American Journal of Primatology 2153-60 (1990) BRIEF REPORTS Discrimination of Chirp Vocalizations in the Cotton-Top Tamar in KIM BAUERS AND CHARLES T. SNOWDON Department of Psychology, University of Wisconsin, Madison Natural exemplars of two types of chirp vocalizations were presented to eight groups of cotton-top tamarins (Saguinus 0. oedipus) in a playback paradigm. The two chirps share many acoustic features in common but are used in entirely different circumstances. The tamarins gave different vocal and behavioral responses to the playbacks of each chirp type, indicating that a discrimination was made. The results suggest that tamarins can discriminate subtle acoustic cues which have communicative significance. In the absence of the nonvocal cues and social contexts that normally co-occur with these vocalizations the tamarins responded with the same vocal responses they would have given in normal social contexts. Significant differences in post-stimulus behavior to the two types of stimuli appeared in the first two blocks of trials, indicating that the playback paradigm can be an efficient means of testing discrimination between different call types. Key words: playback paradigm, vocal communication, natural response measures INTRODUCTION A major question in vocal communication research is whether the various call categories assigned by human investigators are isomorphic with the perceptual classifications used by the animals [Snowdon, 19791. Only a few studies on primates have examined how animals perceive different call types: “coo” calls of Japanese macaques (Macaca fuscata) [Petersen et al., 1978, 1984; Zoloth et al., 19791; loud calls of gray-cheeked mangabeys (Cercocebus albigena) [Waser, 19771; alarm calls of Goeldi’s monkey (Callimico goeldii) [Masataka, 19831; trills of pygmy marmosets (Cebuella pygmaea) [Snowdon & Pola, 1978; Snowdon, 19871; grunts of vervet monkeys (Cercopithecus aethiops) [Cheney & Seyfarth, 19821; screams of rhesus macaque (Macaca mulatta) [Gouzoules et al., 19841. Most of these studies have tested perceptual classification by using a playback Received for publication October 13, 1989; revision accepted December 22, 1989. Address reprint requests to Kim Bauers, Wisconsin Regional Primate Research Center, 1223 Capitol Court, Madison, WI 53715; or Charles Snowdon, Department of Psychology, University of Wisconsin, 1202 West Johnson St., Madison, WI 53706. 0 1990 Wiley-Liss, Inc. 54 I Bauers and Snowdon technique. The value of the playback paradigm lies in its ability to isolate the information conveyed by a vocal signal from the nonvocal cues and social contexts that co-occur with vocalizations in normal social interactions. If there are behavioral measures that differentiate between the two types of vocalizations in a playback paradigm, then one may conclude that the acoustic differences in the structure of the vocalizations are sufficient for discrimination. If the behavioral responses are also appropriate to the situations in which these vocalizations are normally given, then the structure of the call also conveys a semantic message independent of other contextual cues. Some studies have used a n operant paradigm, but the constraints of the operant environment may alter responsiveness to the stimuli, especially if animals must perform tasks that are unrelated to the responses a n animal might exhibit in a natural social environment. For example, Symmes & Newman  used a n avoidance learning task to show that squirrel monkeys could discriminate between isolation peeps of different individuals, and Petersen et al. [1978, 19841 used a lever-releasing paradigm to study discrimination of two different types of coo vocalizations in Japanese macaques and other species. In these studies discrimination was eventually shown, but the squirrel monkeys required several hundred trials and the Japanese macaques appear to have been given upwards of 30,000 trials. Thus the operant technique requires lengthy training and testing of animals. Some primate species are less amenable to extensive handling for operant testing than others, and these methods may be impractical or disruptive with socially housed primate groups. We have studied the discrimination of chirp vocalizations in captive cotton-top tamarins by using playback techniques typically used in field studies. By housing animals in normal social groups, we could observe and record spontaneous responses to stimuli. By studying captive animals we had control over which animals were present for a test, and we could record lower-intensity vocalizations and subtle behavioral changes. The vocal repertoire of the cotton-top tamarin has eight chirp vocalizations (short frequency-modulated calls). Chirps can be described in terms of four acoustic variables, and each type is used in different contexts [Cleveland & Snowdon, 19821. We chose to study the two chirp types most nearly similar in acoustic structure to explore the limits of discriminability by tamarins. F-chirps and G-chirps differ only in mean frequency and in the peak- to end-frequency