Study Site

Yasuní National Park and Biosphere Reserve and the adjacent Huaorani Ethnic Reserve form a 1.6 million ha protected area in Amazonian Ecuador (Fig. 1). While human influences are currently sparse but increasing, most of the park is undisturbed wilderness of lowland tropical rain forest [24]. Rainfall and temperature are aseasonal, with a mean annual rainfall of 2,800 mm and mean monthly temperature of 25-27°C. A 25-ha permanent forest dynamics plot (FDP) is located inside the park ($0^{\underset{\_}{°}}$ 41’ S, ${76}^{\underset{\_}{°}}$ 24’ W, 216-248 m a.s.l. [24]). We use data from the first and third censuses (1994-1997 and 2006-2007).

Study Species

Iryanthera hostmannii (Benth.) Warb. is a small understorey tree in the family Myristicaceae (nutmeg). It is one of 15 species of Myristicaceae that co-occur on the Yasuní FDP [19] and is the 29th most abundant tree species on the plot [24]. Iryanthera hostmannii is dioecious [19]. The fruits are woody capsules, containing a single arillate seed, dispersed by guans (Penelope jacquacu and Pipile pipile) and motmots (Baryphthengus martii) [25, 26]. Flowers are small and yellow-green in colour. They are borne on ramiflorous paniculate inflorescences emerging from leaf axils. Several flowers develop from nodes along the inflorescence. The male perianth is 3-lobed with the filaments fused into a column with three anthers. Female flowers are slightly larger and more fleshy with a single ellipsoid ovary. Pollinators of paleotropical Myristicaceae are usually small nocturnal insects such as beetles that are deceived into pollinating females, although nothing is currently known about the pollinators of Iryanthera [27].

Fig. 1.

Location of study site, Yasuní National Park (YNP), and the Yasuní forest dynamics plot (FDP), in Ecuador.

Data Collection

To quantify investment in flowers and determine the sex of individual trees, we counted the number of inflorescences on all 839 individuals of Iryanthera hostmannii within the Yasuní FDP in July-August of 2002, 2003 and 2004. For every tree we examined at least one inflorescence every year and sexed the tree from flowers. Status (alive/dead) and diameter at breast height (1.3m; DBH) of each tree were taken from the first and third FDP census. Annualized growth rates over 12 years (1997-2009) were calculated as: DBH1 – DBH0 / time1 – time0.

To monitor the phenology of individual flowers, in August 2003 we selected two large male and two large female trees that lay within or close to the FDP, each with many inflorescences. Previous studies have documented that Myristicaceae flowers open in the late afternoon and early evening [28]. We therefore undertook a preliminary census of I. hostmannii. On 13 and 14 August 2003, we monitored inflorescences every two hours from 06:00 to 22:00, and confirmed that flowers opened in the late afternoon. Therefore, we focussed attention on this period for the next 10 nights. From 16:00 hours we censused inflorescences every hour until no more flowers opened. We then recensused all flowers between 06:00 and 08:00 hours the following day to check for flowers that opened during the remainder of the night. We censused 13-25 inflorescences per tree and at every census we recorded the state (closed, open, wilting or dead) of each bud and flower (Table 1, Fig. 2).

In August 2003 we also collected inflorescences from three female and four male trees from outside the FDP. We collected two or three inflorescences per tree and a total of 85 female flowers and 404 male flowers. We removed the flowers from the inflorescences and oven-dried them. The total dry mass of flowers from each individual was calculated, because flowers were too small to weigh individually. We calculated mean flower mass from these group means. The number of flowers and nodes per inflorescence were calculated from the flower phenology data above.

Table 1.

Flowering of two Iryanthera hostmannii male and female individuals over several nights. For each night of observations, we show the number of buds and the number and proportion of buds that opened.

Data analysis

We compared reproductive investment between male and female trees in a variety of ways. We tested for greater mass of individual female flowers using a t-test. We tested for a greater number of flowers per inflorescence node in males using a generalised linear model (GLM) with Poisson errors. We tested for a greater number of nodes per inflorescence in males using a GLM with quasipoisson errors because of overdispersion in the data. Significance of each GLM was assessed using a Chi-squared likelihood ratio (LR) test to compare models with and without sex. We tested for a greater natural logarithm number of open flowers per inflorescence in males and a greater number of open flowers per night in males using ANOVA, and compared models with and without sex using an F-test.

