ArticlePDF Available

Syntactic Structures in the Vocalizations of Wedge-Capped Capuchin Monkeys, Cebus Olivaceus

Authors:

Abstract and Figures

Capuchin monkeys, Cebus olivaceus, combine different types of calls to form compound calls. These compound calls are syntactically organized (i. e. , there is a predictable ordering of call types in the compound call). The analysis of the syntax of animal communication necessarily includes: (a) Accurate classification of types of calls, and a demonstration that calls that occur both alone and in compound calls are structurally similar. (b) Description of the syntactic rules generating compound calls or call sequences, and classification of types of compound calls or call sequences. (c) Examination of the social circumstances or contexts in which both single and compound calls occur. I followed these steps using a large sample of Cebus olivaceus calls, recorded in riparian gallery forest in central Venezuela. The calls were initially screened using a real-time spectrum analyzer, and a group of structurally related call types was selected for further analysis. Temporal and frequency characteristics of 868 of these calls were measured from sound spectrograms produced on a sound sonograph. Call classification involved first defining a very large number of possible call types on the basis of these characteristics, and then using a discriminant analysis to identify which types should be lumped together. A stepwise procedure was followed, using only calls which occurred singly, until five call types (squaws, chirps, trills, whistles, screams) were statistically separable. Two of these types (trills, whistles) showed considerable within-type variation and were further subdivided into four variants each. Discriminant analysis was then used to demonstrate structurally similarity between calls produced singly and those in compound calls. Social circumstances were defined using similarities in the vocalizer's actions, arousal, orientation to and distance from the presumed receiver. Use of call types, when given singly, covaried predictably with social circumstances and presumably with the internal state of the vocalizer. Different calls expressed different internal states on a continuum from contact-seeking to contact-avoiding. Use of trill and whitle variants also covaried with social circumstances. Different variants expressed different states on a continuum from affiliation or submission to aggression. Combinations of internal states, in theory, might be expressed as intergradations or intermediates between different call types. However, such intermediate states are often coded syntactically. Syntactically organized compound calls accounted for 38% of the total sample. The distribution of social circumstances in which compound calls are given was intermediate between the distributions of the constituent call types, which presumably indicates an intermediate internal state. Compound calls are generated by syntactic rules closely analogous to lexical rules of human language. Specifically they act like compounding rules that combine two lexical entries to form a third. They are not analogous to the grammatical rules that generate human sentences.
Content may be subject to copyright.
Syntactic Structures in the Vocalizations of Wedge-Capped Capuchin Monkeys, Cebus olivaceus
Author(s): John G. Robinson
Source:
Behaviour,
Vol. 90, No. 1/3 (Aug., 1984), pp. 46-79
Published by: BRILL
Stable URL: http://www.jstor.org/stable/4534358
Accessed: 27/10/2010 10:08
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=bap.
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
BRILL is collaborating with JSTOR to digitize, preserve and extend access to Behaviour.
http://www.jstor.org
SYNTACTIC STRUCTURES IN THE VOCALIZATIONS OF
WEDGE-CAPPED CAPUCHIN MONKEYS, CEBUS OLIVACEUS
by
JOHN G. ROBINSON')
(Florida State Museum, University of Florida, Gainesville, Florida 32611, U.S.A.)
(With 9 Figures)
(Acc. 16-XII-1983)
Introduction
During close-range social interactions, wedge-capped capuchin monkeys
Cebus
olivaceus
(= nigrivittatus,
following HONACKI
et al., 1982) utter com-
pound calls that are syntactically structured. Different types of calls,
which in other social circumstances are given singly, are compounded
together into mixed doublets, triplets, and occasionally quaduplets. The
ordering of these different call types in compound calls is predictable.
Compound calls are used in social circumstances intermediate between
those of the constituent calls. The rules describing the observed
sequences and their use are generally applicable, in that they both cor-
rectly describe observed sequences and their use, and predict the use of
novel (not observed) sequences in specified social circumstances. This
study characterizes the vocal syntax of these monkeys, and draws
analogies with the syntactical structure of human language.
The study of the zoosyntax, or the syntax of animal communication
(SEBEOK, 1972: 124), is confronted with two difficulties, one
philosophical, the other methodological. Both difficulties result from the
definition of syntax: Syntax is a "system constituted by rules that interact
to determine the form and intrinsic meaning of a potentially infinite
number of sentences" (CHOMSKY,
1972: 69). It is "the principles and
') I gratefully acknowledge and thank T. BLOHM for his enthusiastic permission to work
and live at his ranch: Hato Masaguaral; R. H. WILEY
for generously making available his
lab and the use of the real-time spectrum analyzer; B. HARDY for access to a sound
sonograph machine; and J. EISENBERG,
C. SNOWDON,
J. NEWMAN,
and L. Cox for their
discussion, encouragement, and comments on an earlier version of the manuscript.
Financial support was provided by an IESP grant to J. EISENBERG,
and NIMH research
grant to J. EISENBERG and R. RUDRAN,
and a National Geographic research grant.
SYNTACTIC STRUCTURES IN VOCALIZATIONS OF CAPUCHIN MONKEYS 47
processes by which sentences are constructed in particular languages"
(CHOMSKY, 1957: 11).
The philosophical difficulty is that this definition restricts syntax to
human language. This restriction follows from the common assumption
that animal and human vocal signals differ qualitatively. According to
this view, humans differ from animals in what they communicate.
Vocalizations of animals are "affective", simply representing the specific
behavioral tendencies, motivational state and general arousal of the
animal. In contrast, human language is "symbolic", with signals refer-
ring to objects in the external environment. Humans and animals also
differ in how they communicate. Animal vocalizations simply express the
animal's motivational state, etc. As articulated by CHOMSKY
(1972: 69),
"Every [animal] communication system that is known ... uses one of two
basic principles: Either it consists of a fixed, finite number of signals,
each associated with a specific range of behavior or emotional state, ... or
it makes use of a fixed, finite number of linguistic dimensions, each of
which is associated with a particular nonlinguistic dimension in such a
way that selection of a point along the linguistic dimension determines
and signals a certain point along the associated nonlinguistic
dimension." In contrast, human language is "syntactic", with
organized relations among separate sounds and novel use of different
combinations of sounds. Both symbolism and syntax are considered
necessary, if not sufficient, to define human language, but it is generally
argued that signals that do not symbolize or refer to objects in the exter-
nal environment cannot be syntactically organized. SAVAGE-RUMBAUGH
et al. (1980) state simply that "the issues of combination, syntax, and
novel usage are all subordinate to that of reference". This statement is a
corollary of the acceptance of a dichotomy between affective and
symbolic signals.
This dichotomy derives from the philosophical directions set by
DESCARTES, who reasoned, in Discourse on the Method (1637/1952), that
animals "cannot speak as we do, that is, so as to give evidence that they
think of what they say", and "we ought not to confound speech with
natural movements that betray passions." Two hundred and fifty years
later, HUXLEY
(1893) would still argue that "from the absence of
language, [animals] can have no trains of thoughts, but only trains of
feelings", and JAMES (1890) would state that, "the most elementary
single difference between the human mind and that of brutes lies in this
deficiency on the brute's part to associate ideas with similarity". While
some recent studies (e.g., SEYFARTH et al., 1980a; CHENEY & SEYFARTH,
48 JOHN G. ROBINSON
1982) suggest that some animal signals might also symbolically represent
objects or events, most discussions explicitly (e.g., SMITH, 1977, 1981;
SEYFARTH et al., 1980b) or implicitly (e.g., EISENBERG, 1974; MORTON,
1977, 1982) still retain the distinction between affective and symbolic
signals.
This approach is not followed here. While maintaining the dichotomy
between symbolic and affective signals is perhaps useful in certain
circumstances, it is ultimately artificial, being only the mind-body
dichotomy in another guise (see RYLE, 1949). Most would agree with
GREEN & MARLER (1979: 134-136) that both internal and external infor-
mation are encoded in signals. Signals can have both internal and exter-
nal referents, and different calls presumably can emphasize one or the
other. For instance, the use of alarm calls by vervet monkeys, Cer-
copithecus
aethiops,
depends on both the presence of predators (SEYFARTH
et
al., 1980) and the vulnerability, and presumably fear, of the potential
caller (SMITH, 1981). GREEN (1975) found that knowledge of arousal,
demeanor, social spacing, and orientation more completely predicted the
use of vocalizations in Japanese macaques, Macaca fuscata, than
knowledge of either internal or external circumstances alone. WILEY
(1976) pointed out that in common grackles, Quiscalus quiscula, both
changes in spatial relationships and changes in locomotor tendencies
elicited certain vocalizations associated with spatial relationships. Once it
is accepted that the internal state of an animal is a consequence of the
processing of both internal and external stimuli, the dichotomy between
"affective" and "symbolic" signals need not be maintained. Animal
vocal syntax can be studied independently of the semantic properties of
the signals; syntax is no loger subordinate to reference.
The internal state of an animal, which GREEN & MARLER (1979: 84)
termed the assessment state, may or may not be processed in its turn to
generate the physical signal. Transformational rules govern this selection
of the physical signal, and relate internal state to signal production
(GREEN & MARLER, 1979: 84-95). Syntactic rules are a subset of these
transformational rules. Rather than mapping a specified internal state to
a single vocal signal, as discussed by GREEN & MARLER, syntactic rules
map the specified internal state onto a string or sequence of vocal signals.
Information is encoded in the sequence of vocal signals rather than in any
single call within the sequence.
There remains the methodological difficulty. The syntax of human
speech is determined by deriving a system of rules that assigns structural
descriptions to all existing sentences and can generate acceptable new
SYNTACTIC STRUCTURES IN VOCALIZATIONS OF CAPUCHIN MONKEYS 49
ones (CHOMSKY, 1965: 8). For those interested in the study of the syntax
of animal communication, an analogous exercise would be to derive a
"system of rules that will allow us to predict sequences of signals"
(SNOWDON, 1982). There are, however, problems:
a) How does one discriminate a call from a sequence of calls? Is the
pant-hoot of chimpanzees (MARLER & HOBBET, 1975) a single call or a
sequence? This is a problem if the elements in the sequence are never
produced except in sequences, but does not apply to those cases in which
elements are rearranged to produce different sequences (e.g., titi
monkeys, ROBINSON,
1979), or in which calls are given both alone and in
sequences (e.g., cottontop tamarins, CLEVELAND
& SNOWDON, 1982).
b) How does one determine the beginning and end of a syntactically
organized sequence? A long sequence might not have "punctuation
marks", and changes in call type during long vocal sequences might
reflect gradual and regular change in internal state (see MORTON,
1982).
c) How does one determine whether a sequence of signals is combined
syntactically or whether early calls simply act as contextual modifiers for
later ones (WILEY, 1975; GREEN & MARLER, 1979: 143). This is a concern
in sequences in which calls are not predictably ordered (e.g., Guinea
fowl, MAIER, 1982). Syntax might be indicated if certain combinations
are regularly discernible.
In this study I examined an interrelated system of calls (ROBINSON,
1982), some of which are produced both alone and together in ordered
sequences. This avoids the first and third problems. These sequences of
calls are also uttered as a unit, with continuous transitions between call
types, and this allows punctuation.
Methods
The study area, in the highly seasonal llano intermedio
savanna of central Venezuela, is a
mosaic of shrub woodland, open grassland, palm savanna, and gallery forest bordering
the rivers. The study site is an extensive, continuous gallery forest located in the eastern
part of a large cattle ranch, Hato Masaguaral, owned by Sr Tomas BLOHM. Vocalizations
were recorded periodically from a well habituated, individually identified group of wedge-
capped capuchins, Cebus olivaceus, between February 1978 and July 1979.
