Study Site

We conducted our study in a tropical lowland rainforest next to the field camp ‘Saut Pararé’ (4°02′N, 52°41′W; datum = WGS84) of the French National Center for Scientific Research (CNRS) Nouragues Ecological Research Station within the Nature Reserve “Les Nouragues” in French Guiana (Bongers et al. 2001). The study site was an island of ∼5 ha in area in the Arataye River near the field camp (Ringler et al. 2016b). The island is part of a long-term project and there was no Al. femoralis population on the island before 2012, when tadpoles were relocated to establish a closed population for research (Ringler et al. 2014). Courtship observations took place during the rainy season from 7 February to 2 April 2016.

In addition to the courtship observations, we conducted a concurrent mark–recapture monitoring of the population during daily surveys from 0900 to 1900 h to estimate the males' territory sizes. We attempted to sample all males and females and recorded the locations of each frog on a digital map (Ringler et al. 2016b) on a tablet PC (WinTab 8, Odys, Willich, Germany) using the mobile GIS software ArcPad v10.2 (ESRI, Redlands, CA, USA). We sporadically played a conspecific call from a portable loudspeaker (SRS-M30, Sony Corp., Tokyo, Japan; frequency range = 250–20,000 Hz) simulating a territorial intrusion (i.e., conspecific males calling inside the territory) for short time intervals in order to catch male frogs (phonotactic approach behavior; cf. Ursprung et al. 2009), but not during courtship observations. We used the distinct ventral patterns for individual recognition of all frogs (cf. Ringler et al. 2014) using the pattern-matching software Wild-ID v0.4.5 (Bolger et al. 2012).

Behavioral Observations and Mapping

We located courting pairs by actively searching and listening for courtship calls while walking slowly. To obtain a temporally and spatially unbiased sample, we spent approximately equal time per area while surveying the entire island. After finding a courting pair (i.e., a male producing courtship calls and the female following the male), we carefully observed courtship behavior from a distance of 2–3 m while trying to minimize observer disturbances during courtship activities. We chose the observer distance based on preliminary attempts to observe courtships as a good tradeoff between viewing distance in the rainforest understory and minimal disturbance. We distinguished the male from the female based on the inflated vocal sac during calling.

Observations were paused after 1900 h (nighttime) and resumed the following day from 0630 h (daytime) until oviposition took place. As we interrupted the observations only during the frogs' inactive period from sunset to sunrise, we assumed that we observed the entire sequence of the courtship behavior. At the end of each observation, we caught the male and the female with transparent plastic bags and photographed them for individual identification. We scored courtships ending with a clutch as successful whereas, if one of the frogs stopped courting for more than 60 min and the female–male distance was more than 4 m, we considered the courtship as unsuccessful and we stopped the observation.

For each courtship, we recorded the following parameters: start of observation, end of courtship, time spent during the courtship, time and duration of amplexus, and time when the male and the female left the oviposition site. For spatial analysis, we recorded every position (location and time) where the male stopped for longer than 1–2 min. We did not record movements less than 10 cm. To map the male's courtship trajectory, we drew a hand sketch of the area (branches, logs, leaves, etc.) and of the frogs' movements during the observation. Immediately after the courtship, we marked the vertices of the courtship march, based on the sketches with plastic stakes, and measured the distances between consecutive locations with a laser rangefinder (DLE 50, Bosch, Leinfelden-Echterdingen, Germany). For each vertex, we also measured the distance to two reference points (which were either a tree or a reference marker, both shown in the background map), enabling us to then digitize the exact courtship trajectories in AutoCAD 2017 (Autodesk Inc., San Rafael, CA, USA). For the analysis of the spatial data, we then overlaid the courtship trajectories with the background map of the study site (Ringler et al. 2016b) in ArcGIS v10.4 (ESRI).

Courtship in Relation to Male Territory Size

We calculated central and estimated peripheral territory boundaries of the males based on their locations recorded during the baseline population monitoring. Central territories represented the calling area of Al. femoralis males whereas the estimated peripheral territory represented the area which we expect males to defend from conspecific males (cf. Ringler et al. 2011). We used only those males for which we recorded a minimum of five catching points for this analysis and used only those points where the males were observed calling or approaching the simulated call emitted from the loudspeaker.

