A Report to an Academy, 1980*
H. S. TERRACE
Department of Psychology
Columbia University
New York, New York 10027
T H E FIRST A C C O U N T of an ape who learned to
talk appears to be fictional. In 1917, Franz Kafka wrote a tale about a
chimpanzee who acquired the gift of human language. Recent research
appears to have confirmed Kafka’s sense of what it would take to induce
an ape to speak:
. . . there was no attraction for me in imitating human beings. I imitated them
because I needed a way out, and for no other reason . . . And so I learned
things, gentlemen. Ah, one learns when one needs a way out; one learns at all
costs. [F. Kafka, “A Report to an Academy.7
During the 63 years that have elapsed since the publication of
Kafka’s short story, much has been written about man’s presumably unique capacity to use language and attempts to show that apes can master
some of its features. Linguists, psychologists, psycholinguists,
philosophers, and other students of human language have yet to capture its many complexities in a simple definition. They do agree,
however, about one basic property of all human languages, that is, the
ability to create new meanings, each appropriate to a particular context,
through the application of grammatical rules. Noam Chomskyl and
George Miller? among others, have convincingly reminded us of the
futility of trying to explain a child’s ability to create and understand
sentences without a knowledge of rules that can generate an indeterminately large number of sentences from a finite vocabulary of words.
The dramatic reports of the Gardner~,~
Premack? and Rumbaugh5
that a chimpanzee could learn substantial vocabularies of words of
visual languages and that they were also capable of producing utterances containing two or more words, raise an obvious and fundamental question: Are a chimpanzee’s multi-word utterances grammatical?
In the case of the Gardners, one wants to know whether Washoe’s
The research reported in this article was, in part, funded by grants from the W.T.
Grant Foundation, the Hany Frank Guggenheirn Foundation, and The National Institutes of Health (ROlMH29293). Portions of this article appeared previously in Terrace,
H.S. 1979. How Nim Chirnpsky Changed My Mind. Psychol. Today 1 3 (6):65-76.
94
0077-8923/81/036--0094 $01.75/0 0 1981, NYAS
TERRACE: A REPORT TO AN ACADEMY
95
signing m o ~ de ~ i n kin order to obtain another cup of juice or water bird,
upon seeing a swan, were creative juxtapositions of signs. Likewise one
wants to know whether “Sarah,” Premack‘s main subject, was using a
grammatical rule in arranging her plastic chips in the sequence, Mary
give Sarah apple, and whether “Lana,” the subject of a related study conducted by Rumbaugh, exhibited knowledge of a grammatical rule in
producing the sequence, please machine give apple.
In answering these questions, it is important to remember that a
mere sequence of words does not qualify as a sentence. A rotely learned
string of words presupposes no knowledge of the meanings of each element and certainly no’knowledge of the relationships that exist between the elements. Sarah, for example, showed little, if any, evidence
of understanding the meanings of Mary, give, and Sarah in the sequence, Mary give Sarah apple. Likewise, it is doubtful that, in producing the sequence please machine give apple, Lana understood the meanings of please machine and give, let alone the relationships between
these symbols that would apply in actual sentences.6 There is evidence
that Sarah or Lana could distinguish the symbol apple from symbols
that named other reinforcers. This suggests that what Sarah and Lana
learned was to produce rote sequences of the type ABCX, where A, B,
and C are nonsense symbols and X is a meaningful element. That conclusion is supported by the results of two studies, one an analysis of a
corpus of Lana’s utterances, the other an experiment on serial learning
by pigeons.
Thompson and Church’ have recently shown that a major portion of
a corpus of Lana’s utterances can be accounted for by three decision
rules that dictate when one of six stock sentences might be combined
with one of a small corpus of object or activity names. The decision
rules are (1) did Lana want an ingestible object, (2) was the object in
view, and (3) was the object in the machine. For example, if the object
was in the machine, an appropriate stock sequence was please machine
give; if it was not in the machine an appropriate stock sequence would
be please move object name into machine, and so on.
A recent experiment performed in my laboratory* showed that
pigeons could learn to peck four colors presented simultaneously in a
particular sequence. Such performance is of interest as evidence of the
memorial capacity of pigeons. It does not, of course, justify interpreting
the sequence of the colors (A+B+C+D) as the production of a
sentence meaning please machine give grain.
While the sequences, please machine give g a i n and please machine
give apple, are logically similar, they are not identical. It has yet to be
shown that pigeons can learn ABCX sequences of the type that Sarah
and Lana learned (where X stands for different reinforcers) or that
96
ANNALS NEW YORK ACADEMY OF SCIENCES
pigeons can learn to produce different sequences for different reinforcers. But given the relative ease with which a pigeon can master an
A- B+ C+ D sequence, neither of these problems seem that difficult,
a priori. And even if a pigeon could not perform such sequences or, as
would probably be the case, a pigeon learns them more slowly than a
chimpanzee, we should not lose sight of the fact that learning a rote sequence does not require any ability to use a grammar.
Utterances of apes who were not explicitly trained to produce rote
sequences pose different problems of interpretation. The Gardners
report that Washoe was not required to sign sequences of signs nor was
she differentially reinforced for particular combinations. She nevertheless signed utterances such as more drink and water bird. Before
these and other utterances can be accepted as creative combinations of
signs, combinations that create particular meanings, it is necessary to
rule out simpler interpretations.
The simplest nongrammatical interpretation of such utterances is
that they contain signs that are related solely by context. Upon being
asked what she sees when looking in the direction of a swan it is appropriate for Washoe to sign water and bird. On this view, if Washoe knew
the sign for sky, she might just as readily have signed such less interesting combinations as sky water, bird sky, sky bird water, and so on.
Even if one could rule out context as the only basis of Washoe's combinations, it remains to be shown that utterances such as water bird and
more drink are constructions in which an adjective and a noun are combined so as to create a new meaning. In order to support that interpretation, it is necessary to show that she combined adjectives and nouns in a
particular order. It is, of course, unimportant whether Washoe used the
English order (adjective + noun) or the French order (noun + adjective) in creating combinations in which the meaning of a noun is qualified by an adjective. But it is important to show that all, if not most,
presumed adjectives and that all, if not most, presumed nouns are combined in a consistent manner so as to create particular meanings.
