Group Composition and Monandry in Grizzled Langurs, Presbytis comata, on Java

The genus Presbytis sensu stricto comprises a group of arboreal colobines found in the Indomalayan region, from southern Thailand, south to Sumatra, Java and Borneo and smaller off-lying islands. Over the last 3 or 4 decades, partially aided by molecular phylogenetic analysis [Meyer et al., 2011; Vun et al., 2011] and partially by adopting different species concepts [Groves, 2001], the number of species has increased from 8 [Davies and Oates, 1994] to 17 [Zinner et al., 2013]. While Bennett and Davies [1994] concluded that differences in socioecological behaviour in Presbytis langurs appear to be due to differences in habitat rather than to intrinsic differences between species, when group size, morphology and antipredator behaviour are taken into account, the genus can be divided into 3 distinct groups [Nijman and Nekaris, 2012]. First are the ones that live primarily in large groups, comprising 1 or 2 adult males with several adult females plus offspring, and that are typically brightly coloured and vocal when disturbed (e.g., banded langur, P. femoralis [Curtin, 1980]; red langur, P. rubicunda [D'Agostino et al., 2016]). Second are those species that live in smaller 1-male/multifemale groups, where individuals are typically bicoloured and that are less vocal when disturbed (e.g., Hose's langur, P. hosei [Nijman, 2010], or Thomas's langur, P. thomasi [Steenbeek and Assink, 1998]). Third are species that live in 1-male/1-female groups, and where individuals are dark in colour and that rely on crypsis as an antipredator strategy (e.g., Siberut langur, P. siberu [Tilson and Tenaza, 1976]). The available data for at least 1 species, the white-fronted langur P. frontata, is suggestive of a monandrous social organization (sensu Tenaza and Fuentes [1995]) i.e., 1-male/multifemale and 1-male/1-female groups coexist in a single population¹ [Nijman and Nekaris, 2012]. Too few data are available for a number of species regarding their socioecology to make firm conclusions (e.g., Natuna langur, P. natunae [Lammertink et al., 2003; Nijman and Nekaris, 2012]), or there are conflicting reports or apparent differences in interpretation concerning a species' social organization. The latter is the case for P. comata, the grizzled langur from Java.

Based on data collected in 1976-1977 by Ruhiyat [1983, 1991] at 1 of 2 sites in western Java (i.e., Telaga Patengan, also known as Situ Patenggang; Telaga and Situ meaning Lake in Bahasa Indonesia and Bahasa Sunda, respectively), P. comata is sometimes cited as one of relatively few monogamous Old World monkeys [Watanabe, 1981; Lott, 1991; Davies and Oates, 1994; Yeager and Kool, 2000]. Ruhiyat [1983, 1991] never mentions monogamy, but four 1-male/1-female groups with offspring, alongside two 1-male/multifemale groups, were present at Telaga Patengan during the 14-month study period. Others consider P. comata to occur in 1-male/multifemale groups with offspring [Leutenegger, 1982; Sterck, 2012; Zinner et al., 2013], again largely based on data collected by Ruhiyat [1983, 1991], but this time from his second study site Kawah Kamojang (Kamojang Crater). At Kawah Kamojang, six 1-male/multifemale groups but no 1-male/1-female groups were present over the 24-month study period [Ruhiyat, 1983, 1991].

Reviewing the evolution of social monogamy in mammals, Lucas and Clutton-Brock [2013] found that male-biased sexual dimorphism is found in only about half of the socially monogamous species and that social monogamy only evolved in species in which females are at least as large as males. In primates, Opie et al. [2013] found evidence for linked evolution between social monogamy and discrete female ranging patterns (with little overlap between females), biparental care, and risk of male infanticide. In the approximately 150 species of Old World monkeys, the group to which P. comata belongs, social monogamy has been reported, often with mixed evidence, in De Brazza's monkeys (Cercopithecus neglectus) and P. siberu, whereas monandry has been reported in P. frontata and the simakobu Simias concolor [Watanabe, 1981; Leutenegger and Lubach, 1987; Tenaza and Fuentes, 1995; Nijman and Nekaris, 2012]. Sexual dimorphism is pronounced in C. neglectus and S. concolor, but not in P. frontata or P. potenziani; extensive range overlap is present in C. neglectus and P. potenziani, but not in S. concolor, and direct evidence of infanticide is lacking for all 4 species.

