23
These are not trivial questions, and fueled the
sometimes vitriolic debates
And there arent any clear, clean-cut answers
Although, as well see, Premacks animals seemed
to treat their plastic chips as words
24
an intriguing study by Lenneberg suggested maybe
not
Lenneberg replicated Premacks chip study with
college students,
who, not surprisingly, did extremely well on the
tasks..
BUT, at the end, had no idea they had learned a
language
25
So you need to keep this all in mind as we
discuss the various papers and the various results
Remember, all of these apes were trained in very
different ways, with different techniques
And even Nim and Washoe, both taught ASL, had
strikingly different input
26
So lets go back to part of Premacks paperabout
halfway through, where he starts talking about
Sarahs learning.
After she had acquired a number of labels, she
began to learn more rapidly
and, in fact, connected words and objects just
by having them placed together
27
And, interestingly, we see this difference in our
birds
The first label is learned quickly, as a
generalized gimme
The second and a few subsequent ones are learned
very slowly,
Suggesting that the concept of labeling is what
is being learned
28
And then the animals get it
although naming and requesting are still not
separated
In fact, this separation is what caused many
researchers to use nonreferential food rewards.
The animal supposedly would see the treat as a
reward for naming in general
29
But that didnt happenanimals trained with
nonreferential rewards never really understood
labels
They just learned associations that gave them
treats without connecting the label and the
object as a name
The animals couldnt usually transfer the label
to a similar but not identical item
30
So Premack could show that his animals treated
the plastic chips the way children treated vocal
labels
The real question, still not totally answered, is
what the chipsor labelsrepresent to the ape
Let me try to clarify
31
If I say unicorn to you,
You know exactly what I mean .
Because you have a full mental representation of
the unicorn, even if you had never seen a picture
and I said it was a horse with a horn
But if you needed an association between a label
and a physical object in order to represent it.
32
You might not be able to understand what I meant
And thats what we cant quite separate out for
the animals
When is it just a Pavlovian association, like
with a lemon and salivating, for you and the
animals
or something more??
33
Does real reference require, as Premack suggests,
some kind of theory of mind?
That is, by using the label apple, Sarah
theorizes that you have the same mental
representation that she has?
Tis not at all clear, and has been the basis for
plenty of controversy
34
Because simple systems can be simple associations
without requiring clear representations
That is why Premack used the tests with things
like apple stems, to see if these items
engendered the same response as his plastic chips
The data suggest that the chips may really have
been representational
35
Another intriguing issue that Premack raises is
whether language training changes how apes think
In a sense, any training changes how thinking
occurs.
Thats why programs to train students for SATs,
MCATs, etc. are so popular.
36
If you teach to the test, most studentshuman
or nonhumanwill do better on that test
But they may not do so well on other types of
intelligence test.
So there is a difference between practice effects
and true re-training
37
And, again, that was one of the problems with
Premacks studies
If Sarah finally abstracts the chip that means
color of
And learns brown through a statement brown color
of chocolate
The test of brown must be strong
38
But Sarah was given a brown exemplar along with
three other already KNOWN color samples
So when tested on take brown
She might still have chosen merely on exclusion
I know it isnt red, green, or yellow
39
So she would need to choose from a pile that also
had purple and silver
And, of course, we run into the same issues of
grammar as we did with Hermans dolphins.
Is it just rule-governed behavior, which isnt
exactly grammar?
40
So, for example, Premacks apes did fine with
their usually VERTICAL language with things like
red over blue
But had a terrible time with things like red
under blue
The latter was opposite to the ordering of the
chips on the board
41
Such data presume rule-governed behavior, not
grammar.
So, keep all this in mind when reading about the
language-trained apes that succeeded on certain
tasks
And the nonlanguage-trained apes that failed
42
The task with the cut-up fruits and the tubes of
water is a good example.
Nonlanguge-trained apes knew match-to-sample
And matches were always identical.
So faced with apples and tubes of water
43
It wasnt all that surprising that they were at
chance.
They had no idea that they were matching
proportions.
Only identity, and nothing was an identity
44
The language-trained apes knew match-to-sample
But also had lots of training where they matched
objects to chips
So they understood that the objects didnt have
to have an exact match to be correct
45
To them, the partially-filled tube did not have
to have the same representation as a plastic chip
What the language-trained ape knew was to derive
information from such situations
So it could respond based on the concept of
quarter-ness
46
In a way that the nonlanguage-trained ape could
not.
But language per se was not necessarily the
difference
Only the way the animal had been trained to
respond to certain situations
47
The same issue probably held with the causality
testsi.e., knife with cut item
It wasnt necessarily that the nonlanguage-trained
apes couldnt understand the causality
They didnt know how to express what they might
have known
48
The analogies test are my favorite
because these really ARE a lot like the kinds of
tests given in SATs
And for which training clearly makes a huge
difference in humans
Why not in apes?