To estimate the central territory and the peripheral territory, we used two different methods (cf. Ringler et al. 2012; Fig. 2); (1) Minimum convex polygons (MCP) of encounter locations to calculate the central territories, and (2) Voronoi tessellation to estimate peripheral territories. The minimum convex polygon method connects the outer points of a set of points to contain only convex angles (≤180° between points). We assumed territory shifts when locations of a male were both further than 10 m away from any other catching point for the same male and recorded more than 10 d apart. In this case, we constructed separate MCPs for each consecutive territory and considered the territory occupied only during the respective courtship observation for our analysis. We excluded five courtship observations from this analysis (CS01, CS03, CS04, CS09, CS15) because we had fewer than five catching points.

Fig. 2

Overview map of the central and estimated peripheral territories of two Allobates femoralis males and the estimated peripheral territories of neighboring males (gray) in French Guyana. Central territories represent the calling area of Al. femoralis males whereas the estimated peripheral territory represents the area that males were expected to defend from conspecific males (cf. Ringler et al. 2011). The yellow polygon represents the central territory, surrounded by the estimated peripheral territory of Male i16-m-041 (purple) and Male i16-m-042 (blue).

To estimate peripheral territories, we used Voronoi tessellation. This method partitions an area into regions based on equidistant midlines between pairs of points (Voronoi 1908). To assure temporal relevance, we only used the previous five (territorial) catching points of each male before the respective courtship observation took place (cf. Tumulty et al. 2018). We excluded two courtship observations (CS01, CS09) because the courtship march was not within the Voronoi polygon, which resulted from too few catching points of the respective males and/or a territory shift immediately before the observed courtship. In two cases, the males had a multi-part Voronoi polygon (CS13, CS15)—for these, we kept only the part that contained/ overlapped with the courtship march. Because the shore of the island is steep and not occupied by Al. femoralis, we truncated the outer territories midway between a male's outermost locations and the island shore by adding the vertices of the island outline to the points used in the Voronoi tessellation.

To estimate the area that males potentially show to females during courtship, we used two different methods: (1) We buffered the courtship trajectory on each side by 15 cm, termed “courtship buffer,” by using the “Buffer” function in ArcGIS (see Fig. 3). We chose 15 cm along each side because females were not following the male in exactly the same path, and the buffer approximated the area directly explored by the female during courtship. (2) We calculated MCPs from the vertices of the courtship trajectory, termed “courtship polygon” (see Fig. 3). This calculation represented a wider area potentially perceived by females during the courtship. We correlated both estimations of the courtship area with male territory size.

Fig. 3

A map of an observed courtship march (CM) in French Guyana of the Allobates femoralis Male i16-m-042 on 14 March 2016 and the calculated spatial parameters. The red circle indicates the start and the yellow circle indicates the end of the CM, which is also the end of the observation period. The yellow star depicts previously deposited clutches. The blue circles are the recorded CM points, ranging from light blue (start of the observation) to dark blue (end of CM); the black line presents the CM trajectory; the red triangles represent the territorial observation points of the male subject. The central territory of the male is shown in yellow, the peripheral territory in gray. To quantify the area covered during the courtship, we calculated two CM areas: (1) a buffer of 15 cm on both sides of the CM trajectory (white); and (2) minimum convex polygon of the outer CM points (green).

Courtship in Relation to Clutches

To examine whether or not courtship is related to the locations of previously deposited clutches inside a male's territory, we searched for clutches along the courtship trajectory by carefully flipping leaves on the ground for approximately 60 min (depending on the length of the courtship march) after the courtship. Given that females do not follow the male in the exact same path, we considered all encountered clutches within the courtship buffer. As a control, we searched the male's territory outside the courtship buffer for the same duration. We also recorded the locations of all discovered clutches in the digital background map of the island with ArcPad. To evaluate if courtship was related to previously deposited clutches in each male's territory (e.g., if a male shows his previous clutches to females), we compared the number of clutches found within the courtship buffer to the number found outside.

