Sexual dichromatism in the neotropical genus Mannophryne (Anura: Aromobatidae)

Recent reviews on sexual dichromatism in frogs included Mannophryne trinitatis as the only example they could find of dynamic dichromatism (males turn black when calling) within the family Aromobatidae and found no example of ontogenetic dichromatism in this group. We demonstrate ontogenetic dichromatism in M. trinitatis by rearing post-metamorphic froglets to near maturity: the throats of all individuals started as grey coloured; at around seven weeks, the throat became pale yellow in some, and more strongly yellow as development proceeded; the throats of adults are grey in males and variably bright yellow in females, backed by a dark collar. We demonstrated the degree of throat colour variability by analysing a large sample of females. The red: green (R:G) ratio ranged from ~1.1 to 1.4, reflecting variation from yellow to yellow/orange, and there was also variation in the tone and width of the dark collar, and in the extent to which the yellow colouration occurred posterior to the collar. Female M. trinitatis are known to be territorial in behaviour. We show a positive relationship between throat colour (R:G ratio) and escape performance, as a proxy for quality. Our field observations on Tobago’s M. olmonae showed variability in female throat colour and confirmed that males in this species also turn black when calling. Our literature review of the 20 Mannophryne species so far named showed that all females have yellow throats with dark collars, and that male colour change to black when calling has been reported in eight species; in the remaining 12 species, descriptions of males calling are usually lacking so far. We predict that both dynamic and ontogenetic sexual dichromatism are universal in this genus and provide discussion of the ecological role of dichromatism in this genus of predominantly diurnal, non-toxic frogs, with strong paternal care of offspring.


Introduction
Most anuran amphibians are active mainly at night and intra-specific communication is mainly mediated by acoustic signals [1]. However, as more complexity in anuran behaviour is band of variable tone and width, extending across the ventral surface at the level of the forelimbs). Males possess the collar, but the throat is grey and Wells comments that it is not pulsated during aggressive encounters between males. Bright yellows are often based on carotenoid pigments, regarded as costly to synthesise, and in many taxa, including some anurans, they have been associated with signals of quality (reviewed by Olson & Owens [18]; tree frog example: Richardson et al. [19]).
In this paper, we assess the variability of the yellow pigmentation in a large sample of M. trinitatis females, and report on an experiment where we used escape responses as a proxy for quality. We also document the development of the yellow throat colour as metamorphs grow towards maturity. Finally, we provide field observations on the behaviour of both M. trinitatis and M. olmonae, and review what has been reported on sexual dichromatism in other Mannophryne species.

Colour variability in female Mannophryne trinitatis
A large sample (n = 500) of adult female M. trinitatis was collected from different localities (Fig 1, S1 Table) in Trinidad's Northern Range during June to August 2015 and 2016. Females, recognised by having yellow throats, were caught either by hand or with the aid of small handnets and placed for transport into small polyethylene bags containing a little damp forest leaf litter. They were kept overnight in a holding tank furnished with damp leaf litter, at the University of the West Indies, St. Augustine. The frogs were photographed and measured (snoutvent length [SVL] to 0.1 mm using dial callipers) on the day after collection, and then returned to their collection sites within~20m of capture location. Despite an earlier report of the presence of the potentially deadly fungal pathogen Batrachochytrium dendrobatidis (Bd) in the M. trinitatis population, Greener et al. [20] found the infection to be absent. Capturing and returning the frogs should therefore not risk spreading infection.
To photograph the throat region, each frog was carefully held upside down and the ventral side photographed on a white background beside an X-Rite Colorchecker passport colour rendition chart (X-Rite Inc., Michigan, USA). All photographs were taken using a Sony ILCE-6000 with an attached Sony E-Mount SEL 55-200mm lens and lens hood (2015) or SEL 18-55mm lens (2016). The camera was set with F8.0 aperture, and shutter speed adjusted to give an exposure of 0.0. Photographs were taken in a white light box studio. Constant light was provided by three LED lights positioned on the right, left and above, along with an overhead strip light. All batches (frogs caught that day) of photographs were taken within one hour of each other. As in Stevens et al. [21], RAW format was chosen, as opposed to JPEG-as used by Bergman and Beehner [22]-as this prevents loss of information due to compression, allows for later adjustment, and due to current storage capabilities, file size was not an issue. The photograph from the first frog in each batch of frogs was used to create a colour profile using the software accompanying the X-Rite Colorchecker. A profile allows for colour correction given the same conditions. This profile was then applied to all photographs within the batch in Adobe Photoshop CC 2015, and white balance corrected for each photograph. The area of interest (yellow throat region) was selected, and average blurred before red (R) and green (G) values were recorded. These values were then used to generate R:G ratios. The presence of yellow colouration under the collar was also qualitatively assessed as Yes or No. We modelled the relationship between R:G ratio and SVL, site and presence of colouration under the collar using generalized linear mixed models, including year as a random effect, with statistical significance evaluated by Wald chi-squared testc. All analysis was conducted in RStudio 1.2.5033 environment R3.6.2 [23].