On each selected day recordings were opportunistic. If a situation was likely to yield
vocalizations, I would begin recording, using a Uher 4000-L taperecorder and a
Sennheiser MKH 816T directional microphone. So as to define the social circumstances,
I noted the identity and non-vocal behavior of the vocalizer, its arousal (that I subjectively
placed at one of five levels), its distance from and orientation to the presumed receiver (the
animal to whom the vocalizer was orienting towards or away from), and the identity and
distance of its nearest neighbor. I noted the vocal and behavioral response of the receiver,
its arousal level and orientation, and the identity and distance of its nearest neighbor.
Changes in these variables during an interaction were verbally noted on the tape.
50 JOHN G. ROBINSON
For this paper I restricted the analysis to a small number of interrelated vocal types.
These vocalizations are uttered in close-range social interactions. They appeared
beforehand to be related both structurally, in that transitions and integradations between
the various call types were significantly greater than expected by chance, and contextual-
ly, in that there were similarities and transitions among the social circumstances in which
calls were given. In other words they made up a vocal system (ROBINSON, 1982) that
appeared to regulate the social interactions of animals in the group. Other vocal systems
in the repertoire of Cebus olivaceus
are described in ROBINSON
(1982, and in prep.).
I first examined the recordings on a Spectral Dynamics 301D real time spectrum
analyzer. Amplitude was displayed on the beam intensity axis of a Tectronix 5013N/D10
oscilloscope and frequency on the x axis so that a continuous spectrogram could be
produced on linagraph photographic paper using a Grass Kymograph camera. Film speed
was set at 10 mm/sec and frequency displayed at 10 kHz. Using this record I then
produced spectograms of 868 of these calls with a Kay Sound Spectrograph 7029A, mostly
using the 80-8000 Hz settings. To avoid biasing the selection of calls for analysis I used
every clearly recorded example of these call types for which I had contextual information
taken between February and November 1978. For each call I made 15 measurements
(Table 1) using an acetate overlay to an accuracy of 0.5 mm, so temporal values are
accurate to approximately 5 ms and frequency to approximately 40 Hz.
The derivation of syntactic rules depended on an accurate and replicable classification
of calls, a difficult goal when vocalizations grade into one another. In addition, it was
necessary to demonstrate that calls did not differ structurally between cases in which they
are given alone and those in which they are compounded with other call types. According-
ly I opted to classify calls using the SPSS discriminant analysis (NEI et al., 1975), a tech-
nique used by SMITH,
NEWMAN
et al. (1982) to discriminate among vocalizations used by
different individuals (for a fuller treatment see SMITH,
NEWMAN
et al.'s paper). The aim
was to initially define a large number of vocal categories, and then use the discriminant
analysis to collapse categories until the separation of those remaining was statistically
supportable. Initially I used 23 discriminating variables, the 15 derived from the direct
measurements and 8 more computed from them (Table 1). Discretely distributed
variables like TONE can be included in discriminant analysis if the correlation among
variables is not high (LACHENBRUCH, 1975). By eye I then classified these calls into 14
different categories, sometimes making very fine distinctions. Then discriminant analysis
was used to statistically distinguish among the previously defined vocal categories by
weighting and linearly combining the variables to generate discriminant functions. These
functions statistically separate the vocal categories as much as possible. I used a stepwise
option that enters and removes variables on the basis of their discriminating power. Thus
a variable was considered for inclusion in the discriminant function only if its partial
multivariate F ratio (a measure of its discriminating power taking into account the other
included variables) was larger than 4.0. If it was greater, the variable that resulted in the
largest increase in Rao's V, a generalized distance measure, was entered first. Once
included, a variable could only be exclude if its F ratio fell below 4.0.
The first step was to identify those variables with the most discriminating power. In
each analysis, six or fewer variables contributed disproportionately to the discriminant
functions so the rest were dropped. Reducing the number of variables is desirable to
maintain the ratio of number of variables to number of cases in each group at 1:5 or
greater (see rationale in SMITH,
NEWMAN
et al., 1982). Subsequent analyses relied only on
these six variables. The discrimination analysis generates one less discriminant function
than the number of groups.
The next step, after the discrimination analysis, was to use the discriminant functions
to classify all the cases used to derive the functions. Category membership is assigned
based on each case's values on the discriminating variables. The results indicate what
percentage of cases are correctly classified, and if incorrectly classified, in what other
SYNTACTIC STRUCTURES IN VOCALIZATIONS OF CAPUCHIN MONKEYS 51
category the cases were placed. This then can be used to determine the number and iden-
tity of vocal categories. This is useful when examining graded vocal repertoires in which
discrete vocal types do not exist and classifications must be arbitrary. Initially I made very
fine discriminations and generated many vocal types. I arbitrarily set as a goal the correct
classification of at least 80% of all cases, with at least 60% of cases in each category cor-
rectly classified. When a classification analysis yielded values less than these, I collapsed
vocal categories together and began the discrimination again from the beginning (with all
23 variables). This was continued until the acceptable level of correct classification was
reached. Once the major categories were determined I distinguished variants within two
of the categories (trills and whistles) using the same procedure. The acceptable levels were
slightly lower: correct classification of 75% of all cases, with at least 55 % correctly placed
in each sub-category.
The next stage in the analysis examined whether these vocal types were used in dif-
ferent social circumstances. I distinguished 18 such circumstances using records of the
behavior of the signaler, its arousal, orientation to the presumed receiver, and the
distance to that animal. The statistical analysis of the association of vocal use with social
circumstances followed procedures outlined by GREEN
(1975). Overall heterogeneity of
use of vocalizations with social circumstances was tested with the G-test. As the results
indicated that vocalizations were not used homogeneously in the different circumstances,
I made pairwise comparisons among the vocal types using the G-test. These G-values
provide a measure of the degree of difference in contextual use between each pair of vocal
types, and can be used to order the vocal types. I then constructed a contingency table of
vocal type versus social circumstances. I assumed the existence of a single rule relating
vocal type to social context and placed the ordered group of vocal types along one axis and
then ordered the social circumstances along the other so that the strong associations fell
roughly along the diagonal.
This classification of the vocal repertoire relied exclusively on cases in which the
vocalization was not compounded into a vocal sequence. The final stage of the analysis was
to examine calls that were compounded. I considered only those compound calls in which
the last syllable of the first call was continuous with the first syllable of the second call. In
all cases the vocal types that were compounded also occurred singly. To ensure that the
physical characteristics of these vocalizations were the same as those that were not
compounded, I first calculated the percentage of compounded vocalizations that would be
correctly classified using the discriminant functions derived from the non-compounded
cases. I then examined the compounding among call types. Transition probabilities
among call types were not equiprobable, and within pairs there was a definite order in
which one precedes the other. Finally, I defined types of compound calls and examined
their use in different social circumstances.
Results
Classification of vocalizations.
The first discriminant analysis aimed to categorize the major vocal types
in this vocal system (ROBINSON, 1982), which is only a part of the entire
vocal repertoire. As some of the calls were atonal I restricted the
discriminating variables to those measured on every call (ignoring
characteristics of the dominant band, Table 1). Six of these (TONE,
MINF, DURE, AMPL, RNGE, DISC) contributed disproportionately
to the derived discriminant functions and were used in the final analysis.
52 JOHN G. ROBINSON
TABLE
1. Time and frequency measurements from spectrograph tracing.
Each identified by a four letter code
A. Tonality (TONE) distinguished (1): Tonal: tracings with noise-free, distinct bands, (2)
Atonal: with noisy tracings, no distinct bands, and (3) Tracings with a mixture of
atonal and tonal energy.
B. Distribution of spectral
energy.
(1) Maximum frequency. Highest frequency clearly discernible at -20dB relative to
average peak intensity on spectrogram (MAXF).
(2) Minimum frequency. Lowest frequency clearly discernible at -20dB relative to
average peak intensity on spectrogram
(MINF).
(3) Dominant frequency. Frequency half-way between the start and the end of the
largest, blackest part of the spectrogram
(DOMF).
(4) Range of spectral energy (RNGE) = MAXF - MINF.
C. Characteristics
of the dominant
band
(the largest, blackest
tonal band on the spectrogram).
Applicable only to tonal calls.
(1) Measured Variables.
Position on dominant
band
Measurement Beginning Low High End
Frequency FREB FREL FREH FREE
Duration DURL DURH DURE
(2) Computed variables.
Range of dominant band (RANG)= FREH - FREL
Mean frequency (MEAN) = (FREH + FREL)/2
Change in frequency (CHNG) = FREE - FREB
Relative position of the low point (LOW) = DURL/DURE
Relative position of the high point (HIGH) = DURH/DURE
Rate of change (RATE) = (FREH - FREL)/DURE.
D. Number
of
syllables
(continuous spectrographic tracings
and associated
frequency bands)
in each call (SYLL). A relative measure of continuity within the call (DISC)=
DURE/SYLL.
E. Number
of inflexions
(changes in frequency direction) on dominant band (INFL).
F. Number of tonal bands at midpoint
of dominant
band (NBND).
G. Amplitude
(AMPL) distinguished (1) Loud: vocalizations easily heard by an observer on
the ground
> 20 m away, (2) Quiet: vocalizations not discernible
by an observer
on the
ground
> 20 m away. These categories were very discrete.
Fourteen vocal types were defined at the beginning of the analysis.
This reduced to five (squaws, chirps, trills, whistles, and screams) as a
result of the classification analysis. The final discriminant analysis thus
derived four functions (one less than the number of categories). The first
accounted for 51.3% of the variance and the second for an additional
35.0%, so discussion will focus mainly on these functions.
Measures of signal form characterized discriminant function 1:
TONE, which specifies whether a call is tonal or not, and DISC, a
SYNTACTIC STRUCTURES IN VOCALIZATIONS OF CAPUCHIN MONKEYS 53
TABLE 2. Classification of all five vocal types using a limited list of
variables
Actual vocal types Predicted vocal types
Squaws Chirps Trills Whistles Screams
Squaws 18 0 0 0 0
100%
Chirps 1 35 0 0 0
2.8% 97.2%
Trills 4 17 245 25 6
1.3% 5.7% 82.5% 8.4% 2.0%
Whistles 2 5 21 82 15
1.6% 4.0% 16.8% 65.6% 12.0%
Screams 3 0 2 2 43
6.0% 4
4.0 4.0% 86.0%
Percent of cases classified correctly
= 80.42%.
Standardized
canonical discriminant function coefficients
Variable Function 1 Function 2 Function 3 Function 4
TONE 0.425 -0.210 0.849 0.354
MINF -0.429 -0.186 -0.225 0.553
DURE -0.420 0.032 0.522 -0.832
AMPL -0.184 0.516 0.157 0.606
RNGE 0.276 0.579 -0.519 0.086
DISC 0.651 -0.516 -0.613 0.447
Percent of variance accounted for by function 1 = 51.3%, function 2 = 35.0%, function
3= 11.7%, function 4= 1.9%.
measure of syllable discontinuity, loaded positively, and DURE, the call
duration, and MINF, the minimum frequency, loaded negatively. This
function discriminates whistles and trills from the rest. Two variables,
AMPL, the call amplitude, and RNGE, the range of spectral energy
loaded positively on discriminant function 2. These two variables are
evidently related, for both increase as the vocalizing animal becomes
more aroused. This function divides chirps and squaws from whistles and
screams.
The four discriminant functions were then used to reclassify the cases
from which the functions were originally derived, and I then compared
these results to the original classification. Table 2 presents the final
results. Vocal types were ordered along the axes so that adjacent pairs
were structurally the most similar. About 80% of cases were correctly
JOHN G. ROBINSON
reclassified. The poorest results were for those calls that I initially
classified as whistles, which were incorrectly classified as trills in 16.8%
of cases and as screams in 12.0% of cases. Fig. 1 plots each case using its
values on discriminant function 1 against function 2. Territorial
boundaries between adjacent categories are shown. This scatterplot
assumed that all the functions but the first two are zero. It graphically
shows the whistle cases that have been misclassified.
