The semantics of Cebus olivaceus alarm calls : object designation and attribution


Material Information

The semantics of Cebus olivaceus alarm calls : object designation and attribution
Physical Description:
Norris, Jeffrey Copeland, 1950-
Publication Date:

Record Information

Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 27772003
oclc - 26498495
System ID:

Table of Contents
    Title Page
        Page i
        Page ii
    Table of Contents
        Page iii
        Page iv
    List of Tables
        Page v
    List of Figures
        Page vi
        Page vii
        Page viii
    Chapter 1. Introduction
        Page 1
        Page 2
        Page 3
        Page 4
        Page 5
        Page 6
        Page 7
        Page 8
        Page 9
        Page 10
        Page 11
        Page 12
        Page 13
        Page 14
        Page 15
        Page 16
        Page 17
        Page 18
        Page 19
        Page 20
        Page 21
        Page 22
        Page 23
        Page 24
        Page 25
        Page 26
        Page 27
        Page 28
        Page 29
        Page 30
        Page 31
        Page 32
        Page 33
        Page 34
        Page 35
        Page 36
        Page 37
        Page 38
        Page 39
    Chapter 2. Vocal responses to predators by cebus olivaceus
        Page 40
        Page 41
        Page 42
        Page 43
        Page 44
        Page 45
        Page 46
        Page 47
        Page 48
        Page 49
        Page 50
        Page 51
        Page 52
        Page 53
        Page 54
        Page 55
        Page 56
        Page 57
    Chapter 3. Vocal responses to released snakes
        Page 58
        Page 59
        Page 60
        Page 61
        Page 62
        Page 63
        Page 64
        Page 65
        Page 66
        Page 67
        Page 68
        Page 69
        Page 70
        Page 71
        Page 72
        Page 73
        Page 74
        Page 75
        Page 76
        Page 77
        Page 78
    Chapter 4. Responses to alarm call playbacks
        Page 79
        Page 80
        Page 81
        Page 82
        Page 83
        Page 84
        Page 85
        Page 86
        Page 87
        Page 88
        Page 89
        Page 90
        Page 91
        Page 92
        Page 93
        Page 94
        Page 95
        Page 96
        Page 97
        Page 98
        Page 99
        Page 100
        Page 101
    Chapter 5. The acoustics of cebus olivaceus alarm calls
        Page 102
        Page 103
        Page 104
        Page 105
        Page 106
        Page 107
        Page 108
        Page 109
        Page 110
        Page 111
        Page 112
        Page 113
        Page 114
        Page 115
        Page 116
        Page 117
        Page 118
        Page 119
        Page 120
        Page 121
        Page 122
        Page 123
        Page 124
        Page 125
        Page 126
        Page 127
        Page 128
        Page 129
        Page 130
        Page 131
        Page 132
        Page 133
        Page 134
        Page 135
        Page 136
        Page 137
        Page 138
        Page 139
        Page 140
        Page 141
        Page 142
        Page 143
        Page 144
        Page 145
        Page 146
        Page 147
        Page 148
        Page 149
        Page 150
        Page 151
        Page 152
        Page 153
        Page 154
        Page 155
        Page 156
    Chapter 6. Summary and conclusions
        Page 157
        Page 158
        Page 159
        Page 160
        Page 161
        Page 162
        Page 163
        Page 164
        Page 165
        Page 166
        Page 167
    Appendix A. Narrowband spectrograms of GRRAHS to a 0.68 m BOA
        Page 168
        Page 169
        Page 170
        Page 171
    Appendix B. Narrowband spectrograms of all GRRAH variants
        Page 172
        Page 173
        Page 174
        Page 175
        Page 176
        Page 177
    Appendix C. Definitions of acoustic variables
        Page 178
        Page 179
    Biographical sketch
        Page 180
        Page 181
        Page 182
Full Text






This research could not have been completed without the help of many individuals. Drs. John Robinson, John Eisenberg and Jay Whitehead provided guidance and assistance throughout, for which I offer my deepest gratitude. Sr. Tomas Blohm kindly offered me his hospitality at the ranch. He has been steadfast in his support of this and many other research projects, and a good bit of what is known about neotropical ecology is in his debt. Dr. Bill Hardy generously allowed me to use the equipment in the Bioacoustics Lab of the Florida Museum of Natural History. Lastly, I would like to thank two wonderful women, my mother Mrs. J.D. Folsom and Kim Martin, for helping me through many hard times.



ACKNOWLEDGMENTS ...................................

TABLE OF CONTENTS .................................

LIST OF TABLES .................................... v

LIST OF FIGURES ................................... vi

ABSTRACT .......................................... vii


1 INTRODUCTION ................................ 1

Traditional vs. Modern Perspective
on Animal Communication ................... 1
Previous Studies of Cebus olivaceus
Communication ............................. 4
Linguistics and Animal Communication ........ 6 Research Goals and Rationale ................ 39

CEBUS OLIVACEUS ........................... 40

Methods and Materials ....................... 40
Responses of Cebus olivaceus to Predators ... 46 Further Investigations ...................... 56


Methods and Materials ....................... 58
Results ..................................... 61
Discussion .................................. 75


Methods and Materials ....................... 79
Results ..................................... 88
Discussion .................................. 93


AL.ARM4 CALLS................................. 102

Methods and Materials........................ 102
Results..................................... 116
Discussion.................................... 142

6 SUMMARY AND CONCLUSIONS........................ 157

REFERENCES..................................... .... 160


0.68 m BOA.................................... 168

VA.RIAN~TS...................................... 172


BIOGRAPHICAL SKETCH................................ 180





1 ANIMALS THAT ALARM CALL......................... 14


3 COMPOSITION OF MAIN GROUP....................... 43


5 CONTEXTS OF GRRAH VARIANTS...................... 68










15 DEFINING VARIABLES FOR GRRAHS.................. 136



VOCALIZATIONS................................ 150





1 STUDY AREA .................................. 41

VARIOUS THREATS .......................... 53


4 KEY TO GRRAH VARIANTS ............................ 66

SPECTROGRAM OF JM102 ..................... 109

1111. (FOURIER DERIVED) .................... 110

1111. (LPC DERIVED) .........................


VARIANT 13121 ............................ 118

VARIANT 13121 ............................ 120


13 CLASSIFICATION OF 59 GRRAHS .................... 133





Abstract of Dissertation Presented to the Graduate School of
the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy


Jeffrey Copeland Norris

December, 1990

Chairman: Dr. John F. Eisenberg Major Department: Forest Resources and Conservation (Wildlife and Range Sciences) Wedge-capped capuchin monkeys, Cebus olivaceus, give acoustically distinct alarm calls to different predators. The semantics of these alarm calls were studied in central Venezuela in three stages: 1) through recordings of Cebus olivaceus vocal response to various predators, 2) through release of boa constrictors of three sizes (small, medium, and large) and two quantities (one and two), and 3) through playback of resulting calls. Alarm calls to released snakes were categorized by acoustic features into 15 variants, 2 of which were used solely when a snake was on the ground, and a third when a snake was in a tree. The playback of the locational calls showed that upon hearing a call a capuchin looked in a particular direction, into trees or toward the ground, at the appropriate call. Capuchins use alarm calls


to not only designate objects but also to attribute qualities such as location to those objects.



The wedge-capped capuchin (Cebidae: -Cebus olivaceus=

Cebus nigrivittatus), also known as the weeper capuchin, is a medium sized neotropical primate inhabiting much of South America north of the the Amazon river (Eisenberg, 1989). As its appellation "weeper" implies, it is a notably vocal animal. Among its many calls are a series of alarm calls given to a variety of predators. These calls are the subject of this dissertation--under what circumstances they are used, their variability, and their meaning. This search for meaning is a difficult one for which I am best warned to remember the story of Captain Cook, recounted in Cherry's seminal work, On Human Communication (1978). When the famous explorer first saw a strange hopping animal he inquired of a native as to what it was. The local gentleman, of course, did not understand English and replied "kangaroo," meaning "I don't know."

Traditional vs. Modern Perspective on Animal Communication

The traditional view of animal communication,

universally held until only recently, is that there is a



fundamental dichotomy between language and animal communication. The difference in perspective involves both what is communicated and how it is done. Traditionally animals were believed to communicate only their motivational states: the signal indicates the sender's arousal or intent for action. According to this perspective, animal calls are affective. For example, an alarm call communicates fear. This is in contrast with what humans communicate, where symbols refer to, among other things, objects as well as emotions.

A referent is the designatum of the signal (Green and

Marler, 1979). Reference may be to either an internal state or to an external physical object. Communication about an object may take any of several forms: iconic, indexical, or symbolic. If variations in the object's physical form are transformed along some parallel acoustic dimension then the communication is iconic (Green and Marler, 1979). For example, an alarm call that increases in duration or amplitude in parallel with increased size of the predator is iconically representing the size of the predator. Indexical communication refers to an object through the simple expedient of pointing at it (Green and Marler, 1979). That is, the communicator refers to object by addressing it physically by pointing at it. In the context of an alarm call, a monkey would indexically communicate about a snake by pointing at it. Symbolic communication uses the signal


to represent the object, where the signal itself refers to the object (Cherry, 1978). The alarm call itself refers to the predator because the call designates the object through symbolic representation. It is conventionally held that animals cannot symbolically communicate about referents, internal or external, and that if they want to communicate about an object they must do it either by indexing the object, or by using some iconic signal. Altmann (1967) notes the power of combined affective and indexical communication and how prevalent it is among primates. For example, Menzel and Halperin (1975) show that a chimpanzee's walking gate communicates information to other animals about the quality of food sites.

Animal communication has traditionally been seen as fixed, stereotyped, and simple whereas language was perceived as complex, variant, and open. This perspective began to change with the pioneering investigations of bee communication by Karl von Frisch (1967) where it was shown that a bee could symbolically communicate the distance and direction from the hive to a food source. Subsequent investigations in a wide variety of taxa have shown that the traditional view was inadequate. There is mounting evidence that a new perspective describes the behavior of certain species better than the earlier theory. According to this new view, primates use signals to communicate about objects in a manner very similar to how we use words to identify


objects. I argue below that primates may not only be able to designate objects but also attribute qualities to those objects. Now we examine the previous studies of the wedgecapped capuchin to see how it fits into this framework.

Previous Studies of Cebus olivaceus Communication

Oppenheimer and Oppenheimer (1973) describe, in a brief study, eleven different calls from C. olivaceus living at Hato Masaquaral, Venezuela, the site of this study. They characterize one call, the grrah as being an inter-specific agonistic call. Of the 60 grrahs they recorded, 48 were directed at humans, 4 to overflying birds, and 2 at howler monkeys (Allouatta seniculus). The remaining 6 were used in unknown circumstances. If Qrrahs were directed at animals other than humans, they were often repeated and at variable intervals. If the call was directed at birds, the authors describe the monkeys as giving the call once then "dropping to lower branches and moving deeper inside the tree" (Oppenheimer and Oppenheimer, 1973, pg 422). They also note that grrahs were associated with other qrrahs 92% of the time.

Robinson (1982) described how three calls--huhs, hehs, and arrawhs, mediate spacing within a group. Arrawhs were given in two contexts. Loud arrawhs were given by an animal if it became separated from the group. Group members responded with more huhs and often replied with arrawhs.


Quiet arrawhs were given by an animal if it began lagging behind the group. Robinson concluded that arrawhs acted to reduce spacing between group members.

In a subsequent publication, Robinson (1984) identified five basic vocalizations that were often syntactically combined--chirps, trills, squaws, screams, and whistles. The acoustic parameters that he used to differentiate these social vocalizations are compared to alarm call acoustics in a later section. Robinson first showed that these five calls were used individually and their use varied predictably with social circumstance expressing a continuum of internal states ranging from contact seeking to contact avoidance. The combined calls were relatively common, comprising 38% of the 868 calls recorded. He concluded that "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" (Robinson, 1984, pg 76). This study will be considered more fully below in a discussion of how the syntactic structures of primate vocalizations : parallel in important ways those found in human language.

Before continuing I would like to mention a dilemma facing anyone trying to explain what is occurring when primates vocalize. As we shall see, many primates combine vocalizations; yet in my experience in both South America and Africa, monkeys most often give single, often apparently


repetitive vocalizations. If the single call is referring to an object, why is it given repetitiously? Correspondingly, if calls can refer to objects, to what do combined calls refer? These questions pose methodological problems of vocal classification and procedures for determining meaning of calls.

Linguistics and Animal Communication

Linguistics is an alternative explanatory system to the traditional model of animal communication (see Snowdon, 1982). Linguistics deals with one form of communication, language, at several levels including semantics, phonology, and syntax. Because linguistic considerations are essential to this project, I will define them, discuss how they apply to animal communication, and give examples of pertinent investigations.