Colour development in juvenile Mannophryne trinitatis
A sample of about 60 M. trinitatis tadpoles was collected in early July 2016 by hand netting from a pool in a stream beside the Arima-Blanchisseuse Road in Trinidad's Northern Range mountains. This was a sample of a larger number of tadpoles within the pool. The large number of tadpoles implies that many clutches had been deposited and therefore our sample had offspring from multiple parents. The tadpoles were transferred to the University of Glasgow, Scotland in two-litre polyethylene containers with the tadpoles resting on damp cotton cloth. Downie and Smith [24] showed that these tadpoles survive well under such conditions. In Glasgow, the tadpoles were grown at an initial density of 30 individuals per tank in plastic aquaria 32x18x18 cm in dechlorinated tap water at a depth of 10 cm and with a constant air supply delivered through a submerged air-stone at one end of the tank. The room where the aquaria were located had a 12:12h light/dark cycle and an air temperature of 23-24˚C. The tadpoles were fed daily with aquarium fish food flakes (New Era brand), and the water was changed weekly to avoid the build-up of waste. Each day, the aquaria were checked for tadpoles showing forelimb emergence, the sign of metamorphosis beginning. The first metamorph was found on 8 th July and the last on 16 th September. Each metamorph was caught by hand-net and transferred to an individual translucent 22x15x8 cm polyethylene container with an opaque lid. Each container was provided with a 'shelf' of washed gravel at one end, about 18 cm wide and 1-2 cm deep, and dechlorinated tap water to a depth of about 0.5 cm. Downie et al. [25] have shown that M. trinitatis take 6-7 days to complete metamorphosis and that they hide in gravel during that time.
After completion of metamorphosis, each froglet was provided twice a week with live Drosophila melanogaster as food, supplied in 5 cm long tubes, which the frogs could enter to forage. Froglets were measured (SVL, using callipers accurate to 0.1 mm) and their throat patterns noted and/or photographed, starting about two weeks after forelimb emergence and continuing at approximately every four weeks for 16 weeks. For this purpose, froglets were  Table. Adapted from Greener et al. [20].
https://doi.org/10.1371/journal.pone.0223080.g001 captured by hand, transferred to a transparent 9 cm diameter petri dish, and measured from below. Once the 16-week growth period was complete, each frog was transferred to a communal tank for rearing to adulthood.