Here I present data on the group sizes, group composition and support for a monandrous social system in P. comata based on an island-wide survey and a review of the available literature. I test 4 explicit hypotheses: (1) group sizes and group composition in P. comata vary, but this is not related to altitude, forest area or distance to the forest edge; (2) all-male groups and solitary males, if present, show a similar ranging pattern to that of 1-male/multifemale groups; (3) 1-male/1-female groups, if present, show a similar ranging pattern to that of 1-male/multifemale groups; (4) P. comata groups that form polyspecific associations with other primates, if any, are similar in size to 1-male/multifemale groups, and they show a similar ranging pattern.

P. comata is endemic to the western part of the island of Java, Indonesia [Nijman, 1997b; Groves, 2001] and is listed as Endangered on the IUCN Red List [Nijman and Richardson, 2008]. It is an arboreal species and as such is confined to tall forest that remains scattered along the island. Brandon-Jones [1997] recognizes the eastern populations, or at least those on Mt. Slamet and Mts. Dieng, as a separate species (P. fredericae). Following Nijman [1997a], Groves [2001], and Zinner et al. [2013], I consider the western and eastern populations conspecific.

There is very little sexual dimorphism in P. comata, either in body mass or body size measurements (Table 1). In terms of head-body length, tail length, total length, ear size, canine length and molar size, males are, on average, 2.3% (range 0.7-5.2%) larger than females, but none of these differences is statistically significant (Student t test, all t > 1.44, all p > 0.160). Male feet are 10.5% longer that female feet, but this difference is not significant either (t = 2.14, p = 0.064) (Table 1). Data presented by Smith and Jungers [1997], based on an unpublished 1976 thesis from the Anthropological Institute in Zurich, suggest a slightly reversed sexual dimorphism with females being 0.4% heavier than males (Table 1). However, the provenance of the animals on which this is based is not known and, importantly, they were listed as P. aygula in the 1976 thesis, meaning that they could refer to P. comata, P. hosei(from Borneo), P. thomasi (from Sumatra), or a combination of these taxa (data for P. hosei and P. thomasi in Smith and Jungers [1997] were obtained from a different unpublished thesis). None of the adult wild-caught specimens in the collections of the zoological museums in Leiden (the Netherlands), London (UK), Cibinong (Indonesia), and Singapore I have examined had data on body mass associated with it.

The island of Java (126,500 km²), Indonesia's cultural, political and economic centre, is one of the most densely populated areas in the world. The provinces of Banten, West Java, and Central Java, i.e., the area where P. comata is found, has a human population of over 90 million, at a density of 1,200 people km^-2 (data from 2015 [BPS 2017]). Java is largely deforested, and most of the remaining forest fragments cover (parts of) the numerous volcanoes on the island. Once the island was probably completely covered by tropical forest, but its destruction dates back several centuries, probably to the 10th century AD [Whitten et al., 1996]. In effect most of the lowland forest has been replaced by a mosaic of cities and villages, agricultural land, cash crop plantations, and forest plantations, rendering the natural forest areas, and the primate populations in them, into habitat islands [Nijman, 2013].