CHIRPS
C'J
Z 4-
0 SQUAWS
: TRILLS ®®
O- e
e
0 ^M
A- 0 4·A
Z * a, A A
C] , , <, , / WHISTLES ,
.CREAM
Fig. . Territorial map, showing five regions in which vocalizations
would be classified
as
belonging to one of five vocal types, and plot of individual vocalizations in the sample
using their values on the first two discriminant functions. Closed circles are "correct"
/) -4- SCEM
WHISTLES SCREAMS
-4 0 4
DISCRIMINANT
FUNCTION 1
Fig. 1. Territorial map, showing five regions in which vocalizations would be classified as
belonging to one of five vocal types, and plot of individual vocalizations in the sample
using their values on the first two discriminant functions. Closed circles are "correct"
classifications - my final classification was the same as the region in which that case is
plotted. Open triangles "incorrect" classifications - my final classification was different.
However, the first analysis only relied on a few of the possible
discriminating variables. Accordingly, to confirm the distinctiveness of
whistle and trill categories, I restricted a second discriminant analysis to
the three vocal types which are tonal (chirps, trills, and whistles) and
used all the variables. Six of these were used in the final analysis (NBND,
DURH, DISC, CHNG, HIGH, and RATE).
54
SYNTACTIC STRUCTURES IN VOCALIZATIONS OF CAPUCHIN MONKEYS 55
TABLE 3. Classification of three vocal types using all variables
Actual vocal types Predicted vocal types
Chirps Trills Whistles
Chirps 33 2 1
91.7% 5.6% 2.8%
Trills 2 290 4
0.7% 98.0% 1.4%
Whistles 2 26 97
1.6% 20.8% 77.6%
Percent of cases classified correctly = 91.9%.
Standardized canonical discriminant function coefficients
Variable Function 1 Function 2
NBND 0.851 -0.174
DURH 0.118 -0.528
DISC 0.266 1.044
CHNG -0.275 0.057
HIGH 0.041 0.393
RATE 0.482 0.076
Percent of variance accounted for by function 1 = 69.4%, function 2 = 30.6%.
The first discriminant function accounted for 69.4% of the variance,
and the second for the remaining 30.6% (Table 3). NBND, the number
of bands, loads most heavily on function 1 and separates chirps from the
rest, and DISC loads most heavily on function 2 and effectively separates
trills from whistles. The percentage of cases correctly reclassified
increased to 91.9% and whistles were correctly reclassified in 74.7% of
cases. However, in 20.8% of cases, the whistle calls were still incorrectly
classified as trills.
These results confirm the existence of five vocal types (squaws, chirps,
trills, whistles, and screams) in this selection from the vocal repertoire of
Cebus
olivaceus
(Fig. 2). Between two types, whistles and trills, there are
structural continuities. Indeed, whistles sometimes pass smoothly into
trills and vice-versa (see Fig. 2.Vb or Fig. 8.IIIa), producing calls com-
prising both vocal types. Yet for reasons given below, these were not
classified as compound calls. In these caes I initially assigned the call to a
whistle or a trill category on the basis of the predominant type. Despite
these intergradations, the discriminant analysis indicates that the two
vocal types are discrete enough to maintain a statistical separation. The
la
a.
a.
Ila b
l
o
Illa
'V
( ..
..-......
IVa 0
z
C:
0
tt
0
z
0
z
d
d
amm
Va b c
a--r
a11
ar
a ,
*f*hi
; I - :-- w -wA
Fig. 2. Spectrograms of Cebus olivaceus
vocalizations analyzed with the broad-band filter (effective resolution, 300
Hz), showing the range of variation among five vocal types. Ia, Ib, Squaws. IIa, IIb, Chirps. IIIa, IIIb, Screams.
IVa, Falling whistles; IVb, U whistles; IVc, FM whistles; IVd, Rising whistles. Va, Falling trills; Vb, U trills;
Vc, FM trills, Vd Rising trills. Frequency scale indicates kHz, time scale is 0.5 sec.
qn
j \
4000w" / $I » ;*
. i
4 r it
i f
. j
?f 1
,i I
*
"t'
i, isEt
?f AI
O 1' I
\
tt'
t
A
,
0
,(1'*i"l
SYNTACTIC STRUCTURES IN VOCALIZATIONS OF CAPUCHIN MONKEYS 57
TABLE
4. Classification of all four trill types using all variables
Predicted trill types
Actual trill types Descending U Ascending FM
trills trills trills trills
Descending trills 115 27 2 7
76.2% 17.9% 1.3% 4.6%
U trills 6 44 6 2
10.3% 75.9% 10.3% 3.4%
Ascending trills 0 11 39 0
22.0% 78.0%
FM trills 4 7 4 21
11.1% 19.4% 11.1% 58.3%
Percent of cases classified correctly
= 74.24%.
Standardized canonical discrimination function coefficients
Variable Function 1 Function 2 Function 3
CHNG 0.600 -0.004 0.970
NINF -0.108 0.815 -0.020
LOW -0.718 -0.080 0.905
DURL 0.386 0.627 0.066
RANG 0.413 -0.183 -0.165
Percent of variance accounted for by function 1 = 65.6%, function 2 = 32.0%, function
3= 2.4%.
structural characteristics of the call types are summarized in the
Appendix.
Both trills and whistles show considerable within-type variation in
frequency modulations. Some calls rise from beginning to end, some fall.
Others modulate continuously while still others rise and fall once, or fall
and rise once. Such variation is easily discernible to the human ear. To
see whether this variation fell into discrete categories, a discriminant
analysis was performed on trills alone, the category for which I had the
most samples. I initially defined seven trill variations, but reduced this to
four (falling trills, U ("U" shaped) trills, FM ("Frequency modulated")
trills, and rising trills) on the basis of the first discrimination analysis.
Five variables were used in the final analysis (CHNG, NINF, LOW,
DURL, RANG).
The first discriminant function accounted for 65.6% of the variance
and the second for 32.0% (Table 4). Two derived variables, CHNG, the
JOHN G. ROBINSON
change in frequency of the dominant band from beginning to end, and
LOW, the relative position of the low point, loaded most heavily on func-
tion 1. This function thus separates falling from rising trills. NINF, the
number of inflexions, and DURL, the duration to the low point on the
dominant band, load most heavily on function 2, which effectively
separates FM trills from the rest. Table 4 includes the classification table,
with trill types ordered so that adjacent types are the most similar. About
74% of all trills were correctly classified, and the percentage correct
within each category ranged from 58.3% to 78.0%. Fig. 3 plots each case
C \ FM TRILLS
Z
O 4-
4 ..
~~z~~~~~~~·
z \\
LL
Fo- o-
Z FALLING -
^< TRILLS -e*
\ · *
2~~~~~~~~~~~~~- / \
~RISING
-: / \ TRILLS
O / U TRILLS
co -4-
E
0 4
DISCRIMINANT FUNCTION 1
Fig. 3. Territorial map, showing four regions in which the vocalizations would be
classified as belonging to one of four trill types, and plot of individual trills in the sample
using their values on the first two discriminant functions. Other conventions the same as
in Fig. i.
using its values on the first two discriminant functions. The structural
characteristics of trill types are summarized in ROBINSON, in prep. Similar
variants occur in the whistle vocal types and I distinguished falling, U,
FM, and rising whistles.
58
SYNTACTIC STRUCTURES IN VOCALIZATIONS OF CAPUCHIN MONKEYS 59
Association between vocalizations and social circumstances.
In this analysis I was primarily interested in the association between
vocalizations and the animal's internal state, which can be measured by
the broadcast information of the vocal signal, rather than in the asso-
ciation between vocalizations and changes in the receiver's internal
state, which can be measured by the transmitted information or social
significance of the vocalization. Defined rigorously from the viewpoint of
the external observer, the broadcast information of a vocal signal is the
reduction in the uncertainty concerning the vocalizer's behavior and
internal state after the signal is given, WILEY & RICHARDS
(1982).
Accordingly, social circumstances were grouped using similarities in the
vocalizer's behavioral actions, arousal, orientation to and distance from
the receiver, rather than similarities in the receiver's responses, etc. This
approach generated 18 categories of social circumstances.
TABLE 5. Differences in distribution of use of five vocal types across
18 social circumstances
Chirps Trills Whistles Screams
133.3c 117.9c 107.9 74.7 /
Squaws
h e 0.001
o 0.001 / 0001
0 .00 1 0.
162.2 139.7 107.9 /
Chirps o.001 0.001
o0001
87.4J 119.0
Trills 0.001
.001
58.9/
Whistles J 0.001
Cells contain G values of pairwise comparisons, and significance levels (df= 17, 18
circumstances x 2 types).
Following GREEN
(1975) I examined whe r
ther the use of the five major
vocal types varied across social circumstances. A test of overall
heterogeneity of use was significant (G = 463.3, p< 0.001): Animals use
different vocalizations in different social circumstances. This result
allows pairwise comparisons of use between each pair of vocal types.
Table 5 gives the G statistics and the significance value of the difference
for each pair. I followed GREEN'S
suggestion (that is conservative with
regards to significance) of calculating degrees of freedom using the
60 JOHN G. ROBINSON
number of columns and rows in the entire contingency table even though
some of the pairwise comparisons showed zero observations. The circum-
stances of use differ significantly between each pair of vocal types.
The next question was whether these vocal types are related to social
circumstances by an arbitrary lexicon or whether there are common
elements to the 18 circumstances that covary predictably with vocaliza-
tions. Table 2 and Fig. 1 indicate that structure varies continuously
through the following progression: squaws, chirps, trills, whistles,
screams. If adjacent pairs in this order have the most similarity of use in
social circumstances, then a single rule might relate structural variation
in vocalizations to their use in different contexts. To examine this, I
ordered the vocal types along one axis, and ordered the social circum-
stances along the other so that those cells with the strongest associations
fell along the diagonal. As expected, the ordering of social circumstances
was not arbitrary (Fig. 4). The first six are situations in which the
vocalizer is seeking or maintaining contact with another group member.
In the first two, the animals are in actual physical contact. This is
followed by four categories in which the vocalizer is neutral to or
ambivalent about seeking contact. The last eight categories are
characterized by the vocalizer avoiding contact. The last six of these
involve agonistic interactions: the vocalizer is threatening or chasing
another, or being threatened or chased. They are avoiding contact by
moving away or driving another away. There is a gradation in the
vocalizing animal's behavior, social spacing, and orientation as the
ordered list of social circumstances is traversed. This presumably in-
dicates a gradation in internal states, from contact-seeking to contact-
avoiding. These gradations correlate with the use of different vocaliza-
tions, a result that indicates the existence of a single rule, within this
group of five vocal types, relating internal state to use of vocalizations.
However, within vocal types and even between pairs of vocal types
there appear to be other rules relating structural variation in the vocaliza-
tions to their use in different contexts. Trills, for example, occur across a
wide range of social circumstances and show considerable structural
variation. As before I examined the heterogeneity of use of the four trill
variants by using the G test. The omnibus test was significant (G = 245.2,
p< 0.001), which allowed me to test heterogeneity of use between each
pair of trill variants. Table 6 gives the G statistic and significance of the
difference for each pair. Differences in the distribution of use among trill
variants were not as striking as among vocal types, but with the exception
of that between FM and rising trills, they were significant.