Within Individual Courtship Variation

We observed 7 of the 15 different males repeatedly during courtship (2 males 4 times, 3 males 3 times, and 2 males 2 times), allowing us to analyze the intraindividual variability of courtship. To investigate if the male frog was guiding (different) females along a similar route in subsequent courtships, we used the ArcGIS function “Intersect” to assess the area/percentage of overlap between the courtship march buffers of that male.

Data Analysis

To test for normality of the variables, we used the Shapiro–Wilk normality test. Variables that significantly deviated from a normal distribution (central territory size, estimated peripheral territory size, duration in the morning, courtship buffer size, and courtship polygon size) were log transformed. To test if the length (distance in meters), the total duration, the duration in the morning, the courtship buffer size, or the courtship polygon size are indicators of male territory size (central territory and/or estimated peripheral territory), we used Pearson correlations. To analyze the variation within and between the male subjects, we calculated coefficients of variation (CV) and compared them with analyses of variance (ANOVA) followed by Tukey post hoc tests. Statistical analyses were conducted in R (v3.4.2; R Core Team 2019).

The sample sizes varied depending on the specific analysis because not all parameters were available for each male frog. For the general description of the courtship, we considered only each male's first courtship (n = 15). We calculated Voronoi polygons and MCPs for 10 out of the 15 males, and these males were considered in the correlation of courtship extent (temporal and spatial) to male territory size. The remaining 5 males had too few catching points to calculate reliable territory sizes. For the temporal analysis, we excluded 1 of the 10 males because we could not record some data on account of heavy rainfall during the observation. To relate the courtship to the already present clutches in the territory, we considered 13 out of these 15 males. We excluded 2 males because we had too few points for a reliable calculation of their territory sizes. To analyze the variability within and between the males, we considered 6 of the 7 males, which we observed repeatedly. One male shifted its territory between courtship observations and therefore was excluded from the dataset.

Courtship Success

In total, we observed 29 courtships by 15 different males. We observed 7 of these 15 males multiple times (2–4 times) in different courtships. Twenty-seven out of 29 courtships ended with a successful oviposition whereas 2 courtships were unsuccessful because the female stopped following the male and left the male's territory before oviposition occurred. In one unsuccessful case, the pair was courting in the afternoon, but while the female remained overnight in the male's territory, she did not resume courtship activities on the next morning. After 86 min of courtship calling in the morning, the male stopped and the female left the male. In the second case, the pair was in courtship for 205 min in the afternoon; the female stayed overnight and resumed courting the next morning for another 157 min. Then, shortly after the neighboring male started calling, the female left the first male and directly approached the calling neighbor. The rejected male switched to advertisement calling after the female left. After 24 min of courtship with the neighboring male, they deposited a clutch inside a leaf, 30 cm next to another recent clutch.

General Description

As soon as a female came close to a male (∼30 cm), the male typically turned toward the female and alternatingly emitted advertisement and courtship calls. The courtship calls consist of a broadband burst of pulses produced at a high rate that sounded like a quiet buzz to a human ear (Fig. 1). Each courtship call was approximately 0.5–1 s in duration and 2500–2700 Hz in dominant frequency and sounded different from a tonal advertisement call (see Gasser et al. 2009). The courtship call was softer than the advertisement call and appeared to be used only for short-distance communication between male and female.

As soon as the female jumped toward the male, the male moved slightly away (approximately 10–20 cm) from the female. Then, the male stopped and emitted courtship calls until the female followed. On those occasions when the female did not follow the male within a few minutes, or the female was out of sight, the male switched back to the more prominent advertisement call. This sequence was continued until sunset (between 1845 to 1900 h), when both frogs hid under leaves and interrupted their courtship activities. In 28 out of 29 observations, the female stayed overnight inside the male's territory.

In the morning (between 0630 to 0645 h), the pair resumed their courtship. The male frog always entered the oviposition site first (in the cases where the sequence was observed; n = 9). While sitting inside a folded or covered leaf, the male emitted courtship calls until the female joined him. Sometimes the females jumped outside shortly after entering the site and then the pair continued with courtship. In 5 out of 29 cases, the frogs later returned and mated inside a previously visited leaf. In all other cases (n = 24), the pair mated in a leaf which they had not previously visited.

