Portal de Periódicos da CAPES

Pollinator‐mediated selection on floral traits and size of floral display in Cyclopogon elatus , a sweat bee‐pollinated orchid

2006; Wiley; Volume: 20; Issue: 6 Linguagem: Inglês

10.1111/j.1365-2435.2006.01179.x

1365-2435

Santiago Benítez-Vieyra, A. M. MEDINA, Evangelina Glinos, Andrea A. Cocucci,

Ecology and Vegetation Dynamics Studies

Functional EcologyVolume 20, Issue 6 p. 948-957 Free Access Pollinator-mediated selection on floral traits and size of floral display in Cyclopogon elatus, a sweat bee-pollinated orchid S. BENITEZ-VIEYRA, Corresponding Author S. BENITEZ-VIEYRA †Author to whom correspondence should be addressed. E-mail: sbenitez@efn.uncor.eduSearch for more papers by this authorA. M. MEDINA, A. M. MEDINA Instituto Multidisciplinario de Biología Vegetal (CONICET – Universidad Nacional de Córdoba), CC 495 CP 5000 Ciudad de Córdoba, Córdoba, ArgentinaSearch for more papers by this authorE. GLINOS, E. GLINOS Instituto Multidisciplinario de Biología Vegetal (CONICET – Universidad Nacional de Córdoba), CC 495 CP 5000 Ciudad de Córdoba, Córdoba, ArgentinaSearch for more papers by this authorA. A. COCUCCI, A. A. COCUCCI Instituto Multidisciplinario de Biología Vegetal (CONICET – Universidad Nacional de Córdoba), CC 495 CP 5000 Ciudad de Córdoba, Córdoba, ArgentinaSearch for more papers by this author S. BENITEZ-VIEYRA, Corresponding Author S. BENITEZ-VIEYRA †Author to whom correspondence should be addressed. E-mail: sbenitez@efn.uncor.eduSearch for more papers by this authorA. M. MEDINA, A. M. MEDINA Instituto Multidisciplinario de Biología Vegetal (CONICET – Universidad Nacional de Córdoba), CC 495 CP 5000 Ciudad de Córdoba, Córdoba, ArgentinaSearch for more papers by this authorE. GLINOS, E. GLINOS Instituto Multidisciplinario de Biología Vegetal (CONICET – Universidad Nacional de Córdoba), CC 495 CP 5000 Ciudad de Córdoba, Córdoba, ArgentinaSearch for more papers by this authorA. A. COCUCCI, A. A. COCUCCI Instituto Multidisciplinario de Biología Vegetal (CONICET – Universidad Nacional de Córdoba), CC 495 CP 5000 Ciudad de Córdoba, Córdoba, ArgentinaSearch for more papers by this author First published: 05 September 2006 https://doi.org/10.1111/j.1365-2435.2006.01179.xCitations: 60AboutSectionsPDF ToolsRequest permissionExport citationAdd to favoritesTrack citation ShareShare Give accessShare full text accessShare full-text accessPlease review our Terms and Conditions of Use and check box below to share full-text version of article.I have read and accept the Wiley Online Library Terms and Conditions of UseShareable LinkUse the link below to share a full-text version of this article with your friends and colleagues. Learn more.Copy URL Share a linkShare onFacebookTwitterLinked InRedditWechat Summary 1 Pollinator-mediated selection on traits associated with mechanical fit and attraction of pollinators were investigated through both sexual functions in Cyclopogon elatus (Sw.) Schlechter (Orchidaceae). 2 Only halictid bees, principally Augochlora nausicaa, were observed as pollinators. The pollinarium becomes attached to the ventral surface of the bee's mouthparts (labrum) when the proboscis, which closely matches the length of floral tube, is projected into the flower to reach nectar. 3 We detected directional selection on nectary depth, with deeper nectaries favoured only through male fitness, because this trait affects pollinaria removal but not deposition. Correlational selection was detected through male function between nectary depth and the number of flowers in an individual's floral display. These traits affect pollination in a multiplicative way: flower number is positively related to the number of bee visits, and nectary depth positively affects the effectiveness of pollinaria removal at each visit. 