Introduction

The study of sound signals has played an active role in biological conservation over the last few decades (Laiolo 2010; Pijanowski et al. 2011; Sueur et al. 2008a). This includes traditional applications, such as aiding the taxonomic identification of live specimens (e.g., Ruppell 2010; Thinh et al. 2011) and population surveys (e.g., Dacier et al. 2011; Peck et al. 2011; Savage et al. 2010; Thompson et al. 2010). It also includes more progressive applications, such as measuring home ranges and territory sizes (Kalan et al. 2016; Ringler et al. 2011), aiding decision making in conservation policies (Simões et al. 2014), testing species occupancy trends and differential use of the landscape (Figueira et al. 2015), and estimating the effects of human-induced environmental changes on animal communication and behavior (Laiolo 2010). Analysis of sound signals can also provide insights into animal welfare and be used to evaluate the efficiency of ex situ conservation measures by identifying animals’ positive or negative emotions (Briefer 2012). However, such insights are possible only when reliable records of the behavioral contexts in which signals are given in wild populations are available (Briefer 2012).

Callitrichids often possess diverse vocal repertoires (Bezerra and Souto 2008; Cleveland and Snowdon 1982; Masataka 1982; Moody and Menzel 1976) that are associated with distinct behavioral contexts such as foraging, contact, alarm, and mobbing (Campbell and Snowdon 2007; Elowson et al. 1991; Kirchhof and Hammerschmidt 2006; Schrader and Todt 1993). The pied tamarin (Saguinus bicolor) is a callitrichid with a very small geographic range in central Brazilian Amazonia. Distributed in Manaus, Rio Preto da Eva and Itacoatiara municipalities in the State of Amazonas, Brazil, its current range encompasses a total of 7500 km2 of primary terra firme forests, secondary growth forests, and native forest remnants, many of which are in urban and periurban areas (Gordo et al. 2013; Röhe 2006). Urbanization and other forms of habitat loss in a small natural range restrict individual groups to isolated forest fragments with consequent effects on gene flow between populations, inbreeding depression, and loss of individuals to road accidents (Farias et al. 2015; Gordo et al. 2008, 2012, 2013). Another potential threat includes asymmetric competition with the golden-handed tamarin (Saguinus midas) along the northern boundary of the pied tamarin’s range. Increasing evidence suggests this is reducing the pied tamarin’s geographic range (Ayres et al. 1982; Hershkovitz 1977; Röhe 2006; Subirá 1998). The species is classified as Critically Endangered because of these threats (BRASIL 2014).

Available ethological information on pied tamarins includes space use, diet, cognition, interactions with birds and with golden-handed tamarins, and chemical communication (Azevedo 2006; Egler 1986, 1991, 1996; Epple et al. 2002; Sobroza 2015). A limited number of their sound signals have been described (Egler 1986), but no investigations of the behavioral contexts of their emission. In this study, we aimed to characterize the vocal repertoire of this species. To achieve this, we sampled the acoustic signals of free-living and captive pied tamarins and analyzed their spectrotemporal features as well as the behavioral context in which calls were emitted.

Methods

Study Area

We recorded data from eight free-living, one rescued, and one caged group of pied tamarins. We followed one of the free-living groups, which was habituated to human presence, to collect information about the contexts in which calls are emitted. The habituated tamarins live in Parque Municipal do Mindú, in the city of Manaus, State of Amazonas, Brazil (03°04′54′′S, 60°00′13′′W). This reserve is part of a 46-ha forest fragment surrounded by residential and commercial urban areas. It comprises a mosaic of terra firme (not seasonally flooded) streamside rainforest, and patches of secondary growth forest. The reserve is open to the public all week, 06:00–17:00 h. Thus, tamarins are accustomed to the presence of human observers in their habitat. At least three wild groups of pied tamarins live in the park, comprising three, five, and six individuals. We followed the group of six individuals.

We recorded sounds made by the remaining seven free-living groups during a study of interspecific behavioral interactions (Sobroza 2015). These unhabituated groups live in both secondary and primary growth continuous forests in the municipalities of Manaus (Ducke Reserve: 02°57′27.76′′S, 59°55′26.23′′W; Pau Rosa Road: 02°43′35.44′′S, 60°08′58.52′′W) and Rio Preto da Eva (Carapanã-Açu Farm: 02°37′00′′S, 59°37′56.90′′W).

