Skip to main content
Browse Subject Areas

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Vocal complexity and sociality in spotted paca (Cuniculus paca)

  • Stella G. C. Lima ,

    Contributed equally to this work with: Stella G. C. Lima, Renata S. Sousa-Lima, Rosana S. Tokumaru, Sérgio L. G. Nogueira-Filho, Selene S. C. Nogueira

    Roles Data curation, Formal analysis, Investigation, Writing – original draft, Writing – review & editing

    Affiliation Laboratório de Bioacústica, Departamento de Fisiologia e Comportamento / Programa de Pós-Graduação em Psicobiologia, Universidade Federal do Rio Grande do Norte, Natal, RN, Brazil

  • Renata S. Sousa-Lima ,

    Contributed equally to this work with: Stella G. C. Lima, Renata S. Sousa-Lima, Rosana S. Tokumaru, Sérgio L. G. Nogueira-Filho, Selene S. C. Nogueira

    Roles Supervision, Writing – review & editing

    Affiliation Laboratório de Bioacústica, Departamento de Fisiologia e Comportamento / Programa de Pós-Graduação em Psicobiologia, Universidade Federal do Rio Grande do Norte, Natal, RN, Brazil

  • Rosana S. Tokumaru ,

    Contributed equally to this work with: Stella G. C. Lima, Renata S. Sousa-Lima, Rosana S. Tokumaru, Sérgio L. G. Nogueira-Filho, Selene S. C. Nogueira

    Roles Formal analysis, Investigation, Writing – original draft, Writing – review & editing

    Affiliation Departamento de Psicologia Social e do Desenvolvimento, Universidade Federal do Espírito Santo, Vitória, ES, Brazil

  • Sérgio L. G. Nogueira-Filho ,

    Contributed equally to this work with: Stella G. C. Lima, Renata S. Sousa-Lima, Rosana S. Tokumaru, Sérgio L. G. Nogueira-Filho, Selene S. C. Nogueira

    Roles Formal analysis, Writing – original draft, Writing – review & editing

    Affiliation Laboratório de Etologia Aplicada, Departamento de Ciências Biológicas, Universidade Estadual de Santa Cruz, Ilhéus, BA, Brazil

  • Selene S. C. Nogueira

    Contributed equally to this work with: Stella G. C. Lima, Renata S. Sousa-Lima, Rosana S. Tokumaru, Sérgio L. G. Nogueira-Filho, Selene S. C. Nogueira

    Roles Conceptualization, Data curation, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Validation, Visualization, Writing – original draft, Writing – review & editing

    Affiliation Laboratório de Etologia Aplicada, Departamento de Ciências Biológicas, Universidade Estadual de Santa Cruz, Ilhéus, BA, Brazil


The evolution of sociality is related to many ecological factors that act on animals as selective forces, thus driving the formation of groups. Group size will depend on the payoffs of group living. The Social Complexity Hypothesis for Communication (SCHC) predicts that increases in group size will be related to increases in the complexity of the communication among individuals. This hypothesis, which was confirmed in some mammal societies, may be useful to trace sociality in the spotted paca (Cuniculus paca), a Neotropical caviomorph rodent reported as solitary. There are, however, sightings of groups in the wild, and farmers easily form groups of spotted paca in captivity. Thus, we aimed to describe the acoustic repertoire of captive spotted paca to test the SCHC and to obtain insights about the sociability of this species. Moreover, we aimed to verify the relationship between group size and acoustic repertoire size of caviomorph rodents, to better understand the evolution of sociality in this taxon. We predicted that spotted paca should display a complex acoustic repertoire, given their social behavior in captivity and group sightings in the wild. We also predicted that in caviomorph species the group size would increase with acoustic repertoire, supporting the SCHC. We performed a Linear Discriminant Analysis (LDA) based on acoustic parameters of the vocalizations recorded. In addition, we applied an independent contrasts approach to investigate sociality in spotted paca following the social complexity hypothesis, independent of phylogeny. Our analysis showed that the spotted paca’s acoustic repertoire contains seven vocal types and one mechanical signal. The broad acoustic repertoire of the spotted paca might have evolved given the species’ ability to live in groups. The relationship between group size and the size of the acoustic repertoires of caviomorph species was confirmed, providing additional support for the SCHC in yet another group of diverse mammals–caviomorph rodents.


Comparative ecology and phylogeny analyses have been applied [1] to better understand the relationships between the variation in species’ social behavior and evolutionary history [2]. The evolution of sociality or group living [3] has been extensively investigated in many mammal species to answer many questions including links between sociality and communication [2], and is related to increased predation risk and vigilance [4], diurnal habits [5], body size [6], collective defense [7], burrowing [5, 8], sexual selection [9], mating system [10], food resources [11], habitat [12], and many other ecological factors (for a review, see [2] and references therein). In caviomorph rodents, for instance, it has been suggested that sociality coevolved with increasing body sizes and use of burrows [8].

It is not only natural events that influence the size of social groups in mammals [13]; species group living has also been driven by human-induced rapid environmental change (HIREC) sensu [14, 15], such as climate change, habitat fragmentation, and hunting. These anthropogenic factors have challenged species to adapt to persist in a modified environment [1416]. Therefore, the observation of the behavior of individuals of a given species in human disturbed environments can be misleading and may change under different ecological conditions. This plasticity is particularly evident in caviomorph rodents. Intraspecific variation in spacing, group size, and mating systems in response to many factors including predation, have been reported for this group [17]. The capybara (Hydrochoerus hydrochaeris), for instance, is known to change its behavior in response to environmental pressure, becoming more nocturnal in areas with high hunting pressure [18].

Plastic behavioral traits, such as communicative signals, may respond rapidly to environmental change. In fact, communication and cognition are social traits that differ greatly between social and non-social mammals and therefore, could contribute to understanding the evolution of sociality [2]. Acoustic communication is involved in social bonding, predator defense, cooperative foraging, reproduction, and dominance hierarchy negotiations, to name but a few (for a complete review, see [2]. Complexity in social relationships drives the evolutionary paths of social cognition and communication in parallel to each other [19], as different signals mediate several subtle processes.