Escape responses in Mannophryne trinitatis in relation to female throat colour
Adult female (n = 81) M. trinitatis were captured with the aid of small hand-nets from five sites across Trinidad's Northern Range (Fig 1, S1 Table), during mornings over six weeks, June to August 2018. Sites were visited in rotation, with 5-6 frogs captured at each visit. Frogs were transported to our laboratory at the William Beebe Tropical Research Station (Simla) and housed individually in plastic aquaria furnished with a thick layer of damp forest leaf litter. Frogs were measured (SVL to 0.1 mm with dial callipers; weight to 0.01g using an electronic balance) and any with SVL <16mm were classed as juveniles and excluded from further study. The day after collection, frogs were photographed in the morning around 09.00h and repeated at night around 21.00h: three pictures of the throat were taken using a Canon PowerShot s110 digital camera, in a white light box studio with two LED lamps providing constant light from either side of the light box. The camera was set with F3.2 aperture, and shutter speed adjusted to give an exposure of 0.0, allowing for comparison with throat colour variation measured earlier. Dorsal sides of each frog were also photographed to identify frogs, so as to ensure that no frog was used more than once, following subsequent collections (dorsal patterns are individually variable). Throat colour, as R:G ratio, was measured as described earlier.
Escape responses were assessed two mornings after collection (09.00-12.00h) in a specially constructed outdoor arena set in a shaded area: this had wooden sides about 0.8m high (to prevent frogs escaping) and enclosed an area of short grass 1.5x1.5m. Each frog was put into a 9cm diameter plastic petri dish and placed at the centre of the arena; it was left there for 30 seconds with the lid off to acclimatise. If the frog jumped before 30 seconds were over, it was recaptured and left a further 30 seconds. The frog was then stimulated by a light tap to the rear using a metre stick. Each frog's responses, three times for each frog, were recorded for 20 seconds using a GoPro HERO6 video camera set above the arena. After each response, each frog was given at least 30 minutes to recover before being stimulated to jump again. Air temperature ( 0 C) and relative humidity (%) within the arena were recorded using an ETI pocket thermo-hygrometer at the same times as each set of responses.
Image J (v1.52a) was used to measure the distance of each jump. From the recordings, we calculated maximum and minimum distances of each jump made, mean distance per jump and the total distance travelled in each trial (i.e. different measures of escape performance). Using RStudio 1.2.5033 environment R3.6.2 [23], general linear models were used to test relationships between escape performance and size (SVL), site, humidity, temperature and throat colour (as R:G ratio). The Benjamani-Hochberg procedure was used to correct for multiple testing. We expected escape performance (for example total distance jumped, or initial jump length) to be positively related both to size and throat colour, but to be independent of collection site, humidity and temperature.

New field observations on Mannophryne trinitatis
While assessing the population status of M. trinitatis for Greener et al. [20], we made occasional interesting observations relevant to colour and behaviour, both during the day and at dusk/night, which we present here.

Sexual dichromatism in Mannophryne olmonae
Field observations on M. olmonae were made at several small un-named streams in northeast Tobago in June to August 2014 and 2015. In 2015, female M. olmonae were captured using hand-nets and transferred in individual containers to accommodation in Charlotteville. Here, they were photographed alongside an X-rite Colour Checker rendition chart, under identical lighting conditions, using a Canon EOS Rebel T3i. Throat colour, as R:G ratio, was measured as described earlier and frogs were returned to their collection sites. As in the case of M. trinitatis, Thomson et al. [26] have shown Bd to be absent from the M. olmonae population.

Comparison of sexual dichromatism across the genus Mannophryne
We performed a literature review on male and female colours in life in all Mannophryne species so far described. The black colour in calling males can only be seen when observing males calling in the field; the female yellow throat colour fades in preservative. In some cases, colours in life are not presented in the original species descriptions, but we were often able to find later accounts of colours in life.

Ethics statement
A field research permit (Special Game License) was provided by Government's Wildlife Section for the years of this study (2015, 2016 and 2018) for the months of June-August, no numbers are allocated. This permit allowed for the capture and use of animals in these studies for frogs of all species up to 500 adult individuals per year. Export of tadpoles was granted by Special Export License 001192 (29/6/16). The Special Export License we received from the Trinidad Wildlife Division allowed for capture of animals and the collection of 100 animals, with no distinguishing between adults or tadpoles, it also did not require them to be returned to the wild. No ethical approval was required as studies were non-invasive and very lows stress levels and considered under the threshold of the Animals (Scientific Procedures) Act. All adult frogs (apart from a small number that died in captivity) were returned within a couple of days to their original location. Captured tadpoles were exported live to the UK, where they were reared to metamorphosis until adult. After this, they were donated to an experienced tropical frog breeder. No aspects of our study required anaesthesia, euthanasia or any kind of animal sacrifice.

Throat colour variability in female Mannophryne trinitatis
Larger females were found to be more likely to have a higher R:G ratio on their throat patch (p <0.001) (Fig 2). Two of our sites, Edith Falls and Las Cuevas, were found to have a significant positive effect on R:G ratio (p <0.001). Females with high R:G ratios were found to be more likely to have colour posterior to the collar (p 0.037). However, SVL and its interaction with R: G ratio were not significant predictors of this. We qualitatively observed that the colouration of the throat area differed between individuals not only in R:G ratio but also in the relative size of the throat patch, the intensity and width of the dark collar, and the extent to which the yellow colour extended posterior to the collar (Fig 3).