It is, of course, true that sign order is but one of many grammatical
devices used in sign language and that it is less important in sign language than it is in spoken languages such as English. At the same time,
sign order is one of the easiest, if not the easiest, grammatical device of
sign language to record. It also provides a basis for demonstrating an
awareness of such simple constructions as subject-verb, adjectivenoun, verb-object, subject-verb-object, and so on.
With only two minor excepti~ns,~~*O
the Gardners have yet to publish
any data on sign order. Accordingly, the interpretation of combinations
such as water bird and more drink remains ambiguous. One has too little
information to judge whether such utterances are manifestations of a
TERRACE: A REPORT TO AN ACADEMY
97
simple grammatical rule or whether they are merely sequences of contextually related signs.
PROJECT
NIM
In order to distinguish between these interpretations it is necessary to
examine a large body of utterances for regularities of sign order. The initial goal of Project Nim, a project I started in 1973, was to provide a
basis for deciding whether a chimpanzee could use one or more simple
grammatical rules. Our immediate aim was to collect and to analyze a
large corpus of a chimpanzee’s sign combinations for regularities of sign
order. Such regularities could be used as evidence of a chimpanzee’s
ability to use a finite-state grammar. As puny as such an accomplishment might seem from the perspective of a child’s acquisition of
language, it would amount to a quanta1 leap in the linguistic ability of
nonhumans. As we shall see, showing that a chimpanzee can learn a
mere finite-state grammar proved to be an elusive goal.
Socialization and Training
The subject of our study was an infant male chimpanzee, named “Nim
Chimpsky.” Nim was born at the Oklahoma Institute for Primate
Studies in November 1973 and was flown to New York at the age of two
weeks. Until the age of 18 months, he lived in the home of a former student and her family; subsequently he lived in a University-owned mansion in Riverdale where he was looked after by four students.
At the age of 9 months, Nim became the sole student in a small classroom complex I designed for him in the Psychology Department of Columbia University. The classroom allowed Nim’s teachers to focus his attention more easily than they could at home and it also provided good
opportunities to introduce Nim to many activities conducive to signing,
such as looking at pictures, drawing, and sorting objects. Another important feature of the classroom was the opportunity it provided for
observing, filming, and videotaping Nim without his being aware of the
presence of visitors and observers who watched him through a one-way
window or through cameras mounted in the wall of the classroom.
Nim’s teachers kept careful records of what he signed both at home
and in the classroom. During each session, the teacher dictated into a
cassette recorder as much information as possible about Nim’s signing
and the context of that signing. Nim was also videotaped at home and in
the classroom. A painstaking comparison of Nim’s signing at home and
in his classroom revealed no differences with respect to spontaneity,
content, or any other feature of his signing that we examined. In view of
comments attributed to other researchers of ape language,” that Nim
98
ANNALS NEW YORK ACADEMY OF SCIENCES
was conditioned in his nursery school like a rat or a pigeon in a Skinner
box, it should be emphasized that his teachers were just as playful and
spontaneous in the Columbia classroom as they were at home. (A detailed
description of Nim’s socialization and instruction in sign language can
be found in Nim.12)
Nim’s teachers communicated to him and amongst themselves in
sign language. Although the signs Nim’s teachers used were consistent
with those of American Sign Language (ASL or Ameslan), their signing
is best characterized as “pidgin” sign language. This state of affairs is to
be expected when a native speaker of English learns sign language. Inevitably, the word order of English superimposes itself on the teacher’s
signing, at the expense of the many spatial grammatical devices ASL
employs. An ideal project would, of course, attempt to use only native
or highly fluent signers as teachers, teachers who would communicate
exclusively in Ameslan. There is, however, little reason to be concerned
that this ideal has not been realized on Project Nim, or for that matter,
on any of the other projects that have attempted to teach sign language
to apes. The achievements of an ape who truly learned pidgin sign
language would be no less impressive than those of an ape who learned
pure Ameslan. Both pidgin sign language and Ameslan are grammatically structured languages.
Nim was taught to sign by the methods developed by the Gardners3
and Fouts:13 molding and imitation. During the 44 months he was in
New York he learned 125 signs, most of which were common and
proper nouns; next frequent were verbs and adjectives; least frequent
were pronouns and prepositions.12
Combinations of Two or More Signs
During a 2-year period, Nim’s teachers recorded more than 20,000 of his
utterances that consisted of 2 or more signs. Almost half of these utterances were 2-sign combinations, of which 1,378 were distinct. One
characteristic of Nim’s 2-sign combinations led me to believe that they
were primitive sentences. In many cases Nim used particular signs in
either the first or the second position, no matter what other sign that
sign was combined with.12.14 For example, more occurred in the first
position in 85% of the 2-sign utterances in which more appeared (such
as more banana, more drink, more hug, and more tickle). Of the 348
2-sign combinations containing give, 78% had give in the first position.
Of the 946 instances in which a transitive verb (such as hug, tickle, and
give) was combined with me and Nim,83% of them had the transitive
verb in the first position.
These and other regularities in Nim’s two-sign utterances12 are the
first demonstrations I know of a reliable use of sign order by a chimpan-
TERRACE: A REPORT TO AN ACADEMY
99
zee. By themselves, however, they do not justify the conclusion that
they were created according to grammatical rules. Nim could have simply imitated what his teachers were signing. That explanation seemed
doubtful for a number of reasons. Nim’s teachers had no reason to sign
many of the combinations Nim had produced. Nim asked to be tickled
long before he showed any interest in tickling; thus, there was no reason
for the teacher to sign tickle me to Nim. Likewise, Nim requested various
objects by signing give + X (X being whatever he wanted) long before
he began to offer objects to his teachers. More generally, all of Nim’s
teachers and many experts on child language learning, some of whom
knew sign language, had the clear impression that Nim’s utterances
typically contained signs that were not imitative of the teacher’s signs.
