|Home | About | Journals | Submit | Contact Us | Français|
While pollinators may in general select for large, morphologically uniform floral phenotypes, drought stress has been proposed as a destabilizing force that may favour small flowers and/or promote floral variation within species.
The general validity of this concept was checked by surveying a taxonomically diverse array of 38 insect-pollinated Mediterranean species. The interplay between fresh biomass investment, linear size and percentage corolla allocation was studied. Allometric relationships between traits were investigated by reduced major-axis regression, and qualitative correlates of floral variation explored using general linear-model MANOVA.
Across species, flowers were perfectly isometrical with regard to corolla allocation (i.e. larger flowers were just scaled-up versions of smaller ones and vice versa). In contrast, linear size and biomass varied allometrically (i.e. there were shape variations, in addition to variations in size). Most floral variables correlated positively and significantly across species, except corolla allocation, which was largely determined by family membership and floral symmetry. On average, species with bilateral flowers allocated more to the corolla than those with radial flowers. Plant life-form was immaterial to all of the studied traits. Flower linear size variation was in general low among conspecifics (coefficients of variation around 10 %), whereas biomass was in general less uniform (e.g. 200–400 mg in Cistus salvifolius). Significant among-population differences were detected for all major quantitative floral traits.
Flower miniaturization can allow an improved use of reproductive resources under prevailingly stressful conditions. The hypothesis that flower size reflects a compromise between pollinator attraction, water requirements and allometric constraints among floral parts is discussed.
Flowers and their constituent parts show remarkable morphological constancy within species. Aimed at gamete exchange, stamens, styles and corollas are highly evolved structures that display minimal variation among conspecifics, at least compared to vegetative organs (Berg, 1960; Stebbins, 1970). Morphological constancy is beneficial to pollination because a constant architecture allows for higher precision (i.e. the quality of being reproducible) in pollen exchange, and stabilizing selection has rendered flower form and shape remarkably constant within species. Plant–pollinator interactions have long been recognized to play a major role in driving the evolution of floral size and morphology (Darwin, 1862; Grant, 1949), although additional factors have been proposed such as antagonistic interactions with seed predators (Brody, 1992), water uptake rates (Galen et al., 1999), size/number trade-offs affecting flowers, fruits or seeds (Andersson, 2006), and even pleiotropic relationships with non-floral characteristics (Andersson, 1997; see Strauss and Whittall, 2006, for a review of non-pollinator agents that may exert selection on floral traits). All of these factors may contribute to explain the persistence of flowers of varying sizes in natural populations (Galen, 1989; Campbell, 1991; Herrera, 2005), which does not contradict the high heritability of floral traits (e.g. Mitchell and Shaw, 1993; Campbell, 1996; Galen, 1996; Lennartsson et al., 2000; Motten and Stone, 2000). Additionally, flowers can also show some phenotypic plasticity (Mazer and Schick, 1991; Holtsford and Ellstrand, 1992; Carroll et al., 2001; Elle and Hare, 2002).
Flowers are not isolated, fully independent allocation modules, and this applies particularly if they are borne in inflorescences where positional effects often exist (Herrera, 1991, 2004; Diggle, 1995; Kliber and Eckert, 2004). Despite this, the parts in a flower are invariably better co-ordinated than the flowers in an inflorescence, and therefore it seems justified to consider flowers as the basic units of reproductive investment. The cost of building a flower can be, at least in theory, broken down into components that correspond to specific functions: the cost of properly positioning the flower on the plant pertains to the pedicel; that of protecting flower buds to the calyx; in animal-pollinated species, advertisement costs should be assigned to the corolla; that of rewarding pollinators to nectar; that of capturing and transporting pollen grains respectively to the stigma and style, and so on. This view holds reasonably well when floral parts have clear-cut roles, but cost breakdown can be very difficult to carry out in particular cases. For example, a coloured calyx can contribute both to protection and display; pollen grains may perform their primary role as carriers of male gametes and also act as a reward; and pollination success may benefit from protection/advertisement provided by organs that are not strictly floral, such as bracts. Furthermore, there are indications that static measures of reproductive costs, such as flower biomass, may not always reflect the exact costs to the plant, since dynamic processes associated with reproduction can occur [e.g. photosynthesis can be carried out by some flower parts (Bazzaz et al., 1979; Werk and Ehleringer, 1983), or nutrients can be reabsorbed from senescing flowers (Ashman, 1994)]. Despite these caveats, and even though unravelling costs into discrete components can be difficult in practice, biomass allocation is probably a worthwhile approach to attempt to gain insights into the interplay between form and function in flowers, although it must be kept in mind that it involves many simplifications and approximations. Furthermore, and as demonstrated by manipulative experiments (Andersson, 2006), biomass measurements can be informative regarding the amount of resources allocated to other reproductive functions.