Between the 4th and 14th century AD, Java was ruled by Hindu-Buddhist kingdoms, and from the 16th century onwards Islam became the dominant religion on the island. At present more than 90% of the people living in Java are Muslims. Primates are considered haram (forbidden to be eaten) under Islamic tenets, and hunting of primates for food is far less of an issue on Java than it is in other parts of Indonesia (including, for instance, the Mentawai Islands from where Tenaza and Fuentes [1995] reported S. concolor to be monandrous). On Java, primates are captured to meet the demand for the pet trade, but this is less of an issue for P. comata than it is for species such as slow lorises (Nycticebus spp.) or macaques (Macaca spp.), possibly because the langur is very difficult to keep alive in captivity. Support for this has recently been reported by Nijman et al. [2016] who documented the trade of approximately 2,500 primates in Java and Bali, of which only 2 or 3 were P. comata.

Over the period of 1994-2013, I conducted a survey covering all major large forest areas in Java, during which P. comata was studied in 12 forest areas for a total of approximately 340 days inside the forest (Table 2). The study sites cover the species' entire distributional and altitudinal range (Fig. 1). Groups were located by means of transect walks in the forest or from vantage points overlooking the forest. Almost all groups were unhabituated allowing generally only brief observations, although in some situations (e.g., while seated at watchtowers overlooking the forest) groups could be observed at ease. The best counts were obtained when a group crossed a gap, including small streams or forest trails. As much as possible, groups were counted several times, but are included only once in the analysis. I identified groups either by recognizable individuals or, given that the species have stable home ranges and are territorial, by matching observations [Chapman and Chapman, 2000]. Group ranges are typically stable over time and, especially in the smaller forest areas where few groups were present, it was possible to recognize different groups over time.

Despite the low degree of sexual dimorphism (Table 1), in the field males appeared slightly more robust than females, and adult females could often be sexed by protruding nipples and/or the presence of dependent young. Age classes were identified on the basis of relative size and pelage coloration. Group counts were grouped in 3 classes: (i) known group composition, i.e., number of adult males, adult females, subadults of either sex, juveniles, and neonates; (ii) complete counts, i.e., the total number of individuals without records of all the age and/or sex classes; (iii) minimum counts, where a certain number of individuals was observed, but where in all likelihood the actual number was greater.

A group of P. comata and 1 or more other primate species was considered polyspecific if individuals of the respective species remained in close proximity (less than approx. 10 m) for a prolonged period (>1 h), showed spatial overlap, and if individuals of the different species clearly interacted with each other, either by simultaneous vocalizing, feeding or travelling together.

For each group I recorded the following variables: (i) altitude, recorded with a wristwatch altimeter or GPS, accurate to the nearest 50 m; (ii) estimated distance of the group in the forest interior to the nearest forest edge; (iii) longitude-latitude, recorded with a GPS or taken from forestry maps. For multiple sightings of the same group the estimated centre point of their range was taken. Each record, furthermore, received a climate score, based on the number of rainy days received during the driest 4 consecutive months of the year. These data were taken from Van Steenis [1972]. The dry season is defined as the months June to September, the remaining months comprise the wet season [RePPProT, 1990]. Rainfalls in areas where P. comata was recorded range from a mean of 20-50 mm month^-1 for the dry season to a mean of 150-300 mm month^-1 for the wet season. Size of forest areas (in square kilometres) was taken from van Balen et al. [2001] and my own field data.

Additional data on group sizes in P. comata were obtained from the literature, including unpublished reports and theses. Group counts reported as part of density surveys (along transects or paths) often underestimate true group sizes and were normally not included.

In exploring variations between group sizes with environmental variables and in group comparisons only groups of known composition or complete counts were included. Some groups recorded during different years might have been included twice in the analysis, but these comprise only a maximum of 11% of the total (6/55) with on average 40 months between counts (range 14-77 months). Given the extent of deforestation on Java, there are no forest areas within the range of P. comata that still cover the entire altitudinal range from sea level to montane (Table 2). Hence, in exploring relationships between environmental variations, including the effect of altitude, data from all forest areas were pooled, as the alternative of within-forest comparisons was not an option for most forest areas.