Another explanation of the regularities of Nim’s two-sign combinations that did not require the postulation of grammatical competence
was statistical. However, an extensive analysis of the regularities observed
in Nim’s two-sign combinations showed that they did not result from
Nim’s preferences for using particular signs in the first or second positions of two-sign combinations.15 Finally, the sheer variety and number
of Nim’s combinations make implausible the hypothesis that he somehow memorized them.
The analyses performed on Nim’s combinations provided the most
compelling evidence I know of that a chimpanzee could use grammatical rules, albeit finite-state rules, for generating two-sign sequences. It
was not until after our funds ran out and it became necessary to return
Nim to the Oklahoma Institute for Primate Studies that I became skeptical of that conclusion. Ironically, it was our newly found freedom from
data-collecting, teaching, and looking after Nim that allowed me and
other members of the project to examine Nim’s use of sign language
more thoroughly.
Differences Between Nim’s and Q Child’s Combinations of Words
What emerged from our new analyses was a number of important differences between Nim’s and a child’s use of language. One of the first facts
that troubled me was the absence of any increase in the length of Nim’s
utterances. During the last two years that Nim was in New York, the
average length of Nim’s utterances fluctuated between 1.1 and 1.6
signs. That performance is like what children do when they begin combining words. Furthermore, the maximum length of a child’s utterances
is related very reliably to their average length.16 Nim’s showed no such
relationship.
As children get older, the average length of their utterances increases
1, this is true both of children with normal
steadily. As shown in FIGURE
hearing and of deaf children who sign.14 After learning to make ut-
100
ANNALS NEW YORK ACADEMY OF SCIENCES
CHILDREN
DEAF
0--0 'RUTH'(H SCHLESlNsER
mdalei)
HEARING
o
A
48
-
46
-
44
-
42
-
40
-
38
-
A .EVE* (bOIWJ973)
A 'SARAH'(BROIW.1973)
.---.
'FW~'(KLIMA
+--+
"nLlCE'(HOFNEISTER.1972)
a mwi,
19721
NIM
0
I
CUSSROW SESSDNS
Hew KSSKllS
0-0
x---x
I
0- 4 MOEOTAPE Y Y R E S
I
D
I
A
A
A
6 34 II3 32
u
w
z
a
36
I
0
20
-
18
-
o
A A
12
lo
-
m a
A
- + - - - -+/
-.
2;
4:
-
k
A
iB
32
l4
1
*
/
k*epqfx/
v?
-\, '
:6
/
/
Y
/
Am'
'0-
2b
+,/
/A
4
1;' ' ' '
r
4
+-
,+/'
,/
;
;
+
,
/
'0
/
1614
/
i
I
4
-
/
/
6
22
I
A
A
J
w
I
4
2 24-
a
I
I
A
8 30 28 a
2
W 26 5
I
A
4-A
318
0
4
,n
0
-6 Lfl,
O:
;2;"9
i2
i 4
416
i8
,b
i2
AGE IN MONTHS
terances relating a verb and an object, as, for example, eats breakfast,
and utterances relating a subject and a verb, as, for example, Daddy eats,
the child learns to link them into longer utterances relating the subject,
verb, and object, for example, Daddy eats breakfast. Later, the child
TERRACE:
101
A REPORT TO AN ACADEMY
TABLE
1
MOSTFREQUENT
2- AND 3-SIGNCOMBINATIONS
2-Sign Combinations
play me
me Nim
tickle me
eat Nim
more eat
me eat
Nim eat
finish hug
drink Nim
more tickle
sorry hug
tickle Nim
hug Nim
more drink
eat drink
banana me
Nim me
sweet Nim
me play
gum eat
tea drink
grape eat
hug me
banana Nim
in pants
Frequency
375
328
316
302
287
237
209
187
143
136
123
107
106
99
98
97
89
85
81
79
77
74
74
73
70
3-Sign Combinations
play me Nim
eat me Nim
eat Nim eat
tickle me Nim
grape eat Nim
banana Nim eat
Nim me eat
banana eat Nim
eat me eat
me Nim eat
hug me Nim
yogurt Nirn eat
me more eat
more eat Nim
finish hug Nim
banana me eat
Nim eat Nim
tickle me tickle
apple me eat
eat Nirn me
give me eat
nut Nim nut
drink me Nim
hug Nim hug
play me play
sweet Nim sweet
Frequency
81
48
46
44
37
33
27
26
22
21
20
20
19
19
18
17
17
17
15
15
15
15
14
14
14
14
learns to link them into longer utterances such as Daddy didn‘t eat
breakfast, or when will Daddy eat breakfast?
Despite the steady increase in the size of Nim’s vocabulary, the mean
length of his utterances did not increase. Although some of his utterances were very long, they were not very informative. Consider, for example, his longest utterance, which contained 16 signs: give orange me
give eat orange me eat orange give me eat orange give me you. The same
kinds of run-on sequences can be seen in comparing Nim’s 2-, 3-and
4-sign combinations. As shown in TABLE
1, the topic of Nim’s %sign
combinations overlapped considerably with the apparent topic of his
2-sign combinations (compare TABLE
2).
Of Nim’s 25 most frequent 2-sign combinations, 18 can be seen in his
25 most frequent 3-sign combinations, in virtually the same order in
which they appear in his 2-sign combinations. Furthermore, if one ignores sign order, all but 5 signs that appear in Nim’s 25 most frequent
two-sign combinations gum, tea, sorry, in, and pants appear in his 25
102
ANNALS NEW YORK ACADEMY OF SCIENCES
TABLE
2
MOSTFREQUENT
4-SIGN COMBINATIONS
4-Sign Combinations
eat drink eat drink
eat Nim eat Nim
banana Nim banana Nim
drink Nirn drink Nim
banana eat me Nim
banana me eat banana
banana me Nim me
grape eat Nim eat
Nim eat Nim eat
play me Nim play
drink eat drink eat
Frequency
15
7
5
5
4
4
4
4
4
4
4-Sign Combinations
drink eat me Nim
eat grape eat Nim
eat me Nim drink
grape eat me Nim
me eat drink more
me eat me eat
me gum me gum
me Nim eat me
Nim me Nirn me
tickle me Nim play
Frequency
3
3
3
3
3
3
3
3
3
3
3
most frequent three-sign combinations. We did not have enough contextual information to perform a semantic analysis of Nim’s 2- and
3-sign combinations. However, Nim’s teachers’ reports indicate that the
individual signs of his combinations were appropriate to their context
and that equivalent 2- and 3-sign combinations occurred in the same
context.