Mating behavior.—We observed cephalic amplexus in 15 out of 27 matings, mostly in the early morning (X̄ ± 1 SD = 0757 h ± 68 min; Fig. 4b). In the other 14 cases the pair was hidden, but we could observe both partners crawl under the same leaf and we found a fresh clutch after they had left the nest. At the oviposition site, the pairs initially sat head-to-head while the male emitted courtship calls. Then, the female turned to face away from the male frog. The male climbed on top of the female and grasped her head with his front legs. This cephalic amplexus lasted an average of 2 min and males left the site soon after amplexus (n = 9). During and after amplexus, the female laid her eggs on the leaf and then spent an average of 77.4 min (n = 8) at the oviposition site before leaving the nest.

Fig. 4

Mating sequence of Allobates femoralis in French Guiana: (a) male inside the oviposition site, female in front of it; (b) male and female in cephalic amplexus; (c) female inside the oviposition site on the freshly laid eggs.

In 8 out of 10 males, the central territory was inside the estimated peripheral territory. In the other cases, the central territory was partly outside the peripheral territory. This apparent paradox resulted from the distinct approach used to estimate central and peripheral territories. In the spatial and temporal comparisons, we subtracted the central territory from the peripheral territory and treated them separately in the following results.

Temporal pattern.—Courtship observations started in the afternoon (median ± interquartile range [iqr] = 1715 h ± 53 min) and ended on the following morning (median ± iqr = 0811 h ± 58 min), leading to a total average courtship duration of 15 h (range = 806–1043 min, n = 15). When subtracting the 11 nocturnal hours (1900 to 0600 h) from the total duration, courtships lasted an average of 222 min (range = 116–353 min, n = 15; Table 1). On average, more than half of the duration of all courtships was spent within the respective male's peripheral territory, about 25% of the time was spent in the central territory, and less than 10% was spent outside the male's estimated territory (n = 9, see Fig. 5 and Table 2).

Table 1

Temporal and spatial patterns of the courtship march (CM) of Allobates femoralis in French Guiana.^a

Table 2

Proportions of time spent (n = 9) and distance covered (n = 10) in different areas during the courtship march of Brilliant-Thighed Frogs (Allobates femoralis) in French Guiana. The locations of the oviposition sites are given based on territory parameters for the male frog (n = 9).

Fig. 5

A map of an observed courtship march (CM) in French Guyana of the Allobates femoralis Male i16-m-018 on 30 March 2016. The red circle represents the start of the CM. The clutch produced after the CM is represented by the yellow circle, which is also the end of the observation period. The size of each circle (white) corresponds to the time (in minutes) the male spent at each location. The CM trajectory is shown as the black line; the red triangles represent the territorial observation points of the male. The central territory (minimum convex polygons) is labeled in blue and surrounded by peripheral territories (different shade of gray).

Spatial pattern.—The length of the courtship march ranged from 7.67–41.25 m (X̄ ± SD = 18.69 ± 8.77 m, n = 15). The courtship polygons ranged from 1.50–115.52 m² and the courtship buffer from 1.62–11.21 m² (n = 10; Table 2). On average, more than half of each courtship march was located in the peripheral territory, around 15% was located in the central territory, and around 20% outside the male's territory (n = 10; Table 2). None of the observed ovipositions occurred inside the central territory; 6 clutches were deposited inside the peripheral territory and 3 clutches outside (n = 9; Table 2). On average, the courtship buffer overlapped with 0.01% of the male's central territory (range = 0.00%–25.47%) and 0.98% of the peripheral territory (range = 0.06%–3.19%; n = 10).

Courtship in Relation to Male Territory Size

The peripheral territories ranged from 121.34 to 1116.30 m² (median ± iqr = 259 ± 342.17 m², n = 13) and the central territories from 1.39–44.37 m² (median ± iqr = 11 ± 6.91 m², n = 10; Table 1). The Pearson product-moment correlation model revealed that male territory size (central territory and peripheral territory) was neither predictable by a courtship's path length, the total duration, the duration in the morning, nor the courtship buffer and the courtship polygon.