Data were checked for normality, and data on forest area, altitude, and distance to the forest edge were log-transformed in order to approach a normal distribution. All analyses were checked for differences between populations east and west of 108° longitude (an arbitrarily chosen line running approximately halfway between the areas from where P. (c.) comata and P. (c.) fredericae have been described). No significant differences were found, and thus no further distinction is made between the western and eastern regions.

After checking for colinearity between data, a simple linear multiregression model was used to select the best subset of variables that could explain the observed variation in group size (for all statistical procedures, see Sokal and Rohlf [1995]). Means are reported ± 1 SD. All tests were 2-tailed, and significance was assumed when p < 0.05 although occasionally trends are mentioned when 0.05 < p < 0.10.

For 50 groups containing both adult males and adult females, accurate counts could be made of their group size, averaging 7.1 ± 2.6 individuals (range 2-13 individuals). On 2 occasions, 2 P. comata groups (of 10 and 9 individuals, and 8 and 7 individuals, respectively) remained in close proximity over the course of several hours without any apparent aggressive interaction between them, giving the impression of 2 very large groups. Subsequent observations revealed that they comprised 4 separate groups. None of the groups contained more than 1 adult male and only twice were groups (both comprising 4 individuals) confidently classed as containing only 1 adult male and 1 adult female.

The most frequently recorded group size during the surveys was that of 6 individuals (18% of total). The largest proportion (16%) of individuals in a group was 8, followed by groups of 6 or 9 (Fig. 2). As such, most of the individuals were in groups >6 individuals, and almost half of them in groups of >8 individuals.

Three all-male groups were recorded, and twice a solitary male was observed. Group size of all-male groups averaged 2.2 ± 1.3 individuals (1, 1, 2, 3, and 4 individuals), and these were significantly smaller than 1-male/multifemale groups (Student t test, t = 7.05, p = 0.0001). All-male groups, including the 2 solitary males, ranged significantly closer to the forest edge compared to 1-male/multifemale groups (mean distance to edge 108 ± 167 and 1,007 ± 1,062 m, for all-male and 1-male/multifemale groups, respectively; t = 5.65, p = 0.001), but not at a different altitudinal level (mean altitude 1,010 ± 408 and 1,118 ± 538 m above sea level (a.s.l.), for all-male and 1-male/multifemale groups, respectively; t = 0.55, ns).

Polyspecific associations between grizzled langurs and other primates were restricted to Javan gibbons, Hylobates moloch, or ebony langurs, Trachypithecus auratus. The other diurnal primate occurring on Java, the long-tailed macaque Macaca fascicularis has a preference for edge and often more disturbed habitats, although in some areas both M. fascicularis and P. comata occurred sympatrically. Only once did P. comata associate with both H. moloch and T. auratus at the same time. For 6 associations, group sizes of P. comata were established confidently. Polyspecific groups often remained stable over the course of several hours and even a number of days. The ratio of the observed number of polyspecific groups versus other groups differed significantly between montane forest (>1,500 m a.s.l.) and lowland to lower montane forest (χ² = 5.8, df = 1, p = 0.02), with more polyspecific groups being observed at higher altitudes. Polyspecific groups showed a tendency to be somewhat smaller than other 1-male/multifemale groups (4.8 ± 2.3 individuals; t = 2.24, p = 0.08) and to occur at higher altitudes (1,747 ± 613 m a.s.l.; t = 2.50, p = 0.07), but did not range further or closer to the forest edge (1,277 ± 1,378 m; t = 0.55, ns).

In terms of composition of bisexual groups, based on results of the survey and data presented in the literature (Table 3), it is clear that over 95% of the groups have only 1 adult male. The number of adult females ranges from 1 to 7 with a mean of 2.1 ± 1.4, and the infant-adult female ratio is 0.2 ± 0.3. The overall group size of 6.1 ± 2.8 is somewhat smaller than when combining complete counts with groups with known composition as it is probably easier to confidently identify and sex all individuals in small groups than in large groups.