Though lexically similar to 2-sign combinations, the 3-sign combinations do not appear to be informative elaborations of 2-sign combinations. Consider, for example, Nim’s most frequent 2- and 3-sign combinations: play me and play me Nim. Combining Nim with play me to
produce the 3-sign combination, play me Nim, adds a redundant proper
noun to a personal pronoun. Repetition is another characteristic of
Nim’s 3-sign combinations, for example, eat Nim eat, and nut Nim nut.
In producing a 3-sign combination, it appears as if Nim is adding to
what he might sign in a 2-sign combination, not so much to add new information, but instead to add emphasis. Nim’s most frequent 4-sign
combinations reveal a similar picture. In children’s utterances, in contrast, the repetition of a word, or a sequence of words, is a rare event.
Now that we have seen what Nim signed about in 2-, 3-, and 4-sign
combinations, it is instructive to see what he signed about with single
3, the topics of Nim’s most frequent 25 singlesigns. As shown in TABLE
sign utterances overlap considerably with those of his most most frequent multisign utterances. (Most of the exceptions are signs required
in certain routines, e.g., finish, when Nim was finished using the toilet;
sorry, when Nim was scolded, and so on.) In contrast to a child, whose
longer utterances are semantically and syntactically more complex than
his shorter utterances, Nim’s are not. When signing a combination, as
opposed to signing a single sign, Nim appears to be running on with his
TERRACE: A REPORT TO AN ACADEMY
103
hands. It appears that Nim learned that the more he signs the better his
chances for obtaining what he wants. It also appeared as if Nim made
no effort to add informative, as opposed to redundant, signs in satisfying his teacher’s demand that he sign.
The most dramatic difference between Nim’s and a child’s use of
language was revealed in a painstaking analysis of videotapes of Nim’s
and his teacher’s signing. These tapes revealed much about the nature
of Nim’s signing that could not be seen with the naked eye. Indeed they
were so rich in information that it took as much as one hour to
transcribe a single minute of tape.
Terrace et a l l 4 showed that Nim’s signing with his teachers bore only
a superficial resemblance to a child’s conversations with his or her
parents. What is more, only 12% of Nim’s utterances were spontaneous.
That is, only 12% were not preceded by a teacher’s utterance. A
significantly larger proportion of a child’s utterances is spontaneous.
In addition to differences in spontaneity, there were differences in
TABLE
3
25 MOSTFREQUENT
SIGNS*
Sign
Tokens
hug
play
finish
eat
dirty
drink
out
Nirn
open
tickle
bite
shoe
pants
red
sorry
angry
me
banana
nut
down
toothbrush
hand cream
more
grape
sweet
1.650
1,545
1,103
95 1
788
712
615
613
554
414
407
40 5
372
380
366
354
351
348
328
316
302
30 1
30 1
239
236
*July 5, 1976-February 7, 1977.
104
ANNALS NEW YORK ACADEMY OF SCIENCES
creativity. As a child gets older, the proportion of utterances that are full
or partial imitations of the parent’s prior utterance(s)decreases from less
than 20% at 21 months to almost zero by the time the child is 3 years
0ld.1~When Nim was 26 months old, 38Yo of his utterances were full or
partial imitations of his teacher’s. By the time he was 44 months old, the
proportion had risen to 54%. A doctoral dissertation by Richard
Sanders18 showed that the function of Nim’s imitative utterances differed from those of a child’s imitative utterances. Bloom and her
associates19 have observed that children imitated their parents’s utterances mainly when they were learning new words or new syntactic
structures. Sanders found no evidence for either of these imitative functions in Nim’s imitative utterances.
As children imitate fewer of their parents’ utterances, they begin to
expand upon what they hear their parents say. At 21 months, 22% of a
child’s utterances add at least one word to the parent’s prior utterance;
at 36 months, 42% are expansions of the parent’s prior utterance.
Fewer than 10% of Nim’s utterances recorded during 22 months of
videotaping (the last 22 months of the project) were expansions. Like
the mean length of his utterances, this value remained fairly constant.
The videotapes showed another distinctive feature of Nim’s conversations that we could not see with the naked eye. He was as likely to interrupt his teacher’s signing as not. In contrast, children interrupt their
parents so rarely (so long as no other speakers are present) that interruptions are all but ignored in studies of their language development. A
child learns readily what one takes for granted in a two-way conversation: each speaker adds information to the preceding utterance and
each speaker takes a turn in holding the floor. Nim rarely added information and showed no evidence of turn-taking.
None of the features of Nim’s discourse - his lack of spontaneity, his
partial imitation of his teacher’s signing, his tendency to interrupt - had
been noticed by any of his teachers or by the many expert observers
who had watched Nim sign. Once I was sure that Nim was not imitating
precisely what his teacher had just signed, I felt that it was less important to record the teachers’ signs than it was to capture as much as I
could about Nim’s signing: the context and specific physical movements, what hand he signed with, the order of his signs, and their appropriateness. But even if one wanted to record the teacher’s signs,
limitations of attention span would make it too difficult to remember all
of the significant features of both the teacher’s and Nim’s signs.
Once he knew what to look for, the contribution of the teacher was
easy to see, embarrasingly enough, in still photographs that we had
looked at for years. Consider, for example, Nim’s signing the sequence
TERRACE: A REPORT TO AN ACADEMY
105
me hug cut as shown in FIGURE
2. At first, these photographs (and many
others) seemed to provide clear examples of spontaneous and meaningful combinations. But just as analyses of our videotapes provided
evidence of a relationship between Nim’s signing and the teacher’s prior
2 revealed a previously unnoticed
signing, a re-examination of FIGURE
contribution of the teacher’s signing. She signed you while Nim was
signing me and signed who while he was signing cut. While Nim was
signing hug, his teacher held her hand in the “n”-hand configuration, a
prompt for the sign Nim. Because these were the only photographs
taken of this sequence, we cannot specify just when the teacher began
her signs. It is not clear, for example, whether the teacher signed you
simultaneously or immediately prior to Nim’s me. It is, however, unlikely that the teacher signed who? after Nim signed cut. A few moments
before these photographs were taken the teacher repeatedly quizzed
Nim as to the contents of the cat box by signing who? At the very least,
Nim’s sequence, me hug cut, cannot be interpreted as a spontaneous
combination of three signs.