We obtained additional recordings from two infant male tamarins that were rescued by the Projeto Sauim-de-Coleira at Universidade Federal do Amazonas, Manaus, after the adult tamarin carrying them was killed on a road near the campus (3°4′56.63′′S, 59°57′30.14′′W). The infants were kept in holding boxes, hand-fed by the project staff and received veterinary care. The infants were likely under three-quarters grown, since they still had a fine covering of white hair on their faces (the head is naked in adults; Hershkovitz 1977).

We also recorded seven individuals from a group that had been captured in the forest fragment on campus (03°04′34′′S, 59° 57′30′′W) as part of a radio-tracking study conducted in June 2014 (for details see Gordo 2012). The group was captured with baited Tomahawk TH105 (10 × 10 × 40 cm) live traps placed 1.50 m above the ground. Once the animals were captured they were sedated with 0.2 mg/kg Ketamina® anesthesia. A veterinarian was present during this capture process. We made audio recordings while tamarins were in tomahawk traps, prior to collaring. We covered the cage traps with dark-green cotton sheet, so tamarins were not in visual contact with the researchers or with each other, in an effort to reduce stress.

Behavioral Sampling

We followed the habituated free-living group from October to December 2012, totaling 170 h of fieldwork and 60 h of direct contact with tamarins. When following the habituated free-living tamarins, two researchers observed the group simultaneously. T. V. Sobroza observed and recorded the remaining seven free-living groups from June 2014 to February 2015. We conducted observations from 06:00 h to 17:00 h, when tamarins show the greatest activity (Egler 1986). We used ad libitum sampling to record sounds and focal individual sampling to obtain behavioral data from the habituated group (Altmann 1974). We adopted intervals of 3 min in which we recorded the first sound signal emitted and the behavior simultaneously performed by the caller. We observed focal individuals opportunistically, but avoided sequential observation by changing focal individuals every 3 min.

For each sampling event, we documented the time of recording, a rough description of the sound signal (based on how it was perceived by the human ear), and the behavioral context. We grouped behavioral contexts into the following categories: 1) Foraging; 2) Locomotion; 3) Rest; 4) Affiliative interaction (grooming, play, and mating); 5) Agonistic interaction (chasing, piloerection, and tongue flicking); 6) Alarm; and 7) Others (scent-marking, scratching, autogrooming, and defecating). As infant vocal activity was always induced by the anticipation of feeding, and the captured tamarins remained covered by a dark cotton sheet, we did not sample the behavioral context for these recordings.

We made additional recordings from groups of free-living tamarins outside Parque Municipal do Mindú during behavioral experiments (Sobroza 2015). On these occasions, we performed sound playbacks simulating territory intrusion and recorded the sounds emitted in response, immediately following the broadcastings. We considered behaviors recorded after playbacks to be heavily biased toward territorial interactions, so we did not include these observations in the description of behavioral contexts and their association to specific sound signals. However, we included recorded vocal responses to playbacks in the measurement of acoustic parameters.

Sound Signal Recording and Analysis

At Parque Municipal do Mindú, we recorded tamarin sound signals in uncompressed.wav digital format with a Zoom H1 recorder (16-bit resolution, sampling rate of 44 kHz) attached to a Sennheiser ME 66 directional microphone (frequency response 4–20 kHz). The distance from the microphone to the focal individual varied from 2 to 10 m. We recorded the two rescued infants during feeding conducted daily by the local staff. As distance to recorded tamarins was <1 m, we conducted the recordings using the Zoom H1 recorder’s built-in microphone. We recorded the tamarins in the captured group using a directional microphone Yoga HT-81 microphone (frequency response: 1–16.0 kHz) attached to the same recorder at 5–20 m. We recorded sound signals from free-living tamarins outside the park immediately following playbacks. These groups were unaccustomed to human presence; thus recordings rarely lasted longer than 10 min. We could not determine the contribution of each individual to the final sample, so our results must be considered with caution owing to the risk of pseudoreplication. We were also not able to capture all subjects in order to identify their sex, so we did not examine sex-related differences in sound signals.

We analyzed sound recordings in Raven 1.4 (Cornell University Laboratory of Ornithology) and identified sound signals by visual inspection of spectrograms. For each signal, we obtained the following temporal and spectral parameters: signal duration, upper and lower frequencies, peak frequency (i.e., dominant), presence and number of harmonics, and the number of syllables (i.e., notes) that formed the signal. We measured temporal parameters from oscillograms (Fig. 1). We measured the spectral parameters of the first (fundamental) harmonic in all tonal calls by analyzing power spectra (Fig. 1), built with the following configuration: DFT size = 1024 samples, overlap = 80%, window size = 20 ms, window type = Blackmann. We edited spectrograms representing each type of sound signal using Raven 1.4 (Cornell University Laboratory of Ornithology) and the Seewave package (Sueur et al. 2008b) for R (R Development Core Team 2012).