There was no apparent variation in group sizes along the latitudinal axis of the island. Annual variation in group sizes was not apparent (1-way ANOVA, F_{1, 12} = 1.2, ns) nor was seasonal variation in group sizes (7.1 ± 2.9 and 6.9 ± 2.2 individuals for the dry and wet seasons, respectively; t = 0.26, ns). P. comata was recorded only in the wettest parts of Java, and no difference was found in group sizes between areas with hardly any dry season (>30 rainy days during the 4 driest consecutive months) or with a slightly dry season (10-30 rainy days) (7.1 ± 2.7 and 6.0 ± 2.0 individuals, respectively). Consequently, the effects of season and climate type on group size were not taken into further consideration.

Group size was significantly negatively correlated with altitude (F_{15, 1} = 4.55, R_adj² = 0.09, p = 0.04), and mean group sizes go down roughly by 1 individual with every 600 m rise in elevation (Fig. 3). Mean group size for each of the 12 forest areas was positively correlated with the size of the forest area (log transformed) (F_{10, 1} = 4.3, R_adj² = 0.23, p = 0.07). Group size was positively related to the distance to the forest edge (log transformed), however not significantly so (F_{46, 1} = 2.0, R_adj² = 0.04, p = 0.16).

In a simple linear multiple regression model, altitude and distance to the forest edge (log transformed) proved to be the best subset of predictors for the observed variation in group sizes for grizzled langurs. The equation group size = 4.53 - 0.00167 altitude + 1.51 log distance to the forest edge explained some 17% of the observed variation in group sizes in P. comata (F_{52, 2} = 6.49, p = 0.003).

In terms of composition of bisexual groups, based on results of the survey and data presented in the literature (Table 3) there are clear altitudinal differences. While the number of adult males, subadults, juveniles, and infants does not differ between lowland and hill, lower montane and montane forest, there are clear differences in the total group sizes (1-way ANOVA, F_{2, 63} = 5.15, p = 0.0085) and the number of adult females (F_{2, 63} = 8.06, p = 0.0008); only the differences between groups recorded in the lowland and hill forest and those groups in the montane forest are significant (t = 3.28, p = 0.0019, and t = 4.153, p = 0.0001 for group size and number of adult females, respectively) (Table 3). Number of fewer adult females does not translate itself into smaller infant/adult female ratios as these are 0.18 (lowland and hill forest), 0.19 (lower montane forest) and 0.17 (montane forest).

An overview of group sizes in 8 forest areas where P. comata has been studied in detail (Table 4) likewise shows a negative relationship between the altitude of the study area and mean group sizes, especially when considering the maximum altitudes at which P. comata was observed (data log transformed; minimum elevation: Pearson R = -0.595, p = 0.091; maximum elevation: R = -0.708, p = 0.032).

Only 2 out of 55 groups could be reliably classed as consisting of 1 adult male/1 adult female with offspring. One of these groups was observed in Telaga Patengan in 2000, where Ruhiyat [1983] reported that 4 of his 6 study groups comprised 2 adults plus offspring in 1976-1977. Wedana [1993] reported that 1 of his 4 study groups in Telaga Patengan comprised 1 adult male and 1 adult female, whereas Nurjaman et al. [2002] found 1 of 7 groups in Telaga Patengan to comprise 1 adult male and 1 adult female. Harjenti [1996] studied a population of P. comata at elevations between 2,000 and 2,200 m a.s.l. at Brussel on Mt. Patuha, in a straight line 2 km east of Telaga Patengan. She reported that 4 of her 6 study groups comprised a single adult male/single adult female with offspring (2 adult females were present in the remaining 2 groups).