Water is an essential resource for flower development and function that is often overlooked by studies of resource investment to floral display. Drought stress has been held responsible for the persistence of smaller-flower genotypes in natural populations of Polemonium viscosum and Ipomopsis aggregata (Campbell, 1991; Galen, 1999, 2000), and extensive variations in flower size coupled to long-term precipitation averages have been reported in the sclerophyllous shrub Rosmarinus officinalis (Herrera, 2005). If drought contributes to destabilize floral phenotypes in general, flowers of varying sizes within species should not be rare but a widespread phenomenon in relatively arid places. Mediterranean environments are good candidates to test this hypothesis, as precipitation is highly seasonal and has important year-to-year variation.
The present work addresses the interplay between flower biomass and display in a group of insect-pollinated Mediterranean plants, using both intraspecific and interspecific comparisons. Specific questions are: what is the allometric relationship of flower biomass and linear size across species? Within species, is intrafloral investment as constant as most aspects of flower form, or do conspecifics and populations often show significant quantitative variation? Are relative floral costs constrained by floral architecture, underlying phylogeny, or plant life form? Surveyed species were ecologically and taxonomically diverse, so the results of the present study are expected to apply to the more general case of plants living in semi-arid places.
The study was conducted in the lower Guadalquivir River valley and adjoining mountain ranges in western Andalucia, southern Spain. The climate is typically Mediterranean in this region, with average annual precipitation ranging from 500 to 2000 mm depending on the site. The 18 study localities encompassed all three major habitat types (coast, lowland and mountains up to 1000 m a.s.l.) and were mostly located in nature reserves. The vegetation consisted of sclerophyllous Mediterranean forest dominated by Quercus rotundifolia, Q. suber, Q. faginea or Pinus pinea, as well as scrub with species of Cistaceae, Fabaceae (mostly Genisteae) and Lamiaceae. The maximum distance between any two localities was 180 km. Summary climate data for the study area (and major habitats within it) have been published elsewhere (see table 1 in Herrera, 2005).
For several consecutive years, and during the peak flowering season (February–June), study sites were regularly surveyed for entomophilous (from field observations and floral morphology) plant species in full bloom. Taxa were chosen at random with no regard for family or life form, with the only prerequisites that (1) they were relatively abundant locally and (2) their individual flowers were large enough to attract pollinators (i.e. species with tiny florets arranged into compact inflorescences, such as the Asteraceae, were excluded). The resulting sample included 16 shrub species, 16 herbs, five geophytes and one vine, with a total of 38 species in 16 Angiosperm families (see Appendix; nomenclature follows Valdés et al., 1987). In most cases, a single population per species was sampled, but a few regionally abundant taxa were sampled from several sites to assess site-specific variations (see Appendix). In the case of Cistus salvifolius, a widespread shrub, correlations were sought between floral dimensions and ecological characteristics (average precipitation) of eight populations from three major southern Spanish habitat types (coast, lowland and mountains).