We tested for earlier flower opening times in males using ANOVA adapted to circular data [29]. Significance was determined with a likelihood ratio test [30]. According to the 24-hr clock, 00:01 is at the opposite end of the day from 23:59. In reality, however, there is no natural start or end point to the time of day and thus quantities that vary on a diurnal cycle should be analysed as circular variables, where the time difference between the 23^rd and first hour is recognized to be two hours, not 22 [31]. Thus, the timing of daily peaks in flower opening was quantified as mean angles (0–${360}^{\underset{\_}{°}}$, roughly equivalent to hour of the day) and the dispersion of flowering (i.e. the length of the flowering period) was quantified as circular standard deviations. These two metrics can be displayed on circular histograms as a ‘mean vector’ with a specified length and direction, where vector length is an inverse function of the standard deviation [29].

Fig. 2.

Phenology of flower observations for two male (top) and two female (bottom) Iryanthera hostmannii trees (See Table 1).

We tested for a greater number and mass of inflorescences in males using a GLM with Poisson error structure. We tested for higher growth rates in males over 12 yrs, also using a linear model, and higher survival of males over 12 years using logistic regression. We used a non-linear Poisson error structure where appropriate because of right-skewed count data (e.g. number of flower per node, number of inflorescences), with a large number of low values, or logistic regression for binomial data. The full models for inflorescence number and mass, growth and mortality took the following form: response as a function of sex and DBH, with DBH as a covariate and an interaction with sex. In this way, we tested for a difference between males and females in both the intercept and slope of the response variable as a function of DBH. Significant effects in these models were determined from Chi-squared likelihood ratio tests. The spatial distributions of males, females and non-reproductives were compared using Ripley's K statistic, with 95% confidence envelopes calculated using the ‘envelope’ function in the R-package Spatstat [32]. Significant differences from a 1:1 sex ratio within DBH classes were tested with a G-test [33]. All analyses were conducted in the base statistical package R 2.10.1 [34] unless stated otherwise.

Fig. 3.

Iryanthera hostmannii flowering branches. A) Female, B) Male. Scale bar = c.1cm. Credit: Robin Foster, The Field Museum, Chicago, USA.

Flower and inflorescence structure and investment

Iryanthera hostmanii flower and inflorescence structure were quite different between the sexes (Fig. 3 and 4, Table 2). Female flowers had over twice the mass of male flowers (t-test: df = 2.644, t = 5.75, P = 0.015; Fig. 4A). Female inflorescences also had one more flower per node than male inflorescences (LR χ² = 28.3, P < 0.001; Fig. 4B). Conversely, female inflorescences were shorter than male inflorescences and had fewer than half the number of nodes of male inflorescences (LR χ² = 65.2, P < 0.001; Fig. 4C). However, mean total dry mass per inflorescence appeared slightly greater in females than males (although we could not statistically test this difference because these data were calculated from single mean values per sex; Fig. 4D).

Fig. 4.

Differences in flowering patterns between male (filled) and female (open) Iryanthera hostmannii trees. A) Mass of flowers, B) Number of flowers per node, C) Number of nodes per inflorescence, D) Number of inflorescences per tree, E) Number of flowers opening per inflorescence, F) Number of flowers open per night. Data: A & B: data from inflorescences collected from three female and four male trees, pooled; C, E & F: data from two female and two male trees monitored for flower phenology, treated separately; D: mean values per sex calculated from A, B & C. Significant differences between the sexes are indicated by asterisks: * P < 0.05, *** P <0.001.

At the tree level, there was a significant relationship between the number of inflorescences and tree size (DBH) in all three years (2002: LR χ² = 7241.9, P < 0.001; 2003: LR χ² = 8723.7, P < 0.001; 2004: LR χ² = 6886.8, P < 0.001; Fig. 5). Furthermore, there was a significant interaction between sex and DBH in this relationship: both the intercept and slope of the number of inflorescences as a function of DBH for males were significantly different from those of females (2002: LR χ² = 53.8, P < 0.001; 2003: LR χ² = 201.4, P < 0.001; 2004: LR χ² = 44.3, P < 0.001). In 2002 and 2003 the intercept of male trees was significantly lower than that of female trees, indicating that males produced fewer inflorescences on average. However, in all three years, male trees had a steeper slope than female trees, indicating that male trees produced more inflorescences than females at larger sizes (Fig. 5A-B). In 2004, male trees had a significantly larger intercept, indicating that they produced more inflorescences on average (Fig. 5C).

Table 2.

Investment in flowers by male and female Iryanthera hostmannii. Values: mean ± SE.