FIGURE
2. Nim signing the linear combination, me hug cut, to his teacher
(Susan Quinby). (Photographed in classroom by H.S. Terrace.)
106
ANNALS NEW YORK ACADEMY OF SCIENCES
There remain other differences between Nim’s and a child’s signing
that need to be explored. One important question is how similar are the
contents of a chimpanzee’s and a child’s utterances, whether or not they
are structured grammatically? A cursory examination of corpora of children’s utterances20 indicates that chimpanzees and children differ
markedly with respect to the variety of the utterances they make. While
children produce certain routine combinations, they are relatively infrequent. Nim’s utterances, on the other hand, showed a high frequency of
routine combinations (e.g., play me, me Nirn; see TABLE
1 for additional
examples).
It also remains to be shown that the patterns of vocal discourse
observed between a hearing parent and a hearing child are similar to the
patterns of signing discourse that obtain between a deaf parent and a
deaf child. Videotapes of such discourse are now available from a
number of s0urces.2~-~~
CRITICISMS
OF PROJECT
NIM
Of more immediate concern is the generality of the conclusions drawn
about Nim’s signing. This issue can be approached in two ways. One is
to consider the methodological weaknesses of Project Nim and to pursue their implications. The other is to recognize that, methodological
weaknesses notwithstanding, it is reasonable to ask if all signing apes
sign because they are coaxed to do so by the teachers and how much
sign-for-signoverlap exists between the teacher’s and the ape’s signing?
Consider first some of the questions raised about Project Nim. It has
been said that Nim was taught by too many teachers (60 in all), that his
teachers were not fluent enough in ASL, that terminating Nim’s training at the age of 44 months prevented his teachers from developing
Nim’s full linguistic competence, and that Nim may simply have been a
dumb chimp (e.g., quotes of other researchers studying language in apes
as reported by Bazar”).
Aside from the speculation that Nim may have been a dumb chimp, I
believe that all of these criticisms are valid. And, if questioning Nim’s intelligence is simply a nasty way of asking whether an N of 1 is an adequate basis for forming a general conclusion about an ape’s grammatical
competence, I would readily admit that it is not. A case can be made,
however, that most of the methodological inadequacies of Project Nim
have been exaggerated and, in any event, that they are hardly unique to
Project Nim. Though Nim was taught by 60 teachers, he spent most of
his time in the presence of a core group of 8 teachers: (Stephanie
LaFarge, Laura Petitto, Amy Schachter, Walter Benesch, Bill Tynan,
Joyce Butler, Dick Sanders, and myself). As described in Nim (see
TERRACE: A REPORT TO AN ACADEMY
107
especially Appendix B),12 many of Nim’s 60 teachers served as occasional playmates rather than as regular teachers. Nevertheless, they
were each listed as a teacher. The Gardners have yet to publish a full list
of the teachers who worked with Washoe. Roger Fouts, however, estimates that their number was approximately 40. Of greater importance
is Fouts’s observation that, during Washoe’s 4 years in Reno, Nevada,
she was looked after mainly by a small group of 6 core teachers. Francine Patterson’s doctoral dissertation presents data provided by 20
teachers during the 3 year period in which they worked with “Koko.”
This is hardly surprising. It was difficult for an ape’s teachers to sustain
the energy needed to carry out lesson plans, to engage its attention, to
stimulate it to sign, and to record what it signed for more than 3 to 4
hours a day. A 3- to 4-hour session with Nim also entailed an additional 1
or 2 hours needed to transcribe the audio casette on which the teacher
dictated information about Nim’s signing and to write a report. At least
6 full-time people would be needed to carry out such a schedule on a
16-hour/day, 7-day/week basis. As far as I know, no project has been
able to afford the salaries of such a staff. Accordingly, it is necessary to
make do with a large contingent of part-time volunteers.
Both Patterson and Fouts speak English while signing with their
apes. Films of the Gardners and Fouts signing with Washoe, of Fouts
signing with “Ally” and “Booee” (two resident chimpanzees of the Oklahomas Center for Primate Research), and of Francine Patterson signing
with Koko make clear that none of these researchers use Ameslan. As
mentioned earlier, pidgin sign seems to be the prevalent form of communication on all projects attempting to teach apes to use sign
language.
Washoe is now 15, Koko is 9, and Ally is 9 years old. I know of no evidence that their linguistic skills increased as they became older. An
ape’s intelligence undoubtedly increases after infancy. One must, however, also keep in mind that as an ape gets older, its ability to master its
environment by physical means also increases. An ape’s increasing
strength and its recognition that it can get its way without signing
should result in less motivation to sign and a reduction in its teacher’s
dominance. I am therefore skeptical of conjectures that an ape’s increasing intelligence would manifest itself in a more sophisticated use of
language.
THEGENEFWLITY
OF THE CONCLUSIONS
OF PROJECT
NIM
Whatever the shortcomings of Project Nim, it should be recognized that
they are irrelevant to the following hypothesis about an ape’s use of
signs. An ape signs mainly in response to his teachers’ urgings, in order
108
ANNALS NEW YORK ACADEMY OF SCIENCES
to obtain certain objects or activities. Combinations of signs are not used
creatively to generate particular meanings. Instead, they are used for
emphasis or in response to the teacher’s unwitting demands that the
ape produce as many contextually relevant signs as possible. The validity of this hypothesis rests simply on the conformity of data obtained
from other signing apes.