Floral traits were determined in 20 (ten if the population was sparse) randomly chosen individuals. Two or three freshly opened flowers per plant were picked, kept refrigerated in separate vials, and taken to the laboratory for analysis within 24 h of collection. Whenever flowers were arranged in inflorescences they were consistently collected from the same region to cancel any positional effects. The sampling protocol (and results from previous studies: Herrera, 2005; Herrera et al., 2008) anticipated random and relatively small within-plant variations, so flower values were averaged on a plant basis and these averages were used for the statistical analyses below (data from individual flowers are available on request). In a few herbs with solitary flowers, each plant was represented by the only flower present on the day of sampling.
The quantitative traits examined were as follows. (1) Flower mass, measured as the biomass of the whole flower, including the pedicel but not accessory structures such as bracts. (2) Corolla mass, exclusive of all other flower parts. (3) Linear size, measured as the maximum linear dimension of the organ responsible for visual attraction, this being either the corolla (sympetalous species) or a single petal (choripetalous species). The dissected organ was evenly spread and flattened between two glass slides and its longest dimension measured to the nearest 0·1 mm with digital calipers. (4) Corolla allocation, defined as the ratio of corolla biomass to overall flower mass. Lastly (5), an estimate was devised of how efficient was each species at transforming biomass into visual conspicuousness. This was simply flower biomass divided by linear size, a species-specific ratio (mg mm−1) hereafter termed the ‘transaction cost’ (a concept borrowed from economics, where it describes costs incurred when making an economic exchange). All masses were determined as fresh weight to the nearest 0·1 mg using an electronic balance.
The fact that fresh organ weights have been used throughout this study deserves some justification. This was done on the basis that water represents a valuable resource in the drought-prone habitats where the research was carried out. If drought contributes to destabilize floral phenotypes, then fresh mass addresses the issue more directly than dry matter weight. Obviously, variations reported below include natural variation in water content, and depicting this was a major goal of the present study. It could be argued that water loss during transportation may obscure fresh biomass variation to an unknown extent, but since samples were kept in sealed tubes and refrigerated this source of error should be minimal.
Basic statistics, across-species variability and correlations among traits were examined, along with the relationships of scale between selected characteristics (log-log transformed). This was done with reduced major-axis (RMA) regression, the preferred method whenever variables are, as here, subject to measurement error (Niklas, 1994; p. 17). Computing the regression coefficient or ‘slope’ for a RMA regression (αRMA, also known as the ‘scaling exponent’) is as simple as dividing the least-squares regression coefficient by the correlation coefficient of the data set (Niklas, 1994; p. 17). This empirically determined scaling exponent can then be compared to the expected value which, provided that the two traits are in the same units, is αRMA = 1. In other words, if the confidence interval around the observed value encloses unity, then the ‘dependent’ variable regresses perfectly on the other, and it can be concluded that there is ‘geometric similitude’ (sensu Niklas, 1994; p. 17) between the traits (i.e. albeit differing in absolute size, measured objects share the same geometry and shape). In the particular case that variables are in different units (e.g. a regression of linear size on flower biomass), dimensional scaling must be used to establish the expected value for αRMA. On the basis that biomass is closely related to flower volume, it was assumed that the ratio of dimensions was mm mm−3 (Niklas, 1994; p. 22), so in this case the expected value for an isometric relationship between linear size and biomass should not be 1, but 1/3 = 0·333.
Univariate and multivariate analyses of variance were used to determine whether study variables responded to qualitative characteristics such as phylogeny (family membership), floral architecture (symmetry), and plant growth form (herbs vs. woody perennials; see Appendix for species' assignments). General linear-model analyses were performed with the GLM module in STATISTICA (Statsoft, 2001) on log-transformed variables (arcsin-square-root-transformed in the case of percentage corolla allocation).