Complex communication modulates behavioral responses and is expected to be present in more complex societies [20]. The Social Complexity Hypothesis for Communication (SCHC), states that solitary animals show a lower diversity of communicative signals than species that live in groups with complex social systems [21]. This hypothesis has been tested in some societies of mammals: ground-dwelling sciurid rodents [22], bats [23], non-human primates [24], whales [25], giant otter [26], meerkats, and slender mongooses [27]. In sciurids, for instance, it was reported that different attributes of sociality are linked to different attributes of communication [28]. Therefore, the SCHC may explain the evolution of sociality in caviomorph rodents, which have a relatively resolved phylogeny. Additional information about the acoustic repertoire of the spotted paca (Cuniculus paca) may reveal a different level of sociality that otherwise would not be detected due to HIREC sensu [14, 15].

The spotted paca is the second largest rodent occurring from Southern Mexico to Northern Argentina, and it is widely distributed in Neotropical countries [29]. The species is nocturnal, burrower [3032], and shows moderate sociality [1], mostly reported as solitary, living in monogamous pairs during reproductive periods [3032]. Nevertheless, there are some anecdotal reports of free-ranging spotted paca living in groups of five individuals [3335]. Moreover, farmers easily form groups of paca by gathering post-weaning juveniles to form captive breeding groups [34, 36]. The acoustic repertoire of spotted paca has been qualitatively described and is composed of six types of acoustic emissions, most of them related to contact (low grunt and low whine), alarm (snort and low growl), and aggressiveness (tooth chattering and very loud growl) [37, 38]. However, no data are available on the acoustic parameters of these calls. Thus, we aimed to extend the knowledge on the spotted paca’s acoustic repertoire by describing its acoustic structure and to evaluate if the repertoire of this species adheres to the SCHC predictions. Complex communicative systems predict larger vocal repertoires or features that increase the variability in vocal signals, such as combinations of sounds [39] and formants [40], promoting more efficient information transfer in social groups [41]. We then, aimed to verify the relationship between group size and the vocal repertoire size of caviomorph rodents, to better understand the evolution of sociality in this taxon. Due to reports of spotted paca’s social-living [3335], we hypothesized that the vocal repertoire of this species would show higher complexity in its repertoire size or signal features. Considering the variation of social behavior, social-living and vocal repertoire sizes among caviomorph species, we predicted this group would provide compelling evidence of the positive relationship between the number of call types and group size (SCHC). We also predicted that the size of the vocal repertoire of the spotted paca or its complexity would correspond to those of group-living caviomorph species.


Ethics statement

This work followed the principles of laboratory animal care (NIH publication No. 86–23, revised in 1985) and was approved by the Committee on Animal Research and Ethics of the State University of Santa Cruz, under protocol # 010/11.

Study area and subjects

The study was carried out on a farm where spotted paca are commercially raised, namely Empreendimentos Agropecuários e Obras S/A (EAO), zip code 42841-000, at the town of Camaçari, state of Bahia, Brazil. We recorded vocalizations and behaviors of 51 individuals: 42 adults (26 females and 16 males) and nine juveniles (six females and three males). All individuals were born and raised in captivity. The age of adults ranged from two to four years old and the age of juveniles ranged from 15 days to four months old. Unfortunately, there were no sub-adult animals available to improve our samples. Animals were housed in groups of one male and four females in 20 breeding pens, and four maternity pens with one female and its offspring. Five out of 20 breeding pens had one to two juveniles.

Each pen, including maternity pens, occupied 6 m2 (3.0m long x 2.0m wide), with concrete floor, surrounded by a 2.0m-high wire mesh fence supported by wooden poles. The ceiling was covered with tiles. All pens had two wooden shelters (1.5m long x 1.5m high x 1.0m wide), a water tank (0.6m long x 0.3m wide) and two feeders (1.0m long x 0.3m wide). During observations at night, we used red lights to allow the visualization of animals and to minimize disturbances. Rabbit commercial pelleted feed (200 g per animal) plus seasonal fruits and mineral salt ad libitum were daily furnished. Water was also offered ad libitum.

Data collection

Both calls and behaviors were recorded simultaneously at 1.5 m distance (in maximum) from the animals. The observation sessions usually took place between 4.00pm and 6.00am, the period of the highest activity of the captive spotted paca. We also observed the animals on three other routine occasions: during cleaning of the pen (between 11.00am and 1.00pm), during handling for medical procedures such as parasites control and weighing of animals (between 11.00am and 1.00pm), and when the animals were fed (4.00pm).

We recorded the acoustic signs ad libitum [42] between March and April 2014 using a Sennheiser ME-66 directional microphone (Wedemark, Germany) and a Marantz PMD 670 (Sagamihara, Japan) digital recorder (recorder settings: WAV format, mono mode, 48 kHz sampling rate, and 16-bits resolution). The observer started to record the animals when they were actively emitting sounds and kept recording until any sound emission was produced up to its end or after an interval of 1 minute with no sound emission. The animals were not marked individually; however, it was possible to identify them by natural characteristics. Because the animals are not very vocal, we could not register calls from all individuals housed in the breeding pens. The data collection totalized 90 hours of recording and observations. Just one observer (SGCL) collected and analyzed all behavioral and acoustic data. Two other authors (SLGNF and SSCN) analyzed a subset of the data from this study and using the Kendall’s coefficient of concordance we found inter-observer reliability in coding of these vocalizations (W ≥ 0.77).

To stimulate the emission of isolation or contact calls between a mother and her young, we isolated four mothers from their offspring by using a wooden barrier (0.12m long X 0.75m high X 0.02m wide), which acted as a barrier to physical and visual signals, but did not imply the loss of auditory and olfactory contact. The animals’ keeper was responsible for setting up the barrier between mothers and offspring. The observer started to record the vocalizations just after the keeper isolated the animals. Each mother/pup pair was separated from each other only once. This recording/observation session lasted 30 minutes per pair.