SYNTACTIC STRUCTURES IN VOCALIZATIONS OF CAPUCHIN MONKEYS 61
CALL TYPES
3) U)
SOCIAL CIRCUMSTANCES < -
* w
0 I Cr I )
CY - r -- un
()D t- 3 (n
1. NURSING
xxx xxxxx
2. 9 GROOMING
,xxxxxxxxxxxxx, x
3. GROOMING AND NUZZLING xxxxx,xx CONTACT-
INFANT ON MOTHER xx xxxxxxxxx CONTACT
SEEKING
4. Y APPROACHING OR
FOLLOWING ADULT 0Q xxx xx xx
5. ~ SOLICITING GROOMS
FROM ADULT 0' xxxx xxxxxx
6. APPROACHING ANOTHER xloolx
xxx x
XX )xXXXXX X
X
7 INFANT BEING LEFT BEHIND xxxxxxxx
xxxxxxxx
xxxxxXXXXxxxx
xxx
xxxxxxx
8. PLAYING xxxxxxxx
xxxxxxxx
xxxxxx NEUTRAL OR
8Kxxxxxxxxxxx N
E
U
T R
A
L
AMBIVALENT
9. FORAGING OR MOVING ,xxxxxx x
10. BEING APPROACHED xxx
XXooXXoolx
X
11. MOVING AWAY
12. BEING TEASED xxxxxxXX xxx
13. BEING THREATENED xxxxx
xxxxxooxxx
XXooXoX
X XX
14. THREATENING xxxMO
xxxxxxxx CONTACT-
Kx xxxM ' AVOIDING
15. BEING DISPLACED xx53xxx
<amQXX XX X
16. THREATENING FOLLOWING xXXxxxxo
XXXXXXX'X
xx
DISPLACEMENT a:m8l :xxxxxxx
xxxN
17. BEING CHASED xooO xxxx
xxxxxxx xxxxxxxx xxxxxxxx
xxxxxxx xxu xxxxx
18. CHASING XX xxxxxx XXX XXXXXX'
TOTAL NUMBER CALLS 18 37 290 124 44
Fig. 4. Association of call type with social circumstances.
62 JOHN G. ROBINSON
TRILL TYPES
u,
- U U)
J J
UM IJ NA
. -i j
SOALCIAL
CIRCUMSTANCES j [~ ~- Z
LL I LL
II. MOVING
AWAY xx
5. G SOLICITING
GROOMS
FROM ADULT 0oxxx
4. APPROACHING
AND
FOLLOWING
ADULT xxx xSUBMISSIVE/ AFFILIATIVE
2. ~ GROOMING
3. GROOMING OR NUZZLING
INFANT ON MOTHER x
xxooooo x
6. APPROACHING
ANOTHER XXX
9. FORAGING
OR MOVING
10. BEING APPROACHED ;NEUTRAL
xxxxxxxx *y
7 INFANT BEING
LEFT BEHIND x
xxxxxxxx
JQOOWc JXx,ooo v X
8. PLAYING
17 BEING CHASED uxxx xx x
xxxxo xx ~x AMBIVALENT
15. BEING DISPLACED xxxxooxx
x
XXXXX XOWO
x X x)OOxxxx
13. BEING THREATENED xxxxx
XXX xxxxx XX
12. BEING TEASED XXXX XX XX
16. THREATENING
AFTER x
DISPLACEMENT X xxooo AGGRES
'AGGRESSIVE
14. THREATENING
xxxx xxxx xxx xxxOOOx
18. CHASING x xx
TOTAL NUMBER CALLS 147 57 36 51
Fig. 5. Association of trill type with social circumstances.
SYNTACTIC STRUCTURES IN VOCALIZATIONS OF CAPUCHIN MONKEYS 63
To determine whether there was a rule relating structural variation to
use of trills, I followed the order based on structure along one axis, and
ordered the social circumstances so that the cells with the strongest
associations fell along the diagonal (Fig. 5). Again the ordering of social
circumstances was not arbitrary, though the ordering is different from
the previous analysis when the five vocal types were considered (ordering
can be compared by referring to numbers on the vertical axes). The
ordering indicates the existence, not of a contact-seeking/contact-
avoiding axis, but of a submissive-affiliative/aggressive axis. In the first
six social circumstances, the vocalizer has a submissive demeanor or
shows affiliative behaviour, and falling trills are the most common trills
produced. The animal is moving away or interacting with dominant
animals, often the dominant adult male. As one proceeds down the list,
the next three contexts are relatively neutral with regards to this axis, and
U trills are given most frequently. In particular, infants being left behind
during group progressions give mostly U trills. In the next four social
circumstances, the vocalizer could be characterized as ambivalent. Most
are agonistic situations in which both aggressive and submissive
behaviors were obvious. All variants of trills occur, especially falling and
FM trills. In the last four contexts the vocalizer is most aggressive, and
most rising trills occur in these situations. In most agonistic interactions
animals give a variety of trill types, with rising trills predominating when
the vocalizer is moving towards its opponent, FM trills when the situa-
tion is static, and falling trills when the opponent is advancing. Thus as
the list of social circumstances is traversed, there is a progressive change
from submissive or affiliative to aggressive behaviors, a change
presumably reflected in the internal state of the animal. Synchronously
there is a change in the vocalizer's preference for different trill types,
from falling trills through U and FM trills to rising trills. Therefore,
within trill type, there also appears to be a rule relating internal state to
vocal type, but this rule is not identical to the rule across vocal types.
The structure of both U and FM trills reflect the neutrality or
ambivalence of internal state. U trills are constructed of a single falling
trill followed by a single rising trill. FM trills are composed of alternating
rising and falling trills. Falling trills are produced in affiliative or sub-
missive social circumstances, rising trills in aggressive ones. The relation-
ship of frequency change and social circumstances seems to be paralleled
in Saimiri sciureus,
for NEWMAN et al. (1978) report that dominant animals
incorporate rising syllables in their "twitters" more than other animals.
64 JOHN G. ROBINSON
WHISTLE TYPES
)
w )
w.
TI O) L U)
j H -
SOCIAL CIRCUMSTANCES z L
-
J I 3 z
UI D LL cr
II. MOVING AWAY x
5. ? SOLICITING GROOMS
FROM ADULT 0' wx
4. 9 APPROACHING AND SUBMISSIIVE
FOLLOWING ADULT xxx x
6. APPROACHING
ANOTHER x
10. BEING APPROACHED X J
7 INFANT BEING LEFT NEUTRAL
BEHIND xxxx J
8. PLAYING xx xx XXXXX XXxxx
17. BEING CHASED . AMBIVALENT
XX XXX)QOOO(
XX
15. BEING DISPLACED xxxxxxx
xxxxxxxxxxxx
XX XXXXXX
XXXXXXXX
xx xxxxxxxx
xxxxxxxx,K
12. BEING TEASED XX X
16. THREATENING AFTER
DISPLACEMENT xx VxxXxxxxx,xxx AGGRESSIVE
14. THREATENING
18. CHASING XXX XXXXXX
TOTAL NUMBER OF CALLS 17 12 61 34
Fig. 6. Association of whistle type with social circumstances.
SYNTACTIC STRUCTURES IN VOCALIZATIONS OF CAPUCHIN MONKEYS 65
Finally I considered the relationship between trills and whistles. These
types integrate completely, passing from one to the other, and back again
in the same call. The four trill variants just distinguished were also
discernible within the whistle category (see Fig. 3). A contingency table
of whistle type against social circumstances demonstrates that signal
structure of whistles is related to internal state by a similar rule to that
regulating trills: When axes were ordered identically to the table of trill
types, the cells with the strongest associations again fell along the
diagonal (Fig. 6). In addition, as shown in the following section, the tran-
sition probabilities of trills to other calls in compound calls is similar to
TABLE 6. Differences in distribution of use of four trill types across
17 social circumstances
U trills FM trills Rising trills
114.6 60.9 82.4
Falling trills 0.1 0.001 0.001
65.8 61.0
U trills 0.001 A 0.001
22.0
FM trills ns
Cells contain G values of pairwise comparisons and significance levels (df= 16, 17
circumstances
x 2 types).
the probabilities of whistles to other calls. While pairwise comparison of
the heterogeneity of use did indicate that trills and whistles are used in
different social circumstances, close examination of Fig. 4 indicates that
the differences are more quantitative than qualitative. Whistles are
given in more highly aroused circumstances. This suggests a third rule
relating internal state to vocal type: The probability of an animal giving
whistles rather than trills increases as the animal becomes more aroused.
By examining only the variation within and across five vocal types I
have identified three rules relating internal state to vocal type. These
rules in turn define three axes of internal state underlying the variation in
vocal type: (a) A contact-seeking/contact-avoiding axis, (b) a submissive-
affiliative/aggressive axis, and (c) an arousal axis. These three axes
specify a three dimensional volume. Fig. 7 gives two of these axes; it is
highly diagrammatic and there is no reason to expect that the axes would
be orthogonal or even straight. Any internal state within this volume
66 JOHN G. ROBINSON
0
z
0
0
> SCREAMS RISING
T--- WHISTLES/TRILLS
I-
ZH E WHISTLES/TRILLSL
0
0
HISTLESITRILLS
FALLING
HSUBMISTLESIVTRILL
CHIRPS
0
z
w
w
SUBMISSIVE AGGRESSIVE
Fig. 7. Diagram relating use of different vocalizations to presumed internal state.
might be expressed vocally as intergradations or intermediates between
vocal types. In some cases they are, but often intermediate states are
instead coded syntactically.
Compound calls-classification.
Capuchin monkeys juxtapose calls of different vocal types to form com-
pound calls (Fig. 8). Compound calls are of two types: (a) Transitional
calls that blend two call types, as illustrated by the compounding of trills
and whistles, and possibly whistles and screams. These presumably
express the transition between two internal states. There are no rules
specifying an order or structure to these compound calls, and the transi-
tion probabilities between the pair of vocal types are similar in each direc-
3~~~~~~~~~~~~~~~~~~~~~~~~
. .: >. ...
3
O>~~~~~~~~~~~~~~~~~~~~~~~O
.~ ~ ~ ~~~~~~~~~~~':--?Z ... ...-..
FigR. 8.
:ld
Hz,sown xape f opun
.al. Ia, II
Hz),shoing examlesofchirpoudFalls.I,I,Wisgtrle/Chirps.
VII, Falngtil/hir/Squaw; II,IIc,Huh/Utill./Frquenysaweidcts.
kHz; time scale is 0.5 sec.
68 JOHN G. ROBINSON
TABLE 7. Percentage of vocalizations correctly classified using discrimi-
nant functions when compounded and when not
Vocalization Compounded n Non-compounded n
Squaws 76.7%o 30 100.0% 18
Chirps 89.0% 136 91.7% 36
Trills 100.0% 108 98.0% 296
Whistles 53.3% 30 77.6% 125
Screams 73.7 % 19 86.0% 50
Percent of compounded calls classified correctly
= 87.3%. Percent of calls occurring in
compounded calls = 38.1%.
TABLE 8. Percentage of trills classified correctly using discriminant
function when uttered alone or in a compound call
Vocalization Compound call n Alone n
Falling trills 90.9% 77 76.2% 151
U trills 77.8% 9 75.9% 58
Rising trills 81.3% 16 78.0% 50
FM trills 60.0% 5 58.3% 36
Percent of compounded trills classified correctly
= 86.9%. Percent of trills occurring in
compound calls = 26.6%.
tion. These are not discussed further. (b) Syntactically organized calls in
which there are syntactic rules specifying the order of vocal types in the
compound call. Call type A may precede call type B, but rarely or never
vice-versa.
Presumably the compound call in its entirety expresses a single
internal state.
Of the 868 calls samples in this study, 331 (38.1%) occurred in syn-
tactically organized compound calls. Of the 331, 24 ( = 8 compound calls)
were in triplets, 12 (= 3 compound calls) were in quadruplets, and the
rest occurred as doublets.