P. comata is distributed in the last remaining forest areas of Java, where it can be found as far east as Mt. Merbabu and Mt. Lawu on the border with East Java. Large-scale deforestation in especially the lowlands and hill zones means that many populations are confined to montane areas, and here the species can be found at high altitudes up to 2,565 m a.s.l. Comparisons with the 15 other species of Presbytis langurs (with the exception of the monogamous P. potenziani and the possibly monandrous P. frontata) shows that throughout most of its range P. comata is a rather typical representative of the genus [reviewed in Newton and Dunbar, 1994; Bennett and Davies, 1994; Yeager and Kirkpatrick, 1998; Yeager and Kool, 2000; Sterck, 2012; Zinner et al., 2013]. In terms of the most common social unit (1 male/multiple females with offspring), male emigration (as suggested by the presence of solitary males, all-male groups), group size (means of 6-7 individuals) and habitat use (undisturbed lowland forest preferred over degraded forests and plantations [Nijman, 1997b, 2010]), the species shows no anomalies. Groups with 2 adult males/multiple females have been recorded, but very infrequently, while 1-male/1-female groups do occur in certain areas.

All-male groups seemed to have at least partially different ranging patterns and possibly show different habitat use than 1-male/multifemale groups, as they were significantly more often found close to the forest edge. Differential habitat use or ranging patterns of all-male and bisexual groups are scarcely documented but have been reported for P. thomasi [Steenbeek, 1999] and the pale-thighed langur P. siamensis [Bennett, 1983].

P. comata infrequently forms polyspecific associations with other langurs and gibbons. Polyspecific associations between Presbytis and other langurs (Presbytis, Trachypithecus, Nasalis) have been reported [e.g., Bartels, 1937; Medway, 1970; Curtin, 1980; Ruhiyat, 1983; Wakidi, 2013; Rowe and Myers, 2017] although in South-East Asian rain forest polyspecific associations between primates are far less common than in African or Neotropical forests [Chapman and Chapman, 2000; Heymann and Buchanan-Smith, 2000; Fam and Nijman, 2011]. It is suggested that polyspecific associations are formed under predation pressure [e.g., Struhsaker, 1981, 2000; Noë and Bshary, 1997]. Differences between Africa and the Neotropics on the one hand and Asia on the other have most commonly been explained by differences in predation pressure, in particular the apparent absence of specialized primate-targeting raptors in Asia [Hart, 2007; Fam and Nijman, 2011].

Compared to other Presbytis langurs studied to date, P. comata does occur over a large altitudinal range with resident populations occurring in upper montane forests >2,000 m a.s.l. Polyspecific associations were mostly found at high altitudes, and P. comata groups that were part of such polyspecific associations showed a tendency to be smaller than other groups. Struhsaker [1981] noted a relation between the formation of polyspecific associations and group size, with species with smaller group sizes being more often the initiator of the formation. The relatively high occurrence of polyspecific associations in P. comata at high altitudes may be a response to overcome the negative effects of small group sizes that are imposed by the low productivity of the environment. In order to test this notion, however, more quantitative data are needed about predator pressure, densities and productivity of the forest at the different altitudinal levels.