Finding such data has proven difficult, particularly because discourse
analyses of other signing apes have yet to be published. Also, as mentioned earlier, published accounts of an ape’s combinations of signs
have centered around anecdotes and not around exhaustive listings of
all combinations. One can, however, obtain some insight into the
nature of signing by other apes by looking at films and videotape
transcripts of their signing. Two films made by the Gardners of
Washoe’s signing, a doctoral dissertation by Lyn Miles24(which contains
four videotape transcripts of the two Oklahoma chimps, Ally and
Booee), and a recently released film, “Koko, the Talking Gorilla,” all support the hypothesis that the teacher’s coaxing and cuing have played
much greater roles in so-called “conversations” with chimpanzees than
was previously recognized.
In “The First Signs of Washoe,” a film produced for the “Nova”
television series, Beatrice Gardner can be seen signing What time now?,
an utterance that Washoe interrupts to sign, time eat, time eat. A longer
version of the same exchange shown in the second film, “Teaching Sign
Language to the Chimpanzee Washoe,” began with Gardner signing eat
me, more me, after which Washoe gave her something to eat. Then she
signed thank you -and asked what time now? Washoe’s response time
eat, time eat can hardly be considered spontaneous, since Gardner had
just used the same signs and Washoe was offering a direct answer to her
question.
The potential for misinterpreting an ape’s signing because of inadequate reporting is made plain by another example in both films. Washoe
is conversing with her teacher, Susan Nichols, who shows the chimp a
tiny doll in a cup. Nichols points to the cup and signs that; Washoe signs
baby. Nichols brings the cup and doll closer to Washoe, allowing her to
touch them, slowly pulls them away, and then signs that while pointing
to the cup. Washoe signs in and looks away. Nichols brings the cup with
the doll closer to Washoe again, who looks at the two objects once more,
and signs baby. Then, as she brings the cup still closer, Washoe signs in.
That, signs Nichols, and points to the cup; my drink, signs Washoe.
Given these facts, there is no basis to refer to Washoe’s utterancebaby in baby in my drink-as either a spontaneous or a creative use of
“in” as a preposition joining two objects. It is actually a “run on” sequence with very little relationship between its parts. Only the last two
109
TERRACE: A REPORT T O AN ACADEMY
signs were uttered without prompting from the teacher. Moreover, the
sequence of the prompts (pointing to the doll, and then pointing to the
cup) follows the order called for in contructing an English prepositional
phrase. In short, discourse analysis makes Washoe’s linguistic achievement less remarkable than it might seem at first.
In commenting about the fact that Koko’s mean length of utterances
(MLU) was low in comparison with that of both hearing and deaf children, Francine Patterson speculates, in her doctoral dissertation, that,
“This probably reflects a species difference in syntactic and/or sequential processing abilities” (Reference 2 5 , page 153). Patterson goes on to
observe that, “The majority of Koko’s utterances were not spontaneous,
but elicited by questions from her teachers and companions. My interactions with Koko were often characterized by frequent questions
such as ‘What’s this?’ ” (Reference 25, page 153).
Four transcripts appended to Lyn Miles’ dissertation provided me
with a basis for performing a discourse analysis of the signing of two
other chimpanzees, Ally and Boee. Each transcript presents an exhaustive account of one of these chimps signing with one of two
trainers: Rogert Fouts and Joe Couch. The MLU and summaries of the
discourse analysis of each tape is shown in TABLE
4. TABLES
5 and 6
show exhaustive summaries of two conversations, one from a session
with the hghest MLU and one from a session with the highest percentage of adjacent utterances that were novel. The novel utterances are
very similar to Nim’s run-on sequences. They also overlap considerably
with adjacent utterances that were expansions and with noncontingent
utterances.
In his discussion of communicating with an animal, the philosopher
Ludwig Wittgenstein cautions that apparent instances of an animal using human language may prove to be a “game” that is played by simpler
rules. Nim’s, Washoe’s, Ally’s, Booee’s, and Koko’s use of signs suggests a
TABLE
4
SUMMARYOF ALLYAND BOOEETRANSCRIPTS
Video Taoe Number
Number of utterances
MLU
% Adjacent
% Imitations:
Yo Expansions
% Novel
YO Noncontingent
3
4
5
6
Mean
38
1.63
76.3
13.6
7.89
55.3
23.7
79
1.52
93.7
22.8
7.59
63.3
6.3
102
2.25
77.4
7.84
13.7
55.9
2.6
72
1.93
86.1
8.33
4.16
73.6
3.9
72.75
1.85
83.4
13.03
8.34
62.3
16.6
110
ANNALS NEW YORK ACADEMY OF SCIENCES
TABLE
5
CONVERSATION
NUMBER
4: ROGERAND ALLY
c
Adjacent Utterances (N = 7 4 = 67, ? = 7)’
Novel (N = 50: C = 43, ? = 7)
14
Roger
9
Roger tickle Ally
3
Roger tickle
1
tickle Roger
1
tickle
1
Roger Tickle Ally hurry
2
George
7
Joe
1
string
Expansions (N = 6: C = 6)
1
George smell Roger
I
tickle hurry
1
Roger tickle
Imitations (N = 18: C = 18)
5
Roger
2
tickle
1
Roger tickle Ally
2
shoe tickle
1
shoe tickle
Noncontingent utterances (h’ = 5)
2
Joe
1
you tickle
that
that that box
that shoe
Roger string George
comb
good
food-eat
Ally
2
Roger tickle Ally
Roger tickle Roger comb
Roger tickle Roger
1
1
1
baby
pillow
comb
pull
1
1
2
shoe
pillow
1
1
1
1
1
1
1
2
2
2
* C denotes the number of utterances that followed teachers’ commands; “?”denotes
the number that followed teachers’ questions.
type of interaction between an ape and its trainer that has little to do
with human language. In each instance the sole function of the ape’s
signing appears to be to request various rewards that can be obtained
only by signing. Little, if any, evidence is available that an ape signs in
order to exchange information with its trainer, as opposed to simply
demanding some object or activity.
In a typical exchange the teacher first tries to interest the ape in
some object or activity such as looking at a picture book, drawing, or
playing catch. Typically, the ape tries to engage in such activities
without signing. The teacher then tries to initiate signing by asking
questions such as what that?, what you want?, who’s book?, and ball red
or blue?