Semantics is the study of meaning. In linguistics, meaning is acquired through denotations--how words are defined in a standard text. Words are also understood connotatively through the speakers conception of how a word is used in normal conversation. Words have referents. A word may symbolically refer to a physical object such as a car, whereas other words have internal, abstract referents, for example, anger. Green and Marler (1979) point out that the assessment of any referent is initially an internal


procedure, with the signaller processing sensory and cognitive inputs prior to the generation of a signal. I note this here because in subsequent discussions I will use the terms external referent and internal referent. It is important to remember that all signals possess internal reference to some degree, where neural inputs are transformed into vocal outputs. Objects of external referents are not simply named without several layers of cognitive processing before the signal is generated. Nevertheless, the word car refers to an external physical object whereas anger refers to an internal state.

The concept of meaning is central to linguistics and

communication. The theme of word and meaning is the subject of a pivotal work by Quine (1960). He notes that "meaning, supposedly is what a sentence shares with its translation; and translation at the present stage turns solely on correlation with non-verbal stimulation" (Quine, 1960, pg 32). It is the problem of translation of a language where no translators or dictionaries provide denotative meaning that prompts Quine to coin the term radical translator. This is the problem of a linguist who meets a previously unknown people or a wildlife biologist who is trying to determine the meaning of an animal's vocalizations. Initial steps at translation are more like correlation, where the vocalization is first translated when paired with conspicuous events. First attempts at detecting meaning are


connotative. Quine uses the example of a linguist trying to understand what a native is saying when a rabbit scurries by and the native says gavagai. Is the native in fact signifying food, animal, rabbit, or something else? Meaning or at least a working hypothesis of meaning is made by correlating a reasonable number of stimulus events and the accompanying utterance. The stimulus meaning is the class of all stimulations that prompt that vocalization. If the linguist hears the native say gavagai many times when presented with a rabbit, he may assume that gavagai and rabbit have the same stimulus meaning. Yet the linguist cannot be sure that his translation is correct; there will always be an indeterminacy between stimulus meaning and the denoted meaning. A sort of working dictionary may be compiled but the correspondence of word to stimulus event will be imperfect because this is a mechanism for translating discourse, not single words. Quine goes so far as to state, in the tradition I believe of Turing and G6del, "that rival systems of analytical hypotheses can fit the totality of speech behavior to perfection, and can fit the totality of dispositions to speech behavior as well, and still specify mutually incompatible translations of countless sentences insusceptible of independent control" (Quine, 1960, pg 72). That is to say, given the same text


two translators could come up with internally consistent yet mutually incompatible translations.

I return now to the concept of reference, for it is

here that we confront the subject of what animals mean when they vocalize. Take the situation where a primate sees a predator, a snake for example, and utters an alarm call. To what exactly does this vocalization refer? Let us, for the sake of argument, assume that it does not refer simply to an emotion, but rather to the snake. Is the call referring to some concept of snake or to the snake itself? Does the signaller have a concept of snake or is it simply naming objects? What is the difference? The former would indicate that the monkey uses conceptual modes for organizing its world, while the latter indicates that the animal only names objects and is not conscious of the properties that define a snake no matter how precise and accurate its identifications (see Marler, 1982; Cheney and Seyfarth, 1985, for a discussion of how primates and other animals categorize objects in their world).

To describe the dichotomy between a word and its

underlying conceptual reality I will borrow from set theory and logic the notions of intension (not to be confused with intention) and extension (Quine, 1960; Harre, 1984). "The intension of a class is the concept under which any member must fall, while its extension is the set of members which satisfy that requirement. Intentions are concepts,


extensions are sets of things "1 (Harre, 1984, pg 101). The extensions of a class therefore are the members comprising that class, while intensions are the unifying concepts that define the class. The exclusive use of extensions would therefore mean that the user has no knowledge about what characterizes and differentiates that object from others. Here clearly we are speaking of what the animal knows and intends in its communication.

Quine notes that attributes are monadic intensions, described by the notation "to be an object x such that... X,11 while relations are dyadic intensions, described "to be objects z and y such that . X . y . ."1 (Quine, 1960, pg 164). In other words, for an object to have attributes or be a part of a relationship requires a conceptual grasp--an intensional grasp--of the object. I am therefore distinguishing communication using extensions from communication using intensions. Alarm calls ma~y indicate that the primate is using extensional information, for example, that the predator is a member of a list of predators. Attributing qualities to that predator however proves that the monkey conceptually grasps the defining character of that predator. I would argue that current evidence shows only that primates are able to name an object, x, and there is as yet no evidence to suggest that they are able to talk about such objects. There is still no evidence that nonhuman primates are able to symbolize


something about x such that x has a property. To paraphrase Bertrand Russell, no matter how adept a vervet may be at identifying its mother, it's not able to explain that its parents are poor but honest. If this view is correct, a monkey's lexicon is comprised of extensions of classes. The animal is not able to conceptually describe aspects of the object or to talk about the object, it only names objects. Correspondingly, if it uses intentions of the class of predators it should be able to go beyond simple object naming and attribute qualities, such as number, size, or location, to the object.

To show that an animal has a conceptual grasp requires that we find situations in which the signaller adds specificity about the object, for example, by using adjectives. Adjectives specify attributes to objects. Quine (1960) describes four levels of reference where objects of reference are further specified, from the general to the specific. In the first phase, objects are named through a process of reinforcement, extinction (learning), and ostension. Ostension is the direct experiential association with the object of reference, for example, that a certain object is called snake by other monkeys. Word usage at this phase is of extensions of a class. In the second phase, these general terms are paired with demonstrative singular terms, for example, this snake or that snake. The third phase compounds general terms with


attributes, for example, big snake or ground snake. In the fourth phase, analogy is used to apply relative terms to singular or general terms to form other general terms, for example, larger than a small snake. All phases beyond the first reference phase require conceptual knowledge of the object--using these three further levels requires intensional knowledge.

Attribution therefore involves the addition of

information. This may be accomplished in a variety of ways. In language we achieve this through the use of additional words or modifications of existing words. In any event, attribution involves specifying additional meaning to a word. In linguistics this additional element of meaning may be a phone, morpheme, or clause. These terms, to be defined below, are called recurrent partials. To discover the use and meaning of a partial requires finding recurrent use with another signal to similarly modify it. If capuchins use a signal to indicate the location of a snake, for example, in a tree or on the ground, the necessary and sufficient proof of the meaning of the partial (and therefore whether the animals are attributing qualities to objects) is to find it being used in connection with another object that could also be found in a tree or on the ground. Do the monkeys, for example, indicate whether a tayra is in a tree or on the ground using the same signals that they may use when making the same distinction about snakes? The goal of the


investigation described here is to determine the necessary condition, that is, if _Q. olivaceus use calls to indicate whether they ascribe location to snakes. Subsequent research will be necessary to establish the necessary and sufficient condition of recurrent use of the partial, and therefore whether Cebus olivaceus attribute qualities to objects.

Examples of semantic communication

Ethologists have noted for some time that many birds and mammals use alarm calls, which often are viewed as examples of semantic communication. The studies sited in table 1 are important because they bear on questions of kin selection, evolution, sociality, and animal communication. Two questions are pertinent here : 1) are these vocalizations affective or symbolic?, and 2) is there continuity throughout these taxa in what is referred to by the alarm call. Evidence consistent with the symbolic perspective is the requirement that these signals refer to objects in an arbitrary manner (Sebeok, 1975; Hockett, 1960) and that the signal production is disassociated from its physiological manifestation. A separate view, held largely by those studying various species of ground squirrels, suggests that alarm calls do not classify the predator but rather signify the different time constraints necessary for predator avoidance.


Table 1. Animals that Alarm Call.

Species Reference

chukar partridge Stokes, 1961
red-legged partridge Goodwin, 1953
California quail Williams, 1969
turkey Hale et al, 1969
quinea fowl Maier, 1982
domestic chicken Collias and Joos, 1953
Konishi, 1963
Gyger et al, 1986

black-tailed prairie dog Hoogland,1983
Richardson's ground squirrel Davis, 1984 thirteen-lined
ground squirrel Schwagmeyer, 1980
Belding's ground squirrel S.R. Robinson, 1980, 1981
California ground squirrel Owings and Virginia,1978 Leger and Owings, 1978 Owings and Leger, 980 Leger and Owings, 1978
arctic ground squirrel Melchior, 1971

saddle-back tamarins Bartecki and Heymann,1987
black spider monkey Eisenberg, 1976
black-handed spider monkey Chapman et al, 1990 wedge-capped capuchin Oppenheimer and
Oppenheimer, 1973
Robinson, pers.comm.
Norris, pers.obs.
vervet monkey Struhsaker, 1967
Seyfarth et al, 1980
Japanese macaques Green, 1975
rhesus macaques Chapais and Schulman,1980
chimpanzees Marler and Tenaza, 1977


Alarm calls and other semantic vocalizations have been investigated at length in several primate species, in particular the vervet, Cercopithecus aethiops. Struhsaker (1967) first noted that vervets use a variety of alarm calls, given to at least three different predators. Seyfarth, Cheney, and Marler (1980) further demonstrated through playback experiments that a particular call, for example a martial eagle alarm call, prompted a predictable, adaptive anti-predator response pattern even in the absence of an eagle. Additional adaptive responses were elicited by leopard and python alarm calls, with the response best suited to protect against the differing attack strategies. Each of these three calls was acoustically distinctive. From the perspective of the signaler, particular predators had unique names, while from the receiver's perspective, a certain signal meant a particular predator which required a specific adaptive response.

These signals satisfy the basic requirements of

symbolic communication by being arbitrary, disassociated from physiological manifestation, and non-iconic, (see Altmann, 1967, and Owren, 1985 for discussions of Hockett's defining properties of symbolic communication). The visual image of an eagle, for example, stimulates the vervet and through a process of internal reference to emotive and other dispositions that image is transformed into a signal that prompts an equivalent mental image in a receiver. Cheney,


Seyfarth and Marler (1980) demonstrated that these vocalizations were not simply affective or prescriptive of particular responses. For example, they demonstrated in the playback experiments that when an animal heard an alarm call, it immediately looked in the appropriate direction for the predator symbolized by the call (for example, into the sky for an eagle) and not to the presumptive signaller to indexically find out about what it was calling. As discussed below, I maintain that the receiver's immediate response to a call--whether looking to the signaller or to the presumed external referent--should be the sine aua non test of reference in field experiments.

In addition to alarm calls, several authors have argued that various primates use symbolic communication in food calls and agonistic recruitment calls (reviewed by Marler, 1985). Food calls are known to be used by chimpanzees (Pan troQlodytes) (Wrangham, 1977; Marler and Tenaza, 1977), toque macaques (Macaca sinica) (Dittus, 1984), black spider monkeys (Ateles fuscipes) (Eisenberg, 1976), and wedgecapped capuchin monkeys (C. olivaceus) (Oppenheimer and Oppenheimer, 1973; Robinson, pers. comm.; Norris, personal observation). These authors do not all maintain that food calls are used as symbols for food. More typically the earlier reports view these calls as affective; the animals are indicating their high motivation to eat when presented with particularly desirable or abundant food. Only Wrangham


(1977) and Dittus (1984) explicitly state that the food calls of their study animals are symbolic. In the chimpanzees and toque macaques, and perhaps in other primates, these calls are given when a troop member locates a superabundant food source. Typically, if the food was not superabundant, the call was not given. Calls were apparently nondirected, given to the troop as a whole. In the macaque, when a feeding bout was initiated by a food call, feeding duration was longer than those where no call was given. Dittus concludes these calls "convey information about the presence of a food source, its quantity and location" (Dittus, 1984, pg 476).

Dittus (1984) attempts to show that the macaque food calls are not affective; that they are not communicating only emotional state or probability of future action. He argues that since the stimulus eliciting the call is preci se, specific to the discovery of abundant food in 98% of the calls, the specificity is inherent in the call. I would argue that, on the contrary, since this call is given only when food is abundant, it would unambiguously signal to others that the stimulus, food, is present, regardless of whether the call is referring to the food itself or to the emotional state stimulated by abundant food. In other words, contextual specificity here is not sufficient proof that the signal is symbolic. Dittus has demonstrated that the message of the signal is the presence of abundant food.


He has not demonstrated the meaning of the signal. A cue to the meaning of the signal is the receiver's immediate response upon hearing the signal. Does it look to the signaller or look for the food? The author states, in his description of response to the food call, that "upon hearing a food call, an individual would immediately stop its activity, glance alertly in the direction of the source of the call and run there to feed" (Dittus, 1984, pp 473). Later he states that "upon hearing a food call, animals rapidly and directly approach the call site and feed there" (Dittus, 1984, pp. 476). This seems to imply that signal receivers may be turning to the signaller to find out what stimulated the call rather than turning to the food itself, i.e. that the food's presence is signalled indexically. A more rigorous test for reference would be to determine the receiver's response when the signaller is remote from the abundant food.