Table 1 shows basic statistics for the floral traits studied. Species' averages for flower biomass varied between 5 mg (Lavandula stoechas) and 200 mg (Iris pseudacorus), and the coefficient of variation across species was accordingly high (>200 %). Linear size was more uniform, both across (CV = 68 %) and within species (Fig. 1). Dissimilarities as large as 20 % in the size of the flowers rarely occurred among conspecifics, whereas in biomass this was commonplace. Achieving 1 mm of display (i.e. the' transaction cost') required from 0·5 to 36 mg of flower mass, depending on the taxon (Table 1). The most uniform quantitative floral feature was percentage corolla allocation (CV = 23 % among species).
Most traits correlated positively and significantly with each other (Fig. 2; significance tests use the restrictive Bonferroni correction). The association between flower and corolla biomass was particularly strong (Pearson r = 0·99, P < 0·001, n = 38), and so the latter was omitted from further analyses below. Also remarkable was the direct, marked relationship between transaction cost and linear size (r = 0·61, P < 0·001), indicative of a steep rise in relative mass investment as flowers became larger and more visible.
Regressing corolla mass on flower mass (log-log-transformed species' averages) yielded αRMA = 1 as the slope for the regression line, so the study species were isometrical from the standpoint of corolla allocation. Geometric similitude was implied, suggesting heavier flowers were just scaled-up versions of lighter ones (and vice versa).
The empirically determined scale exponent for a regression of the longest floral dimension on biomass (log-log-transformed species' averages) was αRMA = 0·433 (95 % confidence interval = 0·356–0·510). As the expected value for an isometric relationship was assumed to be 0·333 in this case (see Methods), the hypothesis of isometry was rejected.
The proportion of flower biomass allotted to the corolla was largely uncoupled from other quantitative traits (see Fig. 2), so an attempt was made to find possible qualitative correlates. Families represented by at least three species in the data set (Boraginaceae, Caryophyllaceae, Cistaceae, Fabaceae, Lamiaceae and Scrophulariaceae) were used to test the hypothesis that quantitative floral features depended on family affiliation, with the latter being treated as the main factor (Table 2). In general, family identity had a significant effect on quantitative traits in the MANOVA, although only corolla allocation was significantly family-dependent according to the univariate ANOVA.
Figure 3 depicts the least-square mean values for corolla allocation and pairwise comparisons for the six species-rich families. On average, Fabaceae allotted more to the corolla (70 %) than Cistaceae (50 %) and Caryophyllaceae (45 %). Flowers are elaborately three-dimensional in the Fabaceae whereas simple dish-bowl flowers prevail in the Cistaceae and Caryophyllaceae, so allocation to the corolla might relate to floral architecture or complexity. This was tested in the whole sample of 38 species (not just species-rich families) using the radial or bilateral symmetry of flowers as an indicator of architectural elaborateness and specialization in pollination. The results of the MANOVA in Table 3 show an overall effect of main factor ‘symmetry’ on quantitative floral traits; once again the only feature significantly dependent on symmetry was corolla allocation (univariate ANOVAs). On average, bilateral flowers allocated more to the corolla than radially symmetrical flowers (64·8 ± 2·6 % and 52·8 ± 2·9 %, respectively; n = 18, 20, P < 0·01), and the effect was still significant after excluding the Fabaceae from the analysis (F = 4·661, P = 0·04).
The traits that were studied were unrelated to growth form. Herbs and woody perennials were indistinguishable as regards flower biomass, linear size, corolla allocation and transaction costs (MANOVA analysis: Wilk's λ = 0·859, F = 1·358, which is not significant at P = 0·05).
Individual plant data (Fig. 4) give a picture of trait variation that complements that provided by species' averages. Linear size was a direct correlate of flower mass (either within or among species) in three important families with contrasting floral architectures, with clusters of points (i.e. species) that were often distinct within each family. Conspecifics were usually more alike in terms of flower size or biomass than plants from different species, although large-flowered Cistaceae were exceptional and often overlapped onto the mass–size bivariate surface. On the other hand, relative allocation to display did not show a pattern of change consistent among families: in Cistaceae and Fabaceae (mostly pollen flowers), corolla allocation increased with absolute flower size/biomass, whereas in Lamiaceae (all nectariferous) it decreased.