Acoustic analysis

Spectrograms and oscillograms were generated and analyzed with Raven Pro Software version 1.5 (Cornell Lab of Ornithology, Ithaca NY) using the same settings for all call types: Hann window, 1460 window length, 90% overlap, and 4096 FFT size. First, we categorized putative call types by ear and by visual inspection of spectrograms and oscillograms. The smaller vocal units of the calls were termed elements and were defined as a continuous sound without interruption sensu [43, 44]. For each element, we measured the following parameters: minimum frequency (Hz), maximum frequency (Hz), dominant frequency (Hz), number of harmonics under 1 kHz, and call duration (s) (S1 Table). Sound parameters were measured from spectrograms, oscillograms, and power spectra, and these were used to describe and compare vocalizations. Oscillograms were used to measure signal duration. Spectrograms and power spectrum were used to measure the minimum, maximum frequencies of the sound and the dominant frequency of the selection. Minimum frequency is the lowest frequency boundary and maximum frequency is the upper frequency boundary of the selection [45]. Dominant frequency is the frequency that corresponds to maximum power occurrence within the element selection [45]. To calculate call emission rate per hour (hourly rates) we divided the total numbers of each vocal type by the total data collection (90 h).

We also examined spectrograms for the presence of vocal complexity through structural variability, such as combination of sounds (combination of different vocalization types in sequences) [39], and formants (vocal tract resonance frequencies) [46]. To analyze the combination of sounds of roar and groan vocalizations (S2 Table), we measured the number of elements, the order of elements, the duration of the sequence (s), and the rhythm of the sequence (the number of elements divided by the total sequence duration). To test whether the pronounced horizontal frequency bands visible in roars spectrograms were formants, we compared measured and predicted ‘formant dispersion’, which is the averaged difference between successive formants [47]. For this, we obtained four adult 1-2- year-old male spotted pacas (4.6 ± 0.7 kg) that were destined to be sent to the abattoir at the EAO farm. This part of the study was carried out at the Laboratory of Applied Ethology, Universidade Estadual de Santa Cruz, Ilhéus, Bahia, Brazil. At this site, we individually introduced the animals into cages (1.2 m long X 0.8 m high X 0.8 m wide). The cages were made of metal with a feeder, a drinking trough, and a wooden shelter. After 15 days of habituation to the new conditions, we recorded the acoustic signs ad libitum [42] between August and September 2017, using the same procedures and equipment as described above. Roars were induced by showing to the individuals the handling net. Then we selected 40 roars (10 calls/animal) (S3 Table) and measured the values of the first five formant frequencies using linear predictive coding (LPC: ‘To Formants: Burg method’) command in PRAAT software, version 5.3.06 [48]. The following settings were used: time step: 0.01 s; maximum number of formants: 6; maximum formant value: 11000 Hz; window length: 0.01 s; pre-emphasis: 50 Hz. Each element was measured separately and them we measured the averaged formant dispersion using the equation: (1) where Df is the formant dispersion (in Hz) and N is the total number of formants measured.

We then returned animals to the commercial farm and, after their slaughter, which followed ethical and legal rules; we performed anatomical measurements of the vocal tracts. The vocal tract length (VTL) was determined with a tape measure from the middle of the glottis to the front of the incisors (cm). According to Fitch [47], the measurement of VTL should be associated with formant frequency dispersion, as it follows the path of plane-wave propagation of sound, in a ‘uniform tube’ (a tube that does not vary in cross-sectional area), from the glottis to the oral or nasal radiation site, and the frequency difference between the successive resonances is constant and given by the equation: (2) where i is the formant number, c is the speed of sound (350 m s−1), VTL is vocal tract length (in m) and Fi is the frequency (in Hz) of the ith formant.

Following Ried and Fitch [46] we determined the average of the obtained differences to provide an overall estimate of spectral dispersion of each of the animals. We used the standard deviation of the formant intervals to evaluate the extent to which the uniform tube approximation holds [46].

Statistical analysis

To test the validity of the putative call types that we categorized by ear and by visual inspection of the spectrograms, we conducted a Linear Discriminant Analysis (LDA) based on four of the five acoustic parameters that were measured. We excluded the number of harmonics under 1 kHz in LDA because this parameter showed no variation. We randomly chose 50 elements of each vocal type with high signal quality (low background noise and no overlap among calls from different individuals) from the total 3,341 elements recorded for acoustic measurements, totaling 400 samples (S1 Table). Elements were taken from different recording sections to increase the chances of analyzing different vocal types and calls from different individuals. Moreover, to avoid over-sampling, we randomly selected the data by considering both call types (from three to nine calls per individual) and individuals. We first randomly selected elements for each vocal type from different recording sections, and the individuals were balanced in these samples to increase the chances of analyzing different vocal types from different individuals (see Table 1). Before conducting the LDA analysis we standardized the variables (by subtracting the mean of the variable from each data point and dividing the result by the variable’s standard deviation–the Z score transformation) to avoid the spurious attribution of weights to acoustic parameters measured in different units [49]. To test the significance of the discriminant model we performed a Multivariate Analysis of Variance (MANOVA). Further, to determine whether it was possible to predict each vocal type correctly based on the measured acoustic parameters (independent variables), we performed cross-validation analyses using the ‘leave one sample out’ procedure [50], in which each case was classified by the functions derived from all cases other than the one being classified, reporting the LDA accuracy as the proportion of elements correctly assigned to each vocal type. We performed a binomial test to evaluate whether the proportion of the success of the cross-validation analysis was higher than that expected by chance (12.5%).

Table 1. Mean ± standard deviation (SD) of acoustic parameters measured in spotted paca vocalizations.

Percentage of notes correctly attributed to each call type in the cross validation. The N corresponds to the number of emissions analyzed in each category/number of individuals from which individual calls were selected. The centroids show the spacing of different call types of spotted paca in a two-dimensional signal space defined by the first two discriminant functions (DF1 and DF2). The values in bold indicate the correlation coefficients of the variables that most contributed to the discriminant functions (DF1 and DF2).

Additionally, we ran a General Linear Model (GLM), considering call types and age class as predictors, to verify potential differences between juveniles and adults shared call types. We first calculated the individual’s averages for each acoustic measurement per call type, and then used these means as the dependent variable in the analysis, and the call type as the independent variable. Furthermore, we used the agglomerative hierarchical cluster analysis to investigate whether the combinations of roar and groan would fit into additional call types (S2 Table). This method successively links the most similar objects into larger groups. Average values of duration, number of notes, and rhythm of the different combinations were used as input data. Euclidean distances were used as the measurement of distance between the pair of situations and Ward’s criterion was the linkage criterion. Ward’s criterion minimizes the within-cluster variation. Furthermore, we used the automatic truncation option, based on the entropy and tries, to create homogeneous groups.