Fig. 1
figure 1

Acoustic measurements used in this study. We estimated temporal features such signal duration and number of syllables using oscillograms (a). We analyzed spectral parameters such as upper, lower, and peak frequencies using the power spectrum (i.e., how the energy is distributed through the frequency domain) from the first harmonic (b). We counted the number of harmonics by inspecting the spectrogram (c). We built power spectra with the following configuration: DFT size = 1024 samples, overlap = 80%, window size = 20 ms, window type = Blackmann using Raven 1.4.

Full size image

Statistical Analysis

We classified sound signals a priori into seven categories: a) Call; b) Phee; c) Trill; d) Tsê; e) Cry; f) Chirp; and g) Ploc. We also used hierarchical agglomerative cluster analysis to characterize call groups (N = 766) (McGarigal et al. 2000). We used the mean values of five acoustic properties (number of syllables, duration, upper frequency, lower frequency, and peak frequency), for each vocalization type inferred from field observations and visual classification of spectrograms in a clustering analysis. We analyzed vocalizations of the same type recorded indifferent localities and from different groups separately, taking the mean values of acoustic properties for each. We did not control for recording of multiple calls produced by the same individual, which could potentially bias values of call parameters estimated for each group. We converted values of spectral properties from Hz to kHz for scale adjustment. We conducted cluster analysis in PAST 3.11 (Hammer et al. 2001) using the UPGMA algorithm and Mahalanobis distance as a similarity index. To infer the association of sound and behavior, we used the direct proportion of emission of each sound signal in each behavioral context (N = 863).

Data Availability

The data used for this study are available from the authors upon request.

Ethical Note

Methods for observational data collection were noninvasive and complied with Brazilian laws. The capture and the process of anesthesia, manipulation, and marking of the subjects were authorized by permit number 10286–3 issued by SISBIO/MMA (Ministry of Environment). Authors contributed equally during the present work and have no competing interests.

Results

We identified 12 distinct sound signals from the inspection of spectrograms associated with different behavioral contexts: Call 1, Call 2, Phee, Trill 1, Trill 2, Trill 3, Tsê 1, Tsê 2, Tsê 3, Chirp, Cry, and Moan (Table I; Figs. 2, 3, 4). Call 1 and Call 2 both had ascending (first half) and descending (second half) frequency modulation, giving a ∩ − shaped spectral signature. Call 2 was emitted at a lower sound level than Call 1, usually during locomotion, when tamarins were moving closely and side-by-side. Phee syllables are distinctively longer than the syllables of Call 2. We recorded Phee only from captive infants. Trill 2 was mainly associated with alarm and agonistic behavior and was often emitted while individuals performed other displays, such as lateral head movements and tongue-flicking. Trill 3 and Tsê 2 were emitted at low sound pressure levels and detected only from visual inspection of spectrograms, so we could not attribute these vocalizations to any behavioral context. Chirp was emitted mostly in the context of alarm among free-living tamarins but also by captive infants in a foraging context and during playback trials that simulated territory invasion in free-living groups. Moan, Tsê 3, and Phee were emitted only by infants prior to feeding. In addition to the signals described in the foregoing, we detected a signal that we could not record successfully in both free-living and captive individuals. We named this signal Ploc because it resembled this phoneme to the human ear. It is seldom emitted, but occurs when free-living individuals jump from one tree branch to another, and when captive infants capture live prey.

Table I Summary measurements of the acoustic structure of the vocal signals emitted by pied tamarins recorded in the municipalities of Manaus and Rio Preto da Eva, Brazil, from 2012 to 2015
Full size table
Fig. 2
figure 2

Spectrograms of wild and captive pied tamarin sound signals recorded in the municipalities of Manaus and Rio Preto da Eva, Brazil, from 2012 to 2015. (a) Call 1; (b) Call 2; (c) Phee; (d) Trill 2; (e) Trill 3; (f) Trill 1.

Full size image
Fig. 3
figure 3

Spectrograms of wild and captive pied tamarin sound signals recorded in the municipalities of Manaus and Rio Preto da Eva, Brazil, from 2012 to 2015. (a) Tsê 1; (b) Tsê 2; (c) Tsê 3; (d) Chirp; (e) Cry; (f) Moan.