Quantitative traits were often population-specific (Table 4). Flower biomass varied from one population to another in seven of nine species tested, and linear size and relative allocation to the corolla did likewise (although with lower data availability).
In the widespread shrub Cistus salvifolius, habitat type (coast, lowland or mountain; a rough correlate of precipitation levels) had a significant effect on flower biomass (ANOVA, F = 39·4, P < 0·001), linear size (F = 8·2, P < 0·001) and corolla allocation (F = 22·2, P < 0·001). As shown in Fig. 5, shrubs from the mountains had flowers than were on average larger, heavier and with greater corolla allocation than their counterparts from drier habitats (coast or lowland).
Most species in this study had medium-sized flowers (5–65 mm; median 17 mm) despite the fact that those with very compact inflorescences were excluded. This is in accordance with the modest rewards most of these plants provide on a per-flower basis (Herrera, 1985), and also with the average size of their pollinators. Small- and medium-sized Apidae, Anthophoridae and Halictidae dominate the pollinator fauna in the region (body lengths from 6 to 13 mm; see table 3 in Herrera, 1988) whereas large bees (e.g. Bombus, Xylocopa) and butterflies are comparatively uncommon. Therefore, in general terms, the study system can be described as one of small- to medium-flowered plants that rely necessarily (Herrera, 1987) on small- to medium-sized bees for reproduction.
Flower biomass varied more than linear size across species (CV >200 % and 68 %, respectively), and one may wonder if this was just a byproduct of the contrasting floral architectures represented in the study sample. If, for example, a substantial fraction of the taxonomically heterogeneous species lacked a calyx (presumably a very actively transpiring whorl), then high variation in (fresh) flower biomass would simply reflect differences in transpiration rates among whorls. Yet most of the taxa studied had a photosynthetic calyx, so this is unlikely to be the case.
Furthermore, biomass also varied more than linear size within species (Fig. 1). This has the interesting implication that conspecific individuals achieved appreciable morpho-functional constancy in their reproductive modules despite relatively unsteady allocation of resources (including water) to their flowers. Fresh weights were used throughout the study, so part of the variation in biomass could be due to varying turgidity and/or water content (either as a result of spatially varying soil moisture and/or past natural selection), which obviously would require further research. For example, it would be interesting to know whether variations in mass and linear size are paralleled by variations in cell number, cell size, or both. Galen et al. (1999) found evidence that corolla length correlates positively with cell size but not cell number in Polemonium viscosum, whereas Delph et al. (2004) reported the opposite in Silene latifolia.
The analysis showed that 1 mm of floral display required investing a biomass of 0·5–36 mg, and the ‘transaction cost’ increased with increasing floral linear size. A reasonable proximate cause for this is that flower mass scales as the cube power of flower length, so biomass necessarily will rise at a faster rate than display. In other words, due to allometry it is unavoidable that large, conspicuous flowers pay disproportionately higher costs in terms of biomass (Niklas, 1994). From a more functional perspective, structural elements that provide stiffness are usually essential in large-flowered species, whereas this expenditure can be saved if the flowers are small. Thus, miniaturizing the flower can result in more efficient biomass resource use (and also result in synergistic effects among numerous flowers to attract pollinators due to size/number trade-offs; Andersson, 1999, 2000, 2006).