Are the calls that occur in compound calls structurally the same as calls
that are given alone? Rather than compare all the call parameters
independently, I used the discriminant functions derived from the calls
that were not compounded to classify the calls that were. Table 7 com-
pares the percentage of correctly classified cases in the compounded and
non-compounded categories. I used the functions derived from the first
discriminant analysis to classify squaws and screams, and the functions
SYNTACTIC STRUCTURES IN VOCALIZATIONS OF CAPUCHIN MONKEYS 69
TABLE
9. Compounding among call types
I. Transition matrix
Following
Squaws Chirps Trills Whistles Screams Others Marginal
totals
Begin 1 23 99 24 9 1 157
Squaws -0 1 0 0 2 3
e Chirps 26 -5 1 2 2 36
· Trills 0 85 - -6 9 100
u Whistles 0 24 - -2 0 26
X Screams 0 3 2 5 -0 10
Others 3 1 1 0 0 5
Marginal totals 30 136 108 30 19 14 338
II. Transition probability matrix
Following
Squaws Chirps Trills Whistles Screams Others
Begin 0.006 0.146 0.631 0.153 0.057 0.006
) Squaws 0 0.333 0 0 0.667
.5 Chirps 0.722 - 0.139 0.028 0.056 0.056
Trills 0 0.850 -- 0.060 0.090
Whistles 0 0.923 -0.077 0
a Screams 0 0.300 0.200 0.500 0
Others 0.600 0.200 0.200 0 0
derived from the second to classify the other three call types. For most
call types, the proportion of cases classified correctly was higher in the
non-compounded category, but the differences were not great. Table 8
makes the same comparison with trill variants. The overall proportion of
compounded calls correctly classified was higher than that of calls given
alone, as was the proportion in every category of trill variant. Again the
differences were not great. I conclude that my initial classification of
vocalizations was similar for calls that were compounded and those that
were not, and structurally the two categories are virtually identical.
Compound calls - syntax.
There is an order of vocal types in compound calls (Table 9). Chirps are
frequently followed by squaws, but never vice-versa.
They rarely precede
any other vocal type. Both trills and whistles lead into chirps, but rarely
70 JOHN G. ROBINSON
the reverse. Most cases of triplets began with a whistle or trill, led into a
chirp, and ended with a squaw. Note that I do not include transitions be-
tween whistles and trills (see discussion above). There were 15 instances
of calls composed of combinations of screams with whistles and trills,
which might only be transitional calls. Table 10 shows the frequency of
transition between the trill or whistle variants and other call types. The
proportional representation of trill and whistle variants in compound
calls is similar to the proportion in the overall sample of vocalizations (for
the four trill variants 0.69:0.08:0.06:0.16 versus
0.51:0.20:0.12:0.17).
TABLE 10. Compounding of trills and whistles with other vocalizations
Following call Following call
Preceding trill Chirps Screams Preceding whistle Chirps Screams
Falling trills 59 6 Falling whistle 10
U trills 7 U whistle 2
FM trill 5 FM whistle 9 1
Rising trills 14 Rising whistle 3 1
Compound calls - association with social circumstances.
Are the social circumstances in which compound calls are used indepen-
dent of the circumstances in which the constituent calls are used, or are
they used in circumstances intermediate between those of the constituent
calls? In the first case a novel call would be formed. In the second the
compound call would presumably express an intermediate state between
the internal states expressed by the constituent calls. This question was
examined by comparing the distribution of use of two compound calls,
chirp/squaws and falling trill/chirps, with their constituent calls when
given alone. These compound calls were chosen because they have the
largest sample sizes.
Squaws are given when the vocalizer is in physical contact with
another. They are the only vocalization used by a mother when nursing
an infant. Chirps are given in a wider range of social circumstances.
Vocalizers are seeking or maintaining close physical contact. The
distribution of social circumstances in which chirp/squaws are given
overlaps and is intermediate between the distribution of the two consti-
tuent vocal types (Fig. 9). Falling trills fall in an intermediate position on
SYNTACTIC STRUCTURES IN VOCALIZATIONS OF CAPUCHIN MONKEYS 71
CALL TYPES
O I
u_
0 0
aO_ a. -
0
, n zon z
SOCIAL CIRCUMSTANCES I -
O I C
a
<T
t) U 0 UI- LL
1. 9 NURSING 7 x
2. 9 GROOMING 9 14 I
3. GROOMING AND NUZZLING 2 7 xxx 13
INFANT ON MOTHER xx
4. A
APPROACHING
OR 5 3
FOLLOWING ADULT 0o xxx ,-x--
5. Y
SOLICITING
GROOMS I 4
FROM ADULT 0
6. APPROACHING
ANOTHER 2 II
7 INFANT BEING LEFT BEHIND 6
8. PLAYING 12
9. FORAGING OR MOVING 6 I
10. BEING APPROACHED 8
II. MOVING
AWAY 3
12. BEING TEASED 4
13. BEING THREATENED 13
14. THREATENING 4
15. BEING DISPLACED 2 38
16. THREATENING
FOLLOWING 13
DISPLACEMENT
17. BEING CHASED 12
18. CHASING I
TOTAL NUMBER CALLS 18 24 37 57 147
Fig. 9. Association of compound calls and their constituents with social circumstances.
72 JOHN G. ROBINSON
the contact axis, often expressing tendencies both to approach and to
avoid the animal with which the vocalizer is interacting. When com-
pounded with chirps these calls occur in circumstances characterized by
strong tendencies to seek physical contact. Again the distribution of
circumstances in which the compound call occurs is intermediate.
These results suggest that capuchins can express any state within the
volume defined by the three axes by a syntactic compounding of vocal
types. Aggressive animals that are seeking physical contact might com-
pound rising trills with chirps. Submissive animals seeking contact might
compound falling trills with chirps. If highly aroused they might
substitute whistles for trills. This interpretation is supported by an
analysis of how use of the four trill variants, when compounded with
chirps, varies across social circumstances. The test of overall hetero-
geneity is significant (G = 68.65, df= 30, p < 0.001). Animals used falling
trill/chirps in more submissive or affiliative social circumstances, and FM
and rising trill/chirps in more aggressive circumstances. A similar test
using the four whistle variants is also significant (G = 41.77, df= 21,
p< 0.005).
Discussion
If we define a syntactic rule as a transformational rule that generates a
sequence of calls, then syntactically organized compound calls are not
uncommon in Neotropical primates. They include the "yaps",
"keckers", and "chirps" of Saimiri (WINTER et al., 1966; SMITH et al.,
1982b), the "trill-awk" of Ateles (EISENBERG,
1976), the loud, complex
duets of Callicebus (MOYNIHAN, 1966; ROBINSON, 1979) and Leontopithecus
(MCLANAHAN & GREEN, 1977), and the call sequences of Saguinus oedipus
(CLEVELAND & SNOWDON, 1982). Among apes, candidates include the
"pant-hoot" of chimpanzees (MARLER & HOBBET, 1975) and the great
calls of gibbons (DEMARS & GOUSTARD, 1972; DEPUTTE, 1982). But is this
syntax analogous to the syntactical structure of human language?
MARLER (1977), in an influential discussion of zoosyntactics,
distinguished two types of syntax in animal communications.
Phonological syntax was defined by the capacity of the rules to combine
into sequences either meaningless sounds, or meaningful sounds that lose
their original meaning. GREEN et al. (1977) describe such compound calls
as having elemental openness. MARLER
defined lexical syntax as "the use
of compound signals which derive their meaning from the multiplexing of
the meaning of the components as used separately or in other combina-
SYNTACTIC STRUCTURES IN VOCALIZATIONS OF CAPUCHIN MONKEYS 73
tions." These were described by GREEN et al. (1977) as having secondary
openness. These two levels of syntax are analogous to the duality of
patterning characteristic of human language (HOCKETT,
1960; ALTMANN,
1967): Meaningless phonemes combine to form meaningful morphemes
which are organized to form sentences JAKOBSON
& HALLE, 1956).
The problem with these distinctions for the study of zoosyntactics is
that "meaning" is not defined. If we accept that vocal signals can have
internal referents and thus are audible expressions of the assessment or
internal state of the animal (GREEN & MARLER, 1979: 143; SMITH,
1981),
and define the "meaning" of a vocal signal to the signaler as the broad-
cast information of the signal, we can draw analogies between animal
and human communication systems. Note that this use of the term
"meaning" is not directly equivalent to SMITH'S
(1963, 1965) concept of
message (WILEY &
RICHARDS, 1982). An alternative is to define meaning
pragmatically in terms of the observable behavior of conspecific
receivers: The meaning of a call to a receiver is its change in state on
receiving the signal (CHERRY, 1966: 307), and can be assessed by
studying behavioral responses (MARLER, 1961; ALTMANN, 1967; GREEN &
MARLER, 1979: 140). However, the signal received by a receiver is not
identical to the signal transmitted by the signaler so the original meaning
to the signaler must be defined as "the intended
change of state of the
perceiver" (CHERRY,
1966, italics in the original), a change that is dif-
ficult to quantify. This study examined the covariance between vocaliza-
tions and observable social circumstances, and presumed the internal
state of the animal. Therefore the first approach is used here, and
meaning is equated with the broadcast information of the signal.
If the call types that make up a compound call .re never given except
in a compound call then the syntax governing signal selection is
phonological. The vocal sequences of Callicebus monkeys possibly
represent the fullest elaboration of phonological syntax in non-human
primates. Like phonemes, most Callicebus phrases occur only in
sequences, not alone. Like the rules concatenating phonemes into
morphemes, the rules generating Callicebus
sequences have the capacity
to recombine the phrases into a potentially limitless array of sequences.
And like morphemes, the resulting sequences have a meaning ( = occur in
specific social contexts) that depends on the order of phrases in the
sequence. Callicebus monkeys use different sequences in different con-
texts, respond differently to different sequences, and can be shown to
distinguish sequences solely on the basis of order of phrases (ROBINSON,
1979, 1981). There are differences however between the vocal sequences
74 JOHN G. ROBINSON
of this monkey and human morphemes. Callicebus
phrases are themselves
repetitions of simpler units, the individual calls, and the mechanism
generating both calls within phrases and phrases within sequences
appears to be hierarchically arranged.
Even if animals give the same call type both alone and in sequences,
the syntax of compound calls or sequences of vocalizations would be con-
sidered phonological if the single call is given in social circumstances
unrelated to those in which the sequence or compound is given.
Cebus olivaceus
compound calls, on the other hand, fit the definition of
lexical syntax. The compound calls occur in social circumstances inter-
mediate between those of the constituent calls. SNOWDON (1982, pers.
comm.) suggests that compound call express an "average" of two
internal states (at least when those states can be represented as falling on
a single continuum). This lexical syntax might be analagous to two types
of syntactic rules in linguistic theory: (a) Lexical rules that generate a lex-
icon, dictionary, or list of words. Lexical rules generate combinations of
morphemes, and form new words. For instance, "green" and "house"
can be combined to form "greenhouse". As defined here, "green
house", in contrast, is not syntactically organized, but is an example of
one sound acting as a contextual modifier for a second. (b) Grammatical
rules that generate sentences. Grammatical rules generate combinations
of words to form sentences, such as "The house was green".
Cebus compound calls are generated by syntactic rules closely
analogous to lexical rules, specifically "compounding rules" that com-
bine two lexical entries to form a third. The lexical entries in human
language each have meaning, as does the new compound (see LEECH,
1974). The vocal types in Cebus vocalizations are each given in specified
social circumstances, as is the compound call. New lexical entries retain
the semantic specification of the constituent entries. Cebus compound
calls are produced in circumstances intermediate between those of the
constituent calls. Linguistic lexical rules are only "partially productive"
in that they will not produce an infinite array of new lexical entries. Cebus
syntactic rules will produce only a finite number of compound calls, for
they are limited by a finite set of internal state axes.
The syntactic rules of capuchin monkeys are not analogous to gram-
matical rules generating sentences. First, as stated by BYRNE
(1982), the
"simple addition of meaning", as represented by the compound calls
here, "is a long way from the hierarchical complexity of grammar in
speech." Second, as discussed above, these syntactic rules are not open
in their ability to include additional referents and combine them together
SYNTACTIC STRUCTURES IN VOCALIZATIONS OF CAPUCHIN MONKEYS 75
in an infinite number of new ways. And third, both the simple and com-
pound calls of these monkeys have similar relations to their referents.