P. comata has been considered to be essentially confined to lowland and hill forest below 1,250 m a.s.l. [e.g., Hoogerwerf, 1970; Medway, 1970; Whitten et al. 1996], or alternatively to a small narrow band between 1,200 and 1,800 m a.s.l. [Supriatna et al. 1994; Rowe, 1996], but in fact it occurs from sea level to >2,500 m a.s.l. (Tables 2, 3, 4). In some areas P. comata has been repeatedly recorded at altitudes >2,000 m a.s.l. [Harjenti, 1996; Nijman and van Balen, 1998; Syarifah, 2013; Hidayat, 2016; this study], amidst dwarfed forests, suggesting resident populations. Invariably these groups consisted of only a few individuals. Given the lower primary production of montane forests compared to forests at lower elevations with a different species composition [Whitten et al., 1996] population densities of primates on Java are lower at higher altitudes [Sugarjito et al. 1997; Nijman, 2015]. The difference in structure and species composition between lowland and hill forest and montane forest reflects itself in pronounced differences in the diet of P. comata [Kavana et al., 2015; Yang et al., 2016]. Ruhiyat [1983] found the species to be largely folivorous at altitudes between 1,400 and 1,650 m a.s.l., with 65.2% of the diet comprising leaves and flowers (59.1% young leaves; 5.6% mature leaves, 7% flowers) and only 14.2% comprising fruits and seeds (13.5% fruit, 0.7% seeds). This is similar to that reported by Sujatnika [1992] from similar altitudes (1,250-1,550 m a.s.l.) on Mt. Gede-Pangrango, i.e., 65.0% young leaves, 5.0% mature leaves, 17.0% flowers, 6.4% fruits, and 2.3% seeds. At lower elevations the species is more frugivorous. Sugiarto [2006] studied the species at 700 m a.s.l. on Mt. Pangrango where 38.2% of the diet comprised fruits and seeds and 59.6% leaves. Suryana [2010] studied the species at 500-600 m a.s.l. in Mts. Dieng and found it to feed largely on fruits and seeds (i.e., 24.9% seeds, 17.2% fruits and seeds, 8.9% fruits) and to a lesser extent on leaves and flowers (i.e., 35.8% young leaves, 4.0% mature leaves, and 1.6% flowers). A diet comprised largely of fruits and seeds is very much in line to that which has been reported for the majority of other Presbytis species [e.g., Bennett and Davies, 1994].

The montane environment may impose a selective pressure towards not only smaller group sizes and a different diet, but also a different social system, with ultimately monandry or monogamy as a more favourable evolutionary strategy. The reduction in group sizes along the altitudinal gradient was driven largely by a reduction in the number of adult females that were present in higher-altitude groups, and not by a difference in the number of adult males, subadults, or infants. In non-territorial groups it is probably the dispersion, quality, and size of the food patches within the group's range, rather than the size of the home range, that determine the ability of males to maintain exclusive access to females. Male P. comata are unlikely to defend food sources so there is little advantage for females to associate with more than 1 adult male. In poor-quality habitat, relative to the langur's specific feeding needs, high group cohesion, especially for larger groups, can only be maintained by ranging further, at a greater speed, whilst feeding on fewer and lower-quality food patches [Beckwith, 1995]. The body size of P. comata, its locomotor behaviour and arboreal needs, and its digestive strategy limit these options, but reducing group sizes may allow it to persist in low-quality habitats.

Both males and females disperse in Presbytis [Sterck et al., 2005], and new group formation can be achieved by a number of females splitting from the main group while being joined by an outside male [Davies, 1987]. Females can also join neighbouring groups, with females from large groups often joining smaller groups [Sterck, 1997]. Steenbeek and van Schaik [2001] found that in P. thomasi the upper limit to group size was not set by feeding competition. Instead, they suggested that female reproductive success was maximized in small to mid-sized groups because larger groups show a clear trend to experience a higher risk of take-over. Females can redistribute themselves either by forming new groups or by joining nearby groups during male tenure change, and by doing so adult females can keep the group small [Steenbeek and van Schaik, 2001]. While there are no altitudinal differences in reproductive success for adult females, it appears that in P. comataselective pressures differ with altitude, and possibly females disperse sooner or more frequently forming new, small, groups at higher elevations.

If monandry or social monogamy is expected in P. comata, it is most likely at high altitudes. Small group sizes however do not necessarily imply a monogamous mating pattern. In the present study, several groups were observed to show considerable spatial overlap indicating the potential for intergroup copulations. It has been argued that monogamy has to be defined as a prolonged association (>1 breeding cycle) and essentially exclusive mating relationship between 1 male and 1 female [Fuentes, 2000]; these data are lacking for P. comata.