We used the paired t-test to compare the predicted and measured formant dispersions of the four spotted pacas for which we obtained both anatomical and acoustic data (S3 Table). For this analysis we used the average formant dispersion (Df—measured formant dispersion) determined from vocalisation recordings using Eq 1 and the theoretical values calculated using the VTL obtained and Eq 2 (predicted formant dispersion). All analyses were performed using XLSTAT (Version 2017.4, Addinsoft), with a significance level of α < .05.

Vocal complexity and group size: Phylogenetically based analysis

We determined the relationship between group size and vocal repertoire size by means of independent contrasts analyses. For this, we used the total repertoire size of the selected caviomorph species (see results section), considering only adult repertoire to avoid ontogeny effects. Although different research groups characterized possibly similar call types with different terminology, we assumed that authors, independent of the analyses chosen, discriminated these species repertoire as the maximum. Thus, we used the independent contrasts approach [51, 52] to investigate sociality in spotted paca following the social complexity hypothesis [21], independent of phylogeny.

We used the topology described by Ebensperger and Blumstein [5] to establish major relationships among seven caviomorph rodent species, from which we had information on average group sizes and on the repertoire size of adult vocalizations based on acoustic parameter information. The positions of Spalacopus cyanus and Ctenomys talarum were based on [53]. To calculate independent contrasts, we defined sociality as the log10 of the midpoint of the range of observed group sizes, as the midpoint is an appropriate measure of central tendency, given a wide range of values within a species, following [5] Our continuous independent variables were the log10 of the adult vocal repertoire reported for each species, excluding mechanical sounds and juvenile vocalizations. We transformed mid group size and vocal repertoire to eliminate outliers and to meet assumptions of linear models. As the phylogeny used did not have consistently good estimates of the branch lengths model of evolution we assumed the Nee transformation [54], where the distance from the tips to the focal node was calculated by log10 transforming the number of tips descending from that node. Given these phylogenies, we calculated independent contrasts and determined the relationship between independent contrasts of group size and independent contrasts of vocal repertoire size, in order to investigate sociality in spotted paca following the SCHC [21] through a linear regression analysis. We then used the obtained equation to estimate the group size of the spotted paca through its vocal repertoire. In the linear regression analysis, we also used XLSTAT (Version 2017.4, Addinsoft) with a significance level of α < .05.


Eight putative calls were found based on observations of the animals and visual inspection of the spectrograms (Fig 1; S4 Table). The hourly rate of emissions showed that roar (3.3 calls/s) was the most frequent vocal type, followed by snore (3.2 calls/s), growl (2.6 calls/s), bark (1.8 calls/s), cry (0.7 calls/s), tooth chattering (0.9 calls/s), groan (0.7 calls/s) and click (0.4 calls/s).

Fig 1. Spectrograms and time series of spotted paca vocalizations.

The arrows in the first roar vocalization (box a) indicate five formants. The box f shows combinations of sounds between roar and groan calls, and the arrows (1st and 2nd) of this box indicate the groan and roar calls, respectively.

The LDA, based on the four acoustic parameters measured, discriminated the eight call types initially proposed (roar, growl, bark, tooth chattering, groan, snore, click, and cry) (MANOVA: Wilks λ = 0.10, F 28, 1404 = 44.87; P < 0.0001, N = 400; Table 1). The first two discriminant functions explained 88.9% of variance among the vocalizations. The duration and dominant frequency were the variables that most contributed to the first discriminant function, whereas the minimum frequency was the variable that most contributed to the second function (Table 1). The cross-validation correctly attributed the vocal categories with an accuracy of 66%, which was significantly higher than the 12.5% (or 1/8) expected by chance (binomial test: P < 0.0001). The accuracy of cross-validation for the eight vocal categories ranged from 34% (snore) to 92% (tooth chattering) (Table 1). The spacing of different call types, denoted by the obtained centroids (Table 1) allows confirmation that they are well discriminated based on the factor axes extracted from the original explanatory variables.

The GLM confirmed differences between call types (F20, 122 = 6.3, P < 0.0001). This analysis also showed differences in the acoustic parameters of juveniles and adults (F4, 46 = 18.9, P < 0.0001). However, the statistical model showed no interaction between call types and age class (F20, 122 = 0.87, P = 0.56). Therefore, there were no differences in the acoustic parameters of call types (growl, groan, roar, and bark) shared by juveniles and adults. As just one juvenile emitted tooth chattering we excluded this call type from the GLM.

Besides the bark and groan alarm calls, clicks and cries were the only non-agonistic calls in the species’ repertoire. Clicks were emitted during the feeding context, probably as a contact call (Table 2). During mother-offspring separation, only the cry call (Fig 1) was emitted by offspring, suggesting a contact function (Table 2). During cry emissions by the offspring the mother remained agitated, moving back and forth in the maternity pen, sniffing the air and the wooden barrier between them. The mother could also produce tooth chattering and barks in response to the offspring cry call. On one occasion, we observed one of the mothers performing thumping displays (when the animal beats the ground with its hind legs) during this separation period.

Table 2. Description of context and possible communicative function of the vocalizations of spotted paca associated with age (A: adult and J: juveniles) and sex (M: male and F: female).

The spotted paca emitted groan combined with roar when captured for handling, which apparently seemed to alert conspecifics (Table 2). Although we recorded 14 combinations of both calls, we analyzed just five of these combinations, as these were emitted more than three times from at least three animals. The automatic truncation showed in dendrogram (Fig 2), reveals that from the five analyzed combinations of roar (A) and groan (B) three of them (ABA, ABAB, and ABABA) were acoustically similar, composing one unique vocalization.

Fig 2.

Dendrogram of similarity according to acoustic parameters of five different combinations (C1-C5) of roar (A) and groan (B). C1 (BA), C2 (ABABAB), C3 (ABA), C4 (ABAB), C5 (ABABA). The letters A and B correspond to the sequences of roar (A) and groan (B) appeared in each combination. Hierarchical clustering with Ward’s method was used to construct the dendrogram. Height represents cophenetic distance (dissimilarity) between the combinations. The dotted line represents the automatic truncation, leading to three combination calls of roar and groan.