Courtship in Relation to Clutches

In total, we found six previously deposited clutches inside the courtship buffers of 13 different males. We only found one clutch outside the courtship buffer but still within the estimated peripheral territory of the male. Observed ovipositions never took place in the central territory: in six cases the pair laid the clutch inside the estimated peripheral territory and in three cases outside the territory. Only nine males were included in this comparison because they were the only frogs with sufficient data points to reliably calculate central and estimated peripheral territories.

Within Individual Courtship Variation

We did not find consistency within single individuals, in either length or duration of their courtships (CVs and ANOVA). As the duration depends on the length of the courtship, we calculated residuals (time∼length). We also did not find consistency within the individuals using this metric. The median overlap of the courtship buffer between consecutive courtships by a single male was 0.81%, ranging from 0.24–2.83% (n = 6; see two extreme examples in Figs. 6 and 7).

Fig. 6

Map of four observations of different courtship marches (CM) by the same Allobates femoralis (Male i16-m-003), with little overlap between observations. The red circle represents the start of the CM. The clutch produced after the CM is represented by the yellow circle, which is also the end of the courtship observation. The CM trajectory and the CM buffer are shown in different colors. The central territory is in yellow and the estimated peripheral territory is in gray.

Fig. 7

Map of three observations of different courtship marches (CM) by the same Allobates femoralis (Male i16-m-041) with higher overlap than shown in Fig. 6. The red circle represents the start of the CM. The clutch produced after the CM is represented by the yellow circle, which is also the end of the courtship observation. The trajectory and buffer for each CM are shown in different colors. The central territory is in yellow and the estimated peripheral territory is in gray.

Courtship in Relation to Male Territory Size

We did not find any correlation between courtship area and male territory size. This result indicates that Al. femoralis females cannot use the courtship interaction to assess the size of a male's territory. We observed that the courtship trajectory, and even oviposition, might occasionally occur outside the male's territory. Oviposition happened very rarely in the center of the territory. Previous studies reported that courtship and oviposition of Al. femoralis take place only in the males' territory (Roithmair 1994a; Ringler et al. 2013). In total, 20% of the length (n = 10) and ∼8% of the time (n = 9) of the courtship, and three ovipositions (out of nine courtships observed), occurred outside the male territories (peripheral or central; Table 2). We need to interpret these results with caution because male territories change over time (e.g., change of territory extent, location, or even ownership). It is unlikely that territorial changes and methodological issues account for the large proportion of courtship observed outside the territory, however, because we used two different methods to calculate males' territories. The Voronoi tessellation even takes the time of observation into account and overestimates (rather than underestimating) the size of a territory. Our finding that courtship and oviposition sometimes occur outside the males' territory was unexpected, given that territoriality is a common feature across most dendrobatid frogs and represents an important resource for successful reproduction (Pröhl 2005; Ursprung et al. 2011). In many dendrobatid species, territory characteristics such as size and oviposition sites are important indicators of male quality (Wells 2007). In Am. trivittata (Roithmair 1994a), O. pumilio (Pröhl and Hödl 1999), and Al. paleovarzensis (Rocha et al. 2018b), male mating success was positively correlated with territory size and thus might represent a primary indicator for male quality in these species. Rocha et al. (2018b) suggested that female Al. paleovarzensis use courtship to assess the males' territory size because it correlated positively with courtship duration and numbers of male matings. Possibly, Al. femoralis females do not assess the size of a male territory during courtship but rather make sure that the potential mate is the territory owner.

Courtship in Relation to Clutches

In six cases, the courtship trajectory closely passed freshly laid clutches, but we only found one fresh clutch in the control areas. Previous studies in fish showed that females often copy the choice of other females, and males with eggs are preferably chosen by females (Dugatkin 1992; Pruett-Jones 1992). In Al. femoralis, females might recognize previously deposited clutches inside the male's territory during courtship—our data show that the trajectories of courtship marches often passed close to clutches from previous matings. Interestingly, experiments in the lab showed that female Al. femoralis cannibalize unrelated clutches in the absence of a guarding male (Spring et al. 2019), suggesting that males should avoid passing by previously laid clutches during courtship in order to minimize the risk of cannibalism by females. In these experiments, however, females were kept in close vicinity to the clutches and their motivation to cannibalize unrelated clutches might be different under natural conditions. Further studies are needed to clarify the relation between clutches already present in a male's territory and courtship marches in Al. femoralis.