In terms of inflorescence dry mass, again there was a significant relationship between dry mass and DBH in all three years (2002: LR χ² = 206.5, P < 0.001; 2003: LR χ² = 249.9, P < 0.001; 2004: LR χ² = 194.2, P < 0.001; Fig. 5D–F). However, only in 2003 was there a significant interaction between sex and DBH in this relationship (2002: LR χ² = 1.6, P = 0.209; 2003: LR χ² = 6.06, P = 0.014; 2004: LR χ² = 1.3, P = 0.25). In 2003, females had a significantly higher intercept than males, but males had a significantly steeper slope than females, indicating that females invested more dry mass in inflorescences per tree than males at smaller sizes, but this difference disappeared at larger size classes (Fig. 5E).

Fig. 5.

Differences in inflorescence number and mass between male (filled circles, solid line) and female (open triangles, dashed line) Iryanthera hostmannii trees over three flowering seasons. A-C) Number of inflorescence per tree in 2002, 2003 and 2004. D-F) Estimated total mass of inflorescences (g) per tree in 2002, 2003 and 2004. Significant differences between the sexes (ANCOVA with DBH as a covariate) are indicated by asterisks: ^* P < 0.05, *** P <0.001.

Demography

Monitoring survival and growth over 12 years, rather than six as in Queenborough et al [19], revealed fewer insights than we had hoped. Only seven trees that were reproductive between 2002 and 2004 had died by 2009. However, six of these were female and only one was male. Although this hints that females may suffer higher mortality than males, the sample size remains too small to give a robust estimate.

Fig. 6.

Annualized growth rates of male (filled circles) and female (open triangles) Iryanthera hostmannii trees over 12 years.

In terms of growth, there was no significant relationship between relative growth rate and tree DBH (LR χ² = 0.99, P = 0.19), nor was there a significant difference between the sexes (LR χ² = 0.22, P = 0.54; Fig. 6). Thus, male and female trees of all sizes had equivalent growth rates over 12 years between the two censuses.

Table 3.

Sex ratios of Iryanthera hostmannii over three years on the Yasuní FDP.

Sex ratio

The population of Iryanthera hostmannii in the Yasuní FDP that flowered over 2002, 2003 and 2004 contained more male trees (200) than female trees (170), although this cumulative (total) sex ratio was not significantly different from 1:1 (G = 2.43, P = 0.12). However, flowering sex ratios were significantly male-biased in 2002 and 2004 (Table 2), and also within one specific size class. In the 20-30 mm DBH size class, significantly more male trees flowered than female trees in 2002 and 2004 (2002: G = 8.40, P = 0.004; 2004: G = 5.96, P = 0.015; Fig. 7).

Fig. 7.

Sex ratios in increasing size (DBH) classes in Iryanthera hostmannii in the Yasuní FDP between 2002 and 2004. Total number of males (filled) and females (open) are indicated on each side. Significant differences from a 1:1 sex ratio within a DBH class are indicated by asterisks: ^* P < 0.05, ** P <0.01.

Spatial distribution

Male and female Iryanthera hostmannii were distributed randomly throughout the FDP, indicated by the fact that the observed values of L(r) did not exceed the confidence envelopes (Fig. 8A,B), and as in [19] there was no significant aggregation or over-dispersion of males relative to females (Fig. 8C). There was also no evidence that trees become reproductive only in certain areas of the plot, as reproductive trees were distributed randomly with respect to non-reproductive trees (Fig. 8D).

Fig. 8.

The spatial pattern of male, female and non-reproductive Iryanthera hostmannii trees in the Yasuní FDP. The left panel shows a map of the 25ha plot, with males (filled circles), females (open triangles), and non-reproductives (open diamonds). The size of the symbol is proportional to the DBH of the tree. The right panel (A-D) shows results from point pattern analyses (Ripley's K). 95% confidence intervals are indicated by the shaded bar; if the black line lies out of these it indicates aggregation (above) or over-dispersion (below).

Flower phenology

Iryanthera flowers did not start to open until after 16.00. Most flowers opened between 17.00 and 20.00, with only a few opening later (Fig. 9). Male flowers opened significantly earlier than female flowers, although there was more variation in peak opening time in males (circular ANOVA with likelihood ratio test: df = 1, χ2 = 6.626, P < 0.05). Furthermore, male flowers were much shorter-lived than female flowers, remaining open less than two days before abscising, although again there was considerable variation among trees (ANOVA, sex: df = 1, SS = 4083, MS = 4083, F = 4.9, P = 0.026; tree: df = 1, SS = 102254, MS = 102254, F = 123.8, P < 0.001; Fig. 10). Female flowers, on the other hand, were open and receptive twice as long before either wilting and abscising or beginning to mature into fruit.

Fig. 9.

Flower opening times of two male and two female Iryanthera hostmannii trees. Circular bar charts show the number of open flowers per hour (24 hr clock). Arrows show the mean opening time (angle) and standard deviation (length). Note the log scale.