For four spotted pacas, for which both anatomical and acoustic data were available (Table 3), the predicted and measured formant dispersion were not significantly different (paired t-test: t = 1.33; N = 4; P = 0.27). Standard deviations of formant dispersion were quite stable in spotted paca, ranging from 2.5% to 10.4% of the formant dispersion (mean 7.8%). Thus, spotted paca closely approximated the uniform tube approximation.

Table 3. Descriptive data for anatomical and acoustic variables of adult spotted paca (N = 4) used in formant analysis.

Vocal complexity and group size: Phylogenetically based analysis

Using the available information on average group size and adult vocalizations repertoire size based on acoustic parameter information from seven species of caviomorph rodents (Table 4), as well as the phylogenies of these same species (Fig 3), we determined the relationship between group size and vocal repertoire size across independent contrasts analyses through the equation: Y = 0.27 + 1.59*X (F1, 4 = 10.41; R2 = 0.72; P = 0.03; Fig 4).

Fig 3. Topology of the species of caviomorph rodents used in this study.

Modified from [5] and [53].

Fig 4. The relationship between standardized independent contrasts of log10 vocal repertoire size and standardized contrasts of log10 group size of seven caviomorph rodent species.

The linear regression follows the equation: Standardized contrasts of log10 group size = 0.27 + 1.59*Standardized contrasts of log10 vocal repertoire size (F1, 4 = 10.41; R2 = 0.72; P = 0.03). Gray lozenge indicates the independent contrasts of the original seven caviomorph species, while black circle is the virtual position of the spotted paca based on the equation considering only its adult acoustic repertoire size (N = 6) determined in this study.

Table 4. Data of the seven caviomorph species used to analyze the relationship between group size and the number of vocal types.

Considering the acoustic repertoire size of adult spotted paca determined in this study to comprise six different acoustic vocalizations (Table 2)–excluding the mechanical call and juvenile vocalization–as well as the spotted paca’s position in the phylogeny (Fig 3) we calculated as 0.17 the standardized contrasts of log10 of vocal repertoire size of spotted paca. Then, replacing X by 0.17 (X = 0.17) in the obtained equation (Y = 0.27 + 1.59*X), we calculated as 0.55 the standardized contrasts of log10 group size of spotted paca (Y = 0.55). Using this value and the species’ phylogeny position we calculated the log10 group size of spotted paca as 0.94. Then, using inverse logarithm (anti logarithm), we estimated an average potential group-living size of 8.7 individuals for this species. If we consider the 95% confidence interval the potential group size of the spotted paca is estimated to be 6.0–12.7 individuals.


We provide evidence of the complexity in the vocal repertoire of spotted paca in numbers and features, such as combination of vocal types, which may increase the species’ acoustic repertoire even more. Our results also revealed the expected relationship between caviomorph rodents’ vocal repertoire size and group size (sociality). This relationship corroborates the hypothesis that the acoustic repertoire, together with other ecological factors [55], can be a predictor of sociality in caviomorph rodents. The complexity of spotted paca’s acoustic repertoire described herein, grants this species behavioral capabilities to be included within the more social caviomorph rodents.

The acoustic repertoire of captive spotted paca described here is larger than that previously recorded for free-ranging animals [37], which included six types of calls for the species. In the present study, based on LDA analysis, we confirmed the discrimination of eight acoustic signals from the eight calls initially proposed based on visual inspection of the spectrograms: seven vocal and one non-vocal call types. In an earlier study, Eisenberg [37] did not show the acoustic structure of the calls, which prevented the accurate description of these calls and the analysis of complexity in the repertoire of this species.

The independent contrasts analysis revealed that evolutionary changes in vocal repertoire are a predictor of changes in group-living. Through the linear regression analysis, we verified that the spotted paca has the potential to live in groups of between six and 13 individuals. The relationship between group-living and vocal complexity was also reported for other taxa. For instance, in non-human primates, increases in group size and social grooming, are related to evolutionary changes in acoustic repertoire size [24]. In whales, group sizes are related to differences in acoustic parameters [25], while individuals of Carolina chickadees (Poecile carolinensis) living in larger groups emitted calls with greater complexity than individuals in smaller groups [56]. It is not possible, however, to evaluate if vocal complexity (repertoire size) precedes or follows the increases in sociality (group size) during evolutionary processes [24]. In this context, the vocal complexity in spotted paca explains different sightings of groups ranging from one to five individuals in Peru, Mexico, Colombia and Brazil [3335, 57] and may also explain the success of breeding groups in captivity [36], showing acoustic capabilities to manage social interactions among companions. Moreover, Nogueira-Filho and Nogueira [34] suggested that the solitary habit reported for the species might be a response to the intensity of hunting pressure, which is corroborated by the report that spotted paca changes its burrowing habits depending on hunting pressure [35].

Spotted paca provides the most sought-after game meat in Neotropical countries and suffers high hunting pressure, threatening the species [5860]. This pressure could change the species’ behavior from more social to solitary as a strategy to escape from predation. Although the predatory risk hypothesis predicts that group-living reduces the risk of predation [4, 8], this behavioral pattern seems not to be adaptive for spotted paca, which use burrows to hide from predators [35]. Moreover, there is a correlation between burrow digging and sociality in caviomorphs [5, 8]. Therefore, burrowing behavior could also be an evolutionary trait reinforcing sociality capabilities in this species besides its vocal complexity. Furthermore, studies have showed that prey individuals can change their social strategies according to predation risk [61]. For instance, the African striped mouse (Rhabdomys pumilio) can adjust its social behavior in response to prevailing environmental conditions, in which both sexes switch between solitary- or group-living according to social tactics, to have a greater fitness at a particular time [62].

The occurrence of sociality is expected when the benefits are stronger than costs [7, 63]. This explains the variability recorded in the sociality degrees among and within species as current ecological factors may influence group composition, making it more or less cohesive [15]. Thus, the human-induced rapid environmental change (HIREC) sensu [2] could lead to solitary habits in the spotted paca instead of finding this species living in groups. Complexity features such as vocal combinations can change behavioral responses of vocal receivers. We found at least three different possible types of combinations of the groan and roar calls that were emitted in different contexts than those in which each call was emitted separately. In roar vocalizations we found formant-like structures. Formant structures, which reflect details of individual vocal-tract anatomy and body size [6], usually play an important role in individual recognition. Therefore, the formants may allow receivers to assess the caller’s age, sex and maturity [64, 65, 66], which could help, for instance, in the recognition of the reproductive condition of male/female in the spotted paca. However, further study using playback is still necessary to test the function of vocal combinations and formants in this species.