4 We also detected stabilizing selection on display size. For smaller displays there was a strongly positive association between number of flowers and overall reproductive success, which can be attributed to a simple numerical effect. However, the expected performance of individual flowers is impaired in large displays by pollinator limitation, because bees visiting the display pollinate few flowers per visit and each bee carries, at most, one pollinarium. Introduction Since Darwin (1877), morphological adaptation to pollinators has been invoked as a major factor in explaining flower diversity and evolution. The first direct field evidence of Darwin's hypothesis came from Nilsson's (1983) study on Platanthera, in which selection against short nectar spurs was demonstrated. Finally, experimental shortening of nectar spurs demonstrated this selection unequivocally (Nilsson 1988; Johnson & Steiner 1997). Subsequently, many studies have demonstrated that floral traits are under selection, sometimes strongly (for review see Kingsolver et al. 2001). Flower morphology is thought to be adapted to pollinators, thereby assuring both attraction and contact with the fertile parts once pollinators have arrived, with contact being commonly achieved by a functional fit between flower and pollinator. However, most selection studies have focused either on general floral features such as flower number, flower size and display height, which are likely to be attractive to all flower visitors; or on traits associated with breeding systems (Fenster et al. 2004). Fewer studies have focused on traits that can be attributed to flower specialization for pollinator use, such as traits associated with the mechanical fit between flowers and pollinators (Fenster et al. 2004). Floral traits of specialized plants often experience the effects of natural selection (Johnson & Steiner 1997, 2000), particularly regarding the apparent similarity in length of the spur or nectar tube and the pollinator's mouthparts (Nilsson 1988; Herrera 1993; Johnson & Steiner 1997; Maad 2000; Alexandersson & Johnson 2002; Maad & Alexandersson 2004). Traits associated with the mechanical fit with pollinators are expected to be subjected to directional (Nilsson 1988) or stabilizing (Cresswell 1998) selection. In addition, correlation selection can favour some combinations of character states over others, usually because the characters are functionally related (Sinervo & Svensson 2002; Futuyma 2005). Although studies on flower trait correlation have shown patterns suggesting functional integration (Armbruster et al. 1999, 2000), the above expectation has not been confirmed by empirical evidence from phenotypic selection studies. Very few studies have attempted to find, and even fewer have actually found, correlational selection between floral traits (O'Connell & Johnston 1998; Caruso 2000; Gómez 2000; Maad 2000; Herrera 2001). To date, significant correlational selection gradients that have been found involve only inflorescence traits and phenology (O'Connell & Johnston 1998; Maad 2000). Stabilizing and correlational selection gradients (non-linear) are more rarely detected than directional gradients (linear) and are generally weak (Kingsolver et al. 2001). This is thought to have several methodological reasons, including that investigators often choose to study characters suspected of directional selection and, for a given sample size, it is easier to detect linear rather than quadratic gradients (Conner 2001). Orchids show a broad range of interesting and elaborate pollination mechanisms, exhibiting an extraordinary diversity of striking adaptations to flower visitors (Darwin 1877; Proctor, Yeo & Lack 1996). Most past research has supported the idea that natural selection driven by pollinators is the principal process behind orchid floral evolution (Nilsson 1992; Maad 2000). We focused the present study on an orchid pollinated by sweat bees (Halictidae). Species of several orchid genera are known to be pollinated in South America by these bees (van der Pijl & Dodson 1966; Galetto, Bernardello & Rivera 1997; Singer & Cocucci 1999; Singer & Sazima 1999), most notably species of Cyclopogon, which are known to be pollinated only by halictids (Galetto et al. 1997; Singer & Cocucci 1999; Singer & Sazima 1999). As some flower features apparently fit to these small bees, it has been supposed that most species in this relatively large spiranthoid genus (70 species) should be pollinated by halictid bees (Singer & Sazima 1999). Cyclopogon elatus (Sw.) Schlechter offers a suitable system for selection studies, as sufficiently large populations can