On average, the species studied allocated about half of their fresh floral matter to the corolla, a proportion commonly found in many entomophilous angiosperms (Lovett Doust and Cavers, 1982). It seems safe to assume that pollinator attraction represented the most expensive of all (pre-pollination) floral functions. On the other hand, and in terms of biomass allocation to the corolla, small-flowered species were effectively just scaled-down versions of larger-flowered taxa. This, and the observation that it was family-dependent, seems to indicate that the proportion of biomass allotted to the corolla might be a highly conserved trait [unsteady corolla allocation was observed only in species with extremely reduced flowers (e.g. <1 mg; Fig. 4), and most likely relates to pollination systems in which insect visitation has become relatively unimportant (e.g. facultatively autogamous Tuberaria guttata) or advertisement is no longer performed by the corolla (showy bracts of Lavandula stoechas)]. There is growing evidence that floral parts may not be able to evolve independently of each other due to pleiotropic effects (e.g. Elliott et al., 1996; Krizek, 1999; Delph et al., 2004; Conner, 2006), so evolutionary modifications in the size of the corolla could easily lead to parallel changes in the remaining parts. This would allow flowers to maintain a similar pattern of allocation across a wide range of sizes.
Recent resource/cost hypotheses postulate that reduced corollas can be advantageous for plants living under prevailingly stressful conditions. In Polemonium viscosum, for example, large corollas incur physiological costs because of their greater water uptake (Galen, 1999, 2000). In Epilobium angustifolium (Carroll et al., 2001) plant water status and corolla size are directly related, and Herrera (2005) reported a direct relationship between altitude/rainfall and flower biomass in the sclerophyllous shrub Rosmarinus officinalis (which seems also to be the case for Cistus salvifolius shrubs in the present study). Most southern Spanish sclerophyllous species are massive spring bloomers that take advantage of a short transient window with mild temperatures, moist soil and good insect availability prior to the onset of the harsh Mediterranean summer, and in drought years it is not uncommon that many bloom poorly, if at all. In this ecological context, and as flower production surely represents a major expense for these plants, even the smallest individual differences in floral allocation can become important. Local selective pressures to save resources probably contribute to maintain high levels of intraspecific floral variation (at least across populations), and maybe also to a general reduction in the size of southern Spanish entomophilous flowers.
I thank Dr Stefan Andersson and an anonymous reviewer for comments and suggestions. Dr Peter E. Gibbs kindly revised the English. This work was financially supported by the Spanish Plan Andaluz de Investigación (RNM204) of the Junta de Andalucía, and grant BOS2000-0328 of the Dirección General de Enseñanza Superior e Investigación Científica.
|Family/Ref. no., species (growth form, symmetry)*||No. populations studied||Flower mass (mg)||Corolla mass (mg)||Corolla size (mm)|