The vocal repertoire of spotted paca herein described, provides some reflection regarding the presence of homologies among caviomorph rodents. The cry call, for example, is structurally similar and seems to occur in the same behavioral context (contact/isolation) as the degu’s (Octodon degus) loud whistle [67] and the guinea pig’s (Cavia porcellus) isolation whistle [68]. The emission of tooth chattering, a mechanical signal described for spotted paca, was also present in several caviomorph species, including Hydrochoerus hydrochaeris [44], Cavia sp. [68], Kerodon rupestris [69], Ctenomys talarum [70], and showing similarities in behavioral contexts and acoustic structures [37]. The groan vocalization in the spotted paca is functionally like groans in degu [67] and whine in rock cavy (Kerodon rupestris) [69] and guinea pigs [68], all emitted in aggressive contexts. Growls emitted by spotted paca are agonistic calls and are possibly used as defensive function of the grunt call in tuco-tuco (Ctenomys talarum) [70], rock cavy [69], and degu [67], which provide similar acoustic patterns. Bark calls of spotted paca also show similarities with the capybara’s bark [44] in both function (alarm call) and acoustic features. Click and snore calls are acoustically like those of capybara and emitted during contact and agonistic contexts, respectively [44]. Nevertheless, based on our acoustic analyses, the roar vocalization emitted by spotted paca seems to be unique for this species. These similarities in function and structures, in the acoustic repertoire of caviomorphs, highlight phylogenetic relationships among taxa and urge more studies to use acoustic behavior as characters to rebuild phylogenies and acquire insights into the evolution of sociality in caviomorphs.

The repertoire complexity described here corroborates the hypotheses of sociality for the spotted paca. Our results also support vocal complexity as a predictor for sociality in caviomorph rodents. Nevertheless, these findings show that measurements of complexity in both social systems and communication systems are challenging and liable to arbitrary decisions, depending on how these behavioral features are considered and which metrics are used as highlighted by Fischer et al. [71]. Moreover, we could not include data from wild spotted paca due to the species’ nocturnal habit and defensive behavior. Therefore, we must consider that our data is not a definitive description of the spotted paca’s repertoire, because contexts and motivational states could be restricted in the captive environment. Nonetheless, the data here described urges researchers to continue investigating the species’ vocal repertoire in more diverse contexts, to test the function of the vocalizations already described and to apply the knowledge about vocal communication to the investigation of evolutionary and ecologically relevant questions.

Supporting information

S1 Table. Dataset of acoustic parameters of spotted paca’s vocalizations.


S2 Table. Dataset of combination of roar and groan vocalizations.


S3 Table. Dataset of formants in roar vocalizations.


S4 Table. Dataset of acoustic types of spotted paca.



We thank the commercial spotted paca farm in Bahia state, Brazil for supporting this research. We also thanks Richard Policht and the anonymous reviewer of PlosOne for their comments on the original version of the paper. We also thank the staff of the Laboratório de Etologia Aplicada at the Universidade Estadual de Santa Cruz-UESC, especially Christini B. Caselli, for helpful contributions with the initial acoustic analyses.