easily be found, in comparison with other orchid species. In addition, as in other orchids, the presence of pollinaria enables male and female reproductive success for selection studies to be estimated (Nilsson 1992; O'Connell & Johnston 1998). With regard to its pollinators, a single Agapostemon bee (Halictidae) has been recorded carrying a pollinarium of C. elatus (Galetto et al. 1997). For this orchid species, we describe the pollination and flower functional morphology. We then evaluate the likelihood of selective advantage for phenotypic traits specifically associated to pollinator attraction and mechanical fit. We utilize phenotypic selection methods (Lande & Arnold 1983; Arnold & Wade 1984a; review by Brodie, Moore & Janzen 1995) to help understand the link between floral traits and pollinator specialization. In particular, we investigate whether a flower's nectary depth, alone or in combination with other traits, could influence fitness through pollinator mechanical fit by testing for directional, stabilizing or correlational selection on this trait. In addition, we analyse the effect of display size on total fitness. Materials and methods plant species, floral morphology and study site The orchid C. elatus is a terrestrial herb with 20–60-cm-tall spikes emerging from a basal rosette formed by five to seven leaves and bearing 20–30 tubular flowers (Fig. 1a) (Correa 1955). Plants are self-compatible (Galetto et al. 1997) and can also reproduce clonally by division of short lateral shoots. Sister plants may form dense clusters of up to 14 ramets, although ≈60% of the plants bear only one spike. The flowers are not very noticeable, displaying a greenish to reddish-brown colour and producing a sweet, musk-like scent. They offer sucrose-rich and highly concentrated nectar as a reward to pollinators (Galetto et al. 1997). The 7–11-mm-long labellum (Fig. 1a, inset) is white and violin-shaped, having a broad but short and projecting apical portion, and a longer and shallower-grooved basal portion. The latter portion and the adpressed column form a short nectary tube. Flower architecture is bilabiate (Fig. 1b) with an upper lip covering the column, which is built by the dorsal sepal and the adjoined lateral petals. The lower lip is formed by the labellum's apical part. At the nectary tube, the labellum base bears two conspicuous auriculae involved in nectar secretion (Galetto et al. 1997). Externally, on the labellum, there are two lateral glandular areas identifiable as osmophores, according to anatomical features (A. P. Wiemer, M. Mové, S. Benitez-Vieyra, R. A. Raguso & A. N. Sérsic, unpublished data) and by their positive reaction to neutral red staining (Vogel 1990). The column is short, 3–4 mm long, and bears one erect anther. The rostellum is dorsal and tongue-shaped (Correa 1955). The powdery pollinarium is distally in contact with a diamond-shaped viscidium. The viscidium is covered by a thin membrane which, when touched, breaks dorsally, exposing a glue and fixing the pollinarium to the pollinator's body (Singer & Cocucci 1999). Cyclopogon elatus occurs from Guatemala to Central Argentina, growing in wet and shady places. It flowers in Argentina from mid-August to mid-September, being one of the first nectariferous plants in bloom during late winter. Flowering is simultaneous among most individuals. Figure 1Open in figure viewerPowerPoint (a) Reproductive traits and pollination process of the Cyclopogon elatus spike. Inset: excised labellum. (b) An Augochlora nausicaa bee carrying a pollinarium, with a longitudinal section of a flower showing the correlation between flower parts and the bee's mouthparts. The pollinarium is attached onto the internal surface of the labrum. The dorsal sepal was removed. (c) An A. nausicaa bee with its proboscis introduced into a C. elatus flower. (d) The same bee immediately after carrying a recently removed pollinarium. La, labrum; Ll, lower lip; n, nectary chamber; no, labellum notches; o, osmophore; p, pollinarium; pr, proboscis; s, stigma; v, viscidium. Field observations were carried out in August and September 2004, in a natural plot of ≈1 ha near Cabana (31°12′46″ S; 64°20′52″ W, 729 m), 36 km north-west of Córdoba