49
Sarah would see
?
50
Choice of the correct alternative might be done
just by matching the most number of attributes
The trials were a bit different for the
nonlanguage-trained apes
51
They saw something like
????

52
Tis not exactly sure that I would immediately
understand, if I were an ape and not an adult
human
that I should pick the matching pair
So Im not sure that Premacks tests were
entirely fair
53
Now, it may seem that Im really coming down hard
on Premack
And Im not really.
I do think that his animals learned to reason in
a very advanced way
Im just not sure they learned language
54
So what about an ape like Washoe,
who was not trained in an operant paradigm but
rather like a child?
Well, one problem was that at the beginning the
Gardners were not fluent in sign
So Washoe wasnt really taught ASL
55
But rather a form of signed English
which was neither ASL nor English
Nevertheless, Washoe did learn a lot about the
referent of each sign
A critical issue was that there were no formal
drills like those of Sarah
56
And that Washoe had constant referential
experience with all her signs
Later, younger apes were taught by fluent ASL
people
And thus these apes really were exposed to a
human language
57
And, one of the major issues concerning this work
is that ASL does not have the same grammar as
English
So, as I noted earlier, apes that signed
you-me-eat?
were compared to young speaking children who
might say Mommy, lunch now?
58
Or even older children who might be using full
English sentences
If, however, the apes were compared to ASL
proficient children,
The parallels were much stronger
Even repetitions used for emphasis were similar
59
But much argument centered on even whether
Washoes labels were indeed referential
Did the ASL sign for shoe actually refer to shoes
in general and not to anything on a foot, for
example?
And if the latter, was it clever metaphor or an
error???
60
Of real import was the Gardners descriptions on
their vocabulary tests of getting the subjects to
cooperate
Think about trying to get a somewhat hyperactive
child to sit still through a repetitive test
And you start to see their problems
61
Unlike the dolphins, the apes were not given
their daily rations as part of the test
Mainly because hungry apes are even more
difficult to get to focus
These issues were not trivial because most
critics had worked with pigeons or rats
62
Or with children who were cooperative
In most studies, fussy children are eliminated
from testing
And, overall, the apes did respond appropriately
to most of the exemplars in the vocabulary test
63
Interestingly, many of the errors were small
errors with respect to where a sign was placed or
the action that was used
Placing several fingers to the nose meant FLOWER
and one finger meant BUG
In context, such errors could be made by humans
and be ignored
64
These errors are the equivalent of mucking up
small phonemes in English
Other errors were within category COMB for BRUSH
Objects that might be linked closely in memory
65
The Gardners had other data for referential sign
use
There are videos via hidden cameras of the
younger chimps correctly signing about the
pictures in magazines they are viewing
Or their asking for objects that were lost
66
The Gardners emphasized the strong need for
interaction with their subjects
arguing that communication is, by its very
nature, interactive
and similarly emphasized the need for referential
rewards
67
Their initial papers, published in 1969, strongly
influenced some researchers like myself and Reiss
And their work strongly contrasted with that of
Terrace--
who also used sign, but in a very different manner
68
Terraces ape, Nim, would often be taught signs
in the absence of a referent
under the assumption that the referent would be
distracting and make learning more difficult
So, in many cases, only after a sign was acquired
did the referent appear
69
Not surprisingly, Nim didnt understand why he
was supposed to make these signs.