  1. 1. Rowe DL, Honeycutt RL. Phylogenetic relationships, ecological correlates, and molecular evolution within the Cavioidea (Mammalia, Rodentia). Mol Biol Evol. 2002; 19: 263–277. pmid:11861886
  2. 2. Smith JE, Lacey EA, Hayes LD. Sociality in Non-Primate Mammals. In: Dustin R, Rubenstein P, editors. Abbot Comparative Social Evolution. 2017. pp 284–319.
  3. 3. Rubenstein DR, Abbot P. Comparative Social Evolution. Cambridge University Press; 2017.
  4. 4. Ebensperger LA, Wallen PK. Grouping increases the ability of the social rodent, Octodon degus, to detect predators when using exposed microhabitats. Oikos. 2002; 98: 491–497.
  5. 5. Ebensperger LA, Blumstein DT. Sociality in New World hystricognath rodents is linked to predators and burrow digging. Behav Ecol. 2006; 17: 410–418.
  6. 6. Topping MG, Miller JS, Goddard JA. The effects of moonlight on nocturnal activity in bushy-tailed wood rats (Neotoma cinerea). J Zool. 1999; 77: 480–485.
  7. 7. Alexander RD. The evolution of social behavior. Ann Rev Ecol Syst. 1974; 5: 325–383.
  8. 8. Ebensperger LA, Cofré H. On the evolution of group-living in the New World cursorial hystricognath rodents. Behav Ecol. 2001; 12: 227–236
  9. 9. Emlen ST, Oring LW. Ecology, sexual selection, and the evolution of mating systems. Science 1977; 197: 215–223. pmid:327542
  10. 10. Silk JB. The adaptive value of sociality in mammalian groups. Philos Trans Roy Soc Lond B: Biol Sci. 2007; 362: 539–559.
  11. 11. Foster EA, Franks DW, Morrell LJ, Balcomb KC, Parsons KM, van Ginneken A, et al. Social network correlates of food availability in an endangered population of killer whales, Orcinus orca. Anim Behav. 2012; 83: 731–736.
  12. 12. Chapman CA, Chapman LJ, Wrangham RW. Ecological constraints on group size: an analysis of spider monkey and chimpanzee subgroups. Behav Ecol Sociobiol. 1995; 36: 59–70.
  13. 13. David-Barrett D, Dunbar RIM. Processing Power Limits Social Group Size: Computational Evidence for the Cognitive Costs of Sociality. Proc R Soc B. 2013; 280: 20131151–.20131151. pmid:23804623
  14. 14. Sih A, Ferrari MC, Harris DJ. Evolution and behavioural responses to human induced rapid environmental change. Evol Appl. 2011; 4: 367–387. pmid:25567979
  15. 15. Sih A. Understanding variation in behavioural responses to human-induced rapid environmental change: a conceptual overview. Anim Behav. 2013; 85: 1077–1088.
  16. 16. Price TD, Qvarnström A, Irwin DE. The role of phenotypic plasticity in driving genetic evolution. Proc R Soc Lond B: Biol Sci. 2003; 270: 1433–1440.
  17. 17. Maher CR, Burger JR. Diversity of social behavior in caviomorph rodents. In: Ebensperger LA, Hayes LD editors. Sociobiology of Caviomorph Rodents: An Integrative Approach. 2016. pp 28–58.
  18. 18. Verdade LM. The influence of hunting pressure on the social behavior of vertebrates. Rev Bras Biol. 1996; 56: 1–13. pmid:8731558
  19. 19. Adolphs R. The neurobiology of social cognition. Curr Opin Neurobiol. 2001; 11: 231–239. pmid:11301245
  20. 20. Bradbury JW, Vehrencamp SL. Principles of Animal Communication. 2nd ed. 2011. Sunderland, MA: Sinauer.
  21. 21. Freeberg TM, Dunbar RIM, Ord TJ. Social complexity as a proximate and ultimate factor in communicative complexity. Philos Trans R Soc B. 2012; 367: 1785–1801.
  22. 22. Blumstein DT, Armitage KB. Does sociality drive the evolution of communicative complexity? A comparative test with ground dwelling sciurid alarm calls. Am Nat. 1997; 150: 179–200. pmid:18811281
  23. 23. Wilkinson GS. Social and vocal complexity in bats. In: de Waal F.B.M, Tyack P.L Editors. Animal social complexity: intelligence, culture and individualized societies. Cambridge, MA: Harvard University Press. 2003. pp 322–341.
  24. 24. McComb K, Semple S. Coevolution of vocal communication and sociality in primates. Biol Lett. 2005; 1: 381–385. pmid:17148212
  25. 25. May-Collado LJ, Agnarsson I, Wartzok D. Phylogenetic review of tonal sound production in whales in relation to sociality. BMC Evol Biol. 2007; 7: 136. pmid:17692128
  26. 26. Leuchtenberger C, Sousa-Lima R, Duplaix N, Magnusson WE, Mourão G. Vocal repertoire of the social giant otter. J Acoust Soc Am. 2014; 136: 2861–2875. pmid:25373985
  27. 27. Manser MB, Jansen DA, Graw B, Hollen LI, Bousquet CA, Furrer RD, et al. Vocal complexity in meerkats and other mongoose species. Adv Study Behav. 2014; 46: 281–310.
  28. 28. Pollard KA, Blumstein DT. Evolving communicative complexity: insights from rodents and beyond. Philos Trans R Soc B. 2012; 367: 1869–1878.
  29. 29. Emmons L. Cuniculus paca. The IUCN Red List of Threatened Species. e.T699A22197347. 2016.
  30. 30. Smythe N. The paca (Cuniculus paca) domestic source of protein for the neotropical humid lowlands. App Anim Behav Sci. 1987; 17: 155–170.
  31. 31. Emmons LH, Feer F. Neotropical rainforest mammals. A field guide. 1997. pp 322–323.
  32. 32. Patton JL. Family Cuniculidae GS Miller, Gidley. 1918. In: Patton JL, Pardiñas UFJ, D'Elía G editors. Mammals of South America. University of Chicago Press, Chicago and London. 2015. pp 726–733.
  33. 33. Beck-King H, von Helversen O. Home range, population density, and food resources of Agouti paca (Rodentia: Agoutidae) in Costa Rica: a study using alternative methods. Biotrop. 1999; 31: 675–85.
  34. 34. Nogueira Filho SLG, Nogueira SSC. Criação De Pacas (Agouti Paca). 1. Ed. Piracicaba, Sp, Brasil: Fundação de Estudos Agrários–Fealq. 1999. pp 1–70.
  35. 35. Figueroa-de León A, Naranjo EJ, Perales H, Santos-Moreno A, Lorenzo C. Availability and characterization of cavities used by pacas (Cuniculus paca) in the Lacandon Rainforest, Chiapas, Mexico. Rev Mex Biodivers. 2016; 87: 1062–1068.
  36. 36. Smythe N, Brown de Guanti O. La domesticación y cria de la paca (Agouti paca). Guia de conservación 26 Roma. FAO. 1995. pp 1–91.
  37. 37. Eisenberg JF. The function and motivational basis of hystricomorph vocalizations. Symp Zool Soc Lond. 1974; 34: 211–247.
  38. 38. Francescoli G, Nogueira S, Schleich C. Mechanisms of social communication in caviomorph rodents. In: Ebensperger LA, Hayes LD editors. Sociobiology of Caviomorph Rodents: An Integrative Approach. 2016. pp 147–172.
  39. 39. Arnold K, Zuberbühler K. Call combinations in monkeys: compositional or idiomatic expressions?. Brain Lang. 2012; 120: 303–309. pmid:22032914
  40. 40. Taylor AM, Reby D. The contribution of source–filter theory to mammal vocal communication research. J Zool. 2010; 280: 221–236.