In terms of the number of flowers opening, male trees opened three times more flowers per inflorescence than females (F-test: F = 53.2, P < 0.001), and three times more flowers per inflorescence per night (F = 56.4, P < 0.001, Table 1, Fig. 4E and F).

Fig. 10.

Mean flower life time for two male (filled circles) and two female (open triangles) Iryanthera hostmannii trees in Yasuní. Error bars = SE. Sample sizes (# flowers per tree) indicated at the bottom.

Investment, resource and pollen availability

Previous studies of Myristicaceae have documented significant differences between the sexes in terms of flower size and number [35, 36, 37], such as we found in this study. Males are under selection pressure to effect pollen transfer to females. Competition between males in our case has led to possibly earlier flower opening in the day than females, as well as increased flower number: male inflorescences were longer than female inflorescences and had more flowers, and large males also produced more inflorescences than females. Males that produce more flowers are likely to have greater reproductive fitness than males that produce fewer. However, even small males produced some flowers. Given the short distances that the likely pollinators of this species move [27], as well as the understorey environment in which Iryanthera hostmannii exists, proximity to a female may be just as important as flowering intensity to ensure the successful transfer of pollen [38]. However, given the often orders-of-magnitude difference in flower number between males and females observed in multiple-ovaried flowers of species of other families [39, 40, 41], the difference in flower number in Iryanthera, with the single-ovaried flowers, was not that great. Females can only produce one offspring per flower, and thus may increase flower number to account for this.

An alternative mechanism to increase pollen availability would be for each male flower to produce more pollen, likely leading to larger male flowers. This would entail a whole suite of problems associated with the deceit pollination that occurs in Myristica, and likely Iryanthera as well [27]. Selection on reproductive structures has also been posited to explain dimorphism in canopy structure [42] and leaf size [43]. However, because inflorescences in Iryanthera are ramiflorous, with multiple inflorescences emerging from a single branch, there is little scope for selection pressure to influence these traits here.

Total investment in flowers, as measured by dry mass, was equivalent in males and females in two out of three years of our study. This parity implies that, in general, the difference in reproductive investment between males and females is predominantly in the fruiting stage (at least in terms of carbon). It is likely that only those females with access to abundant resources, or resources stored from previous years, may proceed to set fruit [44, 45]. Resolution of this question will require long-term studies of resource and pollen availability combined with detailed quantification of growth, mortality and reproduction.

Further investigation will also be required to determine whether the same resources limit reproduction in both sexes. For females of many tropical species, carbon is the most likely [46]. However, in the case of the Myristicaceae, each seed is surrounded by a fatty, nutrient-rich aril, and resources such as nitrogen may also limit female reproduction [47]. Thus, dry mass may not be the best measure for comparing resource allocation between the sexes.

Furthermore, resources will only limit reproduction in females if pollen is not limiting [48]. We have previously found evidence of pollen limitation in this species (Queenborough, unpublished data). Determining whether resources or pollen limit any one reproductive event, and thus enabling a robust examination of how resources are allocated, is essential for life history theory.

Long-term costs of reproduction

Evidence for long-term demographic costs of reproduction in woody plants is equivocal [16, 49], apart from the few species of monocarpic tree [50]. Evidence of negative effects of reproduction on growth has come from tree-ring dendrochronological studies or species that produce one fruit per branch, such as some gymnosperms [51], but by definition these studies only include those individuals that survived a long time. Individuals that suffered extremely negative effects of reproduction on growth (i.e. those that died) are lost. Thus, only prospective studies of individuals of known sex will provide robust estimates of the long-term effects of reproduction on growth and survival. Unfortunately, the currently available data cover too short a period of many species' life cycles, and good estimates of long-term growth and mortality are difficult to obtain [49]. There is good evidence from some species, however, that female physiological processes cannot increase output, and that photosynthesis and growth rates decrease with increased costs of fruit production [10]. The true long-term costs of reproduction in many species remain unknown.

Floral phenology

Most previous work on dimorphism in flowering phenology has focused on differences at the whole plant level, i.e. do males flower earlier in the season than females? Less work has been carried out explicitly on daily patterns of phenology at the flower level [22]. Competition between males would suggest that males should flower earlier in both instances, ensuring that pollen is available and is transferred efficiently from males to females. Theory would also predict that males might flower for longer than females, ensuring that they have the potential to pollinate every last female. Evidence for these predictions is equivocal, however, with some studies finding earlier male flowering [9, 52] and some earlier female flowering [53]. In our study, we found a slightly earlier flower opening time in males, but with much overlap between the sexes.