Full size image
Fig. 4
figure 4

Occurrence percentage of different sound types emitted by pied tamarins in each behavioral category at Parque Municipal do Mindú. (a) Call 1; (b) Trill 1; (c) Trill 2, (d) Tsê; (e) Chirp, and (f) Cry.

Full size image

Visual classification was loosely consistent with the results of automatic classification using cluster analysis (Fig. 5). Call 1 samples from all recording sites were grouped in a single cluster, and separated by short Mahalanobis distance values. Signals emitted by captive infants were generally dissimilar to signals emitted by adults. Despite their similarity based on qualitative visual classification, the remaining types of acoustic signals apparently vary in spectral and temporal properties among recording sites. Trills, Chirps, and Tsês were generally scattered on the acoustic distance dendrograms and often clustered with different signals recorded at the same site (Fig. 5), indicating that some vocalization characteristics may be idiosyncratic to the focal group. Additionally, acoustic signals did not cluster according to the behavioral contexts in which they were emitted (Fig. 5).

Fig. 5
figure 5

Dendrogram of acoustic distances (based on Mahalanobis distances estimated from five acoustic traits of each signal) among pied tamarin sound signals recorded in different locations in Manaus, Brazilian Amazonia. Terminals represent the sound signal type and tamarin group (N = number of signal samples used to calculate the mean values used in the cluster analysis). Symbols at some terminals indicate the behavioral contexts of sound signal emission, based on observations of a resident group at Parque Municipal do Mindú. Triangles: locomotion; open circles = rest; closed circles = affiliative behaviors or mating; squares = foraging; open diamonds = agonistic or aggressive interactions; closed diamonds = alarm; stars = sounds emitted only by infant tamarins.

Full size image

Discussion

Among free-living tamarins, most signals were emitted when individuals were at rest or during locomotion. We classified long-distance vocalizations emitted at high sound pressure levels as calls (Call 1 and Call 2). Long calls may be used to advertise the group’s position to neighboring groups, helping to maintain a minimum distance between them (Caselli et al. 2015; Halloy and Kleiman 1994; Wich and Nunn 2002). In other primates, such signals are also considered to be important maintainers of group integration (Waser 1977). During observations of free-living tamarins, we frequently recorded the conspicuous Call 1 from resting and moving individuals. Individuals that lost sight of the group tended to emit it repeatedly, prompting the rest of the group to look for them. Individuals tended to emit louder and longer calls as the distance from their group increased, supporting the hypothesis that calls have a group-binding function. Differences among long calls have been reported for Saguinus oedipus, a closely related tamarin species (Cleveland and Snowdon 1982). In the future, the study of amplitude, duration, and frequency modulation of pied tamarin long calls in response to increasing distances between focal individuals and the core of their group may reveal patterns similar to those described for S. oedipus, i.e. greater number of syllables, with higher duration and different frequencies in some syllables (Cleveland and Snowdon 1982), reinforcing the view that these signals are important mediators of group cohesion.

Snowdon (1993) described the long calls of pied tamarins as consisting of two or three syllables non-modulated in frequency. He may have recorded tamarins that were close to each other and emitting calls we classified as Call 2, since the conspicuous Call 1 is formed of more syllables and is highly frequency modulated. Short calls with variation in the number of syllables occur in other callitrichid species (Bezerra and Souto 2008; Halloy and Kleiman 1994). Future work on multiple taxa would be useful to clarify whether maintaining group cohesion and variation in the number of syllables are common to long-distance calls across the Callitrichidae.

Tsê signals were associated mostly with individual rest and locomotion. Among callitrichids, locomotion is usually related to foraging or browsing for food items (Garber and Porter 2014). Thus, it is difficult to untangle Tsê function in coordinating the spatial locomotion of the group from their role in foraging behavior. However, sound signals associated exclusively with foraging behavior have been described for some tamarin species, such as Saguinus fuscicollis, S. oedipus, Leontopithecus rosalia, and Callithrix jacchus (Benz et al. 1992; Bezerra and Souto 2008; Elowson et al. 1991; Moody and Menzel 1976). In field observations in Parque Municipal do Mindú, Tsê signals were also emitted by stationary pied tamarins after capturing animal prey. Additional observations conducted at feeding platforms (not included in this study) also suggest that Tsê signals are associated with foraging and independent of locomotion.