The majority of primate vocalizations are, as Cheney and Seyfarth (1982) point out, chirps, trills, grunts, and screams typically found in social situations. A variety of studies have attempted to determine the reference and therefore the meaning for several of these calls. Gouzoules, Gouzoules, and Marler (1984) describe five screams given by rhesus macaques during social interactions. The authors assert that these screams were representational signals referring to the social rank of the screamer's


opponent and the likelihood of physical contact in agonistic interactions. For example, where a young animal is signalling that it is interacting with a dominant animal the mother should react more strongly than when her offspring is signalling it is interacting with a lower ranking animal. Through analysis of the probability, duration, and latency of response the authors concluded that the mother responded differentially to four call classes and that the calls referred to the opponent. An alternative hypothesis is that the call refers to the level of fear the animal is feeling during the encounter. The mother's response to a particular call may be a result of, first, her recognizing that the call is from her offspring (through voice recognition) and, second, that levels of fear elicited by the potential opponent's likelihood of harming her infant are encoded in the four calls. For example, one call, one emotion, is stimulated by large dominant animals where the chances of physical contact are high while another call is stimulated by a less dominant animal where contact is unlikely. The stimulus for the call is the same, but the referent is to relative levels of fear, not to the object eliciting the fear.

In a similar study of a series of acoustically similar vervet monkey grunts, Cheney and Seyfarth (1982) demonstrate that vervets may differentiate several grunts that are undifferentiable to humans, and that these grunts may


designate objects and events in the external world. Through a paired comparison experiment in the field, the authors tested the hypothesis that grunts convey specific contextindependent information. Five test stimuli were used; three grunts given during intratroop interactions (when meeting a dominant male, a dominant female, a subordinate female), another grunt used when another vervet group was seen, and the last when the signaller saw a troop member initiate motion into the open. Results indicated a differential response pattern to the grunts. While these experiments may establish differential responses independent of context, they do not prove that the signal was interpreted by the receiver as anything other than the signaller was upset or emoting. In fact, the evidence seems to indicate that the receiver looks to the signaller, the speaker in the playbacks, to see what was causing the signaller to grunt. The appropriate experiment to indicate, for example, that one grunt refers to a dominant animal would be to play a call to an animal in the presence of a dominant animal--as if another vervet is grunting about that dominant animal-and determine if the receiver looks to the speaker or to the dominant animal. Another test would be to determine if an animal is greeted with different grunts before and after changing its social rank relative to the vocalizer.

I conclude that the results of the semantic studies of vervet alarm calls demonstrate that vervets use symbolic


communication. The authors of the studies on food calls and social vocalizations have not, I believe, sufficiently established their claims of symbolic communication. Alarm calls: models for semantic communication

There could be no better subject than alarm calls in a study of semantics. First, if an alarm call is semantic it will refer to an object, a predator, that is remote in space from the animal. This is essential because when a monkey is responding to, for example, a boa, the snake is probably distant from the monkey and not within the group. Testing alarm call semanticity becomes a geometric consideration. The physical remoteness of predator, caller, and receiver makes response quantification relatively easy because it is clear when the monkey is looking at the predator and when it is looking to other monkeys (or speakers). Other semantic studies of primate calls, for example those on food calls, have had the problem of differentiating when a monkey was looking at another monkey and when it was looking at the food the monkey was eating. (This ambiguity would also occur when testing for response to an alarm call referring to a snake close to the caller.) Secondly, by the nature of predation, response to a predator alarm call must be immediate and adaptive. When a predator is sighted there typically is an immediate need to give an alarm call. Likewise when a monkey hears an alarm call there is an immediate need for response. This need for immediate


response and external reference make the study of alarm calls ideal for semantic research. Phonology

Phonology is the study of speech sounds--the phonetics and phonemics of a language. It includes studies of how sounds are made as well as the psychoacoustics of their reception. Where phonetics is used to describe and classify speech sounds, in phonology these descriptions and classifications are used to describe communication systems and explain sound processes (Sloat, Taylor and Hoard, 1978). The basic unit of phonetics is the phone, which simply is any speech sound. The basic unit of phonemics is the phoneme which is the smallest speech unit that distinguishes one linguistic utterance from another.

Acoustic variability may describe phonetic or

phonological differences. The perceived association between the signal and phones is the phonetic quality of the signal. This is in contrast to phonemes which are the abstract class of minimal phonological units consisting of phones (the phonetic unit) that are functionally identical.

If we extend this definition of phonology to the study of all vocal sounds, including those of non-human animals, then we can view each species as potentially having its own phonetic system. Taxonomists of animal vocalizations, lacking (at least initially) a semantic framework to distinguish sounds, face the problem of whether to lump or


split similar sounding vocalizations (for a discussion of this taxonomic question see Marler, 1982). There have been several phonetic studies of primate vocalizations that point to a richly variable repertoire where the variability may be described as phonetic.

The diversity of alarm calls seen in this study prompts the central question of my research, namely is this variability meaningful to the monkeys. Green (1975) faced a similar problem when describing the variability of coo vocalizations by Japanese macaques. He adopted the terms continuous/ discrete to describe morphological variability; a continuous call was one in which the gradations of variability were functional and meaningfully significant, whereas a discrete signal had no functionally intermediate forms (Green and Marler, 1979). Owren (1985) provides the useful insight that it is not the presence of variability alone that determines whether a signal is discrete or continuous, rather it is the function of that variability. Determining function is, of course, a difficult problem and may lead to seemingly contradictory results. For example, Winter (1969 a&b) concluded that squirrel monkey (Saimiri sciureus) used a discrete repertoire whereas Schott (1975) concluded that the repertoire was continuous (=graded). Such difficulties are perhaps not surprising. All vocal signals contain some variability because the production mechanism is a biological organ producing a continuously


varying energy wave that is subject to fatigue, distortion and other perturbations.

Characterization of speech signals in articulatory or auditory terms is described as distinctive features analysis. Research on the distinctive features of human speech has shown that a relatively small set of parameters can define phonemes (for example, see Ladefoged, 1975). This is an important step in the description of variability because distinctive features provide a mechanism for describing how a continuously varying acoustic signal can be described in discrete phonological units. The mechanisms by which humans and other primates perceive their signals is discussed below, but it is clear that many primates distinguish signal variability in a manner similar to that of humans.

Examples of Rhonetic descriptions

Phonetic differences in vocal behavior have been

described in a variety of primates (Table 2). For example, Cleveland and Snowdon (1982) were able to describe eight chirp variants of a tamarin, Saguinus oedipus, each given in a definably different behavioral context. The authors concluded that each call variant represented a particular motivational state, i.e. that while they might have been affective vocalizations, each call was uniquely paired to a particular motivational state. A distinctive features analysis showed that four statistically significant


parameters (presence/absence of a frequency upsweep, difference between peak and end frequency, peak frequency, and duration of frequency downsweep) differentiated the eight calls.

Table 2.

Primates Exhibiting Phonetic Differences
in Their Communication System

Species Reference

Japanese macaque Green, 1975
Talpoin monkey Gautier, 1974
Pygmy marmoset Pola and Snowdon, 1975
Snowdon and Pola, 1978
Cotton-top tamarin Cleveland and Snowdon,1982
Vervet Cheney and Seyfarth, 1982
Seyfarth and Cheney, 1984 Owren and Bernacki, 1988 Owren, 1985
Gelada baboon Richman, 1976
Squirrel monkey Newman et al, 1978
Black spider monkey Eisenberg, 1976
Wedge-capped capuchin Robinson, 1984

In another distinctive features investigation, Owren and Bernacki (1988) were able to define a single acoustic variable, spectral tilt, that correctly classified two Cercopithecus aethiops alarm calls, eagle vs snake, 97% of the time. If the tilt was falling, that is, if the amplitude decreased each successively higher spectral peak, the call was an eagle call, otherwise it was a snake call. Perceptually this means that eagle calls were lower pitched than snake calls because the higher spectral peaks of the snake call were relatively louder. From another


perspective, the monkey stressed the higher frequencies of the snake calls and stressed the lower frequencies of the eagle call.

Spectral characteristics could also differentiate these calls. "If two or three spectral peaks occur but do not differ more than 3 dB in amplitude, or one stronger peak above 1750 Hz occurs, it is a snake call. If two or three spectral peaks occur, one peak below 1750 Hz is at least 3 dB stronger than any other peak, and an identical peak occurs above 1750 Hz, it is a eagle call" (Owren and Bernacki, 1988, pg 1933). These features correctly classified the two calls 89% of the time. These features, while including frequency qualities, still primarily depend on the amplitude characteristics best described using the tilt feature. Spectrograms of these two calls (fig. 1, Owren and Bernacki, 1988) show that the snake alarm call is considerably more wide band than the eagle call. That there appears to be no energy above 4 kHz in the eagle call is an artifact of spectrograms, which have a limited amplitude resolution. The energy above 4 kHz, as seen in the spectra, simply is louder in snake calls than in eagle calls resulting in a greater spectral tilt in the eagle calls.

Seyfarth and Cheney (1984) concluded that vervets may use three acoustic features to differentiate four grunts given in different social situations: frequency of the spectral peak associated with the fundamental (F.),


frequency of the second peak (F,), and increasing frequency in the second spectral peak. These authors did not measure the spectral tilt of these calls, but an examination of the spectra shows that these frequency features are probably the most parsimonious.

Richman (1976) gave perhaps the most detailed

description of a primate vocal distinctive features, describing the vocal features used by gelada baboons (Theropithecus gelada). He presented a number of spectrograms illustrating a variety of calls with complex acoustical morphology. He reported that geladas produce long strings of alternating ingressive and egressive phonations. Through spectrographic analysis he concluded that they were capable of producing both vowel-like and consonant-like calls. He also described place and manner of articulation for these calls.

Geladas appear to be able to vary the relative

frequency positions of the first two formants, either through divergence of the formants like human back vs front vowels or raising or lowering of the first formant--the low vs high vowel contrast. (I should note that there has been some disagreement regarding the use of the term formant when discussing non-human vocalizations. Please see chapter five, the acoustics section). Geladas can also simultaneously lower both formant frequencies--rounding, in acoustic phonetic terms. Lastly, geladas can selectively


dampen and enhance frequency components of the entire spectrum.

Richman further described the distinctions of vocal

onset and finds them similar to those employed by humans in making consonants: gradual, sudden, and fricative. These articulation features produce, in human speech, glides, liquids, stops, and fricatives. He concluded that it appears possible that geladas produce these consonant-like sounds by changes in three places of articulation: labial, a velarlike position, and an intermediate position. The author made no attempt to correlate these acoustic phonetic differences with changes in behavior.

These studies illustrate that calls with similar

acoustic structure often have variants recognizable to the vocalizing animals, though perhaps not to humans. Call variants are often correlated with different behavioral functions. Future taxonomists of animal calls should recognize that phonological variability in animal vocalizations is important yet often very difficult to discern.

Acousticpercerption in Primates

The above examples of how various primates use

linguistic-like modifications of their communication signals assume that the animals perceive these changes. Recently there has been a significant increase in research on animal psycho-acoustics investigating both the signal


characteristics attended to by animals and whether perceptual strategies are similar to those employed by humans in speech. Ideally investigators will identify perceptual sensitivities to particular signal characteristics that are of communicative importance.

Investigators have studied species-specific perceptual processing of communication signals in a limited set of old World primates. The results of various studies by Sinnott (1976, 1985) suggest that 1) primates do not discriminate pure tone changes, the vowel cue, nearly as well as humans, and 2) their discriminatory abilities used in differentiating consonants, where both frequency and intensity cues are crucial, are roughly similar to humans.

Petersen and others investigated the perception of coo vocalizations by japanese macaques (Macaca fuscata) (reviewed in Petersen, 1982) finding, for example, similarities in neural lateralization patterns between primates and humans. They also found that peak frequency location, a factor central to discriminating among the various macaque 'coo' vocalizations, shows a right ear advantage. This is in contrast to how macaques hear pitch changes, which shows a left or no ear advantage. These findings are comparable to similar results in humans. There is a right ear advantage in humans for stop consonants and either a right or no ear advantage for vowels. In regard to selective attention to changes of particular acoustic


parameters, it was found that, when pitch is varied, there is selective attention to peak frequency but when, in the reverse experiment, peak frequency is varied, there is poor attention to pitch. These results indicate that Japanese macaques take a discretely used acoustic cue, peak position, and use it to partition vocalizations where peak frequency position continuously varies. The perceptual processes involved, selective attention, perceptual compensation, and partitioning, closely match those used by humans in speech perception. These results indicate that both human and nonhuman primates apparently detect communicatively salient features in their signals in similar manners. Syntax

Syntax is the study of word order. Syntax describes the ways in which words are put together to form phrases, clauses, and sentences. Syntax describes another layer of meaning, one provided by word order. It is through syntax that we distinguish 'dog bites man' from 'man bites dog'.

The concept of syntax itself is layered, with the

distinctions centered around the use of meaning. At the linguistic level Chomsky defined syntax as a "system constituted by rules that interact to determine the form and intrinsic meaning of a potentially infinite number of sentences" (Chomsky, 1972, pg 69). At a functional level several students of animal communication (Robinson, 1979, 1984; Snowdon, 1982; Cleveland and Snowdon, 1982) have


adopted a more general notion of syntax. To them syntax is simply a system of rules that will generate and predict sequences of signals, thus they view syntax as a probabilistic phenomenon. Altmann (1965) noted for example that in many vocal sequences first-order Markov (predictable) processes describe transition probabilities and are equivalent to grammars.