city, Argentina. This place is a secondary montane dry woodland dominated by trees such as Lithrea molleoides, Acacia praecox, Acacia caven and Kageneckia lanceolata. The population studied included about 300 C. elatus plants. pollination process In three visits to the study site during 2004, involving observations between 10.00 and 17.00 h for 35 h, we observed and took photographs of bees extracting and depositing pollinaria. Individuals of the flower visitor species were caught for later identification. We also recorded the number of flowers visited on each plant and the time spent by pollinators in each flower. Using a stereomicroscope, we measured the anatomical traits of the bees supposedly associated with the mechanical fit with flower morphology in 13 vouchers of the bee species identified as the main pollinator of C. elatus. Six of these vouchers were from within 3 km of the Cabana site, while seven were from locations within 60 km of this site. We measured proboscis length; distance from labrum tip to mid-leg coxas; and distance between distal tips of mid-leg femora when legs are held horizontally and at right angles to the body axis. In order to determine if fruit production is dependent on pollinators, 10 plants were isolated from pollinators and their fruit set was recorded. flower traits and pollinator-mediated selection To estimate pollinator-mediated selection, we measured flowers and recorded male and female reproductive success in 119 individuals. Clone groups were taken as single individual plants. Two newly opened flowers from each plant were harvested and preserved in 70% ethanol to measure morphological traits. The nectary depth, from the rostellum to the column base, and the pollinarium length were measured to the nearest 0·01 mm using a digital calliper. Labella were removed and mounted on microscope slides, then photographed at high resolution with a Leica M420 stereomicroscope equipped with a Nikon Coolpix 5400 digital camera. On the resulting photographs, the width of labellum constriction (distance between notches); labellum maximum width; and length of lower lip were measured. Principal component analysis (PCA) was carried out with three measures (constriction width; labellum width; lower lip length) to obtain a lower lip size factor (Conner, Rush & Jennetten 1996), which was used in subsequent analysis. In tables and throughout the text the first axis of this PCA will be referred as 'size of lower lip'. The software uthscsa image tool ver. 3 for windows (University of Texas Health Science Center) was used for taking all the measurements on the digital photographs, calibrated using a 5-mm reference scale. Finally, the total number of flowers produced per plant was recorded. For selection analysis, all the variables were standardized to zero mean and unit variance. After bearing fruit, spikes of measured plants were collected in order to count the number of fruits produced and the number of flowers that had had their pollinaria removed. The latter was possible as flower parts remain intact in wilted flowers. Number of pollinaria exported and number of fruits set were used as measures of male and female total reproductive success, respectively. Total fitness is a reliable measurement of selection acting on individuals, and can be used to predict evolutionary responses to selection (Arnold & Wade 1984b; Conner 1996). In addition, by dividing the number of pollinaria exported or the number of fruits set by the flower number, we obtained a relative fitness measure per flower, or multiplicative fitness component (Arnold & Wade 1984a; Conner 1996), which is useful for investigating patterns of selection acting at the level of the individual flower. The 'per flower' fitness measures enable the numerical or automatic effect of display size (flower number) to be separated from other possible effects. Finally, the individual values of all fitness measures were divided by the respective population mean. Prior to phenotypic selection analysis, some tests were performed on a subset of six randomly chosen plants to ensure morphological variation was greater among than within individuals. All the above traits were measured in all completely developed flowers (usually six or seven), which had been preserved in 70% ethanol. One-way anovas were used to determine the degree of variation in each trait within–among individuals (Medel, Botto-Mahan & Kalin Arroyo 2003). Only traits satisfying the condition of being significantly different between individuals were included in the phenotypic selection analysis. The total potential or opportunity for selection (I) acting in the population was estimated as the variance of total fitness (Arnold & Wade 1984a; Brodie et al. 1995). To assess the magnitude of natural selection acting on the phenotypic characters, the model proposed by Lande & Arnold (1983) was followed. The selection differentials, which represent the direct and indirect selection on a specific trait, were estimated. For directional selection, these values were estimated as the covariance between each trait and relative fitness. For disruptive or stabilizing selection, the differentials were estimated as the covariance between squared deviations of the trait and relative fitness. Selection gradients were calculated to reveal the magnitude and direction of selection on a specific trait or combination of two traits (correlational selection; Sinervo & Svensson 2002) independently from the indirect effect of other traits. These were estimated by means of a multiple regression using the above-mentioned model. The sequential Bonferroni correction (Rice 1989) was used to evaluate the table-wide significance (P < 0·05) of selection differentials and gradients. The results obtained represent the best linear or quadratic approximation (Brodie et al. 1995) to the relationship between fitness and a given trait or combination of traits. To avoid assumptions about the shape of the relationship and other limitations of this model, the cubic spline non-parametric regression was applied to depict univariate associations with fitness for traits that were shown to be significantly affected by phenotypic selection (Schluter 1988). This procedure allows for the estimation of complex functions with multiple peaks and valleys (Brodie et al. 1995), because it provides a local fit to the data instead of a global fit, as in an ordinary regression. A smoothing parameter (λ) was identified, which approximately minimizes the sum of squared deviations between the estimate and the true fitness surface (Schluter 1988). Results pollination process This orchid depends on pollinators for sexual reproduction, as none of the isolated plants produced any fruits. Four bee species were recorded visiting the flowers of C. elatus: Apis mellifera Linnaeus, 1758 (Apidae); Augochlora nausicaa (Schrottki, 1910) (Fig. 1c,d); Pseudoagapostemon jenseni Friese, 1908; and Temnosoma metallicum Smith, 1853 (Halictidae). Of these, A. mellifera and A. nausicaa were the most frequent (13 and 12 records, respectively), while only two individuals of each of the other two species were recorded. The main visitor activity was recorded between 11.00 and 14.00 h. Apis mellifera visited more flowers per plant than A. nausicaa (6·82 ± 25·76 and 2·00 ± 6·73 flowers, respectively; Mann–Whitney U-test: U = 16·5; P = 0·001). In contrast, the average time spent per flower was greater for A. nausicaa than for A. mellifera (5·60 ± 11·26 and 3·68 ± 0·86 s, respectively; Mann–Whitney U-test: U = 52·5; P = 0·011). Augochlora nausicaa and P. jenseni were the only bees observed carrying pollinaria (Fig. 1d). Augochlora nausicaa bees landed on the flower by embracing the labellum with their mid-legs. The lower lip, which acts as landing platform, is not large enough for the bee to settle completely on the flower, with the abdomen and most of the thorax remaining outside it. To reach nectar, the bee introduced its head below the upper lip by pushing it upwards and then fully extended its mouthparts into the nectary tube (Fig. 1b,c). While doing so, the bee pressed the ventral surface of its labrum against the dorsal surface of the flower's viscidium, and the pollinarium became attached to the ventral surface of the labrum (Fig. 1b,d). After leaving the flower, the bee retracted its proboscis and labrum, thus folding back the pollinarium to a protected position under the head. When a bee that carried