range. The two calls are given in different behavioral contexts and elicit different responses from cagemates when each is emitted spontaneously. The G-chirp is typically given by animals that appear to be relaxed and engaging in environmental exploration. The F-chirp alerts other animals and is given in response to conspecific groups heard from other rooms. In this study we monitored the behavioral and vocal responses of tamarins to playbacks of each chirp. METHODS Subjects and Housing Eight pairs of adult cotton-top tamarins were tested. Three of these pairs had one or more offspring. All but one of the animals were laboratory-born descendants of a colony established before the species was designated as endangered. The animals were housed in large cages (0.9 x 1.5 x 2.3 m to 1.8 x 3.1 x 2.3 rn) fitted with nest boxes and numerous branches and ropes to provide arboreal pathways. Additional details of housing and husbandry are presented in McConnell & Snowdon [19861. Chirp Vocalizations in Cotton-Top Tamarins / 55 TABLE I. Acoustic Parameters of Chiru Vocalizations (kHz, Mean ___ ~~ Parameter Initial frequency End frequency Freauencv range Initial frequency End frequency Frequency range ~ f S.D.)* ~ F-chirD all subiects F-chin, stimulus 4.7 ? 0.6 3.6 & 0.4 1.5 ? 0.3 4.4 2.9 1.4 Single or first of two G-chirps Second G-chirp G-chirp stimulus 8.9 & 0.5 5.8 ? 0.3 3.1 ? 0.4 6.9 ? 0.3 5.1 ? 0.5 1.8 ? 0.4 6.9 5.0 1.9 *The parameters are based on 12 individuals for the F-chirps and 14 individuals for the G-chirps. Acquisition and Preparation of Stimuli Chirp vocalizations were recorded from each adult in the study by using a JVC KD 1636 cassette recorder with a Sennheiser ME 80 microphone with a K3U power module. Several examples of each chirp that were recorded from each of the eight pairs were measured to provide structural parameters for the population. For the playback experiments we used chirps produced by a single animal living in a group that was not tested. Thus, each subject was tested with identical stimuli, and each animal was unfamiliar with the calls of the animal that provided the stimuli. Table I presents the structural characteristics of the stimuli used in the experiment and of the population of calls recorded. We elected to use a G-chirp that matched the parameters of the second G-chirp in a series to minimize the acoustic differences in comparison with the F-chirp. Thus animals were presented with the F-chirp and G-chirp stimuli that were most similar to each other. If a discrimination is made between these two stimuli, then more diverse exemplars of each class of calls should be discriminable as well. Figure 1presents the sound spectrograms of the stimuli used. Stimuli were processed through a signal-processing software program (VOCAL) to attenuate low-frequency noise and to adjust the peak amplitudes to match those in spontaneously given calls. The stimuli consisted of a sequence of two identical chirps separated by a 600 ms inter-chirp interval. This interval was chosen based on inter-chirp intervals observed when chirps were produced in sequence spontaneously. Since the stimuli were very brief, we wanted to be certain that if one stimulus were masked by cage noise or other vocalizations, the second would be likely to be heard. Experimental Procedure The JVC recorder was connected via a Realistic amplifier to one of two speakers (Audax HD 13D34 with frequency response to 15 kHz) that were mounted in each cage. A Sony TC-158SD cassette recorder and Sennheiser ME 80 microphone were used to record the vocal activity of a pair. Each of the eight pairs was presented with four trials within each block, two with the F-chirp and two with the G-chirp. Within each test session there were two playback trials using the same stimuli separated by 15 min. The order of presentation of F-chirps and G-chirps was alternated from one session to the next. Four blocks of trials were conducted for a total of 128 trials. No pair was tested more often than twice a week. All of the subjects were habituated to the observer. They ceased to attend visibly to the observer within 5-6 min of her entry to the room. Five minutes of 56 I Bauers and Snowdon G CHIRP F CHIRP 500 MS Fig. 1. F-chirp and G-chirp stimuli (Kay Sonagraph Model 6061 B, 160-16,000 Hz frequency range, narrowband filter). blank tape ran prior t o the auditory stimulus. A variety of behavior and all vocalizations of the animals were recorded for 5 min before and 5 min after the test stimulus. Vocal activity was recorded continuously. Because it was not always possible to determine unequivocally which animal produced a call, the pair was considered as a unit in evaluating vocal response. A focal animal point-sampling system in which the male and female alternated as focal animals was used t o score non-vocal behavior at 30 s intervals. Behavior Analysis The 30 s point samples from the 2.5 min preceding and following the stimulus presentation were used for analysis. An immediate change in visual orientation was expected to occur immediately after presentation of either