|Mean (s.e.)||n||Mean (s.e.)||n||Mean (s.e.)||n|
|01, Leucojum trichophyllum (H, R)||1||87·8 (4·1)||26||75·3 (5·3)||14||18·0 (0·6)||14|
|02, Anchusa calcarea (H, R)||1||35·6 (1·2)||20||17·7 (0·8)||20||14·2 (0·4)||20|
|03, Cerinthe gymnandra (H, R)||1||75·1 (2·2)||21||49·7 (1·6)||21||20·0 (0·1)||21|
|04, Echium plantagineum (H, B)||1||110·1 (3·1)||21||72·0 (2·8)||21||28·9 (0·4)||21|
|05, Brassica barrelieri (H, R)||1||16·2 (0·8)||20||8·4 (0·6)||10||12·0 (0·4)||10|
|06, Malcolmia lacera (H, R)||1||14·5 (0·7)||20||9·9 (0·9)||10||14·5 (0·3)||20|
|07, Lonicera implexa (W, B)||1||93·5 (6·5)||10||78·2 (5·2)||10||43·4(1·5)||10|
|08, Dianthus broteri (H, R)||1||210·9 (6·1)||20||88·0 (3·6)||20||49·1 (0·8)||20|
|09, Dianthus inoxianus (H, R)||1||378·7 (40·7)||10||172·4 (18·6)||10||51·0 (1·2)||10|
|10, Silene colorata (H, R)||1||70·2 (5·7)||20||33·5 (2·6)||20||17·7 (0·7)||20|
|11, Cistus albidus (W, R)||7||341·7 (9·5)||140||192·3 (9·4)||12||28·1 (0·8)||53|
|12, Cistus ladanifer (W, R)||1||843·0 (37·5)||11||525·0 (22·9)||11||40·7 (0·9)||11|
|13, Cistus salvifolius (W, R)||8||314·3 (5·7)||143||177·2 (6·1)||53||24·7 (0·3)||69|
|14, Halimium commutatum (W, R)||1||25·5 (1·0)||20||10·1 (0·6)||20||13·1 (0·2)||20|
|15, Tuberaria guttata (H, R)||1||7·4 (0·4)||20||3·1 (0·2)||20||8·3 (0·2)||42|
|16, Erica ciliaris (W, B)||1||20·4 (0·5)||20||9·8 (0·3)||20||10·2 (0·1)||20|
|17, Erica umbellata (W, R)||1||5·7 (0·1)||26||2·9 (0·2)||6||4·8 (0·1)||6|
|18, Chamaespartium tridentatus (W, B)||1||28·4 (0·8)||20||20·4 (1·0)||10||12·2 (0·3)||10|
|19, Cytisus grandiflorus (W, B)||2||143·0 (3·9)||44||112·2 (5·0)||16||21·2 (0·4)||20|
|20, Genista hirsuta (W, B)||1||17·9 (0·5)||38||11·6 (0·5)||8||10·7 (0·3)||10|
|21, Genista triacanthos (W, B)||2||8·1 (0·2)||40||5·4 (0·1)||10||5·8 (0·1)||7|
|22, Lupinus luteus (H, B)||1||61·7 (1·7)||10||43·3 (1·3)||10||16·2 (0·2)||10|
|23, Gladiolus italicus (H, B)||1||436·8 (21·4)||20||349·3 (17·0)||20||49·0 (1·2)||20|
|24, Iris pseudacorus (H, R)||1||2371·9 (74·6)||10||979·6 (37·1)||10||64·6 (1·7)||10|
|25, Cleonia lusitanica (H, B)||1||24·7 (1·0)||20||15·8 (0·8)||20||27·9 (0·5)||20|
|26, Lavandula stoechas (W, B)||3||4·5 (0·1)||40||2·9 (0·1)||20||8·4 (0·2)||20|
|27, Phlomis purpurea (W, B)||6||127·8 (1·9)||132||67·8 (1·6)||32||22·8 (0·2)||32|
|28, Rosmarinus officinalis (W, B)||9||20·9 (0·3)||237||15·5 (0·6)||10||10·6 (0·1)||106|
|29, Teucrium fruticans (W, B)||5||68·2 (1·6)||84||40·5 (1·3)||20||12·6 (0·2)||20|
|30, Asphodelus ramosus (H, R)||1||125·7 (3·8)||23||58·2 (1·7)||23||16·7 (0·2)||23|
|31, Myrtus communis (W, R)||7||86·0 (2·3)||144||33·2 (3·6)||27||9·0 (0·3)||52|
|32, Serapias lingua (H, B)||1||178·7 (7·4)||20||129·7 (4·3)||20||24·7 (0·6)||20|
|33, Anagallis monelli (H, R)||1||15·8 (0·9)||10||10·2 (0·6)||10||9·2 (0·3)||10|
|34, Anemone palmata (H, R)||1||225·2 (10·0)||20||156·7 (7·2)||20||18·8 (0·4)||20|
|35, Ranunculus bulbosus (H, R)||1||101·8 (7·5)||14||39·1 (2·7)||14||12·3 (0·4)||14|
|36, Linaria viscosa (H, B)||1||21·8 (0·7)||20||16·2 (0·5)||20||20·0 (0·4)||20|
|37, Scrophularia frutescens (W, B)||1||7·9 (0·3)||19||4·2 (0·2)||19||5·8 (0·1)||19|
|38, Verbascum barnadesii (H, B)||1||313·3 (6·0)||20||179·5 (4·1)||20||23·7 (0·3)||20|
* H, herb; W, woody perennial; R, radial flowers; B, bilateral flowers.