Flowers adaptively face down-slope in 10 forest-floor herbs


†Author to whom correspondence should be addressed. E-mail:


  • 1An animal-pollinated plant living on a slope should orientate its flowers down-slope towards the more open space if by doing so it receives more pollinator visits and thereby achieves increased reproductive success.
  • 2We measured flower orientation relative to slope direction on individuals of 10 species of forest-floor herbs in cool temperate forests in Japan. For one of these species, Erythronium japonicum, we also manipulated flower orientation to test experimentally for its effects on both male and female reproductive function.
  • 3In all 10 species, flowers were preferentially orientated down-slope. This pattern was more pronounced in plants growing on steeper slopes.
  • 4Our manipulative field experiment in Erythronium japonicum demonstrated that pollen dispatch was highest in flowers orientated down-slope. Additionally, flowers orientated up-slope may have achieved a lower seed set on steep slopes.
  • 5We conclude that down-slope orientation of flowers was a general phenomenon among the species that we studied, and that this behaviour was adaptive in enhancing plant fitness through pollination.


Because they are sessile, most flowering plants require pollen vectors for sexual reproduction. Animal-pollinated flowers attract their pollinators from a distance using showy petals and/or fragrance. Attraction size, which involves both flower size and number, often affects the frequency of pollinator visitation (de Jong & Klinkhamer 1994; Conner & Rush 1996; Harder & Barrett 1996; Ohashi & Yahara 2001), indicating that pollinators use showy petals as a kind of flower trademark. To make flowers more attractive for pollinators, the orientation of attraction-related petals towards the pollinators would be important to some species (Patino et al. 2002). In this sense, flower orientation is one aspect of floral attraction.

Flower orientation is considered to have evolved under pollinator-mediated selection and/or under natural selection due to abiotic environmental factors, such as rainfall and temperature (Hocking & Sharplin 1965; Kevan 1975; Kudo 1995; Neal et al. 1998; Giurfa et al. 1999; Huang et al. 2002; Galen & Stanton 2003; Ushimaru & Hyodo 2005; see also Patino et al. 2002). Flower orientation includes two dimensions: flower angle (Neal et al. 1998; Tadey & Aizen 2001; Huang et al. 2002; Ushimaru & Hyodo 2005) and flower direction (Kudo 1995; Zavala-Hurtado et al. 1998; Kita & Wada 2000; Patino et al. 2002). Flower angle is the vertical direction of orientation that the flower faces from its point of attachment to the plant (e.g. upwards, vertical, oblique or downwards), whereas flower direction is the geomagnetic direction that the flower faces (e.g. east, south, west, north). In some species, flower angle and direction are closely related. For example, angles of heliotropic flowers are usually upwards (or inclined) to increase internal flower temperature, which in turn promotes pollinator attraction, pollen tube growth and seed maturation (Hocking & Sharplin 1965; Kevan 1972, 1975; Kudo 1995; Patino et al. 2002; Galen & Stanton 2003).

A plant on a slope experiences a heterogeneous spatial environment (see Ishii & Higashi 1997). When a plant grows vertically, a larger space occurs on the down-slope compared with the up-slope side of the individual (Fig. 1). Assuming that pollinators are distributed uniformly in space, we expected that the number of pollinators distributed on the down-slope side of individuals would be larger than that on the up-slope side. Thus, the probability of pollinators approaching the flowers would be higher from the down-slope side than from the up-slope side, especially on a steeper slope. If a plant with a vertical flowering stem has flowers that face toward a pollinator-abundant open space to increase pollinator visitations, we can predict the following flower-orientation patterns on a slope: most flowers will face down-slope, and this pattern should be most common on steeper slopes; moreover, flowers that face up-slope will have lower reproductive success.

Figure 1.

Measurements of flower direction and slope direction and angle. The angular distance between the slope and flower direction (ADSF) was calculated as in this figure.