In the present survey 2 out of 55 groups could be reliably classed as 1-male/1-female with offspring, but no data were obtained on whether these small groups persist over long periods of time. Data from the forests of Telaga Patengan and adjacent Brussel collected intermittently between 1976 and 2002, however, suggest that in the forest of Mt. Patuha 1-male/1-female groups are consistently present alongside 1-male/multifemale groups. The five 1-male/multifemale groups observed in Telaga Patengan averaged 6 individuals, and this is only slightly below the average for Java as a whole (Table 3). Ruhiyat [1983] drew attention to the unusual group composition, the high population density, the unusual high overlap between groups, and the frequent splitting of 1 of the larger groups (10 individuals including 2 adult males) into 2 smaller groups (each containing 1 adult male). He attributed this to the recent isolation and strong reduction of the forest area around Telaga Patengan (due to an enlargement of the surrounding tea plantation and due to logging) and suggested that at least 2 of the groups had only very recently moved into his study area [Ruhiyat, 1991, p. 165]. The atypical situation at Telaga Patengan was the reason to move his study to another, less disturbed site [Ruhiyat, 1983, 1991]. Group sizes in P. comata are positively correlated with forest area and distance to the forest edge and are furthermore negatively related with altitude. Smaller than average group sizes are hence expected in small or highly fragmented high-altitude forest areas, of which Telaga Patengan (1,650 m a.s.l., 10-15 km²) is a prime example.

For 3 decades after Ruhiyat [1983] had published the results of his fieldwork, his findings were often the only ones used for P. comata in comparative studies [e.g., van Schaik and Hörstermann, 1994; Bennett and Davies, 1994; Wright and Willis, 2012; Tsuji et al., 2013]. Other studies that were published on P. comata in mainstream scientific journals over this period that contained primary data dealt with distribution, conservation, or taxonomy [Weitzel and Groves, 1985; Melish and Dirgayusa, 1996; Nijman, 1997a, b; Nijman and van Balen, 1998; Meyer et al., 2011; Supartono et al., 2016]; the few ecological studies that were conducted either appeared in Indonesian journals, internal reports or remained hidden in unpublished theses [Sujatnika, 1992; Wedana, 1993; Nurdiana, 1997; Sugarjito et al., 1997; Nurjaman et al., 2002; Heriyanto and Iskandar, 2004; Suryana, 2010; Sawitri et al., 2010; Syarifah, 2013; Hidayat, 2016]. As such, the populations of P. comata that were present in Kawah Kamojang and Telaga Patengan in the 1970s have become the standard for the species, ignoring (perhaps unintentionally) the socioecological intraspecific variation that is present. The present study showed that the variation in group sizes and group composition in P. comata can, at least in part, be explained by environmental variables such as altitude and forest characteristics. Within primates, there does not seem to be a single prime correlate of social monogamy. Instead, social monogamy may result from an array of selective pressures that act alone or in combination, some more important, some less important in a given environment at a given time [cf. Thalmann, 2001]. The role of environmental factors (climate, productivity, disturbance) in shaping the social system within species seems to be rarely quantified, and this may be a promising avenue along which to proceed with further research.

The surveys were conducted in co-operation with the Directorate General for Forest Protection and Nature Conservation, the Ministry of Forestry and Estates Crops and under the sponsorship of the Indonesian Institute for Sciences. Financial support was received from Stichting het Kronendak, the Netherlands Commission for International Nature Protection and the Society for the Advancement of Research in the Tropics. For help and information I would like to thank Dewi M. Prawiradilaga (LIPI, Cibinong), Budiman (IPB, Bogor), Iwan Setiawan (PILI, Bogor), Sujatnika (Aksenta, Bogor), I Made Wedana Adi Putra and Asep R. Purnama (Konus, Bandung), Kunkun J. Gurmaya and Nurdiana (Padjadjaran University, Bandung), Andi Prima Setiadi (YPAL, Bandung), Resit Sözer (Cikananga Wildlife Rescue Centre, Sukabumi), and Bas van Balen (Basilornis Consultancy, Arnhem). Two reviewers made helpful suggestions that improved the paper.

There is no conflict of interest to declare.