The more rapidly the ape signs, the more rapidly it can obtain what it
wants. It is therefore not surprising that the ape frequently interrrupts
the teacher. From the ape’s point of view, the teacher’s signs provide an
excellent model of the signs it is expected to make. By simply imitating
TERRACE: A REPORT TO AN ACADEMY
111
TABLE
6
CONVERSATION
NUMBER
6: JOE AND BOOEE
Adiacent utterances IN = 79: C = 31. ? = 47).
Novel (N = 57: C = 21, ? = ?5)
food-eat Booee that
3
food-eat Booee
3
food-eat
3
1
you-food-eat Booee that
Booee food-eat more Booee
1
food-eat Booee hungry
1
1
food-eat me fruit Booee
1
food-eat fruit Booee
1
that food-eat Booee
1
fruit food-eat you
1
food-eat fruit hurry
1
gimrne food-eat
1
girnrne
1
fruit more Booee
girnme fruit hurry
1
1
there more fruit
4
hurry
1
fruit
that Booee over there
1
that
4
Expansions (N = 14: C = 4, ? = 10)
hurry tickle Booee hurry Booee hurry
1
1
Booee you Booee hurry girnrne
1
tickle Booee girnme
1
me tickle
I
me tickle hurry
1
food-eat Booee that
Imitation (N = 8: C = 6, ? = 2)
4
that
2
YOU
Noncontingent Utterances (N
tickle Booee
tickle
more Booee
hurry that tickle
tickle gimrne
food-eat
food-eat Booee
=
more that Booee
that Booee
more Booee
Booee
you Booee
more
1
1
1
2
Booee hurry
there you girnrne
you Booee you Booee hurry
more Booee hurry
tickle Booee
tickle you me
you there
hurry tickle Booee hurry
you there that there
that Booee there Booee
baby that
more baby
1
1
3
1
1
1
1
3
1
1
1
1
1
1
1
that Booee
that more baby
there that there
that tickle
there you
1
1
1
1
1
tickle
baby
1
more fruit there
that
that that Booee gimme
gimrne
Booee
hurry gimme hurry
that there
1
1
1
YOU
1
23)
6
2
I
1
1
1
1
2
1
1
1
* C denotes utterances that followed teachers’ commands; ”?”denotes utterances that
followed teachers’ questions.
a few of them, often in the same order used by the teacher, and by adding a few “wild cards”-general purpose signs such as give, me, Nim, or
more- the ape can produce utterances that appear to follow grammatical rules. What seems like conversation from a human point of view is
actually an attempt to communicate a demand (in a nonconversational
manner) as quickly as possible.
112
ANNALS NEW YORK ACADEMY OF SCIENCES
FUTURE
RESEARCH
It might be argued that signing apes have the potential to create sentences but did not do so because of motivational rather than intellectual
limitations. Perhaps Nim and Washoe would have been more motivated
to communicate in sign language if they had been raised by smaller and
more consistent groups of teachers, thus sparing them emotional upheavals. It is, of course, possible that a new project, administered by a
permanent group of teachers who are fluent in sign language and have
the skills necessary for such experiments, would prove successful in getting apes to create sentences.
It is equally important for any new project to pay greater attention to
the function of the signs than to mastery of syntax. In the rush to
demonstrate grammatical competence in the ape, many projects (Project Nim included) overlooked functions of individual signs other than
their demand function. Of greater significance, from a human point of
view, are the abilities to use a word simply to communicate information
and to refer to things which are not present. One would like to see, for
example, to what extent an ape is content to sign flower simply to draw
the teacher’s attention to a flower with no expectation that the teacher
would give it a flower. In addition one would want to see whether an
ape could exchange information about objects that are not in view in
order to exchange information about those objects. For example, could
an ape respond, in a nonrote manner, to a question such as, What color
is the banana? by signing yellow or to a question such as Who did you
chase before? by signing cat. Until it is possible to teach an ape that signs
can convey information other than mere demands it is not clear why an
ape would learn a grammatical rule. To put the question more simply,
why should an ape be interested in learning rules about relationships
between signs when it can express all it cares to express through individual signs?
The personnel of a new project would have to be on guard against
the subtle and complex imitation that was demonstrated in Project
Nim. In view of the discoveries about the nature of Nim’s signing that
were made through videotape analyses, it is essential for any new proiect to maintain a permanent and unedited visual record of the ape’s
discourse with its teachers. Indeed, the absence of such documentation
would make it impossible to substantiate any claims concerning the
spontaneity and novelty of an ape’s signing.
Requiring proof that an ape is not just mirroring the signs of its
teachers is not unreasonable; indeed, it is essential for any researcher
who seeks to determine, once and for all, whether apes can use language
in a human manner. Nor is it unreasonable to expect that in any such
TERRACE: A REPORT T O AN ACADEMY
113
experiment, ape “language” must be measured against a child’s sophisticated ability. That ability still stands as an important definition of the
human species.
While writing “A Report to an Academy,” Kafka obviously had no
way of anticipating the numerous attempts to teach real apes to talk
that took place in this country and in the U.S.S.R.3-13.25-28
Just the
same, his view that an ape will imitate for “a way out” seems remarkably
telling. If one substitutes for the phrase, “a way out,” rewarding activities such as being tickled, chased, hugged, and access to a pet cat,
books, drawing materials, and items of food and drink, the basis of
Nim’s, Washoe’s, Koko’s, and other apes’ signing seems adequately explained. Much as I would have preferred otherwise, a chimpanzee’s
“Report to an Academy” remains a work of fiction.
REFERENCES
1. CHOMSKY,
N. 1957. Syntactic Structures. Mouton, The Hague.
2. MILLER,
G. A. 1964. The psycholinguists. Encounter 23 (1): 29-37.
3. GARDNER,
B.T. dr R.A. GARDNER.
1969. Teaching sign language to a chimpanzee. Science 162: 664-672.
4. PREMACK,
D. 1970. A functional analysis of language. J. Exp. Anal. Behav. 4:
107- 125.
5. RUMBAUCH,
D.M., T.V. GILL6r E.C. VON GLASERSFELD.
1973. Reading and
sentence completion by a chimpanzee. Science 182: 731-733.