These patterns are more likely to be observed in plants with vertical and oblique oriented flowers because these flowers are considered to restrict the approach direction of visiting insects, i.e. pollinators should always approach these flowers from the front (Neal et al. 1998; Ushimaru & Hyodo 2005). Furthermore, plants with flowers near the ground should exhibit these patterns more strongly than flowers on taller plants because flowers growing at greater heights may be less influenced by spatial heterogeneity on a slope.

In this study, we examined the above predictions of the relationship between slope conditions and flower direction in 10 herbaceous species with vertical or oblique flowers to test and to generalize our hypothesis. We then examined our prediction on the relationship between flower direction and reproductive success in a natural Erythronium japonicum Dence. population by conducting a field experiment in which flower direction was artificially manipulated. Based on our results, we discuss the evolution of flower orientation in angiosperms on slopes.

Materials and methods

study species

We examined flower orientation on slopes in the following 10 species that have vertical or oblique-oriented flowers: E. japonicum, Hosta sieboldii (Paxton) J. Ingram (Liliaceae), Asiasarum sieboldii (Miq.) F. Maek. var sieboldii (Aristlochiaceae), Viola eizanensis (Makino) Makino (Violaceae), Shortia uniflora (Maxim.) Maxim. var. kantoensis Yamazaki (Diapesiaceae), Corydalis lineariloba Sieb. et Zucc. var. lineariloba (Papaveraceae) in the Ogawa Forest Reserve (OFR) and Calanthe reflexa Maxim. (Orchidaceae) in the Ashu Experimental Forest and Monotropastrum globosum H. Andres at Mt. Higashiyama, M. globosum H. Andres f. roseum Honda at Mt. Kirishima and Monotropa uniflora L. at Yada Hills and Iwakura (Table 1; data of these Monotropaceae species are after A. Imamura & A. Ushimaru, unpublished). Species were deliberately selected to include both monocots and dicots and both actinomorphic and zygomorphic flowers. Individuals of E. japonicum, A. sieboldii and S. uniflora var. kantoensis have a single-flowered inflorescence, whereas those of V. eizanensis, M. globosum H. Andres, M. globosum f. roseum and M. uniflora often have several (sometimes more than five) single-flowered inflorescences. Calanthe reflexa and C. lineariloba have many flowers (usually more than five) per inflorescence. H. sieboldii usually has one or two opening flowers per inflorescence.

Table 1.  List of plant species that were investigated
Family speciesFlower angleFloral symmetryaSiteLocationNo. of slopesSlope directionbNo. of flowers (individuals)Date
  • a

    Ac = actinomorphic, Zy = zygomorphic.

  • b

    North = 0° (360°), east = 90°, south = 180°, west = 270°.

 Erythronium japonicumDeclineAcOgawa Forest Reserve36°56′ N, 140°35′ E13 30–350°146 (146)April 2002
 Hosta sieboldiiVerticalZyOgawa Forest Reserve36°56′ N, 140°35′ E 1 80–120° 33 (30)August 2002
 Calanthe reflexaVerticalZyAshu experimental forest35°18′ N, 135°43′ E 2 20–300°170 (36)August 2001
Aristlochiaceae Asiasarum sieboldii var. sieboldiiIncline–declineAcOgawa Forest Reserve36°56′ N, 140°35′ E 3  0–340° 66 (66)May, 2002
Papaveraceae Corydaris lineariloba var. linearilobaVerticalZyOgawa Forest Reserve36°56′ N, 140°35′ E 4 50–290°150 (30)April 2002
Violaceae Viola eizanensisVerticalZyOgawa Forest Reserve36°56′ N, 140°35′ E 3 20–280° 50 (32)April 2002
Diapensiaceae Shortia uniflora var. kantoensisVerticalAcOgawa Forest Reserve36°56′ N, 140°35′ E 1320–10° 30 (30)April 2003
Monotropaceae   Yada Hills34°40′ N, 135°43′ E 1 70–350° 38 (26)September 2000
Monotropa unifloraVertical–declineAcIwakura-hanazono-cho35°04′ N, 135°48′ E  1 30–60°  5 (3)September 2000
 Monotropastrum globosumVertical–declineAcMt. Higashiyama34°59′ N, 135°47′ E 3  0–360°243 (73)April 2001
 M. globosum f. roseumVertical–declineAcMt. Kirishima31°55′ N, 130°50′ E 1110–350°111 (55)June 2002