Colour development in juvenile Mannophryne trinitatis
Forty-four of the tadpoles reached metamorphosis. Of these, 25 developed long enough for their sex to be distinguished by throat colour differences, 20 as females and five as males. This sex ratio is significantly biased towards females (chi squared = 15.8; p< 0.001). Table 1 shows size and colour development data for all froglets recorded beyond 90 days post metamorphosis (omitting two females that escaped at around 80 days). Dark collars developed in both males  and females. Although these were variable in width and shade, we saw no consistent difference between males and females. Throat colour started and remained grey in five individuals throughout the observation period and these were classified as males (we did not check sex by examining gonads). In individuals developing as females, throat colour started as grey, became pale yellow at around seven weeks post metamorphosis, and either remained pale or became more brightly yellow around nine weeks (Fig 4). Since we did not assess throat colour weekly (too frequent disturbance could be stressful, and risked escapes), we cannot tell precisely when the yellow colour first became apparent. However, throats were pale grey in all individuals at the first set of observations (2-3 weeks after metamorphosis began) and remained grey at 5-6 weeks in individuals that developed as females. These results indicate that throat colour is ontogenetically sexually dichromatic, developing in the juvenile phase, well before female maturation (mature females are around 20mm SVL; the yellow throat was distinguishable at around 15.5 mm).

Escape responses in Mannophryne trinitatis in relation to female throat colour
Responses to stimulation were quite varied. Some frogs made only a few jumps before stopping on the grass; others jumped to the edge of the arena and climbed some way up the wall (Mannophryne have adhesive toe pads); the direction of jumping was also variable but most tended to maintain more or less the same direction once they set off.
Comparison of morning and night photographs of frog throats indicated that the colour was stable, with no diurnal variation. Table 2 shows female frog sizes, colour variability and escape performance. There were no significant differences between collection sites and escape performance (p>0.05 for all measures). Air temperature during the trials had a range of only 1.8˚C (26.9-28.7˚C), but humidity varied more widely (53-86%). There were no significant relationships between temperature or humidity and escape performance (all p>0.05). Also, there was no significant relationship between collection site and throat colour (R:G ratio: p>0.05). However, there were strong and significant positive relationships between R:G ratio and frog size (weight: F = 18.42, p<0.001; SVL: F = 19.07, p< 0.001). Weight and SVL were also strongly correlated with one another (not shown).
We found a positive significant relationship between R:G ratio (throat colour) and the maximum distance travelled in a single jump (F = 11.55, p<0.01). One observation of throat colour (R:G = 1.61) had a noticeably larger R:G ratio than the rest (Fig 5), though this difference was not alarming as the relationship between maximum distance travelled and throat patch colour was still significant whether this datapoint was included or not for statistical analysis (p<0.01 in both cases). There were no significant relationships between R:G ratio and the other measures of escape performance: minimum distance in a single jump; mean distance per jump; initial jump distance; total distance travelled (p>0.05 in all cases).
Interestingly, we found no significant relationships between SVL and any of the escape performance measures (p>0.05 in all cases). S2 Table summarises the statistical results on escape performance.

New field observations on Mannophryne trinitatis
We observed M. trinitatis sites at dusk and during the hour after sunset. Some male Mannophryne remained active, calling, feeding and transporting tadpoles. Although numbers were fewer and calling less frequent than during the day, frogs were out in the open and not hard to find. During a day-time survey, we noticed a male in normal non-calling colouration calling occasionally i.e. not at the usual high frequency. As we watched, this frog began to change to black and over the same time, its calling frequency increased, the full change taking about 20 minutes.