and the occasional food reward just acted to
confound the label to be learned with the food
reward
He did eventually learn a number of signs
70
And initially, Terrace claimed full
referencewhich he later retracted
He argued in particular that not only Nim, but
all apes failed in putting the labels into
sentences
And that therefore they really hadnt learned
very much at all
71
Interestingly, when Terraces students did
interact with Nim on a less formal level
labeling actions and objects
they didnt do it in a systematic way, so Nim was
stuck with gavagai like situations
72
Up might be used when he was lifted to his high
chair, but not when he was lifted in other
circumstances
leading to lots of confusion as to the meaning of
labels
The Gardners, in contrast, were very good about
using a label in a number of contexts so meaning
could be abstracted
73
Now, as you may have noted, the Gardners take a
hit at Terrace in their article
mainly about experimental design
Because Terrace had published an article in
Science a few years earlier
that had completely trashed them
74
Terrace had taken some of the Gardners video and
done a frame-by-frame analysis
which isnt exactly fair for a fluid language
like ASL
And argued that Washoe really didnt respond to
questions but just repeated her trainers
75
The arguments were unfounded, but led the
Gardners to lose funding and close their lab
Remember that ASL has a different grammar from
English sometimes where in space a sign is
placed is critical
And lots of arguments ensued
76
The Gardners gave their apes to their student,
Roger Fouts, who had space and funding in Oklahoma
And, as expected, continued to collaborate
Even after the apes were moved to Washington State
77
The Gardners were particularly interested in how
the apes used their signs interactively
because that was one of Terraces real
argumentshence the 2000 report
tho not clear when it was performed
78
As you can see from the transcripts, some of the
material was a true dialogue
Other transcripts were a bit surreal
Some of the criticisms, however, would also be
leveled at young human signers
79
Specifically, issues of emphasis and repetition,
or motivation for a particular item
or, in particular, interest in something other
than what the experimenter was targeting
The latter could really mess up their trials,
suggesting the apes werent understanding the
signs
80
Often times the apes would stick to the original
topic despite the experimenter changing the
subject
That seemed particularly common when the ape was
signing about food and the experimenter ignored
what could have been a request
81
But even if the topic was not food, if the
experimenter stuck on topic
the ape was more likely to stick as well
Sometimes expanding the utterancebut often just
copying it
82
However, some of the copies were meaningful for
emphasis
Or, when just one more sign was added, the ape
wasbecause it was ASL, not English
actually requesting something of the experimenter
83
And one has to wonder at what the apes thought
about the non-sequitor type interactions
Did the apes think that the humans were weirdly
off-topic and should be brought around?
We cant immediately fault the animals for their
actions
84
Of course, even when the experimenter was on
topic, it isnt clear as to what is happening
85
Nevertheless, the apes are responding in a
statistically valid manner
IF you compare them with young signing children.
And that was really one of the major issues
86
There were also ASL projects with an orangutan
(Lyn Miles) and a gorilla (Penny Patterson)
but none of their data were ever published in
peer-reviewed journals
which is why we arent examining their material
87
The other major players in the ape language
debate were the Rumbaughs
As mentioned earlier, they first worked with
Lana, lexigrams, and a computer
and Lana had almost no social interaction with
humans
88
The idea was that she could not be cued in any
way by humans
And that her data would be collected by a
computer 24/7 therefore be totally above reproach
Note that she did learn to use the lexigrams in a
standard manner to request food, drinks, movies
89
And their ideas about keeping humans out of the
picture didnt quite work out.
video
90
She even learned to erase a sentence given to
her that didnt make sense.
That is, the machine could give food but made
movies.
and shed detect the incorrect use of a verb
91
So Lana had really, really good production.
And, because in children, comprehension usually
precedes production, Rumbaugh assumed that Lana
comprehended her labels
But that was a real mistake
92
When given a lexigram and asked to choose the
item to which it referred, she was often at chance
Suggesting that she had learned particular
patterns that got her what she wanted
but only in the particular context in which it
had been trained
93
They taught their second apes, Sherman and
Austin, quite differently in order to avoid such
problems
These apes interacted more with each other than
their trainers although they were also taught
with lexigrams
And the Rumbaughs stopped color coding the
lexigrams as cues
94
Sherman and Austin did acquire a lot more
reference, but in their case much of it was in
context
So they understood that a key was something to
open a box, but had some trouble identifying it
as a sole object
simply because that was the way they were trained
95
There are a large number of studies with these
two apes
All of which are fascinating but we dont have
time to go through them.
One issue of interest was that although the
experimenters talked with these apes,
96
the Rumbaughs claimed they didnt understand
English
Now, privately, Boysen disputed this issue, and
Fouts had shown that other apes could associate
ASL signs and English
Given how many animals do understand human
speech, one would expect apes to be among them
97
Interestingly, data show that chinchillas and
even quail actually parse sounds just as humans.
that is, they exhibit categorical perception
Such that sounds like /p/ and /b/ have distinct
cut offs in VOT and dont blend into one another
98
So, if nothing else, chimps are likely to have
the hearing and the neurological underpinnings
to be able to distinguish what we call minimal
pairs
pea/tea, cork/corn
99
About this time, the Rumbaughs decided to try to
work with a pygmy chimp
Under the assumption that the pygmy chimps might
by even more closely related to humans
And thus more likely to acquire language-like
behavior
100
They started with an adult, Matata, who was still
caring for her infant, Kanzi
Despite many, many months of instruction, Matata
didnt get very far in using the lexigrams
But Kanzi, who had been watching most of his
mothers training
101
in a way reminiscent of our using two humans to
train Alex,
Actually began to pick up quite a bit of
understanding about what the symbols meant
video
102
He even did fairly well with complicated
sentences,
Even when novel and somewhat nonsensical, whether
with lexigrams or in English
video
103
Now, of course, Kanzi was far better than Sherman
and Austin
But was it because of his species or his training?