For E. japonicum, we examined in further detail the relationship between flower orientation and slope; we therefore provide some further details on its floral biology as follows. E. japonicum occurs on the forest floor in cool temperate forests of Japan. This perennial herb rarely exhibits clonality and reproduces sexually by ant-dispersed seeds (Ohkawara et al. 1996). In April, a sexual individual of E. japonicum produces a single flowering shoot, which has a pinkish, bisexual flower that lasts c. 2 weeks (Ishii & Sakai 2000). A flower bud, which emerges with two leaves, faces toward just above the scape and then its peduncle bends to orient the flower face obliquely. The species is self-incompatible and requires insect vectors for pollination (Ishii & Sakai 2000). This species is mainly pollinated by relatively large bees, such as Xylocopa, Tetralonia, Nomada and Bombus species (Utech & Kawano 1975). However, we have observed that E. japonicum flowers are most frequently visited by small bees in the OFR (Ushimaru et al. 2003). Flowers are adichogamous, and pollen receipt is saturated within 3 days after corolla opening in E. japonicum, whereas pollen dispatch continues for 10 days (Ishii & Sakai 2000). Fruits of this species mature approximately 1 month after anthesis. In 2001, we conducted an experimental study on an E. japonicum population in an old-growth deciduous forest in the OFR. The blooming season began in late March and lasted until late April.

measurements of slope and flower conditions

Herbaceous plants are probably affected by microscale sloping landforms. We therefore measured slope direction and angle in the microhabitat (20 cm × 20 cm) of each individual and flower orientation for each flower using a clinometer (Showa Sokki Co., Tokyo, Japan; Fig. 1). We then calculated the angular distance between the slope and flower directions (ADSF; 0° = ADSF = 180°) for each flower (Fig. 1). We measured two to five flowers from the bottom of the inflorescence and five flowers from the top of the inflorescence in C. reflexa and C. lineariloba, respectively (we selected early blooming flowers of each inflorescence). When an individual had more than one flower, we calculated the mean ADSF of flowers as the ADSF for individuals of V. eizanensis, H. sieboldii, C. lineariloba, C. reflexa and Monotropaceae species.

First, flowers were categorized into four ASDF classes: A, ADSF less-than equal-to or greater-than 45°; B, 45° < ADSF less-than equal-to or greater-than 90°; C, 90° < ADSF less-than equal-to or greater-than 135°; and D, 135° < ADSF less-than equal-to or greater-than 180°. We counted the number of individuals in each class and compared the natural distribution pattern of ADSF to a uniform distribution using a chi-square test for each species. Our hypothesis predicts that the number of individuals is highest in class A and lowest in class D. We then examined correlations between slope angle and ADSF to test our hypothesis in 10 species.

relationship between adsf and female reproductive success in erythronium japonicum

We investigated the relationship between reproductive success and natural variation in ADSF in 146 flowers whose ADSF was measured in the above examination. One- or 2-day-old flowers were tagged on 13 April 2002. We checked fruit set for 146 tagged flowers on 11 May 2002. The relationship between ADSF (0–180 in the field) and fruit set was examined using a logistic regression analysis with fruiting (1) and non-fruiting (0) as a dichotomous response. We also collected fruits of these flowers, and the number of seeds was counted for each fruit in the laboratory. The relationship between the number of seeds per fruit and ADSF was examined using simple linear regression.

direction-change experiment in erythronium japonicum

To experimentally examine the effect of down-slope flower orientation on female reproductive success in E. japonicum, we conducted a direction-change experiment in the OFR. We arbitrarily chose and tagged 108 flowers whose ADSFs were < 20° on five slopes (one north-, two east-, one south- and one west-facing slope). The direction change to up-slope orientation was achieved by fixing the flower stalk with a wire. Flower stalks of controls were also fixed with wires. We changed the ADSFs of 56 flowers to more than 160° (up-slope direction) and left the remaining 52 flowers intact as controls on 1–3 April 2002. The slope angle of each flower was also measured. We monitored fruiting and seeding of these 108 flowers to check female success.