Marler and Tenaza (1977) differentiated two forms of syntax: phonological (called phonetic by Snowdon, 1982) syntax and lexical syntax. Phonological syntax is a system for arranging communicative components into more than a single pattern, each with a distinctive function. In language it is analogous to word formation through phonemic rearrangement. In lexical syntax "compound signals derive their meaning from the multiplexing of the meanings of the components as used separately or in other combinations" (Marler and Tenaza, 1977, pg 25). Linguistically it is analogous to phrase formation through word combination so that the product is the sum of the meanings of the individual elements.

Several additional points about syntax are pertinent. First, in parallel to the semantic classes of agent, patient, instrument, and object there are syntactic classes such as noun phrase, verb phrase, and determiners. These syntactic classes are necessary for modeling human language. Premack (1985) makes the valuable point "that syntax cannot


be derived from semantics. No metamorphosis has been demonstrated for turning the semantic caterpillar into the syntactic butterfly: agent, recipient, and the like, no matter how abstractly construed, will not turn into noun phrase, verb phrase, etc." (pg 284). If primates can, as I suspect, name objects and are able to differentiate agent from patient, there still is no evidence to suggest that they are capable of the purely linguistic differentiation of noun phrase from verb phrase. Secondly, Chomsky's conception of syntax has progressed beyond the definition given above. He now views mental representations, traditionally within the purview of semantics, as being a form of syntax.

The study of the relation of syntactic structures
to models, "pictures," and the like, should be regarded as pure syntax, the study of mental representations, to
be supplemented by a theory of the relation these
mental objects bear to the world . Thus, the shift
-towards a computational theory of mind encompasses a
substantial part of what has been called "semantics" as
well. (Chomsky, 1986, pg 45).

Additionally, Chomsky views universal grammars as being endowments due to innate, biologically determined principles. This is our language faculty, our "language acquisition device, an innate component of the human mind that yields a particular language through interaction with presented experience, a device that converts experience into a system of knowledge attained: knowledge of one or another languagell (Chomsky, 1986, pg 3). In this new formulation I


perceive a weakening in his position that animals are categorically incapable of a language; instead, an animal 'language' will be the product of that species' own innate language faculty. All animals convert experience into a system of knowledge, some perhaps using vocal behavior to differentiate objects. Without begging the definition of language, if we accept that some species now use 'words' to identify objects in their environment, the separation between human language and animal communication becomes smaller, though still very real. The difference between our language and the communication systems used by animals may be a difference in degree, not kind. Alarm call syntax

If certain alarm calls are used to indicate

attributions about an object, for example, the location of a snake, they are used in some sense like adjectives. one would expect then that these calls might be associated with other calls. For example, if one vocalization means 'snake' and another means 'ground snake' there must be some additional segment to the latter signal carrying the additional information, either as another call or as an infix in the original signal. In English we modify the meaning of a word both ways, either by using another word, an adjective, or by internally modifying the original word with prefixes or suffixes, e.g. usual and unusual.


Investigating such matters requires knowing the phonetic structure of the communication system. All biological signals have variability but only some of it is communicatively meaningful. The alarm call classification scheme described below parses the various alarm calls into a number of variants. Without knowing the phonetic structures of the communication system, I am unable to determine if C. olivaceus alters its signals with internal modifications. I am, however, able to determine whether calls are combined. I will limit this study to the description of associations between grrahs alone because to accurately determine the association between particular alarm calls and other calls, huhs for example, requires knowing how the other calls vary. Examples of syntactic communication

A wide assortment of primates use syntactic

combinations of different calls--long calls of gray-cheeked mangabeys (Cercocebus albigena) (Waser, 1975), chimpanzees (Pan troglodytes) (Marler and Hobbet, 1975), various gibbon species (Tembrock, 1974; Tenaza, 1976; Marshall and Marshall, 1976); alarm calls of cotton-top tamarin (Saguinus oedipus) (Cleveland and Snowdon, 1982); intragroup calls in the pygmy marmoset (Cebuella Dvgmea) (Snowdon, 1982) and wedge-capped capuchin monkey (g. olivaceus) (Robinson, 1984) and "singing" in titi monkeys (Callicebus moloch) (Moynihan, 1966; Robinson, 1979). Some of these combinatory signals use lexical syntax to impart additional information to the


signals, while in others the combined signal elements have a separate meaning--phonological syntax.

Cleveland and Snowdon (1982) demonstrated lexical syntax in cotton-top tamarins (Sacruinus oedinus). They found that two frequently used calls, an alarm chirp and a low arousal alerting call, were combined such that animals responded to this combined call as if it was the sum of the constituent elements.

Robinson (1979) described the remarkably complex vocal behavior of titi monkeys (Callicebus moloch). He found an elaborate hierarchical system where calls are repeated to form phrases, which may then be variably combined to form sequences. He distinguished six types of sequences by their different structure, as disclosed by transition probabilities, their situational context, and the sex of the caller. For example he describes duetting as "an alternation of pant and bellow phrases follows the introductory moaning phrase. As the sequence continues the animals begin to add pumps, either after pants or bellows, and insert honks between pant phrases and after pumps.. Honking usually ends the sequence" (Robinson, 1979, pg 393).

The driving force behind sequencing, and the effect sequences have on meaning remains unknown; without such knowledge we cannot determine if these syntactic patterns are lexical or phonological. Robinson (1979) notes that the hierarchies found in titi monkey calls are similar to those


derived by phrase structure grammars. Further analysis of sequencing is severely hampered by our failure to understand what the vocalizations mean; we lack a semantic context to place the calls. Chomsky differentiated phrase structure grammars from more simple sequential models by the characteristics of nested dependencies: 'if. . then' structures. Syntactic structure may depend on the occurrence of particular words, with independent order for intervening words (see Robinson, 1979). Alternatively, responses to different sequences may be due to different proportions of phrases and not to their syntactic order. Robinson (1979) points out, however, that titi monkeys clearly distinguish phrase types and concludes that variation in phrase order is important and has communicatory significance.

Robinson (1984) has also investigated the syntactic

structures in vocalizations of Q. olivaceus. He found that these monkeys use lexical compounding rules to generate new sequences, where the new compound call is produced in contexts intermediate between those of the constituent vocalizations. Additionally, he found a second, more simple sequence type where two call types are blended to form a transitional third call. These new calls are without apparent syntactic structure. He concluded that these lexical syntactic rules are analogous to the rules we use to generate words from morphemes, they are not analogous to the grammatical rules of language.


The studies of syntax in both Callicebus and C. olivaceus by Robinson are important because they best illustrate that at least some primates use rudimentary ordering schemes in vocal production and these syntactic patterning rules may be similar to those used by humans.

Several problems emerge from the above description of linguistics and primate communication. I have posited that primates are capable of naming objects but to date there is no evidence that they speak about them. If this is the case, syntax poses a particular problem because syntax provides another layer of meaning upon the individual lexical elements. Is syntax a linguistic operator requiring one form for extensional terms and another for intensional? The linguistic elements syntactically combined may imply differences in conceptual understanding. Evidence for this would be how elements are syntactically combined. Syntactic combinations of extensional terms would involve combinations of simple names and an operator, whereas a combination of intensional terms would involve, for example, analogy or other mechanisms requiring a conceptual grasp of the combined words. I believe there is a conceptual difference between the syntax of 'man bites dog' vs 'dog bites man' and that occurring when water is lexically combined with bird to form waterbird (where waterbird is used in a context separate from water and other birds to refer to a duck). The former construction requires only an ability to name two


objects and an action, whereas the latter seems to require a conceptual grasp of the terms.

The above duality suggests a novel hypothesis about how animals communicate about emotions and objects using both affective and symbolic communication. This communication system would use, first, purely affective vocalizations to communicate about emotions and combine those signals to refer to intermediate emotions, for example, the lexically and phonologically combined vocalizations in the capuchin and titi monkeys, and, second, use symbolic terms to refer to objects, for example, alarm calls. According to this hypothesis primates would combine calls to form other calls referring to intermediate emotions using lexical syntax and restrict their use of symbolic terms for named objects. Sequences would be restricted to combinations of affective vocalizations, where combining the emotive signals acts to fluidly transpose the emotions into a third transitional emotion that would have its own subjective reality. Primates would give symbolic calls singly (or repetitions of the same 'word') because they refer to an identifiable object and they would not use them in a sentential form because sentences cannot refer simply to objects. Quine (1960) argued that only words have reference and that sentences are identified by their logical truth. If my hypothesis is correct, it may explain much of the primate vocal behavior described above. They have, first, words for


objects and, second, calls and sequences for emotions, each reflecting an underlying reality of objects and feelings. The remainder of this study will address representational communication by g. olivaceus to determine whether they have words for objects.

Research Goals and Rationale

The goal of this research is twofold: 1) to establish whether . olivaceus semantically uses alarm calls to refer to objects and, if so, 2) whether it is then able to attribute qualities to those objects.

In this first chapter I reviewed the literature and

described the problems currently prominent in the study of primate communication. In Chapter 2 1 describe the alarm calls of C. olivaceus. In Chapter 3 1 describe the first of two experiments on the semantics of a particular alarm call type, the snake alarm call. The first experiment involves release of different sizes and numbers of snakes. In chapter

4 1 describe the second experiment where I play recordings of the alarm calls given to the released snakes back to the monkeys. The purpose of the first experiment is to determine if C. olivaceus uses alarm calls as names for particular predators and in the second experiment whether it attributes qualities to them. In Chapter 5 1 describe the acoustic structure of C. olivaceus alarm calls. I finish with a general summary of findings.



Methods and Materials

Research Site

This project was conducted at Hato Masaguaral in three periods: preliminary observations from May-July, 1986; snake release experiments from March-August, 1988; and playback experiments from April-August, 1989. Hato Masaguaral is an active cattle ranch as well as wildlife refuge owned by Sr. Tomas Blohin and is located 146 miles south southwest of Caracas in central Venezuela (8*34'N, 670351W). The study site was a 4 kin2 gallery forest (figure 1). It was bordered to the east by a seasonal river, the Cafio Caracol and, respectively, to the north and south by two recently deforested ranches, Finca Torres and Hato Flores Morades. To the west the gallery forest gradually became a more open forest. The study site seasonally flooded from approximately May to October after which it dried and the deciduous canopy became increasingly open. Rainfall in the region averaged 1450 mm/ year. The mean temperature was approximately 28*C. See Troth (1979) for a more detailed description of the ranch.







.S te'si

Figure 1. Study area showing adjoining ranches, the
bordering streams, and the trail system. Marks on trails
are at 25 m intervals (Robinson, 1986).


Recordings and Data Analysis Earuipment

Two recording systems were used. Initially, from May to July, 1986, a Sony TC-D5M cassette recorder and a Sennheiser MKh-805 directional microphone were used ad libitum to record monkey calls. Recording distances were less than 15m. The system's frequency sensitivity was flat from approximately 50-15,000 Hz. These recordings were used to determine the monkeys lexicon.

A Sony 8mm CCD-V9 video camera with a Sennheiser MK-415 directional microphone was used during the snake releases and alarm call playbacks as both the auditory and visual recording device. The camera image sensor produces 380,000 pixels and the audio dynamic range is >80 dB over a frequency range of 30-15,000Hz. The camera has multiple shutter speeds which allowed for high speed filming. Video recordings were played back on a 19"1 Sony Trinitron TV. Study Animals

The C. olivaceus studied here were part of a population of approximately 300 animals that inhabit Hato Masaguaral. This population contains approximately 12 groups, 4 of which were studied--Main, White, Red, and Splinter. Most work was done with Main group, which contained 21-27 animals throughout the study: a single adult male, 5-8 adult females, and assorted juveniles and infants. Table 3 shows its composition at the beginning of the 1989 field season. These groups have been continuously studied by Robinson


(1981, 1982, 1984, 1986) from 1976 to the present time. All animals in the four groups were individually identifiable by name, with identification made possible by age/ sex, size, and individual markings. All animals, particularly Main group, were habituated to my presence. This made it possible to study the animals at extremely close range, usually 210m. Familial relationships within these four groups have been known for at least ten years.

Table 3.
Composition of Main Group Males Females
Name Age* Name Age*

White Beard >10 Mo 18
Jefe 8 Amelia 17
Griffin 8 Beaut 17
Stu 7 Whitey 12
Mani 5 Pointy Face 12
Babas 4 Alexandra 8
Winston 3 Onica 7
Mike 1 Margo 7
Hanna 5
Mali 5
Puffy 4
Amanda 3
Modem 3

age as of 1989

Predators of Cebus Olivaceus at Hato Masaguaral

As Cheney and Wrangham (1987) note, predation on

primates is rarely observed. There have been no confirmed instances of predation on C. olivaceus at the ranch. The Harpy eagle (Harvia harovia) and the boa constrictor


(Constrictor constrictor) are the only confirmed predators of _. olivaceus (Rettig, 1978; Chapman, 1986). Boas are common on the ranch, but this eagle species does not occur there. Aside from human predation through hunting which is negligible at the ranch, presumed predators are aerial, terrestrial, or arboreal.