Sexual dichromatism in Mannophryne olmonae
Our field observations confirm that M. olmonae males are black when calling. On a late afternoon in June 2014 we first observed a group of three males on a rock by a stream near the Charlotteville-Bloody Bay road (Tobago). All three were calling and all were black. One soon hopped away. The other two called facing one another until one other hopped away, leaving the 'victor' of the encounter (Fig 6A and 6B). We made many similar observations over the next four weeks, but did not actually observe the colour transformation from brownish to black, although we did see a calling male that was mainly brown dorsally and possibly at the start or end of the transition (Fig 6C and 6D). Of 47 calling males observed, 68% were on rocks, 17% on leaves and only 15% in crevices. In a further visit in 2015, we captured 12 adult females and photographed their throats for colour analysis. All had yellow throat patches with a narrow brownish collar and an R:G ratio ranging from 1.03 to 1.104. Table 3 shows the results of our literature search for evidence of both dynamic sexual dichromatism in males and sex differences in throat colour, assumed to be ontogenetic. We found that in 8 (maybe 9) out of 20 species, dynamic colour change has been documented. In a few cases, colour change in males has not been observed, but this is generally in species where calling has not been seen. We also found that in all cases, females displayed at least some yellow colouration in the throat region. The table includes any information found on the speed of colour change in males, but this has only been reported in a few species so far. All data used within paper can be found in S3-S5 Tables.

Discussion
In this paper, focused on the Trinidad stream frog Mannophryne trinitatis, with additional observations on the Tobago stream frog M. olmonae, and a review of the colour descriptions of the other Mannophryne species so far identified, we show that a) dynamic sexual dichromatism  is widespread in the genus, and b) that ontogenetic sexual dichromatism, principally involving the development of a bright yellow throat patch occurs throughout the genus in females. This highlights another example of frogs exhibiting both dichromatisms, this time across a whole genus. We show that the yellow throat patch is highly variable in M. trinitatis and M. olmonae, and that colour development begins soon after metamorphosis in M. trinitatis. We also provide a test that supports our hypothesis that throat colour provides a signal of female quality. Reviews of the occurrence of conspicuous colouration in frogs [2,3] emphasise two general cases. First, aposematic (warning) signals to other species indicate that these frogs are well protected by toxins. A complication may arise where harmless species evolve to mimic the toxic species, gaining protection without incurring the costs of producing toxins. Second, to protect the frog from drawing the unwelcome attention of predators, the conspicuously coloured element is either temporary or concealed, except from the intended receiver.

Dynamic sexual dichromatism in male Mannophryne
Bell et al. [6] found a relationship between dynamic sexual dichromatism and explosive breeding aggregations in hylids, bufonids and some other groups, and suggested that colour change may assist mate recognition in such situations. However, they also noted that dynamic dichromatism occurs in the absence of breeding aggregations in some species, and related it to intraspecific competition, such as territory defence, in these cases. Colour change in frogs is generally found to be slow (hours to days) and mediated by hormones. However, Kindermann et al. [39] found that a dorsal change from brown to yellow in amplexing male Litoria wilcoxii took around 5 minutes, and that it could be induced in non-amplexing males by epinephrine injection, implying a neuroendocrine mechanism. Wells [17] reported that male M. trinitatis begin calling and then change to black within 1-10 minutes; we also noticed that males start to call before changing to black and that the change occurred over minutes. In other Mannophryne species, La Marca [28] reported M. cordilleriana changing to black a few seconds after starting to call, and Rojas-Runjaic et al. [15] made a similar observation on M. molinai. In other species where a change has been seen (8 out of 20 species, plus probably M. riveroi: Table 3), the rate of change has not been noted. In a few accounts, the return from black to brown has been reported to be similarly fast. Although no research has been reported on the mechanism of colour change in male Mannophryne, the speed implies a neuroendocrine process, and a study to test this is needed.
The function of the male change to black is unclear. Mannophryne trinitatis is not an explosive breeder: frogs are distributed along stream sides, with females holding long-term territories. Breeding can occur throughout the long wet season, presumably dependent on females having ripe egg clutches [16,17]. Mannophryne, although closely related to dendrobatids, lack toxic protection and are generally cryptic in colouration and behaviour. Their usual habitat is the margins of rocky streams in tropical forest, and they are mainly active during the day. Light levels are low, and the habitat provides abundant shaded crevices where frogs with mottled dorsal colouration of browns, greys and blacks are well concealed. Wells [17] found that males with normal brownish colouration never attacked other males, nor were they attacked by calling males. However, aggressive encounters between black calling males were common. Calling males did not appear to be particularly territorial in their behaviour, often changing Table 3. Reported occurrences of ontogenetic and dynamic dichromatism in living specimens of Mannophryne species.