This is not true of words and sentences. QUINE (1960) points out that
words refer to or designate external or internal referents, but sentences
do not have referents, they are identified by their logical truth or
equivalence. For example, the words "house" and "green" designate
specific referents, but the sentence "The house was green" (but is now
red) does not refer to green houses.
There is a duality in the patterning of the syntax in Cebus
vocalizations.
Syllables, which never occur alone, are combined to form calls, which are
given in specific circumstances. Calls are combined to form compound
calls, which occur in circumstances intermediate between those of the
constituent calls. This second class of syntactic rules is analogous to the
lexical rules of human language, which generate combinations of
meaningful morphemes into words. But there is as yet no evidence of
syntactic rules in Cebus vocal communications analogous to the gram-
matical rules of human language.
Summary
Capuchin monkeys, Cebus
olivaceus,
combine different types of calls to form compound
calls. These compound calls are syntactically organized (i.e., there is a predictable
ordering of call types in the compound call). The analysis of the syntax of animal com-
munication necessarily includes: (a) Accurate classification of types of calls, and a
demonstration that calls that occur both alone and in compound calls are structurally
similar. (b) Description of the syntactic
rules generating
compound calls or call sequences,
and classification of types of compound calls or call sequences. (c) Examination of the
social circumstances
or contexts in which both single and compound calls occur. I followed
these steps using a large sample of Cebus
olivaceus
calls, recorded
in riparian
gallery forest
in central Venezuela.
The calls were initially screened using a real-time spectrum analyzer, and a group of
structurally
related call types was selected for further analysis. Temporal and frequency
characteristics of 868 of these calls were measured from sound spectrograms
produced
on
a sound sonograph. Call classification
involved first defining a very large number of possi-
ble call types on the basis of these characteristics,
and then using a discriminant
analysis to
identify which types should be lumped together. A stepwise procedure
was followed, using
only calls which occurred singly, until five call types (squaws, chirps, trills, whistles,
screams) were statistically separable. Two of these types (trills, whistles) showed con-
siderable within-type variation and were further subdivided into four variants each.
Discriminant analysis was then used to demonstrate structurally
similarity between calls
produced singly and those in compound calls.
Social circumstances were defined using similarities
in the vocalizer's actions, arousal,
orientation to and distance from the presumed receiver. Use of call types, when given
singly, covaried predictably
with social circumstances and presumably with the internal
state of the vocalizer. Different calls expressed different internal states on a continuum
from contact-seeking to contact-avoiding. Use of trill and whitle variants also covaried
with social circumstances. Different variants expressed different states on a continuum
76 JOHN G. ROBINSON
from affiliation or submission to aggression. Combinations of internal states, in theory,
might be expressed as intergradations or intermediates between different call types.
However, such intermediate states are often coded syntactically.
Syntactically organized compound calls accounted for 38% of the total sample. The
distribution of social circumstances in which compound calls are given was intermediate
between the distributions of the constituent call types, which presumably indicates an
intermediate internal state. Compound calls are generated by syntactic rules closely
analogous to lexical rules of human language. Specifically they act like compounding rules
that combine two lexical entries to form a third. They are not analogous to the gram-
matical rules that generate human sentences.
References
ALTMANN,
S. A. (1967). The structure of primate social communication. - In: Social
communication among primates (S. A. ALTMANN,
ed.). Univ. of Chicago Press,
Chicago.
BYRNE, R. W. (1982). Primate vocalizations: structural and functional approaches to
understanding. - Behaviour 80, p. 241-258.
CHENEY, D. L. & SEYFARTH, R. M. (1982). How vervet monkeys perceive their grunts:
field playback experiments. - Anim. Behav. 30, p. 739-751.
CHERRY, C. (1966). On human communication: A review, a survey, and a criticism. -
M.I.T. Press, Cambridge, Mass.
CHOMSKY, N. (1957). Syntactic structures. - Mouton, The Hague.
- (1965). Aspects of the theory of syntax. - M.I.T. Press, Cambridge, Mass.
-- (1972). Language and mind. 2nd ed. - Harcourt, Brace &Javanovich, New York.
CLEVELAND, J. & SNOWDON, C. T. (1982). The complex vocal repertoire of the adult
cotton-top tamarin (Saguinus oedipus oedipus). - Z. Tierpsychol. 58, p. 231-270.
DEMARS, C. & GOUSTARD, M. (1972). Structure et regles de deroulement des emissions
sonores des Hylobates (Hylobates concolor). - Bull. biol. Fr. Belg., 106, p. 177-191.
DEPUTTE, B. L. (1982). Duetting in male and female songs in the white-cheeked gibbons
(Hylobates
concolor
leucogenys).
- In: Primate communication (C. T. SNOWDON, C. H.
BROWN & M. R. PETERSEN, eds). Cambridge Univ. Press, Cambridge.
DESCARTES, R. (1952). Discourse on the method. - In: Great books of the western world
(R. M. HUTCHINS, ed.). Univ. of Chicago Press, Chicago. (Originally published
1637).
EISENBERG, J. F. (1974). The function and motivational basis of hystricomorph vocaliza-
tions. - Symp. zool. Soc. Lond. 34, p. 211-247.
-- (1976). Communication mechanisms and social integration in the black spider
monkey, Atelesfusciceps
robustus,
and related species. - Smithson. Contrib. Zool. 213,
p. 1-108.
GREEN, S. (1975). Variation of vocal pattern with social situation in theJapanese monkey
(Macaca fuscata): a field study. - In: Primate behavior, Vol. 4. Academic Press,
New York.
-- & MARLER, P. (1979). The analysis of animal communication. - In: Handbook of
behavioral neurobiology, Vol. 3 (P. MARLER &
J. G. VANDENBERGH, eds). Plenum
Press, New York.
- et al. (1977). Comparative aspects of vocal signals including speech - Group report.
- In: Recognition of complex acoustic signals: report of the Dahlem workshop (T.
H. BULLOCK, ed.). Abakon Verlagsgesellschaft, Berlin.
HOCKETT, C. F. (1960). The origin of speech. - Sci. Amer. 203, p. 86-96.
HONACKI,J. H., KINMAN, K. E. & KOEPPL, J. W. (1982). Mammal species of the world:
a taxonomic and geographic reference. - Allen Press and the Association of Sys-
tematic Collections, Lawrence, Kansas.
SYNTACTIC STRUCTURES IN VOCALIZATIONS OF CAPUCHIN MONKEYS 77
HUXLEY, T. H. (1893). Collected essays. - Macmillan, London.
JAMES,
W. (1890). Principles of psychology. - Henry Holt, New York.
JAKOBSON, R. & HALLE,
M. (1956). Fundamentals of language. - Mouton, The Hague.
LACHENBRUCH, P. (1975). Discriminant analysis. - Hafner, New York.
LEECH,
G. (1974). Semantics. - Penguin, Harmondsworth.
MAIER, V. (1982). Acoustic communication in the Guinea Fowl (Numida meleagris):
structure and use of vocalizations, and the principles of message coding. - Z.
Tierpsychol. 59, p. 29-83.
MARLER, P. (1961). The logical analysis of animal communication. -J. Theor. Biol. 1,
p. 295-317.
- (1977). The structure of animal communication sounds. - In: Recognition of
complex acoustic signals: report of the Dahlem workshop (T. H. BULLOCK,
ed.).
Abakon Verlagsgesellschaft, Berlin.
- & HOBBETT,
L. (1975). Individuality in a long range vocalization of wild chimpan-
zees. - Z. Tierpsychol. 38, p. 97-109.
MCLANAHAN,
E. B. & GREEN,
K. M. (1978). The vocal repertoire and an analysis of the
contexts of vocalizations in Leontopithecus
rosalia. - In: Biology and conservation of
the Callitrichidae (D. G. KLEIMAN,
ed.). Smithsonian, Washington D.C.
MORTON,
E. S. (1977). On the occurrence and significance of motivational-structural
rules in some bird and mammal sounds. - Amer. Nat. 111, p. 855-869.
- (1982). Grading, discreteness, redundancy and motivational-structural rules. -
In: Acoustic communication in birds, Vol. 1 (D. E. KROODSMA,
H. OUELLET
& E. H.
MILLER,
eds). Academic, New York.
MOYNIHAN, M. (1966). Communication in the titi monkey Callicebus. - J. Zool.,
Lond. 150, p. 77-127.
NIE, N. H., HULL, C. H., JENKINS, J. G., STEINBRENNER, K. & BENT, D. H. (1975).
SPSS. - McGraw Hill, New York.
NEWMAN, J. D., LIEBLICH, A. K., TALMAGE-RIGGS, G. & SYMMES,
D. (1978). Syllable
classification and sequencing in twitter calls of squirrel monkeys (Saimiri sciureus).
Z. Tierpsychol. 47, p. 77-88.
QUINE, W. V. 0. (1960). Word and object. - M.I.T. Press, Cambridge, Mass.
ROBINSON, J. G. (1979). An analysis of the organization of vocal communication in the titi
monkey Callicebus moloch. - Z. Tierpsychol. 49, p. 381-405.
-- (1981). Vocal regulation of inter- and intragroup spacing in the titi monkey Callicebus
moloch. - Primates 22, p. 161-172.
-- (1982). Vocal systems regulating within-group spacing. - In: Primate communica-
tion (C. T. SNOWDON,
C. H. BROWN
& M. R. PETERSEN,
eds). Cambridge Univ.
Press, Cambridge.
RYLE, G. (1949). The concept of mind. - Hutchinson, London.
SEBEOK, T. A. (1972). Perspectives in zoosemiotics. - Mouton, The Hague.
SEYFARTH, R. M., CHENEY, D. L. & MARLER, P. (1980a). Monkey response to three
different alarm calls: Evidence of predator classification and semantic communica-
tion. - Science 210, p. 801-803.
,-- & -- (1980b). Vervet monkey alarm calls: semantic communication in a
free-ranging primate. - Anim. Behav. 28, p. 1070-1094.
SMITH, H. J., NEWMAN, J. D. & SYMMES, D. (1982). Vocal concomitants of affiliative
behavior in squirrel monkeys. - In: Primate communication (C. T. SNOWDON,
C. H. BROWN & M. R. PETERSEN,
eds). Cambridge Univ. Press, Cambridge.
-- - -,
- HOFFMAN, H. J. & FETTERLY,
K. (1982). Statistical discrimination
among vocalizations of individual squirrel monkeys (Saimiri sciureus). - Folia
primatol. 37, p. 267-279.
SMITH, W. J. (1963). Vocal communication of information in birds. - Amer. Nat. 97,
p. 117-125.
78 JOHN G. ROBINSON
-- (1977). The behavior of communicating. - Harvard Univ. Press, Cambridge,
Mass.
-- (1981). Referents of animal communication. - Anim. Behav. 29, p. 1273-1275.
SNOWDON,
C. T. (1982). Linguistic and psycholinguistic approaches to primate com-
munication. - In: Primate communication (C. T. SNOWDON,
C. H. BROWN
& M.
R. PETERSEN,
eds). Cambridge Univ. Press, Cambridge.
WILEY, R. H. (1975). Multidimensional variation in an avian display: implications for
social communication. - Science 190, 482-483.
- (1976). Communication and spatial relationships in a colony of common grackles.
- Anim. Behav. 24, p. 570-584.
-- & RICHARDS, D. G. (1982). Adaptations for acoustic communication in birds: sound
transmission and signal detection. - In: Acoustic communication in birds, Vol. 1
(D. E. KROODSMA,
H. OUELLET
& E. H. MILLER,
eds). Academic, New York.
WINTER, P., PLOOG,
D. & LATTA, J. (1966). Vocal repertoire of the squirrel monkey
(Saimiri sciureus), its analysis and significance. - Exp. Brain Res. 1, p. 359-384.