The largest raptor found at Hato Masaguaral is the

ornate hawk-eagle (Spizaetus ornatus). This bird is almost certainly capable of taking gbus as prey. Other potential avian predators are the spectacled owl (Pulsatrix perspicillata), great horned owl (Bubo virginianus), savanna hawk (Heterospizias meridionalis), zone-tailed hawk (Buteo albonotatus), black collared hawk (Busarellus nigricollis), collared forest falcon (Micrastur semitorquatus), laughing falcon (Herpetotheres cachinnans), and the road-side hawk (Buteo magnirostris).

The only strictly terrestrial predator likely to take C. olivaceus at Hato Masaguaral is the dog, Canis familiaris. All other potential predators are capable of following g. olivaceus into the trees. There are four felids at the ranch; jaguar (Felis onca), cougar (Puma concolor), ocelot (E. pardalis), and jaguarundi (f. yagouaroundi). Tayra (Eira barbara) have been seen chasing capuchins (Robinson, pers.comm.). In my experience, the relationship between tayra and C. olivaceus is apparently context sensitive; among the same individual animals during


some interactions the monkeys chase the tayra, while during others the monkeys appear quite alarmed, and during still others they apparently ignore each other.

The boa constictor (Constrictor constrictor) is the only large snake living in the forest at Hato Masaguaral that is likely to take C. olivaceus. Other potentially dangerous snakes are anacondas (Eunectes murinus) and a rattlesnake (Crotalus durissus) which is rarely found inside the forest.

Estimated predation levels on C. olivaceus at Masaguaral are moderate to low. In seven years of observations on approximately 175 animals Robinson estimated that 15 animals were taken by unknown predators, an estimated annual predation rate of 3% (Robinson in: Cheney and Wrangham, 1987, pg. 232). Predation rates were highest on smaller animals; of those presumed taken 70% were infants and 20% juveniles with the remainder split between adult males and females. Reuter (1986) discusses the influence of predation on foraging behavior of C. olivaceus.

I should emphasize that intraspecific aggression

appears to be a major source of mortality among the study animals, and is probably greater than predation. For example, in 1988 a new male, WB, entered Main group and within 6 weeks the four newborn infants had all disappeared, with one confirmed case of infanticide by that new male (Valderrama et al, 1990).


olivaceus appear to be predators on bird eggs and hatchlings during the nesting season. Such behaviors typically elicit active nest guarding. I observed Main group attack, kill, and eat a fledgling yellow-knobbed curassow (Crax daubentoni) and attempt to attack a nestling giant potoo (Nyctibius grandis), which was actively defended by an adult. Pairs of road-side hawks called and swooped at C. olivaceus most often during the nesting season. Active nest defense may be the reason that capuchins gave alarm calls at these relatively small raptors. Responses to attacks by nest guarding hawks presumably do not markedly differ from predation events.

Responses of Cebus olivaceus to Predators Anti-Predator Behavior

Responses to predators, both vocal and otherwise, varied according to the predator and the context of the interaction. The description of crrah usage by Oppenheimer and Oppenheimer (1973) is different from my experience in several ways. First, I consider qrrahs to be specifically alarm calls, not simply agonistic calls. Secondly, the Oppenheimers differentiated only a single alarm call, apparently lumping all alarm calls as grrahs. Lastly, I rarely saw capuchins take more vigorous anti-predator actions than just giving alarm calls. Also, in many interactions between C. olivaceus and howler monkeys


(Allouatta seniculus), I never heard an alarm call. In the case of raptors, Oppenheimer and Oppenheimer (1973) write that when a bird swiftly flew over the animals one or more C. olivaceus would give a single Qrrah then drop to lower branches or move deeper into the trees. I differ with the Oppenheimers by distinguishing alarms calls to birds as being acoustically distinct from other alarm calls, with waahs given to avian threats and Qrrahs given to terrestrial and arboreal threats.

In the presence of avian threats the monkeys typically gave a single waah, looked up, and on rare occasions moved away from the perceived threat. A variety of birds elicited waahs from C. olivaceus: ornate hawk eagles, road-side hawks, zoned tailed hawks, forest falcons, laughing falcons and various species of vultures. I should note that with some birds the monkeys waahed only after they appeared to be surprised as the birds flushed in front of them or as they flew low overhead: chachalacas (Ortalis ruficauda), currosows (Crax daubentoni), egrets, ibis, muscovy ducks (Oxvura dominica), and macaws (Ara macao). I will recount, from field notes, two examples of interactions between C. olivaceus and hawks.

On one occasion a road-side hawk attacked a juvenile
female crossing a tree gap on a vine. As the hawk
approached, other animals waahed then the hawk hit the
monkey's dorsal side without apparently hurting it.

On another occasion an adult female carrying a newborn was moving on a branch when a laughing falcon (that had been displaced by a harrassing monkeys) flew 2-3 m over


it. Other animals waahed and she imediately swung
beneath the branch to interpose the branch between the falcon and herself. It was not clear whether the bird was attacking the female or simply flying away from the
harassing individuals.

This was the only observed active defense, other than vocalizations, against avian predators. When the physical threat was nonavian the situation was quite different. Mobbing behavior, rarely directed towards raptors, was regularly used against non-avian threats. Q. olivaceus have been observed mobbing jaguar, puma, ocelot, tayra, boa constrictors, and humans. Additionally, they arrahed at donkeys, cows, and deer, though this appeared to happen only when the animal ran through the forest, presumably frightening the monkeys.

In a typical interaction, multiple animals cluster

near the potential predator and give many cirrahs. At least in the case of boas, which often remain still and therefore cryptic around monkeys, the animal discovering the snake often remains near it through much of the interaction, giving the most arrahs while other troop members move by also emitting alarm calls. Under these conditions when. a predator is mobbed there may be hundreds of crrrahs given.

The emission of grrahs is a highly directed behavior. Typically the alarming animal faces the threat and gives numerous calls in a clearly agitated manner. These may or may not elicit calls from other monkeys. Grrahs appear to function most clearly as an alarm, warning other animals of


the presence of the threat. That they serve to warn other animals of the presence and even location of the predator is the subject of the experiments described below.

On one occasion I observed White group mob an ocelot (Felis pardalis). The ocelot was walking on the ground while the monkeys followed it in the trees, crrrahing as they went. The monkeys appeared to be highly agitated whereas the cat appeared quite indifferent. A similar incident was witnessed with Red group. In both cases the monkeys were already in trees while the cat was on the ground, many animals vocalized, and the ocelot showed little interest in the monkeys.

On a variety of occasions I observed tayra (Eira barbara) move among White, Red, and Main group. For example, just as I was preparing to conduct a playback experiment with Red group I heard a grrah and turned to see a young tayra approach the area on the ground. I was standing below an adult male, Finger, and several other monkeys, less than 10m distance. The tayra saw me and approached to investigate. The tayra appeared to ignore the monkeys who casually watched it as they continued to feed. At another time, for two weeks Main group inhabited a region that was also frequented by a tayra. Sometimes the two species would amicably occupy the same tree while on other occasions the monkeys would chase the tayra out. on still another occasion I saw an adult tayra move through the same


trees as capuchins without incident. Lastly, I observed Main group approach a mother and infant tayra in a palm tree. When the adult tayra saw me it tried to leave by coming to the ground and running away, however the infant wouldn't leave and when the mother appeared to abandon it the C. olivaceus threatened the young tayra. This prompted the mother to return, which caused the Cebus to back off. The monkeys gave few qrrahs but gave many other threatening calls during the interaction. J. Robinson (personal comm.) reported seeing an adult tayra chasing C. olivaceus through the trees. I conclude that tayra are a potential threat to C. olivaceus, but the context of the interaction was important as to whether the monkeys reacted.

The response of animals hearing an alarm call is

typically immediate and marked. For example, I was filming a subadult male, GR, as he sat near me on a branch. Main group was spread out around us. A monkey approximately 60m away saw a tayra and Qrrahed. Griffin immediately turned to look at the caller and began scanning that same distant area.

Van Schaik and van Noordwijk (1989) described the role of male Cebus apella and C. albifrons in predator avoidance. They measured differences in vigilance and related behaviours between sexes of both species and, finding a difference, interpreted it as a special male role in predator avoidance. The same anti-predator behaviors of


~olivaceus were not measured here, specifically vigilance for predators. It was my impression however that males were not more likely to give alarm calls or mob a predator. As discussed below, in vocal responses to released snakes, females gave many more alarm calls than males, which is different from what van Schaik found in the other Cebus species.

The differences in response by C. olivaceus to avian

and terrestrial predators, particularly the number of calls given, is explained well in a geometric model of alarm call behavior (Taylor et al.., 1990). The authors show that slow moving predators should elicit more calls than fast moving predators. This was found to be the case here where waahs were typically given once while many grrahs were often given.

Alarm Calls

Cebus olivaceus give alarm calls to a variety of

animals, as is clear from the above descriptions. While I divide these calls into two types, waahs and Qrrahs, there appears to be great variability within these call types.I will examine in greater detail how C. olivaceus respond to snakes and the variability within the snake alarm calls in subsequent chapters. Here I will provide examples of arrahs and waahs given to a variety of predators and apparent threats. These calls can be compared to other C. olivaceus calls presented in Robinson (1982, 1984).



Graahs, in general, can be distinguished from other C. olivaceus calls by two characterstics: 1) falling frequency and 2) a breathy aspiration sound at the end. Other C. olivaceus calls are downswept, chirps and hehs are good examples, but the terminal aperiodic breathy sound is distinctive to arrahs. The amount of frequency drop and number of energy bands is highly variable, as will be seen in later chapters.

Grrahs to snakes. Figure 2a illustrates narrow band spectrograms of two qrrahs given by an adult male, SH, in Red group to a boa constrictor (Constrictor constrictor). Note that they have 3-4 continuously falling energy bands with the fundamental frequency beginning at approximately 1 kHz. Note also that the calls are less periodic, more noisy at the end. Figure 2b is another spectrogram of a qrrah to a boa by Quay, an adult male in Splinter group. In this call the pitch remains relatively steady until the end where it first climbs then precipitously falls. Note again the broad band noise at the end, particularly in the second formant. The acoustics of Qrrahs to snakes is the subject of chapter 5 which provides examples of many other grrah variants. Notice that the calls just described are similar to some of those given by Main group.

Graahs to humans. Unhabituated and, on occasion,

habituated capuchins would emit grrahs at humans. Figure 2c



a: 0


$4J 0)0
to 010
41 00)


14.41 0 4

k0) Owl I go


is a grrah at me given by an adult female, Irma, in White group. Note that it is noisy at the end and downswept.

Graahs to other animals. Figure 2d illustrates two Qrrahs to an ocelot (Felis pardalis) from an unidentified animal in White group as it watched the cat walk under it. Both calls have two energy bands and drop discontinuously. The terminal aspiration is slight. The pure tone at approximately 1.5 kHz is an artifact.

On several occasions C. olivaceus would arrah at

seemingly anomalous objects. Figure 2e provides two qrrahs to a donkey given by unknown animals. These calls are quite distinctive with a long fundamental having multiple energy bands only in the middle of the call after which the frequency drops.


Waahs are, in general, longer duration, slightly lower in frequency and less forceful than grrahs. Like cirrahs the pitch falls, however since waahs typically have longer duration the slope is less. They may also be quite breathy or noisy. While the sample size for waahs is much smaller than for grrahs it is quite apparent there is also significant variability among waahs. The semantic utility of this variability is unknown.

Waahs to Road-side hawks. Waahs are most often

given to the road-side hawk (Buteo magnirostris). Figure 3a illustrates two waahs by adult females Gertrude and Irma



O 1




WI ~~ 4J

8 4


from White group. These calls typically have multiple energy bands, which like arrahs, fall in the later half of the call. Notice however that the higher energy bands in some calls may be stressed.

Waahs to vultures. Figure 3b presents two waahs by

the dominant male in Red group, Finger, to a vulture, probably a turkey vulture (Cathartes aura). Again these are relatively long duration calls that fall in pitch through the majority of the call.

Waahs to caracara. Figure 3c is a waah to a crested

caracara (Caracara cheriway) by the adult male Brow from White group. This call is shorter in duration, with the second formant stressed.

Further InvestiQations

The anti-predator behavior of C. olivaceus suggests that, like the vervet (Cercopithecus aethiops), the alarm calls described above are used in a semantic manner, with the call referring to an object, in this case a predator, and not simply to an emotion. Demonstrating that C. olivaceus use these calls in a semantic manner requires investigating how a monkey responds to an alarm call. I therefore undertook a series of experiments to verify that C. olivaceus use their alarm calls to refer to objects. I first obtained a series of recordings of known animals calling at a predator and then played those calls to other


known animals to determine where they looked upon hearing the signal. These experiments replicate and extend Seyfarth, Cheney, and Marler's (1980) experiments on vervets. I hypothesized that if capuchins named objects, e.g. snakes, they might also attribute qualities to those objects. For example, would they vocally distinguish one snake from two, small from large, or provide information on a snake's location? To test these hypotheses, I released different numbers and sizes of snakes then recorded the monkeys responses. Following that I played to them selected calls to determine if a listening animal responded appropriately.