So the Rumbaughs decided to raise a pygmy chimp
and a regular one together so training wouldnt
be an issue
104
Of course, the Rumbaughs could not know if there
were any individual differences in these apes.
Supposedly the apes were chosen with respect to
birth-dates alone
But that doesnt totally exclude such differences
105
And, of course, although they claim that the
animals were treated equivalently,
obviously every trainer knew which ape was which.
as well as the expectations of the Rumbaughs
about the bonobos
106
There are plenty of studies with humans
suggesting that children will live up to
expectations
whether good or bad
Substitute teachers were told that the same class
was either exceptionally bright or slow
And students met both criteria
107
If you look at the transcripts, both apes seem to
be functioning at similar, if not equivalent,
levels
And, of course, we dont know if what was chosen
was truly representative of the sessions
And, yes, Im being quite critical
108
Its very clear from the reported material that
the apes were treated somewhat differently
were the caretakers truly responding to the
levels of the apes?
Differences could also be in how well the
subjects attended
109
And that could be a species OR an individual
difference.
We really cant know with only two subjects
The arguments about differences in frontal lobes
are compelling
particularly with respect to keeping a string of
labels in mind
110
But dolphins do just that, and an ape that makes
a termite poking tool or eating certain nettles
has to have memory for various steps
So Im not sure that I buy into that explanation
111
It would be very interesting to see if the
bonobos and chimps have different MNs
Of interest, too, is that in terms of production,
Both species of ape engaged in very non-ape like
pointing and eye contact
112
which suggests that avoidance of eye contact
might be learned or, obviously, at least
unlearned
And that pointing can be easily acquired in a
fully enculturated ape
And, again, the differences in production could
be species-related or not
113
Did the chimpanzee figure that what she had was
good enough (like my youngest parrot)
so that it wasnt really worth bothering to learn
the additional lexigrams?
Could be, given that she was slightly the junior
114
The really critical issue was that both species
were given what we would call immersive input or
full enculturation
And that their development far outpaced that of
Sherman and Austin who had not had such input
or that of Lana
115
Taken together, these studies suggest that
whatever communication skills apes do achieve
they, like children, need an interactive
environment that includes
reference, functionality, and socialization
116
So, the next question is why most of these
studies have ended
not only did the ape language work force us to
examine the competencies of nonhumans
but it also forced us to examine human language
more closely
117
Because if linguists are insisting that human
language is unique
we have to understand what differentiates it from
that of nonhumans
Not just argue that language is whatever it is
that apes cant achieve
118
The bottom-line issue was that all these studies
were first attempts
And the glare of publicity made sure that any
errors were blown out of proportion
because many humans didnt want to accept that an
ape could indeed communicate with humans
119
We now look at the work with the knowledge of all
the incredible things that animals can indeed do
And maybe wonder at the fuss
But few of the studies weve read had yet been
done, and animals were not considered capable of
much
120
So claiming language-like abilities, even if
quite primitive
was more than most scientists could bear
Just witness the current flap over the starling
grammar at present to get an idea of what was
happening in the 1970s and 1980s!
121
Briefly, Chomsky, Hauser and Fitch have argued
that what makes human language unique is
recursion..
being able to parse and understand embedded
phrases
This is the cat that ate the rat that lived in
the house that Jack built
122
So they exposed tamarins to different types of
grammars
Some heard sets that were ABAB and others heard
sets that were AABB
Something like put book take key versus put
book key pail
123
Note that the tamarins werent given anything
meaningful like verbs and nouns
Just various syllables produced either in male or
female voices
And they werent specifically trained, just
exposed to the sounds
124
Then they tested whether the tamarins could
detect violations of the familiar grammars
Those exposed to ABAB detected when they were
given AABB instead
But those exposed to AABB did not respond to ABAB
as a violation
125
Humans, given the same nonsense syllables did
catch the differences
So the argument was that these orderings were
representative of a recursive grammar
And that a nonhuman primate failed to understand
recursion, and hence the language difference
126
Last week, a paper came out on starlings, who
were trained to recognize either of these
grammars.
And, interestingly, the experimenters used notes
from their learned songs
rattle and warble motifs, not random syllables
127
They didnt learn very quickly, but once they did
learn
It didnt matter what they had learned they
distinguished it from the other grammar
And easily transferred to new collections of
rattles and whistles
128
So, did they succeed because birds need to be
able to recognize vocal patterns
or because they were trained over thousand of
trials and their exemplars were natural?
We wont know til the relevant experiments are
done
129
But that doesnt stop the debate that this paper
has engendered in the literature
And that is not a bad thing
Bringing up all sorts of alternative
possibilities that must be examined serves to
clarify the experimental design
130
Sodo animals have language?
Most have their OWN language which we are mostly
too stupid to decipher
And some species can map what we teach them onto
what they have
To learn some elements of ours