a pollinarium visited another flower, it extended its proboscis again and the pollinarium was pressed against the stigmatic surface, depositing pollen clumps. A single pollinarium was able to pollinate several flowers. Pseudoagapostemon jenseni showed exactly the same behaviour as A. nausicaa on its visits to flowers. Apis mellifera behaved as a nectar thief: its proboscis, which is longer than that of halictid bees, does not have a suitable dorsal structure at a corresponding length from the tip for the pollinaria to become attached to. The infrequent visits of T. metallicum could not be observed in enough detail to describe its behaviour. However, it is unlikely that it can act as a pollinator; due to its small size, it is difficult for it to pick up and carry a pollinarium. Regarding the pollination process described, some flower morphological traits appear to fit to anatomical parts of A. nausicaa, in particular the mean nectary length fits the proboscis length of A. nausicaa (t-test: df = 126; t = −0·04; P = 0·966; Table 1; Fig. 1b). On the other hand, the lower lip, onto which the bee must alight and grip, is small compared with the bee's respective parts: the labellum constriction is smaller than the distance between the bee's femora tips (t-test: df = 128; t = −10·85; P < 0·0001; Table 1); the length of the lower lip is also smaller than the distance from the bee's head to the mid-legs (t-test: df = 127; t = 15·02; P < 0·0001; Table 1). Table 1. Mean and SD of flower characters and fitness measures of Cyclopogon elatus, and possible related anatomical traits of its pollinator Augochlora nausicaa C. elatus traits (n = 119) A. nausicaa traits† (n = 13) Description Mean (SD) Description Mean (SD) Nectary depth 4·25 (0·42) Proboscis length 4·23 (0·90) Labellum constriction width 1·99 (0·20) Distance between femora extremes 2·14 (0·20) Length of lower lip 1·61 (0·16) Distance from head to mid-legs 4·24 (0·49) Labellum maximum width 3·42 (0·35) Pollinarium length 2·42 (1·27) Flower number 42·8 (30·2) Pollinaria exported 14·6 (12·4) Fruits set 38·2 (27·9) Pollinaria exported/flowers 0·36 (0·18) Fruits/flowers 0·87 (0·17) All floral characters measured in mm. † Anatomical traits probably associated with flower characters. Descriptive statistics of floral traits and fitness measures are shown in Table 1. For subsequent analysis, the labellum variables were summarized in the first axis of a PCA, which explains 55% of the variance and represents the size of the lower lip, because all the eigenvalues of this axis were positive. Pollinarium length was excluded from phenotypic selection analysis, as the variation among individual plants was similar to the variation in a single plant (anova: df = 40; F = 2·07; P = 0·093). All other traits differed significantly among individuals. Variables related to lower lip are all correlated with each other (Table 2). From the traits included in the selection analysis, only flower number and nectary depth show a significant correlation (Table 2). Table 2. Pearson correlation coefficients for Cyclopogon elatus traits Parameter Nectary depth Pollinarium length Size of lower lip Labellum constriction width Labellum maximum width Length of lower lip Flower number 0·210* −0·057 −0·044 0·024 0·022 −0·153 Nectary depth 0·256** −0·103 −0·026 0·158† −0·466*** Pollinarium length 0·141 0·201* 0·053 0·048 Size of lower lip 0·791*** 0·822*** 0·592*** Labellum constriction width 0·476*** 0·198* Labellum maximum width 0·279** n = 119 individuals. †, P < 0·1; *, P < 0·05; **, P < 0·01; ***, P < 0·001. phenotypic selection Opportunity for selection was higher through male than through female function (I = 0·72 for number of exported pollinaria; I = 0·54 for fruit set). With respect to total fitness, significant directional selection differentials were found acting on nectary depth and flower number, while stabilizing selection was acting on flower number through the male component of reproductive success (Table 3). Through the female component of reproductive success, the results are similar, but selection acting on nectary depth was weaker and non-significant after Bonferroni correction. Regarding fitness contribution per flower, directional selection