Data Analysis Methods

The alarm calls recorded during snake releases were digitized at 20,000 samples/ second using a Digital Translation DT2821 analog/digital- digital/analog board in a Compaq Portable II computer using the RDA and LDA routines in the ILS spectrum analysis system. Where necessary these calls were digitally filtered using ILS. These signal were then played into and analyzed on a Kay Elemetrics model 7029A spectrograph. The spectrograms were narrow band (resolution= 45Hz), 80-8,000 Hz bandwidth, with FL1 shaping.

An essential element to the analysis of these calls was the identity of the caller. This was determined by

1) identifying an animal as it calls on video,

2) identifying the caller in the field and calling out
its name while recording,

3) determining who was in the area during the
interaction and attributing caller identification
by listening, a process of elimination,

4) where calls sounded the same to me, I visually
identified the group of animals that was present
and calling and then compared the fundamental
frequency for similar calls by those same animals
at other times. This attribution was made by
comparing fundamental frequency of calls.



Where none of the procedures worked, I discarded the

call. Most of the attribution were done with the first two methods. Using these procedures I identified the caller for 190 of 279 arrahs recorded during releases.

The following data were input into the Q & A data base management program to correlate the various calls with the contexts of the snake releases,

1) Caller distance to snake and nearest neighbornear, medium, far

2) Identity of nearest neighbors

3) Number of snakes

4) Snake location (tree or ground) and size.

Additional information included vocalization number, call class (variant type, see below) and caller. Release Methodology and Release Schedule

Recordings of arrahs to snakes were made, with a single exception, by releasing a snake or snakes in front of Main group. In the exceptional case, the monkeys discovered a boa in a tree, I video recorded the interaction, then captured that snake for a later controlled release. In a controlled release, I approached the monkeys, stayed with them for an appropriate time to make sure my presence was not alarming, then removed the snake or snakes from a cloth bag in front of the monkeys. In some situations where I wanted to obtain recordings from a particular animal, I made sure it was alone or had few companions before releasing a


snake in front of it. Usually the snake began moving away, often into trees. I pulled the snake down to the ground if it began to move beyond my reach. Typically, the monkeys immediately began calling and mobbing the snake when it was released. Sometimes many other animals approached while at other times only the animals that first saw the snake called. On a few occasions the monkeys did not apparently see the snake and left without vocalizing. On still other occasions they obviously saw the snake but did not immediately begin to call.

There were eight boa releases in the following order, listed by snake length: medium, large, medium, large and medium, small, large and medium, large and medium (to a lone monkey), medium and small. Additionally, a second snake species, Drvmarchon corais, was used in a single release. The number of cirrahs recorded for the nine snake releases ranged from 9-58.

Snake Descriptions

Two species of snakes were used: boas (Constrictor

constrictor) and Drvmarchon corais, a yellow phase indigo snake. The ]2. corais, a snake very similar to the indigo snake, was used to determine if C. olivaceus used alarm calls only at boas or whether they used a general snake call. I used four boas of three sizes (table 4) in combinations of small, medium, large, small and medium, medium and large, and medium and medium.


Table 4.
Snakes Used in Controlled Releases

Species Length (m) Weight (kg)
----------------------------------------Boa 1.95 4.65
Boa 1.60 2.00
Boa 1.58 1.94
Boa 0.68 0.20
D.corais 1.80 2.50


Responses by Monkeys to Snakes

I will first describe the release of a 0.68 m boa and provide examples of the grrahs recorded in response to it. The transcription below of my commentary on the video recording shows that while the calls are diverse, several animals give quite similar calls, and that some calls are given in only certain contexts. The commentary also provides a background for understanding the resulting investigations. The vocalizations referred to in the text are presented as spectrograms in appendix A. There were 87 calls during the release, of which 29 were arrah variants. It is clear from the spectrograms that these qrrahs are highly diverse, with many configurations. Five animals called during the 24 minutes interaction.

Snake release commentary

I release the small, 0.68m boa in front of Jefe as
he is sitting alone 4m above me eating a butterfly
larva. For the first 1 minutes he continues eating
without calling even though he appears to see the snake. once he finishes eating he begins calling,


JEl0l is his first grrah. He then gives a long series of huhs interspersed with a single grrah after a minute.
By the second minute other animals begin arriving and while other animals in the distance continue giving huhs an arriving animal grrahs, JE 118. By then there are three agitated monkeys over the snake, giving few grrahs. At 2:02 minutes into the interaction, Jefe gives cirrah JEll9, the first ground snake grrah variant (for a discussion of arrah variants, see below). For the next two minutes a juvenile female, Hanna, sits over the snake watching it without giving alarm calls. Jefe then gives another grrah, JEll7, which appears to prompt Hanna to intently look around, otherwise there is little action. Hanna returns to lying on a branch over the snake content in watching it. Other animals continue to huh in the background. overall the troop is relatively placid. Eventually Hanna and another small female moved.
Four minutes into the interaction Jefe gives
another ground snake arrah, JEl2l. Then, after another lag in action, Hanna gives another type of ground snake call, HAlO2, this one a loud brief call like a bark which she repeats 45 seconds later, HA 105. By now the snake, which was released on the ground, movs into water and neither the monkeys nor I can find it. Hanna gives arrah HA122 as she moves lower in a bush, intently trying to locate it.
A minute and a half later the snake begins to swim away, prompting Hanna to give cirrah HA 107. Other monkeys move in and give several desultory arrahs (that are too poor for analysis). Hanna then gives three calls, a loud cirrah, HA 108, followed immediately by HA 109, then another call like the loud call, HA 111. Next she gives four ground snake calls, JE 123, HA112, and JE 124 (one is masked by bird calls so it isn't analyzed), like Jefe's earlier ground snake call, JE119. The interaction is now 10 minutes long. The middle call, HA 112, prompts another monkey to intently look down. The boa then begins to climb into a vine. Hanna gives grrah HA125 followed by two similar calls, JE126 and HAllS. Notice the wide diversity in call types used by Hanna in these last 14 calls.
Over the next 2 minutes there are few calls, yet
Babas, a juvenile male, agitatedly brakes a branch near the snake. They do not appear to know where the snake is. The juvenile monkeys left after two minutes without more calling.
Three minutes after the last grrah Mo approaches, stopping frequently to look down to the ground where the boa had been before it moved into the vines. She is quite evidently looking for something. For two


minutes she looks for the snake, at which time I pull
it out of the vine and put it on the ground in front of her. She watches the boa for a minute without calling.
It is now seventeen minutes into the interaction.
Finally she gives a quiet grrah (not analyzed) and
another animal gives a huh. Her next call is MM101, a call quite similar to RA122 by Hanna. A half a minute later she becomes agitated, jumps, and then gives arrah
MM102 while looking down at the boa. Another animal huhs and she responds with MM103 as the snake swims
below her. Fifteen seconds later she gives arrah MM106, an alarm call used for snakes in a tree. I
cannot confirm the snakes location at that time, though
1h minutes later the boa was I meter up a vine. The
last calls recorded from Mo are the two similar calls, MM107 and MM108. At this last call, 30 seconds after
the tree snake call, she does appear to be looking
obliquely at the snake instead of down, i.e. the snake may be in a vine. She then begins to DILh at Hanna who
then rapidly moves down a vine emitting hehs. Hanna appears to direct the calls at the snake rather than
Mo. White Beard, the adult male of the group then
passes below her unperturbed by either the snake or Hanna's aggressive hehs. (On other occasions other
monkeys emitted hehs at boas).
Two minutes after the last grrah, Hanna is lying
on a vine and Mo is self grooming, both monkeys can
plainly see the small boa as it tries to climb a palm.
The interaction ends with Mo moving away, five minutes
after the last grrah.

I conclude from this interaction that 1) C. olivaceus

gave a variety of acoustically distinct alarm calls to

snakes, 2) the alarm call variability was apparently not

random, instead different animals gave apparently similar

calls, 3) the alarm calls Mo heard informed her of the

snake's presence, 4) Mo may have called even before she

herself saw the snake, and 5) the presence of a snake does

not automatically prompt a monkey to call.

By all appearances Mo was looking for the snake. I

should note that this snake release occurred during the


early rainy season, August 3, 1988, and there is standing water in many places. The monkeys rarely come to the ground at this time of year so it is highly unlikely that she was looking for food or other monkeys when she scanned the ground. This is noteworthy because the snake had probably not been visible for four minutes.

I interpret Mols behavior to indicate that she understood the various arrahs to refer to a snake. Alternatively she could have seen the snake earlier or seen the other monkeys looking down. Call Types

The fact that different animals made similar calls suggests that the variability represents meaningful differences to the animals. Testing this hypothesis requires a classification schema. I therefore devised a key based on seven arrah features:

1) Format frequency drops/ increases
2) Frequency drop is continuous/ discontinuous 3) Frequency drop is extensive/ not extensive
4) Number of formats
5) Which format was stressed
6) Stressing at end was above/ below the tail
7) Extensive frequency drop in formats other than F1
8) Call duration

These seven features use all three acoustic domainsfrequency, amplitude, and duration. I concentrated on the frequency domain because the calls appeared to vary the most in this domain, and, a Priori, appeared to be the most robust domain to differentiate the calls. Call duration was


clearly variable yet duration variability did not appear, in advance of the analysis, to discriminate the calls as well as frequency components. The only amplitude component examined was the relative location of stressing, that is, the position of maximum amplitude. I could not compare calls by their overall amplitude, because the calls were neither recorded at a standard distance nor did I use a calibration tone when recording.

The grrah variant classification schema was defined solely on the basis of the appearance of the spectrograms rather than how the calls sounded to me, however there is, of course, agreement between a spectrogram's appearance and its sound. I assume there are acoustically relevant features that my hearing may not attend to yet may be important to the animals. I further assume that these features will be evident on a spectrogram.

Using these features I parsed the 279 calls recorded during ten snake releases into 15 grrah variants. The key, including the number of grrahs in each variant, is presented in figure 4. Two calls were found only once each and were excluded from further analysis. Initially, as a control on individual variation of call variants, only those calls from known vocalizers were classified, then once a robust classification system was devised all suitable qrrahs were included. Spectrograms of representative samples of each grrah variant are presented in appendix B. While there is


l..0.0.0 Formants only drop

1.1..0.0 Formant level then drops

1.1.1..0 Freq. drop cont. and extensive One formant or F1 stressed27 F2 stressed 7

1.1.2..0 Freq. drop discont.and extensive 32

1.2..0.0 Formants drop continuously

1.2.1..0 One formant 17

1.2.2..0 Multiple formats F1 stressed 22 F2 stressed 8

1.3..0.0 Formant drop not extensive

1.3.1..0 F1 stressed Tail stressed below 36 Tail stressed above 17 F2 extensive drop 34 Fother extensive drop 6 1.3.2..0 F2 stressed, tail stressed below 27 2..0.0.0 Formants rise and fall 13

2.1..0.0 Formants rise and fall discontinuously 3 3..0.0.0 Others 18 Short duration 10

Total 277

Figure 4. Key to grrah variants by spectrographic configuration with totals for each variant.


some variability within call types, this schema does appear to group calls into similar appearing classes. Context of the Call

Once the grrahs were classified the specificity of

grrah usage was examined in three contexts: snake location, snake number, and snake size. Table 5 lists the context where each call was given, the number of different releases in which the call was given, and the number of different animals giving the call. The following call variants appear related to certain contexts:

Call Snake Context: On the Ground

1111 26/27 calls by at least 5 animals in 5/8
possible situations, i.e. a snake was on the ground in 8 releases, during 5 of which the
monkeys gave qrrah variant 1111. In the
single exception there were two snakes--one in a vine and the other on the ground. The
call was given by an animal off camera so it was not possible to determine to which snake
the call was directed, i.e. the snake may
have been on the ground.

3001 10/10 calls by at least three animals in 4/8
possible situations.

Overall, when a snake was on the ground 184 calls of 14 types were used by at least 13
different animals.

Snake Context: In Trees

2000 11/13 calls by at least 4 animals in 4/4
situations. In one exception, by Mo, the
snake may have been in a vine, it certainly
was soon afterwards.

13122 5/6 calls by one animal in 1/4 situations.

In total 83 calls of 12 types by at least 10
animals are given to arboreal snakes.