Zusammenfassung
Kapuzineraffen, Cebus
olivaceus,
verbinden verschiedene Rufarten um Mischrufe zu
bilden. Diese Mischrufe sind syntaktisch
organisiert (d.h., es gibt eine Ordnung der Ruf-
arten im Mischruf, die vorhersagbar
ist). Die Analyse der Syntax der Tierverstandigung
beinhaltet: (a) Sorgfaltige Einteilung der Rufarten und eine anschauliche Darstellung,
dass Rufe die allein und in Mischrufen
vorkommen, strukturmassig
ahnlich sind. (b) Be-
schreibung der syntaktischen Regeln auf denen Mischrufe oder Rufsequenzen aufbauen
und Einteilung der Arten von Mischrufen und Rufsequenzen. (c) Untersuchung der
sozialen Verhailtnisse
oder Zusammenhange unter denen sich beide, Einzel-und Misch-
rufe, ereignen. Ich habe, diesen Schritten folgend, eine grosse Anzahl von Cebus olivaceus
Rufen in an Wasser gelegenen Galeriewaldern
in Mittelvenezuela aufgezeichnet.
Die Rufe wurden anfanglich mit Hilfe einer Echtzeit-Spektrumanalyse
sortiert, und
eine Gruppe strukturmissig verwandter Rufarten wurde zur weiteren Analyse ausge-
wahlt. Zeit- und Frequenzcharakteristiken
von 868 dieser Rufe wurden von Schallspek-
trogrammen
gemessen, die auf einem Schallsonographen
erzeugt wurden. Die Rufeintei-
lung brachte
zuerst die genaue Definition einer sehr grossen Zahl mdglicher
Rufarten auf
der Grundlage dieser Charakteristiken
mit sich, und dann eine Diskriminantenanalyse,
um festzustellen, welche Arten zusammengefasst werden sollten. Ein schrittweises Ver-
fahren wurde angewandt, wobei nur Rufe die einzeln auftraten benutzt wurden, bis
fiinf Rufarten (squaws, chirps, trills, whistles, screams) statistisch trennbar waren. Zwei
dieser Rufarten (trills, whistles) zeigten betrichtliche Abweichungen innerhalb ihrer Art
und wurden weiter,
jede in vier Varianten, unterteilt. Ich habe dann die Diskriminanten-
analyse benutzt um strukturelle
Ahnlichkeiten
zwischen Rufen die einzeln und in Misch-
rufen auftraten aufzuzeigen.
Soziale Verhiltnisse wurden definiert, indem Ahnlichkeiten in des "Sprechers"
Handlungen, seiner Erregtheit, seiner Orientierung und Entfernung zu einem
mutmasslichen Empfinger benutzt wurden. Die Anwendung von einzelnen Rufarten
veranderte sich vorhersagbar mit den sozialen Verhaltnissen und vermutlich mit dem
inneren Zustand des Sprechers. Verschiedenen Rufe driickten verschiedene innere
Zustande in einem Kontinuum von kontaktsuchend bis kontaktvermeidend aus. Der
Gebrauch von Triller- (trill) und Pfeif- (whistle) Varianten verinderte sich ebenfalls mit
den sozialen Verhiltnissen. Verschiedene Varianten druickten
verschiedene Zustande in
einem Kontinuum von Angliederung oder Unterwerfung bis zu Aggression aus. Verbin-
dungen innerer Zustainde diirften theoretisch als Zwischenstufen oder Mittellagen
SYNTACTIC STRUCTURES IN VOCALIZATIONS OF CAPUCHIN MONKEYS 79
zwischen verschiedenen Rufarten ausgedriickt werden. Solche Zwischenzustinde sind
jedoch oft syntaktisch
verschliisselt.
Syntaktisch aufgebaute Mischrufe machten 38% in der gesamten Untersuchung aus.
Die Verteilung sozialer Umstinde in denen Mischrufe benutzt werden lag in der Mitte
zwischen den Verteilungen der Grundrufarten, was einen inneren Mittelzustand ver-
muten lasst. Mischrufe werden mit syntaktischen
Regeln erzeugt, die annahernd analog
zu den lexikalischen
Regeln der menschlichen
Sprache
sind. Genaugenommen wirken sie
wie zusammengesetzte Regeln die zwei lexikalische
Eingaben kombinieren um eine dritte
zu bilden. Sie sind den grammatischen Regeln die menschliche Satze erzeugen nicht
analog.
... The wedge-capped capuchin monkeys (Cebusolivaceus) produce different types of calls in different social circumstances (Robinson 1984). This species of capuchin monkeys is also able to produce these different types of calls in a compound structure of doublets, triplets and sometimes quadruplets. ...
... The ordering of these compound calls is predictable, and they are 'syntactically' structured. Following Robinson (1984), we refer to this structure as 'syntax'. This syntactic structure is not the same as the human syntax, which is demystified with "organized relations among separate sounds and novel use of different combinations of sounds," but instead consists of a finite number of signals that "do not symbolize or refer to objects in the external environment" and cannot be syntactically organized in the same way humans use syntax (Robinson 1984, 47). ...
... Taking this definition of syntax, the syntactically structured compound calls are not uncommon for the capuchin monkeys -around 38% of the total sample in Robinson's (1984) study. There is a duality of patterning seen in the syntax of Cebus vocalizations¬ -syllables are combined to form calls made in specific circumstances. ...
Article
Full-text available
The paper is a critique of the existence of protolanguage based on some personal reasoning given the findings of previous research. This paper focuses on the nature of semantic compositionality and its existence (and therefore manifestations) in animal communication systems as evidence of the existence of protolanguage. This compositional state that started to color human language has paved the way to hierarchical syntax and thus has helped in the emergence of a recursive, fully complex language. On the other hand, non-human animal communication systems, including those examined in this paper, have not managed to progress beyond the holophrastic state, and thus remained highly confined and unproductive. This is explicated by the observation that these systems do not employ discrete meaningful units that can be used recursively in different linguistic contexts, a necessary condition for a system of communication to be compositional. This is interesting in the study of language evolution as it could suggest that human language could not have evolved from a rudimentary, intermediate stage called protolanguage. Speculating about the existence of protolanguage subsumes convergent evolution (e.g. songbirds). The lack of semantic compositionality in non-human communication system suggests that convergent evolution is hard to prove, which puts the existence of protolanguage on the line. This thesis is, however, far from being established and requires a lot of further research to prove its validity. Received: 21 September 2020 / Accepted: 3 November 2020 / Published: 17 January 2021
... Furthermore, numerous studies have suggested that some communication systems of wild and some domestic bird, whale, and primate taxa exhibit syntactic properties, i.e., rules for combining different structures or structure types (Berwick et al. 2011Griesser et al. 2018;Suzuki et al. 2018 Ficken et al. 2000), tits (Paridae; Hailman et al. 1985), chickadees (Poecile;Freeberg 2008;Freeberg and Lucas 2012), true finches (Fringillidae; Riebel and Slater 2003), New World sparrows (Passerellidae; Rose et al. 2004), estrildid finches (Estrildidae; Abe et al. 2011;Beckers et al. 2012;Honda and Okanoya 1999;Katahira et al. 2007;Leonardo 2002;Sturdy et al. 1999;van Heijningen et al. 2009), nightingales and relatives (Luscinia; Hultsch and Todt 1989;Hultsch 1996, 1998), starlings (Sturnidae;Gentner et al. 2006), gulls (Laridae;Beer 1976), bats (Chiroptera; Bohn et al. 2013), mice (Chabout et al. 2015), mongooses (Herpestidae; Fitch 2012), humpback whales (Megaptera novaeangliae; Mercado et al. 2005;Payne and McVay 1971;Suzuki et al. 2006), tamarins (Saguinus;Fitch and Hauser 2004), capuchins (Cebinae;Robinson 1984), and gibbons (Hylobatidae; Clarke et al. 2006;Haimoff 1985;Inoue et al. 2017Inoue et al. , 2020aTerleph et al. 2018). Hurford (2012: 97) comments: "Despite serious underexploitation of combinatoriality, […] whalesong and much birdsong exhibit a hierarchically layered structure formally similar to the hierarchical structure of human syntax." ...
Preprint
Full-text available
Recent discoveries of semantic compositionality in Japanese tits have enlivened the discussions on the presence of this phenomenon in wild animal communication. Data on semantic compositionality in wild apes are lacking, even though language experiments with captive apes have demonstrated they are capable of semantic compositionality. In this paper, I revisit the study by Boesch [1991 (Hum Evol 6:81–89] who investigated drumming sequences by an alpha male in a chimpanzee (Pan troglodytes) community in the Taï National Park, Côte d’Ivoire. A reanalysis of the data reveals that the alpha male produced semantically compositional combined messages of travel direction change and resting period initiation. Unlike the Japanese tits, the elements of the compositional expression were not simply juxtaposed but displayed structural reduction, while one of the two elements in the expression coded the meanings of both elements. These processes show relative resemblance to blending and fusion in human languages. Also unlike the tits, the elements of the compositional expression did not have a fixed order, although there was a fixed distribution of drumming events across the trees used for drumming. Because the elements of the expression appear to carry verb-like meanings, the compositional expression also resembles simple verb-verb constructions and short paratactic combinations of two clauses found across languages. In conclusion, the reanalysis suggests that semantic compositionality and phenomena resembling paratactic combinations of two clauses might have been present in the communication of the last common ancestor of chimpanzees and humans, not necessarily in the vocal modality.
Thesis
The presence of musical behavior in several phylogenetically distinct taxa in the animal kingdom and throughout every society in the human species is one of the greatest unsolved mysteries in biology. I have incorporated knowledge of human musical universals into a measurable working (rhythm absent) definition of melody: essentially variation and repetition of (tonal) intervals. I hypothesized that repetition and variation in melody have evolved as honest signals of underlying quality of precision and range (respectively) of distance gauging abilities. I predicted that melodic display would be positively correlated with remote targeting behaviors such as projectile locomotion (jumping, brachiation, or arm-swinging). Quantitative locomotor distributions and spectrographic vocal repertoires of primates were collected and the spectrogram analysis program, gramparse, was used to split, match, and score each vocalization type. Four linear least squares regression models of (percent) projectile locomotion vs melodic score (as measured by the number of repeated interval groups per vocalization) were run on both species level data and the corresponding phylogenetically controlled independent contrasts with and without control from additional behavioral and environmental variables. Percent projectile locomotion was statistically significant at the 0.01 level but failed to explain more than 20% of the variance in melodic display. This work is an important first step in both investigating possible selection pressures in a currently neglected corner of music origins research and, more generally, quantifying the formerly qualitative task of assessing musicality. The origins of music is one of the most perplexing mysteries in biology. Music is highly pervasive and, apparently, independently evolved in numerous taxa. There also appears to be strong evidence for specialized neurological circuitry for musical processing in both humans and other animals. And yet, from Darwin’s “mating call theory” (Darwin 1871) on, scientists have struggled with a way to structure compelling adaptationist solutions and most have conceded that the mystery remains unsolved or that existing solutions are highly problematic.
Preprint
Full-text available
I read with great interest the study by Leroux et al. [(2021) Anim Behav 179, 41–50] who investigated the nature of pant-hoot–food-call combinations in a community of wild chimpanzees (Pan troglodytes schweinfurthii) at the Budongo Conservation Field Station, Budongo Forest, Uganda. The authors propose, among others, that they reveal the first evidence that wild chimpanzees are able “to combine meaning-bearing units into larger structures” – i.e., that they are capable of semantic compositionality and, by extension, syntax. Their analysis represents an important addition to a growing body of research and discussions on communicational combinatoriality in wild primates and specifically apes, and, by extension, extinct hominins. Incidentally, I have recently published a paper in Animal Cognition in which I also suggested, based on a reanalysis of existing data, that wild chimpanzees can display semantic compositionality and syntax, i.e., are able to combine meaningful units [Gabrić (2021) Anim Cogn, online ahead of print]. In the present commentary, I argue that Leroux et al.’s (2021) interpretation of the data is ungrounded given that (1) unlike for food calls, there is currently very little if any indication in the scientific literature that pant-hoots have semantic content, i.e., are meaningful, (2) Leroux et al. (2021) did not investigate their a priori assumption that the observed pant-hoots are in fact meaningful/semantic, (3) they did not report on recipients’ behaviors in association with neither the individual nor combined calls, and (4) they did not compare the callers’ behaviours in association with the individual calls vs. combined calls. Since pant-hoots feature prominently in the chimpanzee vocal repertoire and the debate on their eventual meaningfulness/semanticity is still wide open, this represents a fine opportunity to revisit this issue in the context of Leroux et al.’s (2021) study. Their paper further raises several other less significant questions. Notwithstanding, their paper brings important novel insights into communicational combinatoriality in wild chimpanzees and supports the notion of using linguistic methods in wild animal communication research.