stimulus. Based on prior observations we expected the behavioral categories of “freezing, visual scanning, and piloerection”; “scent marking and nose rubbing”; and “visual orientation to other groups” to increase after an F-chirp playback. The combined categories of “eating, drinking, and foraging” and of “grooming, huddling, and sitting” indicate relaxed and affiliative contexts and were not expected to change following an F-chirp. Other than a brief change in orienting response, no change in the frequency of any behavior or vocalization was expected in response to the G-chirp stimulus. The proportion of scan samples in which an animal moved was used as an index of activity. Vocal Analyses The frequency of emission of all major classes of calls in the adult repertoire was analyzed for the 2.5 min before and after stimulus presentation. The record- Chirp Vocalizations in Cotton-TopTamarins I 57 ings from each session were input to a Spectral Dynamics Real Time Analyzer (Model SD301D-C) with an analysis range of 0-20,000 Hz at a sampling rate of 60,000 Hz. The output of the analyzer was displayed on an oscilloscope screen and filmed on a Grass CR4 Kymograph Oscilloscope Camera. From the continuous strip of film it was possible to identify all calls by using the vocal categories defined by Cleveland & Snowdon [19821. Based on observations of vocal activity after spontaneous F-chirps, we predicted increases in F-chirps and in the combined category of “squeaks and flat whistles.” Increases were also expected in “normal and quiet long calls.” Three types of calls were predicted to remain at the same level after both stimuli: “modulated multi-level calls,” “initially modulated whistles,” and G-chirps. These are locational or contact calls to which no response is given. Because the G-chirps of juveniles could not be discriminated from those of adults, all G-chirps were counted. The total number of vocalizations given in the 30 s before and after the stimulus was analyzed to detect any change in the animals’ overall rate of vocalizing following a playback. The Wilcoxon Test was used to evaluate differences in response to F-chirp and G-chirp stimuli. RESULTS The results from each of the four blocks were evaluated separately. By the third block there was habituation to the test stimuli. Thus, the data analyses are based on Blocks I and 11, the mean of four trials with each stimulus for each subject pair. There was a significant change in orienting responses to both stimuli at the stimulus playbacks in comparison to both 30 s pre-stimulus and post-stimulus levels (P‘s < 0.005 for each, Wilcoxon Test). The orienting response rarely occurred except immediately following stimulus presentation. The orientation response was greater to an F-chirp stimulus than to a G-chirp stimulus (82% of trials vs. 47% of trials, P < 0.005). The combined category of “freezing, scanning, and piloerection” indicative of arousal and vigilance occurred significantly more often after an F-chirp than a G-chirp stimulus (P < 0.01; Wilcoxon Test, 71% vs. 22% of trials). Visual orientation toward the hallway or ceiling also increased significantly following an F-chirp stimulus (P < 0.025; 10% to 18%),and there was an increase in movement following an F-chirp (P < 0.02, 8% to 17%).There were no differences to the two stimuli in the remaining behavioral measures. Following an F-chirp presentation there was a significant increase in the number of F-chirps (P < 0.005, Fig. 2). The rate of F-chirps following an F-chirp stimulus increased 358%in contrast t o the pattern following a G-chirp stimulus (P < 0.01). There was also an increase in the number of “squeaks and flat whistles” following an F-chirp playback (P < 0.0051, and in the total number of vocalizations (P < 0.01). “Normal long calls and quiet long calls” occurred too infrequently during test sessions to evaluate statistically. The three categories of calls that were not predicted to change as a function of stimulus presentation (“modulated multilevel calls,” “initially modulated whistles,” and G-chirps) were produced a t constant rates throughout the tests. DISCUSSION The cotton-top tamarins discriminated between exemplars of F-chirp and Gchirp stimuli in a playback paradigm. The responses agreed with our prior observations of the behavior associated with spontaneous production of these calls. These brief stimuli differed in initial frequency and peak- to end-frequency range 58 I Bauers and Snowdon RESPONSE TO G - CHIRPS (4 4 1 F - CHIRPS 3i . I SQUEAKS / FLAT WHISTLES IT 2 1 n TOTAL CALLS PRE POST DIFF PRE POST DIFF Fig, 2. Vocal responses to chirp stimuli. A. F-chirp responses. B. "Squeaks and flat whistle" responses. C: Overall frequency of vocalization. Histograms represent the mean number of calls in the pre-stimulus interval (left),and post-stimulus interval (center).The right histogram in each set presents the pre-stimulus to post-stimulus difference scores. but shared many other acoustic features (Fig. 1).The acoustic stimuli alone, in the absence of contextual cues which might otherwise elicit these