The effect of direction change on male reproductive success was also examined. Male reproductive success has often been measured by counting the number of pollen grains remaining on anthers, which is assumed to be negatively correlated with siring success (Ishii & Sakai 2000). We collected a single anther in each flower from 29 and 27 control and 30 and 29 direction-changed flowers at 3 and 10 days after anthesis, respectively. Single anthers from 20 newly opened flowers were also sampled. These anthers were stored in 1·0 ml of 70% ethanol. Pollen grains easily detached from anthers in the solution. We estimated the number of pollen grains per anther by counting the pollen number in a 5·0-µl drop under a microscope three times per solution. We multiplied the estimated number by six to calculate the number of remaining pollen grains per flower.

We tested the effects of the treatment and slope angle on fruit set using a generalized linear model (GLM) with binomial errors and a logistic link function, in which experimental treatment (control and direction changed), slope angle and their interaction were independent variables. The effect of treatment, slope angle and their interaction on the number of seeds per fruit was examined using a GLM with Gaussian errors. We also used a GLM with a Gaussian error structure to test the effect of treatment and days after anthesis (3 and 10 days) on the number of pollen grains remaining per flower. In this model, we did not use the slope angle and interactions between the slope angle and other variables as explanatory variables, according to the result of preliminary analyses based on model selection using the Akaike Information Criterion (AIC).


distribution of adsf and relationship between adsf and slope angle in herbs

We found a general trend among 10 herb species that most flowers faced down-slope. In E. japonicum, A. sieboldoii, S. uniflora, V. eizanensis and H. sieboldii, 50·0–76·7% of flowers faced down-slope within a 45° deviation (Fig. 2). These distribution patterns differed significantly from a uniform distribution. Most C. reflexa, C. lineariloba and Monotropaceae individuals had a < 90° ADSF of the inflorescence (Fig. 2), which also differed significantly from the uniform distribution. Thus, species whose individuals always had one or two flowers showed a strong down-slope orientation, whereas others showed weak trends.

Figure 2.

Frequency distribution of the angular distance between the slope and flower direction (ADSFs) for 10 forest herb species. Flowers were categorized into four ASDF classes: ADSF less-than equal-to or greater-than 45° (A), 45° < ADSF less-than equal-to or greater-than 90° (B), 90° < ADSF less-than equal-to or greater-than 135° (C) and 135° < ADSF less-than equal-to or greater-than 180° (D).

Although the ADSF of single flowers or the ADSF of the individual varied largely for each species, they were negatively correlated with slope angle in all species (Fig. 3). For all species, flowers (plants) with an ASDF > 90° were rarely found on steep (> 25° angle) slopes (Fig. 3).

Figure 3.

Relationship between slope angle and the ADSF in 10 forest herb species. For multiflowered individuals, plots indicate the ADSF of an individual. The sampling number is shown in parentheses.

reproductive success in erythronium japonicum on slopes

About 65% of tagged flowers (96 flowers) produced fruits in the field. Logistic regression analysis revealed no significant relationship between the ADSF and fruiting success (df = 1, χ2 = 0·863, P = 0·35). The average seed set for 146 tagged flowers was 30·2 ± 6·68 (SE). The number of seeds per fruit was not significantly related to the ADSF (coefficient = −0·014, r = 0·000, P = 0·95).