0 1


0 40
r-4 >4 f- IV1 1

N4 (n4 Ni

0.1 -4 1I

(44 N~ H -4 H- N4 N1 q C N rj N I
0 1
5. I4 N LA 0 V 0% N to r- %D 0o0 1-P4

Rl oi N H4 N N .4

-4 -q 4 4 -4 -4 -4 -4I


r4 m .4I C-4 m C w H It - NI
tv I4 1r4 u I N
1 0 r-4 m4 U, 9-4 w 0 m % (4 10 N 4 Cl 14
1IN N "-I r-4
m) + + + + + + I
0 o<~ N ra t 0 %D -w m in w m N- C m NI1
cEnIN N1 ml v N N 4W fn -W m N en ml en1
.~ 0 I -4
-4 -WIr P4 N o-4 V U, -w m4 V4 '0 O 0 %0 1r4
3 I 4 r4 r4 9-I '-4 N N4 r4

U) *++++ + + + + + I

M I- En 1l N N 1 4 m in I Cl N Cl l I m z I I

> I'0 U, 0 U, r N 0 v H4 N IV in r4 m.
I -4-4 v-4 f-4 -.4 I
.1+ + +
r4 I N r-4 N C4 l U, U, in Nn HI
1.4 I
004 N -4 .-4 ml ml ml N4 m -I ml m
54 E-4 II~
0 enI N Cl C Oc0 C 0
(44 4 v-4 H
1.1 + + + ++ +1
LA* VI i~ .4 %0 0 CN %0 ml %D H- (N ml ml 41U 1I I -4I

Q) 0I' %D r4 v %D U, N r, NV o '0 H. N I 4
1m I I
r. ~ H' U, 0 to Cl co w N CA m t-4 .-4 N 0 014 0 0 4ftI1N N r-4 i-I N .4 N O .1
U U 1 1
1". 4I N1r- r N Nl N CI %0 N N- V %D m (n co 011 0 0I N Cl -4 N Hl.- N ml .4 H- r.4I1
E-4 I4 00V-.4 N~
'-4 H .4 1.-H N 0 0 .-4 N- r-4 N -4 N N 0 0 0 H41 (U .0~ R r I-4 I N H. CN N H4 H N H- r-4 0 -4 0 0 1 41 (a t E>,,II H 4 H -I N1 N N C) (n ml ml m 0 H4 0 0 10 E4C U E4I r- I H -4 H- H4 r-4 -4 r-4 H- 9-4 1I N N mC m CIE-4


Snake Context: Near the Vocalizer

1222 7/8 calls by 4 monkeys in 3/5 situations.

Overall, 107 calls of 13 types by at least lo
animals. 3/8 calls given when snake was on the ground and 5/8 when snake was in a tree.

Snake Context: One Snake

13122 6/6 calls by 2 animals in 2 situations.

Overall, 151 calls of 14 call types by at
least 12 animals.

Snake Context: Two Snakes

2100 3/3 calls by 1 animal in 1 situation.

Overall, 121 calls of 14 types by at least 9

Snake Context: Medium Sized Snake

13122 6/6 calls by 2 animals in 2 situations.

In total, 101 grrahs were given by at least
13 different animals.

The evidence for other calls being given in particular

contexts is less compelling.

1122 5/7 calls to a snake on the ground, all by
unknown vocalizers in a single release; 2/7
in tree, by whitey in one release and another
by an unknown animal in another release. It
was also given 5/7 when a snake was near.
Its specificity is suspect because of small
sample size and it was given by several
animals with snakes in both locations.

1130 28/32 calls to a snake on the ground by at
least 6 animals in 4/8 situations, however
3/32 calls by 2 different animals in 2
situations were given to snakes in a tree.

1210 14/17 calls to a snake on the ground by at
least 10 animals in 6/8 situations. 3/17 calls were to a snake in a tree, all by a
single animals in one situation.


Calls by Age! Sex Class

No call given by more than one caller was given by only a single age/ sex class. There was, however, a significant skew to overall age/ sex class of all callers: 175/ 190 calls were given by adult females, subadult males and juvenile females. The remaining 15/ 190 calls by identified callers were given by adult males (2 calls), juvenile males (9 calls), and infants (4 calls).

There was no significant correlation (Kendall r= -.40, p=0.33) between dominance rank of adult females (the only age class for which sufficient data were available) and number of crrrahs given to released snakes. Dominance was defined operationally by social grooming patterns (T. O'Brien, pers. comm.).

Grrah Diversity

There was no clear pattern to the diversity of grrah usage. Females used a more diverse grrah vocabulary than males, with the Shannon diversity H Ifemaes=l.l3, whereas H'Ixtes=*9l (table 6). Overall diversity was H'=l.12. Notice that females used all variant types whereas males did not use five variants, including one indicating snake location, variant 3001. Notice also that males gave many fewer alarm calls than females, 41 vs 149. Two males, Stu and Griffin, produced almost half of the male data set. These two males are subadults, both born outside the group and approximately second and fourth in dominance among


Table 6. Grrah diversity by individuals.

Females Males
C all-- - - - - - - - - -
Type AHa MdMl MoPf PuWh Ba GrJe Mn St WB cf 9
---------------------------------------------------------1111 6 1 2 1 3 3 10
1112 1 1
1120 3 4 3 2 2 2 12
1210 3 5 1 1 1 1 1 1 1 4 11
1221 8 3 1 1 1 1 1 14
1222 2 3 1 2 2 6
1311 5 2 2 7 4 2 1 3 20
1312 2 3 3 2 2 7 5
13121 3 2 9 3 2 5 2 9 17
13122 5 0 5
1321 7 2 4 2 3 2 2 3 1 8 18
2000 8 1 1 1 2 9
2100 3 0 3
3000 3 2 5 1 0 11
3001 3 3 1 0 7
----------------------------------------------------------Tot. 15 22 1 26 47 10 7 21 2 7 9 4 7 10 2 41 149

H'= .45 0 .94 .67 0 .66 .47 0 .91
.87 .83 .65 .91 .56 .45 .59 1.13


males. White Beard, the dominant adult male, gave only two alarm calls. This low production of alarms was typical of him and apparently of other adult males, both for alarm calls and other vocalizations. Grrah Syntax

A matrix of grrah variants indicating call pairs is provided in Table 7. The data set is from all identified callers, with call variants modified from that provided in figure 4: variant 3000 includes 3001 and both 1112 and 2110 are excluded due to small sample size. One grrah followed another 158 times, approximately 78% of the time, so the most likely association is between arrahs. The most common association was call duplication, where the same call variant was simply repeated. This happened significantly more often than by chance (X2= 52.37, p= 2.3 E-7, d.f.= 11). This same statistic was not possible with other pairings because of the numerous null cells in the matrix, but it is clear that there are no strong associations between any calls. I conclude that except for duplicating calls C. olivaceus do not combine grrah types in a syntactic manner.

If C. olivaceus do not syntactically combine different qrrahs, do they combine calls to form a new call with intermediate form? This intermediacy in form could indicate that 1) the meanings of the two calls were combined, using phonological syntax or 2) that the signals are graded. If this is phonological syntax, I am suggesting that the


O 4 -4N -4N NIV Q Nr-4 N

0 W4., 4
0 1-4


N ~ L H N H


41 C

0 -r4. ~ Wr4 r
'-4 N Nm Hr4Cj c

N~. V-4-4H'-4 r


4i fn C N 14~

o -4

.-I N N V-4 4-q 4N

H 0 HN-4 NN -40
H1- N I r4H f4 r- I rdC% r-4 H r-4 H 1-4 r4 N


monkeys are combining elements of different calls to form a new signal, with a new meaning. This would be analogous to word formation in human language. For these calls to be graded (=continuous) requires that the new form follow a continuum where the meaning of the signal is intermediate between the meaning of the two calls from which it is formed. This raises questions that are essentially epistemological, for if these alarm calls are referring to objects, what could be the intermediate meaning, e.g. what is the intermediate meaning for the word snake? For this reason, I doubt whether the calls are graded. Notice however that the physical evidence for both positions, phonological syntax or graded signal, is the same--calls whose spectrographic configuration is intermediate between two other calls. Examining appendix B shows that several variants appear intermediate in form, for example, variant 1130 seems intermediate between variants 1111 and 1210. Even if they were intermediate in form, I have no data to indicate whether they were used in contexts that were in some way intermediate between the contexts in which the other calls were used, that is, that they had an intermediate meaning.

Grrahs to Other Snakes

In the single release in which another snake species was released, two monkeys gave three different call types, including one call, AD11, of qrrah variant 3001, the ground


snake call. It seems that g. olivaceus do not restrict alarm calls only to boas.

other Responses

There were other responses to released snakes besides

grrahs. For example, on four occasions monkeys hit a branch while they or another animal were calling and each time a different call was used: variants 2000, 1221, 1210, and 3001. At least twice monkeys knocked the snake out of a tree by hitting the branch it was on.


I conclude from the above data that C. olivaceus emit certain arrah variants when snakes are in particular locations: grrah variants 1111 and 3001 when a snake is on the ground, crrrah variant 2000 when it is in a tree, and variant 1222 when the snake is near the caller.

There is, at present, no convincing evidence that C. olivaceus communicate about snake numbers or size. I found the associations between grrah variants and number of snakes unconvincing, primarily because the sample size for each call was small and the calls were given by few animals. Grrah variant 13122 was given 6 times, of which 5 were given by Whitey, while variant 2100 was given only 3 times, all by Mo. A larger sample from more animals is needed before I am convinced of their specific utility. I feel that the specificity apparent in the usage of cirrah variant 13122 is


due to its limited application and we can conclude nothing about its usage.

Perhaps one of the most surprising aspects to the

capuchin's reactions to snakes was their calling at small boas in the same manner as they did larger snakes. For some reason the monkeys apparently felt threatened by a .68m snake. Thirty grrahs of nine different variants were given at it, twice as many as were given to the largest snake, which was almost three times as long and over twenty times as heavy.

Grrah variant 1130 was the second most commonly used grrah and apparently used almost exclusively to snakes on the ground. I remain skeptical, however, of concluding its specific usage because it was also used by two animals in two releases when snakes were in trees. Likewise, variant 1210 was used mostly toward snakes on the ground, but was also given to snakes in a tree.

The production of many of the other grrah variants was separated almost equally between one and two snakes or snakes on the ground and snakes in a tree.

Examination of these results shows that the above call variants are not always given when a snake is in the prescribed location. For example, call variant 1111 was given in only 5/8 situations when a snake was on the ground. The snakes location is apparently not a necessary condition for the utterance of a particular call. On the other hand,


the emission of a call is sufficient for determining the location of a snake. That is, whenever a monkey gives a arrah, any other monkey hearing it would know where a snake was located. Stated another way, the combined probability of grrah variant 1111 being given when a snake is on the ground is 5 of 8 whereas the probability that a snake is on the ground when arrah variant 1111 is given is 26 of 27.

The three vocational calls were not common; never more than 14% of grrahs given in a particular context. of course, meaning should not depend on the frequency of the call; a call given once would be sufficient to indicate location.

There were fewer calls to snakes in trees than to

snakes on the ground but the average number of calls of any variant per situation was approximately the same. For example, grrah variant 1311 was given 27 times in 7 situations (3.9 calls/ situation) when the snake was on the ground and 9 times in 3 situations (3 calls/ situation) when the snake was in a tree.

Examination of the various grrah variants in appendix B shows that the classification schema is robust, yet there is some overlap; differences between some variants is slight. These similarities may be due to their being syntactic combinations of separate calls or they may be, in fact, distinct calls representing different objects or emotions.


The next step in this translation process is to find out how a monkey will respond to these vocational call variants. For example, if grrah variant 3001 is always given to a snake on the ground, and in fact means 'ground snake' then we would expect a monkey to look down to the ground when it hears this call. The experiments described below will serve to confirm the above results.


Playback, Data Recording. and Analysis EcTuir)ment

These experiments required both playback and recording devices. Calls were played to the monkeys on a Sony TC-D5M audio cassette recorder over two Sony SRS-30a self powered speakers at the end of a 15m cable. The speakers were typically hidden in a green bag placed under leaves in a tree. The recording equipment was the same as that used in the snake releases: Sony video camera and Sennheiser directional microphone. The responses of the monkeys were scored in the field by watching the video recording of the playback through the camera (as a camcorder it was capable of both playback and recording). Results were verified by watching the playbacks on a larger TV.

Immediately after each playback a map was drawn of the experimental scene, indicating the location and height of the monkey, position of other animals, and the distance, angle and height of both the speaker and camera relative to the monkey. This map made it possible to judge where the monkey looked upon hearing a call: down to the ground, up



toward the sky, into foliage, toward the camera, toward the speaker, or toward other animals. Playback Methodology and Schedule Test stimuli

The schedule, call type, call, caller, and receiver for 21 alarm calls of four variants used in playbacks are presented in table S. Each call was selected for clarity from the set of calls recorded during snake releases. Test trials were designed to test the response of single animals to single alarm call playback. Control trials involved either playing 0.5 seconds of noise or setting up the equipment and running a trial with no signal as if a signal were played.

The playback tapes were produced by taking digitized examples of the chosen calls and recording them onto 10 second tape loops. Amplitude of each call was equilibrated by making the maximum amplitude of each oscillogram equal. A single call was presented in each playback.

The calls used in the playbacks were the two ground snake calls, grrah variants 1111 (5 replicates), 3001 (6 replicates), and the tree snake call, graah variant 2000 (7 replicates).

Playback methodology

The major goal in a playback experiment was to present as realistic a stimulus as possible. verisimilitude was essential.