  41. 41. Freeberg TM, Dunbar RIM, Ord TJ. Social complexity as a proximate and ultimate factor in communicative complexity. 2012; 367: 1785–1801. pmid:22641818
  42. 42. Altmann J. Observational study of behaviour: sampling methods. Behav. 1974; 49: 223–265.
  43. 43. Feng AS, Riede T, Arch VS, Yu Z, Xu ZM, Yu XJ, Shen JX. Diversity of the vocal signals of concave-eared torrent frogs (Odorrana tormota): evidence for individual signatures. Ethology. 2009; 115:1015–1028.
  44. 44. Barros KS, Tokumaru RS, Pedroza JP, Nogueira SSC. Vocal repertoire of captive capybara (Hydrochoerus hydrochaeris): structure, context and function. J Ethol. 2011; 116: 83–93.
  45. 45. Charif RA, Waack AM, Strickman LM. Raven Pro 1.3 User’s Manual Ithaca, New York: Cornell Laboratory of Ornithology. 2008.
  46. 46. Riede T, Fitch T. Vocal tract length and acoustics of vocalization in the domestic dog (Canis familiaris). J Exp Biol. 1999; 202: 2859–2867. pmid:10504322
  47. 47. Fitch WT. Vocal tract length and formant frequency dispersion correlate with body size in rhesus macaques. J Acoustic Soc Am. 1997; 102: 1213–1222.
  48. 48. Boersma P, Weenink D. Praat: Doing phonetics by computer version 5.5. 2012. Available from:
  49. 49. Noy Meir I, Walker D, Williams WT. Data transformations in ecological ordination. II. On the meaning of data standardization. J Ecol. 1975; 63: 779–800.
  50. 50. McGarigal K, Cushman S, Stafford S. Multivariate statistics for wildlife and ecology research. New York, Springer. 2000. pp 1–279.
  51. 51. Felsenstein J. Phylogenies and the comparative method. Am Nat. 1985; 125: 1–15.
  52. 52. Felsenstein J. Inferring phylogenies. 2003. Sunderland: Sinauer Associates. Pp 1–645.
  53. 53. Tomasco IH, Boullosa N, Ho FG, Lessa EP. Molecular adaptive convergence in the α-globin gene in subterranean octodontid rodents. Gene. 2017; 628: 275–280. pmid:28735726
  54. 54. Purvis A. A composite estimate of primate phylogeny. Philos Trans R Soc Lond B: Biol Sci. 1995; 348: 405–421.
  55. 55. Ebensperger LA. A review of the evolutionary causes of rodent group-living. Acta Theriol. 2001; 46: 115–144.
  56. 56. Freeberg TM. Social complexity can drive vocal complexity. Psychol Sci. 2006; 7: 557–561.
  57. 57. Aquino R, Deyber G, Etersit P. Aspectos ecológicos y sostenibilidad de la caza del majás (Cuniculus paca) en la cuenca del río Itaya, Amazonía peruana. Rev Peru Biol. 2009; 16: 67–72.
  58. 58. Leuchtenberger C, Tirelli FP, Mazim FD, Peters FB, de Oliveira ÊS, Cariolatto L, et al. New records of Cuniculus paca (Rodentia: Cuniculidae) in a temperate grassland dominated landscape of Pampas region of Brazil and Uruguay. Mammalia. 2016; 0: 1–4.
  59. 59. Queirolo D, Vieira E, Emmons L, Samudio R. Cuniculus paca. The IUCN Red List of Threatened Species. Version 2014.3. 2008. Available from:
  60. 60. Gallina S, Pérez-Torres J, Guzmán-Aguirre CC. Use of the paca, Cuniculus paca (Rodentia: Agoutidae) in the Sierra de Tabasco State Park, Mexico. Rev Biol Trop. 2012; 60: 1345–1355. pmid:23025103
  61. 61. Morrell LJ, Ruxton GD, Richard J. The temporal selfish herd: predation risk while aggregations form. Proc R Soc Lond B: Biol Sci. 2011; 278: 605–612.
  62. 62. Schradin C, Lindholm AK, Johannesen JES, Schoepf I, Yuen CH, Koenig B, Pillay N. Social flexibility and social evolution in mammals: a case study of the african striped mouse (Rhabdomys pumilio). Mol Ecol. 2012; 21: 541–553. pmid:21883591
  63. 63. Krause J, Ruxton GD. Living in Groups. Oxford University Press, New York. 2002. pp 1–189.
  64. 64. Owren MJ, Rendall D. Sound on the rebound: bringing form and function back to the forefront in understanding nonhuman primate vocal signaling. Evol Anthr. 2001; 10: 58–71.
  65. 65. Reby D, McComb K. Anatomical constraints generate honesty: acoustic cues to age and weight in the roars of red deer stags. Anim Behav. 2003; 65: 519–530.
  66. 66. Rendall D, Owren MJ, Weerts E, Hienz RD. Sex differences in the acoustic structure of vowel-like grunt vocalizations in baboons and their perceptual discrimination by baboon listeners. J Acoust Soc Am. 2004; 115: 411–421. pmid:14759032
  67. 67. Long CV. Vocalisations of the degu Octodon degus, a social caviomorph rodent. Bioacoustics. 2007, 16: 223–244.
  68. 68. Monticelli PF, Ades C. The rich acoustic repertoire of a precocious rodent, the wild cavy Cavia aperea. Bioacoustics. 2013; 22: 49–66.
  69. 69. Alencar Jr RN. O repertório acústico de um especialista de rochedos da Caatinga, o mocó. Master Thesis; Universidade de São Paulo. 2011. Available from:
  70. 70. Schleich CE, Busch C. Acoustic signals of a solitary subterranean rodent Ctenomys talarum (Rodentia: ctenomyidae): physical characteristics and behavioural correlates. J Ethol. 2002; 20: 123–131.
  71. 71. Fischer J, Wadewitz P, Hammerschmidt K. Structural variability and communicative complexity in acoustic communication. Anim Behav. 2016; 124: 1–9.
  72. 72. Lima, SGC. Comportamento acústico e complexidade social em caivioidea. Master Thesis; Universidade Federal do Rio Grande do Norte. 2016. Available from: S5 Text.
  73. 73. Veitl S, Begall S, Burda H. Ecological determinants of vocalization parameters: the case of the coruro Spalacopus cyanus (Octodontidae), a fossorial social rodent. Bioacoustics. 2000; 11: 129–148.
  74. 74. Busch C, Malizia AI, Scaglia OA, Reig OA. Spatial distribution and attributes of a population of Ctenomys talarum (Rodentia: Octodontidae). J Mammal. 1989; 70: 204–208.
  75. 75. Dubost G. Ecology and social life of the red achouchy, Myoprocta exilis; Comparison with the orange-rumped agouti, Dasyprocta leporina. J Zool Lond. 1988; 214: 107–123.
  76. 76. Asher M, Spinelli O, Sachser N. Social system and spatial organization of wild guinea pigs (Cavia aperea) in a natural population. J Mammal. 2004; 85: 788–96.
  77. 77. Ebensperger LA, Chesh AS, Castro RA, Tolhuysen LO, Quirici V, Burger JR, et al. Burrow limitations and group living in the communally rearing rodent, Octodon degus. J Mammal. 2011; 92: 21–30. pmid:22328789
  78. 78. Nowak RM. Walker's mammals of the world. Johns Hopkins Press. 1999. pp 1–1921.
  79. 79. Begall S, Burda H, Gallardo MH. Reproduction, postnatal development, and growth of social coruros, Spalacopus cyanus (Rodentia: Octodontidae), from Chile. J Mammal. 1999; 80: 210–217.
  80. 80. Mones A, Ojasti J. Hydrochoerus hydrochaeris. Mamm Species. 1986; 264: 1–7.