Males Females
M. caquetio (1,8) Dark grey abdomen and throat; no report of change to black when calling Black, well defined collar; no mention of yellow throat in (1) calling site, usually a conspicuous position such as a rock or log (however, we have often seen males calling from shaded crevices). Since both calling and black colouration make the males conspicuous, it is unclear why both signals are needed, especially when they likely increase the risk of predation. Mimicry is unlikely to be at work here. Although some toxic frogs are conspicuously black (for example, dendrobatids of the genus Ameerega), their ranges do not appear to overlap with those of Mannophryne [40]. The key to the dynamic black signal in males may lie in the unusual territorial behaviour of female Mannophryne (see later). In order to attract a female, the males may need to demonstrate their own quality by having visibly successful encounters with other males, or by being conspicuous (colour and sound) for an extended period.
Bell et al. [6] checked 13 species of Mannophryne for the occurrence of colour change in the males and found that only M.trinitatis fitted their criteria, which required photographic evidence. However, we read original species descriptions and some later reports. Since the occurrence of the black colour is transient and not found in preserved specimens, definitive sightings require field observations of calling males. These are often lacking in reports and colour change is often only briefly referred to, since it cannot be used as a species identification criterion in preserved specimens. Of the 12 species reviewed by Bell et al. [6] as lacking colour change, we found colour change descriptions in three cases: If male colour change occurs in some Mannophryne but not others, a possible explanation is phylogeny. Manzanilla et al. [41] analysed mitochondrial DNA sequences from 13 of the 15 Mannophryne then known. They identified three clades of five, one and three species respectively. More recently, Grant et al. [13] analysed 14 species and essentially confirmed Manzanilla et al.'s clades, though they noted some anomalies in the species so far identified. Male colour change while calling has been reported in species belonging to both of the larger clades. Given that all Mannophryne species appear to live in similar habitats with similar behavioural ecology, our hypothesis is that colour change in calling males is likely to occur throughout the genus, and that those species where it has not been reported have not yet been adequately observed in the field.

Ontogenetic sexual dichromatism in Mannophryne
Amongst the anurans, hyperoliids (reed frogs) are the group previously shown to commonly undergo ontogenetic sexual dichromatism, with females developing a bright colour while the males remain dull (found in 35 of 215 species [5,42]). As we have shown, ontogenetic dichromatism in female throat colour appears to occur throughout Mannophryne. Females have yellow throat patches of varying size and shade, backed by a dark collar of variable width and tone; the yellow patches sometimes extend beyond the collar on to the abdomen. Males may have a collar, but their throats are invariably grey to black, never yellow. Some species accounts refer to the colour patterns of juveniles, but our study presents the first account of ontogenetic changes in throat colour from the end of metamorphosis to near maturity, and it is clear that the adult sexual dichromatism is ontogenetic. Wells [17] found that female M. trinitatis aggressively defend their territories and that the signal used to denote a territory holder is pulsation of the throat with the head held high so that the yellow throat patch, pulsation and dark collar are clearly visible to any approaching conspecific. Durant and Dole [43] described similar behaviour in female Colostethus (now Mannophryne) collaris. Exposure of a coloured signal that is normally concealed has been infrequently reported in frogs. One other example is the brightly coloured foot webbing in the foot-flagging frog Staurois parvus, where males extend and rotate their legs, displaying the colours, during social interactions [44].
To our knowledge, no one has previously suggested that variation in the yellow colour acts as a signal of female quality. However, across the animal kingdom, bright yellow patches are often used in this way, related to the cost of synthesising the carotenoids on which yellow colours are often based, and to the role of carotenoids in immune system function [18]. In anurans, a relevant example is the orange-coloured (carotenoid-based) vocal sac of chorusing tree frogs, where females prefer males with colourful compared to pale vocal sacs [19,45].
Our data show considerable individual differences in female throat colour in both M. trinitatis and M. olmonae. To test whether throat colour differences act as signals of female quality, it would be best to recover the winners and losers after territorial encounters and measure their throat colours. However, in practice, the combination of the need to observe from a distance (to prevent disturbance) and the nature of the terrain (rocks with abundant deep crevices used by frogs for concealment) made this unfeasible. We used a proxy for quality instead. Royan et al. [46] reported experiments where the escape response trajectories of captured M. trinitatis were measured using an outdoor arena. They found that angle of escape was variable, indicating a degree of unpredictability, which could help individuals to escape potential predators. We reasoned that differences in escape response might provide as good a measure of quality as the results of territorial encounters. Our results showed a positive relationship between R:G ratio and two measures of escape performance, providing evidence that throat patch colour is indeed a signal of quality. However, the signal emitted by females is more than simply the colour of the throat. Our measurements over a large population of female M. trinitatis and a small sample of M. olmonae show variation in throat colour, patch size, the width and colour of the collar, and the extent of the yellow patch posterior to the collar so the quality of the signal may include all these components.