The percentage of fruiting flowers was 66·7% and 73·0% for direction-changed and control flowers, respectively. The mean number of seeds per fruit was 19·8 ± 18·4 (SD) for direction-changed flowers and 24·1 ± 18·6 for control flowers. GLM analyses revealed that the experimental treatment had a marginally significant effect on fruit set and the number of seeds per fruit (Table 2). In contrast, the interaction between treatment and slope angle significantly affected both fruit set and the number of seeds per fruit (Table 2, Fig. 4). This indicates that fruit and seed set decreased with increasing slope angle in direction-changed flowers, but not in control flowers (Fig. 4).

Table 2.  Effects of experimental treatment, slope angle and their interaction on female reproductive success analysed and the effects of treatment and days after anthesis on male reproductive success by generalized linear models (GLM)
 Response factor coefficientStandard errorz-valueP
Fruit set
Treatment 3·0071 1·5435 1·95 0·051
Slope angle 0·0450 0·0444 1·01 0·311
Treatment × slope angle−0·1230 0·0566−2·17 0·030
Intercept−0.1719 1.1725−0·15 0·883
No. of seeds per fruit
Treatment19·806211·7399 1·69 0·092
Slope angle 0·4591 0·3275 1·40 0·161
Treatment × slope angle−0·9172 0·4249−2·16 0·031
Intercept11·8168 9·0896 1·30 0·194
Remaining pollen number
Treatment−4 307·81992·2−2·16 0·031
Days after anthesis−863·6 284·6−3·03 0·002
Intercept17 209·82300·5 7·48< 0·001
Figure 4.

Relationship between slope angle and the number of sired seeds per fruit in experimental Erythronium flowers: direction-changed flowers, black circles and black regression line; control flowers, open circles and grey line. The number of seeds per fruit significantly decreased with increasing slope angle in direction-changed flowers (see text and Table 2).

Direction-changed flowers contained more pollen grains than controls both at 3 days and 10 days after anthesis, and the number of remaining pollen grains decreased over time (Fig. 5). Both experimental treatment and days after anthesis had significant effects on the number of remaining pollen grains (Table 2). This indicates that pollen dispatch was lower in direction-changed flowers throughout anthesis. At 3 days after anthesis, 23·1% of the pollen grains remained on the anthers in direction-changed flowers, while 13·8% remained on control flowers.

Figure 5.

Effect of direction-change treatment on pollen dispatch in Erythronium japonicum. The open circle indicates the mean number of pollen grains on intact flowers. The black circle and square indicate the number of remaining pollen grains on direction-changed (up-slope-oriented) flowers and on control (down-slope-oriented) flowers, respectively.


trends in flower orientation of herb species on slopes

Flowers of all investigated herb species exhibited a downslope flower orientation independent of slope direction. We rarely observed flowers whose angular distance between the slope and flower directions (ADSF) was more than 135°. This result verifies our predictions that most flowers face down-slope and that this trend is common among forest-floor herbs, although some deviations (more than 30% of individuals had ADSFs > 45°) were found in every species.

These deviations could be explained by slope angle. On gentle slopes, ADSF varied largely, but flowers rarely had ADSFs > 90° on steep slopes (angle greater-than equal-to or less-than 25°). We found a significant negative correlation between slope angle and ADSF in seven species. This result is concordant with the prediction that down-slope orientation should be most common on steeper slopes. Thus, our field data support the generality of our hypothesis that flower direction of vertical or oblique oriented flowers is influenced by the slope direction and angle. Natural variation in ADSF did not affect female success in 146 intact E. japonicum flowers in the field, suggesting that the natural ADSF may be more or less adapted to each microslope environment. More than half of the C. reflexa and C. linearioba individuals had ADSFs > 45°. These species have many flowers on each inflorescence, so that early blooming flowers may reduce free space for other flowers of the same inflorescence, i.e. the down-slope side is not available for late-blooming flowers. Furthermore, on gentle slopes, the orientation varied substantially among flowers in the same inflorescence. On a flat plain, a multidirectional orientation would be more attractive than a single-directional orientation for species with many flowers because pollinator approaches can be expected equally from every direction on flat land, unlike on slopes. The adaptive significance of multidirectional flower orientation on flat ground should be investigated in light of the evolution of floral display.