Article
Full-text available
Recent discoveries of semantic compositionality in Japanese tits have enlivened the discussions on the presence of this phenomenon in wild animal communication. Data on semantic compositionality in wild apes are lacking, even though language experiments with captive apes have demonstrated they are capable of semantic compositionality. In this paper, I revisit the study by Boesch (Hum. Evol. 6:81–89, 1991) who investigated drumming sequences by an alpha male in a chimpanzee ( Pan troglodytes ) community in the Taï National Park, Côte d’Ivoire. A reanalysis of the data reveals that the alpha male produced semantically compositional combined messages of travel direction change and resting period initiation. Unlike the Japanese tits, the elements of the compositional expression were not simply juxtaposed but displayed structural reduction, while one of the two elements in the expression coded the meanings of both elements. These processes show relative resemblance to blending and fusion in human languages. Also unlike the tits, the elements of the compositional expression did not have a fixed order, although there was a fixed distribution of drumming events across the trees used for drumming. Because the elements of the expression appear to carry verb-like meanings, the compositional expression also resembles simple verb-verb constructions and short paratactic combinations of two clauses found across languages. In conclusion, the reanalysis suggests that semantic compositionality and phenomena resembling paratactic combinations of two clauses might have been present in the communication of the last common ancestor of chimpanzees and humans, not necessarily in the vocal modality.
Article
Abstract: Human language is combinatorial: phonemes are grouped into syllables, syllables are grouped into words, and so on. The capacity for combinatorial processing is present, in different degrees, in some mammals and birds. We used vibrational insects, Enchenopa treehoppers, to test the hypothesis of basic combinatorial processing against two competing hypotheses: beginning rule (where the early signal portions play a stronger role in acceptability) and no ordering rule (where the order of signal elements plays no role in signal accept-ability). Enchenopa males use plant-borne vibrational signals that consist of a whine followed by pulses. We tested the above hypotheses with vibrational playback experiments in which we presented Enchenopa females with stimuli varying in signal element combinations. We monitored female responses to these playbacks with laser vibrometry. We found strong support for combinatorial processing in Enchenopa: in brief, females preferred natural-combination signals regardless of the beginning element and discriminated against reverse-order signals or individual elements. Finding support for the combinatorial rule hypothesis in insects suggests that this capability represents a common solution to the problems presented by complex communication. Keywords:communication, mate choice, signal processing, phonology, syntax, vibrational signaling
Thesis
This dissertation explores the social, behavioral, and habitat based influences on the evolution of elaborate vocalizations in primates. The goals of this project were to merge and test hypotheses derived from both the primate morphological origins literature and the musical behavior origins literature. Theoretical elements from the leaping and arboreal hypotheses from the primate origins literature were brought to bear on vocal display data from extant primates and on theories on the function of song in animals and music in humans. Specifically, I used computational statistical techniques to test the dual hypotheses that precision limb landing, as required in navigating complex arboreal topologies, could have not only shaped primate morphology, but also evolutionarily shaped their vocal signaling behaviors. I also tested other theories from the literature on both animal communication (e.g. the acoustic adaptation hypothesis) and human music (e.g. coalition signaling as well as group and pair bonding). I collected behavioral data, for as many species as possible, from the primary primate behavioral literature, including vocalization, locomotion, and socio-ecological regression control variables. Data were then analyzed in three separate substantive chapters. First, in chapter 2, spectrographic vocalizations were scored manually along ethnomusicologically universal acoustic parameters and subsequently reduced into a univariate acoustic reappearance diversity index [ARDI], reflecting call elaborateness. Second, in chapter 3, quantitative locomotion activity data (e.g. leaping and swinging percentages) were harvested from the positional behavior primary literature in order to both estimate its ancestral influence on morphology as well as to create better (non-binary) predictor variables for testing vocal display origin theories. Third, in chapter 4, I tested for my hypothesized co-evolutionary relationship while controlling for phylogeny. ARDI was highest in smaller, socially monogamous, and forest living groups, as well as positively associated with swinging and leaping. These results are consistent with the dual hypothesis that both elaborate vocal displays and vision changes in primates could have evolved as a result of increased demand for precision limb landing locomotion. In chapter 5, I discuss the limitations and drawbacks as well as the theoretical significance, scientific relevance, broader implications, and ideas for future research.
Article
Language has a complex and hierarchical structure, which includes context-free grammar and phrase embedding. However, there have been no reports of a phrase-embedding structure in animal vocal communication, although there are several reports of combinational animal sounds that reference different objects and emotions. The songs of male Mueller’s gibbons in our study area consist of two notes, “wa” and “oo,” and combinations of these are flexible. There are various types of note orders in their song phrases. When the animals sing songs by combining acoustic elements flexibly, a complicated syntax may emerge. When phrase “N” is inserted within another phrase “AB,” the generated phrase is shown as “ANB.” We named this structure a phrase-inserting structure. If the phrase “N” itself has an inserting structure elsewhere, “ANB” is shown as “AABB.” This structure is considered a phrase-embedding structure, as defined by Abe and Watanabe (Nat Neurosci 14:1067–1074, 2011). We hypothesized that there is a phrase-inserting structure in male Mueller’s gibbons’ songs. We analyzed 70 songs by a single male and found note orders that suggested a phrase-inserting structure. Among them, there were four examples of phrases with structure within phrases, suggesting phrase embedding. Phrase inserting might be considered a sort of precursory level to recursion. Although our sample size was small, our results show that male Mueller’s gibbons may be able to produce recursive sequences. This provides an important insight into the study of language evolution.
Conference Paper
Full-text available
Every wildlife species has a language or 'signal' for communication both within and without. In addition, the interaction between wildlife species and their natural habitat is usually a balanced system. But it is always the entry of man and his insatiable exploitative appetite into this union that is the cause of problems in this natural ecosystem. The case of land use and land use change comes into focus particularly the ones that lead to wildlife habitat fragmentation. A lot of pioneer ecological studies have looked at the effect of landscape modification on the life of wildlife. Very few, if any, have attempted to assess the inaudible voices from the wildlife habitat as a response to the said modifications. This paper used Narration Review (NR) method to analyze previous elephants' behavior observation study projects in Tsavo and Amboseli ecosystems in Kenya. A total of five papers were identified for sysnthesis.All the five papers discussed the response of elephants in terms of usage of wildlife crossing structures (CS) that were installed during the construction of Mombasa-Nairobi leg of Standard Gauge Railway (SGR) project that was commissioned in 2017. It was established that understanding the behavior of elephants in their local rangeland can help position crossings for landscape connectivity. Secondly, the SGR project's operation has come with an increasing trend of elephants' mortality in some of its corridor locations. Interestingly however, it was reported that elephant's high cognitive intelligence facilitates their faster learning of use of crossing underpasses compared to the rate of response by other wildlife species. The paper recommends for the initiation of more animals' behavior observation projects to help in development planning decision on their ethical treatments. This study contributes to the entrenchment of sustainability thinking in wildlife conservation efforts in the face of piquing demand for infrastructure development, especially in the sub Saharan Africa region which still has a huge gap.
Article
Full-text available
Investigations of the display behavior of tyrannid flycatchers suggest that the meaning of a display which is used in different contexts may be different in each context. In the species studied vocalizations seem to be more important than visible displays, and most of the vocalizations are variable in a number of aspects. It seems possible that the different modes of variation of a display are capable of modifying considerably the information content of that display.
Article
Full-text available
Reviews the debate over the kinds of referents that signals can have and how this confusion hampers understanding of the semantic contributions of signals to interactional behavior. This can be alleviated by (1) defining the behavioral, situational, or other correlates in each signal and the extent to which they change; (2) outlining "external" referents; (3) recognizing that there is no justification for excluding behavioral referents; (4) realizing that a signal may have several referents; and (5) distinguishing among different sources of information. (8 ref) (PsycINFO Database Record (c) 2012 APA, all rights reserved)
Chapter
In the pattern classification approach to fault diagnosis outlined in Chapter 3, it was described how the dimensionality reduction of the feature extraction step can be a key factor in reducing the misclassification rate when a pattern classification system is applied to new data (data independent of the training set). The dimensionality reduction is especially important when the dimensionality of the observation space is large while the numbers of observations in the classes are relatively small. A PCA approach to dimensionality reduction was discussed in the previous chapter. Although PCA contains certain optimality properties in terms of fault detection, it is not as well-suited for fault diagnosis because it does not take into account the information between the classes when determining the lower dimensional representation. Fisher Discriminant Analysis (FDA), a dimensionality reduction technique that has been extensively studied in the pattern classification literature, takes into account the information between the classes and has advantages over PCA for fault diagnosis.
Article
An analysis of song-spread displays performed in social communication by carib grackles, Quiscalus lugubris, revealed that two components of this motor coordination, beak elevation and wing elevation, varied independently. Variation in each depended strongly on the performer's social context, an indication that these components simultaneously reflected different behavioral tendencies of the performer. Independent variation in the elements of a display can permit transmission of more information per display about the performer's behavioral state.
Article
This paper describes a concept, not altogether new but largely neglected, that should lead to a greater understanding of the information contained in certain classes of vocal communication signals of birds and mammals. The concept is based on empirical data, first pointed out by Collias (1960, p. 382), showing that natural selection has resulted in the structural convergence of many animal sounds used in "hostile" and "friendly" contexts. Simply stated, birds and mammals use harsh, relatively low-frequency sounds when hostile and higherfrequency, more pure tonelike sounds when frightened, appeasing, or approaching in a friendly manner. Thus, there appears to be a general relationship between the physical structures of sounds and the motivation underlying their use. I hope to develop the idea that this relationship has had a far greater influence on the evolution of animal communication systems than has hitherto been discussed. I will discuss the idea that there exist motivation-structural rules (MS) governing the physical structure of close contact sounds in animal communication systems. The greatest value of the MS concept is that it provides the opportunity to compare the evolution of vocal communication in any species against an abstract concept. The adaptive nature of communication systems against varying backgrounds of environment, social system, and competition will appear in clear relief.
Article
Research on primate vocal behaviour has shown it to be varied, complex and intricate; selected developments of particular interest are reviewed, and the methodological advances which allowed them are described. Primate repertoires are extensively graded, but this continuous variation may be perceived discretely by the animals: some species at least show categorical perception of their calls, which means that the lists of "calls" in the primate literature are largely misleading. In addition, some species show complex patterning which has yet to be properly understood. A rudimentary form of syntax and simple compounding of call meanings has been shown. It is now known that calls can convey population dialect and individual identity, and in certain cases can "refer" to classes of objects in a fashion analogous to that of words in speech. Calls normally give information on the caller's location, but may be specialised to avoid doing so or to do so better, where these are advantageous. Specialisation is also shown in perception, with neurally lateralised apparatus for processing of conspecifics' calls. Long distance propagation of certain calls is enhanced both by signal form and time and manner of delivery. Such calls are well suited for experimental study, and are now among the best understood. Function turns out to be more labile than form in evolution of vocalisations, with homologous calls serving different functions in close relatives; in addition, calls apparently identical in form may mediate several unrelated functions in a single species. Long-range calls are employed especially to coordinate movements where groups are dispersed in areas of low visibility, and in various different ways as components of intergroup spacing systems.