Conclusion and a hypothesis
The adaptive significance of female sexual dichromatism in Mannophryne seems clear: it is associated with territorial defence. Territorial behaviour in dendrobatids is common, but mainly involves males [2]. Long-term defence of a territory by female frogs, as occurs in Mannophryne (extrapolating from the species where it has been demonstrated, such as M. trinitatis and M. collaris), is very rare. The obvious suggestion is that females are defending resources, most likely food. Frogs captured in late afternoon showed that females had significantly fuller stomachs (small insects, arachnids and occasional snails) than males [47]. It is not known whether food resources are patchily distributed or simply related to area: a study on territory size in relation to throat colour would be helpful. If we are correct in concluding that throat colour differences in females provide a measure of quality, then it is likely that the throat pulsing observed by Wells [17] during female to female intraspecific confrontation provides a signal, potentially avoiding a physical confrontation. Additionally, it would be beneficial to the males to choose the best possible quality of mate, demonstrated by their visually striking territorial defence. Females also need to select the best possible mate: males guard the eggs throughout incubation and then transport the hatchlings to a suitable body of water. Downie et al. [48] found that transporting males may take several days to locate a suitable pool, ideally one lacking predators. This post-hatching transportation phase probably makes it unfeasible for male Mannophryne to guard multiple egg clutches, as occurs in some other clutch-guarders, such as glass frogs [49]. Males therefore have to be sure that the quality of the eggs is high enough to justify the considerable investment in time involved in incubation and transportation. Males can demonstrate their quality by calling for long periods, and by turning conspicuously black, both hazardous and possibly energetically expensive activities. The mating dances the males perform also contribute [17]. Our hypothesis therefore is that the occurrence of sexual dichromatism in both sexes derives from the resource-based territoriality of the females, using throat colouration as a signal in female to female conspecific confrontations, and strong selection for quality in both sexes. Males avoid the predation costs of conspicuousness by their colour signal being temporary and quickly turned off and on; for females, the colour signal is concealed, except when used against conspecific receivers. Follow up studies on the reproductive success, or even just territory defensive success, of females with different throat colouration would be very interesting. However, we are aware of how difficult this would be. Further studies on the development of the female colouration across multiple species and over a longer time period may lead to some interesting results. Clarification of the presence or absence of the dynamic colour change for males of species from the rest of the genus would also be useful.
Supporting information S1  Table. Data from escape responses in Mannophryne trinitatis in relation to female throat colour. All data collected and used for analysis in escape responses in Mannophryne trinitatis in relation to female throat colour study. (XLSX)

Acknowledgments
The fieldwork upon which the results reported here are based was carried out on University of Glasgow expeditions to Trinidad, and to Tobago over the years 2015-18, with many students helping make observations and catch frogs. Fieldwork permits were provided in Trinidad by the Government's Wildlife Section, Forestry Division, in St. Joseph. In addition, Special Export License 001192 (29/6/16) gave permission for export of the tadpoles used in the rearing experiment. Thanks to Anna Bandoo and Romano McFarlane for facilitating these permits. In Tobago, Angela Ramsay of the Tobago House of Assembly's Department of Natural Resources and the Environment provided fieldwork permits. Multiple agencies supported the expeditions financially, notably the University of Glasgow's Chancellor's Fund, Glasgow Natural History Society, the Gilchrist Educational Trust, the Thriplow Trust, and Dennis Curry's Charitable Trust. A Percy Sladen Memorial Fund grant funded the camera used in the escape response experiments. The University of the West Indies and the Asa Wright Nature Centre provided laboratory accommodation in Trinidad, and Pat Turpin in Tobago. We thank Dr James D.