The prolonged courtship behavior in dendrobatids is important to locate a suitable oviposition site (Lima et al. 2002; Montanarin et al. 2011; Forti et al. 2013; Souza et al. 2017; Rocha et al. 2018a). In Al. femoralis, oviposition sites are rain-protected leaves in the leaf litter. We observed males going under a leaf, followed by the female, who exited shortly after entering. This sequence was repeated (up to four times) at different locations, until both frogs stayed under a leaf and oviposition took place. This behavior indicates that the female evaluates the oviposition site presented by the male (cf. Switzer 1997). Showing previous clutches to females during courtship might likewise serve to demonstrate high quality, which is a common strategy in fish (e.g., Unger and Sargent 1988). Thus, the prolonged courtship in Al. femoralis is likely to serve as a mechanism to locate suitable oviposition sites.

We suggest that female Al. femoralis use the courtship sequence to explore and learn the area, especially with relation to the exact location of the clutch. Male Al. femoralis are predominantly responsible for tadpole transport, but females perform compensatory tadpole transport in cases where the respective fathers are missing (Ringler et al. 2015). Recent studies have shown that Al. femoralis females use indirect spatial cues to discriminate between their own and an unrelated clutch (Ringler et al. 2015, 2016a). From the male's perspective, locating multiple clutches close together facilitates defending, tending, and recalling their locations whereas females should rather prefer oviposition sites that are not too close to clutches of other females to avoid accidentally transporting unrelated clutches in case they have to perform this task (cf. Ringler et al. 2015, 2016a). Prolonged courtship might help in learning the location of the oviposition site which, in turn, would be useful for subsequent tadpole transport. Future studies should focus on the relationship between female vs. male parental care and courtship duration in different poison frog species.

All observed females stayed overnight in a male's territory before mating, indicating that females might need time to reach the physiological condition for oviposition. As Al. femoralis are prolonged opportunistic breeders, and environmental conditions vary, females would benefit from having the ability to control the timing of ovulation and to wait for optimal conditions. When followed in the field, female Al. femoralis laid clutches at variable intervals (5–12 d, n = 7; M.-T. Fischer, personal observation). Previous studies of other frog species suggest that prolonged courtship is important for stimulating ovulation (e.g., An. stepheni [Juncá 1998] and Al. caeruleodactylus [Lima et al. 2002]). Roithmair (1994a) and Montanarin et al. (2011) proposed that female Al. femoralis need to experience the courtship call and/or behavior to reach the necessary physiological condition for oviposition. In An. stepheni, the male amplexes the female two times and each amplexus lasts around 50 min. The first amplexus is without oviposition but after approximately 5 h, the male amplexes the female again and oviposition occurs. Juncá (1998) demonstrated that the first amplexus stimulates ovulation and that the females need the time between the two bouts of amplexus to ovulate. Female Ranitomeya vanzolinii had ovulated eggs during courtship, but not when they were without a male (Caldwell and Oliveira 1999). Females in Al. femoralis might thus require prolonged courtship to stimulate their physiological condition for ovulation.

Comparison to other Populations across South America

Our results show differences in the reproductive traits compared to other Al. femoralis populations across South America (Peru and Brazil). Although some differences across populations could be attributed to observation protocols and analytic procedures, the overall courtship behavior observed from our study population in French Guiana differed in many aspects from populations in Peru and Brazil. For example, compared to the other populations, we observed that courtship interactions between male and female frogs last longer, that tactile or visual signaling is absent during courtship, and that courtship culminates in cephalic amplexus (see Table 4). The variation of courtship and reproductive behaviors in different Al. femoralis populations across South America is ideal to investigate the functions of different courtship components, but more-standardized, methodological approaches to studying poison frog courtship are needed.

Table 4

Comparison of the reproductive traits in three different populations of Allobates femoralis: (1) Panguana Biological Station, Peru (Roithmair 1992, 1994a); (2) Reserva Florestal Adolpho Ducke, Brazil (Montanarin et al. 2011); and (3) this study. Table modified after Montanarin et al. 2011.