6. TERRACE,
H.S. 1979. Is problem-solving language? J. Exp. Anal. Behav. 31:
161-175.
7. THOMPSON,
C.R. dr R.M. CHURCH.
1980. An explanation of the language of a
chimpanzee. Science 208: 31 3-314.
8. STRAUB,
R.O., M.S. SEIDENBERG,
T.G. BEVERdr H.S. TERRACE.
1979. Serial
learning in the pigeon. J. Exp. Anal. Behav. 32: 137-148.
9. GARDNER,
B.T. 6r R.A. GARDNER.
1974. Comparing the Early Utterances of
Child and Chimpanzee. pp. 3-23. University of Minnesota Press, Minneapolis, Minn.
10. GARDNER,
B.T. dr R.A. GARDNER.
1974. Teaching sign language to a chimpanzee, VII: Use of order in sign combinations. Bull. Psychonomic SOC.4:
264-267.
11. BAZAR,
J. 1974. Catching up with the ape language debate. Am. Psychol.
Assoc. Monitor 11: 4-5, 47.
12. TERRACE,
H.S. 1979. Nirn. A. Knopf, New York, N.Y.
13. FOUTS,R.F. 1972. Use of guidance in teaching sign language to a chimpanzee. J. Comp. Physiol. Psychol. 80: 515-522.
14. TERRACE,
H.S., L.A. PETITTO,R. J. SANDERS
6r T.C. BEVER.
1979. Can an ape
create a sentence? Science 206: 891-902.
15. TERRACE,
H.S., L.A. PETITTO,R.J. SANDERS
6r T.G. BEVER.1980. O n the
Grammatical Capacity of Apes. Gardner Press, New York, N.Y.
114
ANNALS NEW YORK ACADEMY OF SCIENCES
16. BROWN,R. 1973. A First Language: The Early Stage. Harvard University
Press, Cambridge, Mass.
17. BLOOM,L.M., L. ROCISSANO
6r L. HOOD. 1976. Adult-child discourse:
Developmental interaction between information processing and linguistic
knowledge. Cogn. Psychol. 8: 521-552.
18. SANDERS,
R.J. 1980. The Influence of Verbal and Nonverbal Context on the
Sign Language Conversations of a Chimpanzee. Ph.D. dissertation, Columbia University.
19. BLOOM,L., L. HOODdr P. LICHTBOWN,
1974. Imitation in language development: If, when, and why. Cogn. Psychol. 6: 380-420.
20. BLOOM,L. 1970. Language development: Form and function in emerging
grammors. M.I.T. Press, Cambridge, Mass.
21. BELLUCI,
U. dr E.S. KLIMA.1976. The Signs of Language. Harvard University
Press, Cambridge, Mass.
22. HOFFMEISTER,
R.J. 1978. The Development of Demonstrative Pronouns,
Locatives and Personal Pronouns in the Acquisition of American Sign
Language by Deaf Children of Deaf Parents. Ph.D. dissertation, University
of Minnesota.
23. MCINTIRE,
M.L. 1978. Learning to take your turn in ASL. Working paper.
Department of Linguistics, UCLA.
24. MILES,H.L. 1978. Conversations with Apes: The Use of Sign Language by
Two Chimpanzees. Ph.D. dissertation, University of Connecticut.
25. P A ~ E R S OF.G.
N , 1979. Linguistic Capabilities of a Lowland Gorilla. Ph.D.
dissertation, Stanford University.
26. KELLOGC,
W.N. 1968. Communication and language in the home-raised
chimpanzee. Science 182: 423-427.
27. HAYES,C. 1951. The Ape in Our House. Harper dr Row, New York, N.Y.
28. TEMERLIN,
M.K. 1975. Lucy: Crowing up Human: A Chimpanzee Daughter
in a Psychotherapist’s Family. Science and Behavior, Palo Alto, Calif.
Rubric for Discussion Assignment #2
Can non-human animals learn language?
Summer 2016
Assignment
Instructions for completing and submitting the assignment
Read the Terrace paper.
Follow the steps on each slide. Instructions are typed in red.
Delete the red text and replace it with your answers.
Upload your completed PowerPoint file to the discussion forum
Evaluate at least one other student’s proposed experiment (see
slide 7) and comment on it
Your response to their post should be at least 100 words
Consider (Concept Map of Introduction)
Read the Introduction section of the paper and construct a
concept map by defining key terms and creating appropriate
links between them
Fill your concept map with keywords that help you to
understand the concepts in the introduction:
Make a note of the concepts that you need to review or study
Define new issue(s) to be addressed. What was new about the
research?
Locate relevant variables and determine their relationships
Design your concept map by going to the insert tab, clicking on
shapes, selecting boxes or bubbles, and lines to connect them.
Type the concepts that you would like to relate into each box or
bubble
Here is a link to an example of a concept map:
https://library.usu.edu/instruct/tutorials/cm/CMinstruction2.ht
m
Delete this text and place your concept map here
Read (Methods and Results)
Read the methods and results sections
Describe how the data were obtained (type here)
Define unfamiliar words here (type here)
Create a simple visual depiction (flowchart, diagram, etc.)
of the methods for the study (type here)
Elucidate Hypotheses
Describe the major hypothesis (type here)
Describe what is being examined in Figure 1 (type here)
Create your own title to describe the figure (type here)
Describe what is being examined in Table 4 (type here)
Create your own title to describe the table (type here)
Analyze and Interpret Data
Determine what the data mean. Relate the results to the
hypothesis and purpose of the research (type here)
Make at least 4 bullet points to summarize the
conclusions (type here)
Read the discussion and conclusions sections and identify
any differences between your conclusions and those
presented in the paper (type here)
Can Nim Chimpsky and other animals learn language?
Why or why not?
Construct a logical argument to support your opinion
using information from the Herbert Terrace paper and the
Savage-Rumbaugh video.
Think of the Next Experiment
Imagine that you are an author of the article just
analyzed. What experiment should be done next?
Design a distinct study to follow up:
Describe the purpose (type here)
Describe the hypotheses (type here)
Describe the methods (you can make a flowchart here if you
wish)
Purchase answer to see full
attachment