We recorded Chirp signals in both locomotion and alarm contexts, which were generally not independent; i.e., we recorded signals while individuals moved away from a perceived threat. Escape is the most common defensive behavior among primates threatened by predators, and sound signals may be important to alert other group members to a threat (Friant et al. 2008). Alert sounds present clues about the spatial position of the threat, to coordinate group movements and provide group defense (Fischer and Zinner 2011). Chirps were also associated with stereotyped aggressive movements such as lateral movements of the head, tongue-flicking, and scent-marking (Hershkovitz 1977). These results suggest that agonistic and alarm signals of pied tamarins might have a multimodal nature and are composed of acoustic, chemical, and visual cues.

We encourage a more refined study of Chirp calls, since tamarins may have a referential alarm system (Kirchhof and Hammerschmidt 2006) and the study site where we collected behavioral data probably has fewer predators than more pristine locations. Such absence of predators may have reduced group’s vocal repertoire or affect the frequency at which some call types are emitted. Another interesting line of research involving tamarin predators is that of acoustic mimicry. The margay cat (Leopardus wiedii), a predator not found in the urban part of the study area, can attract pied tamarins by mimicking their calls (Calleia et al. 2009). Experiments with signal processing and sound manipulations may help elucidate the capacity of pied tamarins to discriminate between mimicked and actual pied tamarin calls. Information about the pied tamarin vocal repertoire may also provide means to investigate behavioral interactions with the golden-handed tamarin (Saguinus midas), a potential competitor that occurs in sympatry on the northern edge of the pied tamarin’s range (Ayres et al. 1982; Röhe 2006; Sobroza 2015). By recording and classifying aggressive sounds from both species, playback techniques can be used to investigate asymmetries in the aggressive responses of the two species, along with their contact zones, allowing tests of hypotheses concerning interference competition and dominance behavior (Sobroza 2015).

Pied tamarins employed a diverse signal repertoire during intra and intergroup agonistic interactions, including Cry, Trills, Chirps, and Calls. This supports the hypothesis that behavioral excitation stimulates the emission of a more variable sound repertoire (Morton 1977; Owren and Rendall 2001). These effects can be expressed as the emission of a greater number of different sounds, a shorter latency to reach peak frequency and intensity, and an increase in harmonic structure (Briefer 2012; Morton 1977; Ordóñez-Gómez et al. 2015). Behavioral excitement (triggered by imminent feeding) also elicited the emission of multiple types of sound signals by infant pied tamarins. The emission of a diverse sound repertoire by infants is common in callitrichids and has been attributed to behavioral arousal (Roush and Snowdon 1994).

Knowledge of the behavioral contexts associated with sound signals can provide valuable information for conservation. For example, monitoring protocols targeting biodiversity conservation use automated sound recordings to collect ecological data based on the detection and quantification of acoustic signals from spectrograms (Blumstein et al. 2011; Kalan et al. 2016; Spillmann et al. 2015). Software can aid researchers to pinpoint signals of interest during long recording sessions with variable levels of background noise, but this requires adequate description of species’ sound repertoires. Moreover, conservation measures for endangered species characterized by small populations inhabiting fragmented habitats often include translocation of individuals between natural habitat remnants or ex situ breeding programs (Halloy and Kleiman 1994; Price et al. 2012; Richard-Hansen et al. 2000). Steps such as acclimatization, feeding behavior, and group formation can be monitored via vocalizations such Chirp, Tsês, and Trills, for example. This monitoring may prove to be important for the pied tamarin because it is very territorial (Gordo et al. 2008, 2013) and an indiscriminate release of tamarin individuals in nature or captive colonies may result in death.

Finally, cluster analysis suggested the existence of idiosyncrasies in the quantitative properties of sound signals from different groups. Cluster analyses of primate vocalizations are extremely sensitive to the number of call parameters used, and many dozens of acoustic measurements may be necessary to classify the most complex signals correctly (Wadewitz et al. 2015). However, Call 1 proved to be ubiquitous and highly stereotyped among groups of pied tamarins occupying different forest sites. Other acoustic signals were typically emitted solely by infants (Phee, Moan). Call 1, Phee, and Moan could be used in monitoring protocols to assess site occupancy and demographic parameters of the pied tamarin across the species’ geographic range. These results can be integrated with current conservation efforts that involve population monitoring, genetic sampling, and radio-tracking to evaluate the species’ ecological response to environmental changes and intra- and interspecific interactions. Furthermore, the present study contributes to the roll of Callitrichid vocal repertoire studies, which allows multitaxa comparative analyses and aids understanding of vocal communication in Neotropical primates as a whole.