Table 8. Responses to Alarm Call Playbacks.
a) Responses to call variants 2000 and 1111 Responses
Date Call Sp.Rec. Up/Sky Down Fol. Spkr. Res. Other
-----------------------------------------------------------ARBOREAL CALLS-2000
4/9 HQ125 WH HA X
4/9 HQ125 WH MI X
4/9 HQ125 WH STU X
4/26 MO126 MO AM X
4/26 MO126 MO BA X
4/26 MO126 MO GR X
4/26 MO129 MO Am X
4/27 MO129 MO WH X
4/27 MO129 MO BU X
4/30 MO126 MO ON,Am X
4/30 MO126 MO JE,STU X X
5/22 MG113 WH AM X
5/22 MG113 WH AM X
5/22 MG113 WH AL X
5/31 AB16 MI BA X
5/31 AB16 MI BU X
6/8 MO128 MO WH X
6/8 MO128 MO STU X
6/8 MO128 MO STU X
6/21 MO128 MO STU X
6/21 MO128 MO STU X
6/29 AB67 MO HA X
6/29 AB67 MO WB X
6/29 AB67 MO MO X
----------------------------------------------------------Totals 0 2 7 4 11 1

Ground Calls-lill
4/16 JM102 HA MD X
4/19 BB122 PU GR X
4/19 BB122 PU ML X
4/19 BB122 PU ML X
5/6 AF40 MD WH X
5/20 JE124 HA STU,MN XX
5/25 JE119 JE PF X
5/25 JE119 JE PF X
5/25 JE119 JE STU X
6/10 JE124 HA PF,ML X X
6/10 JE124 HA STU X
6/15 JE124 HA AM X
6/15 JE124 HA WB X
6/15 JE124 HA ML X
6/15 JE124 HA AM,Am X X
Totals 0 8 0 3 5 2-----------------------------------------------------------Totals 0 8 0 3 5 2


Table 8-continued.
b) Responses to call variant 3001.

Date Call Sp. Rec. Up/Sky Down Fol. Spkr. Res. Other
---------------------------------------------------------------Ground Calls-3001
4/22 HA105 HA ML X
4/22 HA105 HA ML X
4/22 HA105 HA MO X
5/4 BB117 PF STU X
5/4 BB117 PF STU X
5/17 BB117 PF BA X
5/17 BB117 PF ON X
5/29 AF27 WH ON X
5/29 AF27 WH AM X
5/29 AF27 WH PU X
6/13 HA105 HA MO X
6/13 HA105 HA WH X
6/13 HA105 HA WB X
6/13 HA105 HA STU X
6/13 HA105 HA WH X
6/17 BB118 PF AM X
6/17 BB118 PF AM X
6/17 BB118 PF BA X
6/17 BB118 PF ML X
6/23 HA105 HA MG X
6/23 HA105 HA BU X
6/26 HA105 HA STU X
6/26 HA105 HA STU X
6/26 HA105 HA WB X
7/6 AD12 HA STU X
7/13 AD12 HA WB X
7/13 AD12 HA AL X
7/15 AD12 HA PF X
7/15 AD12 HA ON X
7/15 AD12 HA ? X
7/15 AD12 HA AM,ON X
7/15 AD12 HA MO X
7/15 AD12 HA MO X
7/15 AD12 HA MO X
7/17 BB126 PF MG X
7/17 BB126 PF AL,ON X
7/17 BB126 PF AL X
Totals 1 6 1 0 22 7----------------------------------------------------------------Totals 1 6 1 0 22 7


Table 8-continued.
c) Responses to waahs and control trials

Date Call Sp. Rec. Up/Sky Down Fol. Spkr. Res. Other

4/24 WA104 ? STU X
5/2 WA101 BR GERT X
5/18 WA104 ? FI X
5/18 WA104 ? FI X
5/22 WA103 GERT JUV d X
5/23 WA103 GERT FIL X
5/30 WA101 BR FF,GERT XX
5/30 WA101 BR IR X
6/12 WA104 ? 9 X
6/16 WA101 BR JUV d X
6/18 WA101 BR JUV C X
6/18 WA101 BR JUV d X
7/14 WA104 ? Juv X
-------------------------------------------------------------Total 8 0 0 3 3 0

Control Trials
4/16 EQ. MAIN X
4/22 EQ. MAIN PU X
6/7 EQ. MAIN X
7/14 NOISE RED 9 x
---------------------------------------------------------------Total 0 0 0 6 10 2
----------------------------------------------------------------Up/Sky= animal looks up or to open sky; Down= animal looks down; Fol.= animal looks to foliage; Spkr.= animal looks to speaker; No Res.= No response; Other= Other responses.


once the subject animal for a playback was chosen, I followed that animal until the conditions for a playback were appropriate. I tried to set up each trial before a monkey was near, though this was at times difficult and the trials could not always be run. Speakers were always set in trees, never on the ground, because this was the typical position of a calling monkey and signals broadcast better off the ground. I ran a trial when the experimental subject

1) was relatively near the speaker,
2) in clear view and roughly facing the camera,
3) the speaker was out of sight of the monkey,
4) the animal producing the signal (recorded the
previous year) was out of sight.

If these conditions were met I then started filming the animal and then played the tape loop and continued filming for at least 30 seconds after the call. This playback protocol was developed in 1988, with Red group. Data taken during that period were not used in this analysis.

Playbacks often were to only one or a few animals,

which meant that other animals would not hear the playback. Thus, in some cases it was possible to do several playback experiments in a single day. I never repeated the same trial with the same animal on the same day.

For procedural reasons dealing with interpretation

difficulties, grrah variant 1222, the call given when snakes were near, was not tested. When considering a playback experiment it is instructive to think in geometrical terms. Playback experiments form a triangle, the three points being


the monkey, the alarm source (speakers), and the recordist. Ideally where the monkey looks--the response variable--is isolated from either of*the other two triangle points. For example, in a playback of a ground snake alarm call the monkey looks down, which is a different direction than either to the recordist or sound source. Unfortunately, in a playback of a near snake call, this condition could not be met. When the monkey hears this call, it should look near the sound source and determining whether it was looking at the sound source or near it for a snake would be difficult. For this reason, I did not test this call.

Playbacks of grrahs were done to 20 of the 21 available animals in Main group, the lone exception being Winston, a juvenile male. Playbacks of waahs were done to eight different animals in Red, White, and Main groups. Response Variable

This set of experiments tested where a monkey looked upon hearing an alarm call. The measured response was the direction the animal looked upon presentation of the stimulus. Response criteria were:

1) Up or to the sky- The animal looks toward the open
sky, either directly above it or towards a break
in the canopy.

2) Down- The animal looks down- either directly below
it or towards the ground?

3) Foliage- The animal looks to adjacent foliage,
either by looking around itself or looking up into

4) Speaker- The monkey looks toward the speaker.


5) No Response- The animal did not react to the
signal. Did it change its behavior,particularly
where it was looking, immediately upon
presentation of the signal?

6) Other- Were there other responses, such as looking
towards other monkeys, at the camera (me), or at
other animals?

Experimental Limitiations

There were several limitations to this experimental

design. First, regarding experimental replication there is a requirement that each event be independent. To produce independent replicates of these tests I could repeat them with another group of monkeys. Playing the calls used in these experiments back to another group of monkeys however might produce the confounding factor of the animals recognizing the vocalizations as not being from a member of their own group and responding to the stranger's voice rather than the message of the signal. Simple repetitions of the same stimulus only increase sample size. Replicates within the monkey group were accomplished by using a number of different stimuli given to a number of different animals. This design therefore avoided pseudoreplication. The playback design used here most closely follows design 1D among those described by Kroodsrna (Table 1, pg 601, 1989).

Another limitation was the inability to quantify a control response: responses to natural alarm calls. The monkeys did not always respond to the stimulus. Did they respond at a different rate to alarm calls given by actual


monkeys than to playbacks, or were they habituating to the presentations? To answer this requires a video record of the monkeys response to actual alarm calls. I have only one such record, where Jefe immediately looked in the direction of a monkey that called at an approaching tayra. I have no records of animals ignoring natural alarm calls. Lacking such records, I must judge whether these experiments represented realistic situations. Did the animals respond as if another animal had seen a snake and then give a call in alarm? Based on the behavior of responding animals compared to responses to actual predators I feel these responses were real, that the animals understood the calls and responded as if another monkey was giving alarm calls at a snake.

When a monkey did not respond, was this an indication that it had not heard the call? Was it not concerned about the snake? Lack of concern could be due to habituation or as a result of only a single call being heard rather than several. Lack of response is a difficult problem, particularly when I have no measure of its frequency in natural situations. I will assume that the specimen is the authority and have used as my data set only those responses where the experimental subject was apparently looking for a snake. I have not used responses where monkeys looked toward the speaker, other responses, or no responses.



The results of the 111 playbacks are listed in table 8 above. There were adaptive responses to waahs and various grrahs 33 times, after discounting the "no responses", "looks to speaker" and "other responses." Grrahs

Responses to playbacks of the three grrah variants

indicating snake location are given in table 9. A Fisher's exact probabilities test indicates there is a highly significant difference in response probabilities (p< 0.00214), indicating that the monkeys respond differentially to each alarm call.

Table 9.
Responses to Playbacks of Grrahs

Responses Calls
2000 1111 3001
------------------- ------------------------Look Down 2 8 6
Look to Foliage 7 0 2
------------------- ------------------------Totals 9 8 8


The question arises as to whether response differences are due to factors related to signal configurations such as signal duration, and intensity, or to other factors such as the age, sex, familial or social relationship of the caller and respondent. I found no correlation between signal


duration and percentage of correct responses (T=0.1521, Kendall rank correlation). Since the signal amplitudes were equilibrated there is no influence of amplitude differences on responses.

As for response rates to the various replicates,

within-call variant mean response rates for playbacks (for calls used more than once) are provided in table 10.

Table 10 '
Response Rates for Calls and Individuals Calls Individuals
------------ -----------Mean Range Mean Range

Grrah 1111
39% 33-50% 40% 33-55%

Grrah 3001
24% 0-50% 26% 17-33%

Grrah 2000
38% 0-67% 39% 33-50%

66% 40-100% 66% 40-100%

Variation by caller. I will first examine response variations relative to the context of the caller. There were insufficient data to statistically compare differential responses by sex, familial relationship, or social dominance of the caller.

There were only five playbacks of calls from males,

which is too small for further analysis, whereas there were


73 playbacks of calls by females. Within the sample of female callers, there were no response differences by caller age-class, with 11/25 correct responses to juvenile females and 11/25 correct responses to adult females. (As a matter of definition, correct is used here to mean that the animal appropriately responded to the alarm call, which typically was to look for the predator.) There do not appear to be differential responses to females according to social dominance, though the sample is small. Of the 73 playbacks using female callers, 68 were done with the calls of four monkeys, three adult females (Whitey, Mo, and Pointy Face), and a single juvenile (Hanna). Whitey and Hanna are dominant within their age class, while Mo is intermediate between Whitey and Pointy Face, the lowest ranking adult female. Animals correctly responded to calls of the two dominant animals in 30% of the playbacks, 35% to the intermediate animal and 27% to the lowest ranking monkey.

There was only a single playback with related animals. Mo did not respond to a call by her youngest son, Mike. of the four females, Mo had the most offspring in the troop, Margo, Malli, Mike, and Modem, whereas Whitey had only a single son, Winston present. Pointy Face's only surviving offspring was a newborn in 1989, while Hanna had not reproduced.


Variation by respondent. Next we examine the

character of variations among respondents, by sex, ageclass, and familial and social relationships.

There was a significant difference (p=0.08, two-tailed Mann-Whitney test) in correct response rates between sexes of respondents by age class (adult, subadult, and juvenile). That is, males in three age classes responded correctly significantly more often than did females. There was no such relationship by sex of individuals outside of age-class.

The question next arises as to whether certain animals with many relatives in the group responded correctly more often than animals with few relatives. For example, did the members of Mols family respond differently than the unrelated group of males? Mols family contained five members, none of which were fathered by a male currently in the group. When Mols families rate of response is compared to that of the four presumably unrelated males (White Beard, Jefe, Griffin, and Stu) we find that there is no significant difference (p=0.53, two tailed Mann-Whitney) in response rate. This result is not surprising given that there was only the single test of an animal responding to a related animal's call. This explanation also assumes that the monkeys recognize each others' voices. A better test would compare data from responses to calls from related animals to those from unrelated animals. However if animals do not recognize each others' voices this result is somewhat


surprising, since one would expect that related animals would respond more often to each others' calls. This result can, by extension, be construed to further indicate that animals do perhaps recognize each others' calls.

Lastly, if we examine the relationship between social

dominance and rate of correct response within the five adult females we see that there is no correlation (r=0.4122, Kendalls rank correlation).


Another series of playback experiments was undertaken using waahs. The same overall playback protocol was followed. Fourteen playbacks were done using three calls given by animals in two different groups (Table 8). In eight of the fourteen playbacks the respondent reacted by looking up to the open sky, three times they looked towards the speakers, and three times there was no response. I conclude from these experiments that when the monkeys responded to these calls as alarms they reacted in every case, 8/8 times, by looking up for an avian predator. ResRonse variations

Response variations by sex of the respondent is

interesting. In the eight calls played back to males, they responded by looking to the sky on three occasions, whereas females responded each of the five times. I was unable to sex one respondent. When the caller was a male, monkeys correctly responded 4/7 times while for females it was 2/2