effect of down-slope orientation on reproductive success in erythronium japonicum

Our field experiment revealed that the direction-change treatment (towards up-slope) did not strongly reduce female success in E. japonicum. However, we found a significant effect of the interaction between the direction change and slope angle on female reproductive success, such that up-slope flower orientation significantly decreased fruiting and seeding as the slope steepness increased, whereas no such trend was found for down-slope-oriented flowers. This means that the disadvantage of up-slope orientation increases with slope angle. In E. japonicum, about 11 pollen grains per stigma are enough to produce a fruit (Ishii & Sakai 2000). It is likely that up-slope-oriented flowers rarely received pollen from pollinators on steep slopes.

Pollen dispatch was significantly lower in up-slope-oriented flowers than in down-slope-oriented flowers. The direction-change treatment also diminished male reproductive success in Erythronium flowers on the slopes. Our experimental results are consistent with our prediction that flowers facing up-slope should have lower reproductive success. Thus, the down-slope flower orientation of E. japonicum may be an adaptation to spatial heterogeneity produced by slope environments.

down-slope flower orientation

In this study, we showed that down-slope orientation was a general phenomenon among herbs with vertical or oblique flowers and that this trend was more conspicuous on steeper slopes. The field experiment revealed that down-slope flower orientation influenced male and possibly female success in E. japonicum. These results together support our postulate that down-slope orientation in vertical (or oblique) flowers has evolved to enhance pollen receipt and dispatch mediated by pollinators under heterogeneous spatial conditions on slopes. Down-slope orientation in vertical flowers would facilitate pollinator visits in several ways. In terms of finding flowers, a down-slope orientation may increase the attractiveness of a given flower by facing toward open space, consequently increasing pollinator visitations. In terms of approaching flowers, a vertical flower orientation limits the pollinator approach course (Neal et al. 1998; Ushimaru & Hyodo 2005), which may be secured by down-slope orientation. A change in flower orientation may affect the approach course and, consequently, reduce the attractiveness of a given flower because the advertisement area recognized by pollinators can be affected by the approach course (Dafni 1994). A down-slope flower orientation may also help pollinators assess the flower and facilitate landing. Pollinators usually land on flowers after they assess pollen and/or nectar availability (see Lunau 2000). An up-slope orientation may make it more difficult for pollinators to access anthers and other floral organs, so that they cannot determine food availability on flowers. This would lead to a rejection of flowers by pollinators before landing. Furthermore, because landing points are controlled by vertical orientation (Ushimaru & Hyodo 2005), a down-slope orientation may make landing points easily accessible by pollinators. These behaviours were not examined in our experiment because of infrequent pollinator visitations to Erythronium flowers. Future studies should attempt to examine this issue using flowers with more frequent pollinator visitations. In particular, more attention should be paid to whether attraction size (advertisement area) is affected by flower orientation on the slope. Most heliotropic flowers attract pollinators using heat as a floral reward (Hocking & Sharplin 1965; Kudo 1995), and there is evidence that flower orientation may affect flower temperature (e.g. Patino et al. 2002) and hence pollinator attraction.

Another question remains to be answered; i.e. how do flowers face down-slope? Kita & Wada (2000) reported that E. japonicum tends to orient its flowers toward brighter areas (e.g. toward canopy gaps) in a forest on a flat plain. Artificial shading on the down-slope side changes flower direction up-slope in some individuals of M. globosum (A. Imamura & A. Ushimaru, unpublished data). These results suggest that flower orientation is determined using a light cue. Thus, flowers may orient toward brighter areas, in this case, down-slope.


The authors thank Hiroshi S. Ishii and Kensuke Nakata for their advices to accomplish field works and statistical analyses, James E. Cresswell and anonymous reviewers for their critical comments and suggestions during reviewing process and Shoko Sakai, Akiko Fukui, Masatoshi Yasuda, Chikako Ishida and Mitsue Shibata for their field assistance and Tohru Nakashizuka for his financial support.