was found to be acting on three traits: individuals having smaller-sized lower lips were favoured through both sexual functions; those with deeper nectaries through male fitness; and those with a larger flower number through female fitness. Table 3. Univariate phenotypic selection on three flower traits through male and female function in Cyclopogon elatus Sexual function Fitness measure Character i S i (SE) C ii (SE) Male Number of exported pollinaria Nectary depth 0·343 (0·072) *** 0·067 (0·053) Flower number 0·713 (0·043) *** 0·253 (0·039) *** Size of lower lip −0·096 (0·062) −0·036 (0·034) Exported pollinaria/flowers Nectary depth 0·179 (0·040) *** 0·013 (0·053) Flower number 0·054 (0·043) −0·028 (0·025) Size of lower lip −0·069 (0·034)* −0·005 (0·019) Female Number of fruits Nectary depth 0·166 (0·068)* 0·032 (0·047) Flower number 0·720 (0·012) *** 0·301 (0·028) *** Size of lower lip −0·044 (0·055) −0·030 (0·030) Fruits/flowers Nectary depth 0·007 (0·018) 0·001 (0·012) Flower number 0·042 (0·018)* −0·001 (0·010) Size of lower lip −0·028 (0·014)* −0·009 (0·008) n = 119 individuals. †, P < 0·1; *, P < 0·05; **, P < 0·01; ***, P < 0·001. Standardized directional selection differentials (Si), stabilizing/disruptive selection differentials (Cii) and standard errors (SE) are indicated. Selection differentials in bold were significant after sequential Bonferroni correction. In the multivariate selection analysis (Table 4), directional selection was detected through male function on nectary depth and flower number for the overall fitness measure. Also, a tendency to directional selection was detected for lower lip size. Stabilizing selection was found to act on flower number. Correlational selection was found only between nectary depth and flower number. Selection acted differently through female function: directional selection did not act on nectary depth, was present only in flower number, and was weak in size of lower lip. Stabilizing selection was also found to act on flower number. With respect to fitness contribution per flower, the patterns of selection on nectary depth and size of the flower lip were the same as those found through overall fitness, but stabilizing selection on flower number was more intense and correlational selection was absent. Table 4. Multivariate phenotypic selection on three flower traits through male and female function in Cyclopogon elatus Sexual function Fitness measure Character i βi (SE) γii (SE) Character j Flower number γij (SE) Size of lower lip γij (SE) Male Number of exported pollinaria Nectary depth 0·199 (0·040) *** –0·086 (0·066) 0·137 (0·039) *** 0·008 (0·036) Flower number 0·655 (0·038) *** −0·220 (0·066) ** −0·043 (0·034) Size of lower lip −0·060 (0·031)† −0·020 (0·032) Exported pollinaria/flowers Nectary depth 0·168 (0·041) *** −0·013 (0·076) 0·004 (0·044) >0·001 (0·041) Flower number 0·015 (0·041) −0·203 (0·074)** −0·013 (0·038) Size of lower lip −0·055 (0·032)† −0·019 (0·037) Female Number of fruits Nectary depth 0·011 (0·013) −0·011 (0·022) 0·009 (0·013) 0·002 (0·012) Flower number 0·717 (0·012) *** −0·069 (0·022) ** 0·013 (0·011) Size of lower lip −0·018 (0·010)† −0·008 (0·011) Fruits/flowers Nectary depth −0·005 (0·018) −0·007 (0·033) 0·011 (0·019) −0·005 (0·018) Flower number 0·041 (0·018)* −0·092 (0·032)** 0·017 (0·016) Size of lower lip −0·027 (0·014)† −0·016 (0·016) n = 119 individuals. †, P < 0·1; *, P < 0·05; **, P < 0·01; ***, P < 0·001. Standardized directional selection gradients (βi), stabilizing/disruptive selection gradients (γii) correlational selection gradients (γij) and standard errors (SE) are indicated. Selection gradients in bold were significant after sequential Bonferroni correction. Combinations of directional and stabilizing selection on flower number were evident as approximately quadratic fitness functions, but with their maximum displaced from the mean values (Fig. 2b,f), or as monotonic increasing curve functions (Fig. 2c). In the other cases, cubic splines approximated very well to the linear (Fig. 2a,d) or quadratic functions (Fig. 2e) of Lande & Arnold's (1983) parametric model. A selection surface was constructed to show correlational selection (Fig. 3), which demonstrated that plants