responses, appear t o provide sufficient information for animals to discriminate between chirps and to make semantically appropriate responses. Natural response measures, appropriately documented, can be an effective means of assessing the validity of categories that the human observer recognizes. Chirp Vocalizations in Cotton-Top Tamarins I 59 The tamarins produced increased F-chirps and “squeaks and flat whistles” and the total number of calls increased following the F-chirp stimulus. Tamarins also showed more orientation and more vigilant behavior such as “scanning, freezing, and piloerection” as well as greater locomotor activity after a n F-chirp stimulus. The fact that orientation a t the stimulus was significantly increased to both Fchirps and G-chirps indicates that G-chirps were perceived by the tamarins even though there were no other changes in behavior or vocal activity after a G-chirp stimulus. While it was possible to show a significant discrimination of call types with only four trials per stimulus, there was also a rapid habituation of response to the playback trials. By the third block of trials, there was a reduced tendency to orient to the test stimuli. Thus, while a playback technique is efficient compared with traditional operant techniques, there are potential drawbacks with playback techniques. Nonetheless, the playback procedures used here allow the evaluation of behavioral responses to biologically relevant sounds by unrestrained captive primates in a normal social environment. This technique is very useful for understanding how other species partition their communicative sounds and helps human observers validate their understanding of the contexts in which various sounds are used. CONCLUSIONS 1. Two forms of chirp vocalizations with few acoustic differences (F-chirps and G-chirps) were presented in a playback paradigm to eight pairs of cotton-top tamarins. 2. The tamarins oriented to both types of stimuli but showed predicted increases in specific vocalizations (F-chirps and “flat whistles”) and increased vigilant behavior following playbacks of F-chirps. These behavioral changes did not occur following G-chirp playbacks. 3. The behavioral and vocal responses to the playbacks of each chirp type were similar to the behavior and vocalizations following production of spontaneous chirps. This indicates that the acoustic information in each chirp may be sufficient to elicit appropriate responses in the absence of other contextual cues. ACKNOWLEDGMENTS This research was supported by a National Science Foundation Graduate Fellowship to K.B. and by USPHS Research Grant MH 29,775 and Research Scientist Award MH 00,177 to C.T.S. We are grateful for Jayne Cleveland’s long-term observations and analyses of the cotton-top tamarin vocal repertoire. We thank Kevin Colwell for electronic assistance and Jack P. Hailman and anonymous reviewers for critical reading of the manuscript. REFERENCES Cheney, D.L.; Seyfarth, R.M. How vervet monkeys perceive their grunts: Field playback experiments. ANIMAL BEHAVIOUR 30:739-751, 1982. Cleveland, J.; Snowdon, C.T. The complex vocal repertoire of the adult cotton-top tamarin (Suguinus oedipus oedipus). ZEITSCHRIFT FUR TIERPSYCHOLOGIE 58:231-270,1982, Gouzoules, S.; Gouzoules, H.; Marler, P. Rhesus monkey (Mucucu muluttu) screams: Representational signalling in the recruitment of agonistic aid. ANIMAL BEHAVIOUR 32:182-193, 1984. Masataka, N. Categorical responses to natural and synthesized alarm calls in Goeldi’s monkey (Cullirnico goeldii). PRIMATES 24:40-51. 1983. McConnell, P.B.; Snowdon, C.T. Vocal interactions between unfamiliar groups of cap- 60 I Bauers and Snowdon tive cotton-top tamarins. BEHAVIOUR 97: 274-296,1986. Petersen, M.R.; Beecher, M.D.; Zoloth, S.; Green, S.; Marler, P.; Moody, D.B.; Stebbins, W.C. Neural lateralization of vocalizations by Japanese macaques: Communicative significance is more important than acoustic structure. BEHAVIORAL NEUROSCIENCE 98:779-790, 1984. Petersen, M.R.; Beecher, M.; Zoloth, S.; Moody, D.; Stebbins, W. Neural lateralization of species-specific vocalizations by Japanese macaques (Macaca fuscata.) SCIENCE 202:324-327,1978. Snowdon, C.T. Response of nonhuman animals to speech and to species-specific sounds. BRAIN, BEHAVIOR AND EVOLUTION 16:409-429, 1979. Snowdon, C.T. A naturalistic view of categorical perception. Pp. 332-354 in CATE- GORICAL PERCEPTION. S. Harnad, ed. New York, Cambridge University Press, 1987. Snowdon, C.T.; Pola, Y.V. Interspecific and intraspecific responses t o synthesized pygmy marmoset vocalizations. ANIMAL BEHAVIOUR 26:192-206, 1978. Symmes, D.; Newman, J.D. Discrimination of isolation peep variants by squirrel monkeys. EXPERIMENTAL BRAIN RESEARCH 19:365-376,1974. Waser, P.M. Individual recognition, intragroup cohesion and intergroup spacing: Evidence from sound playback to forest monkeys. BEHAVIOUR 60:28-74, 1977. Zoloth, S.; Petersen, M.R.; Beecher, M.; Marler, P.; Green, S.; Moody, D.; Stebbins, W. Species-specific perceptual processing of vocal sounds by Old World monkeys. SCIENCE 204370-873,1979.