1. Most dimorphic angiosperm species produce chasmogamous flowers adapted for allogamy and cleistogamous flowers adapted for autogamy. Plant size is an important internal environmental factor and several studies have suggested an adaptive strategy whereby cleistogamous flowering ensures a basic seed output while larger plants can spare resources to make additional investment in the more expensive but presumably more fit outcrossed seed produced by chasmogamous flowering. This study tests the effect of plant size on the percentage of chasmogamy in a species, Mimulus nasutus, where both the cleistogamous and chasmogamous flowers are autogamous and so seed from chasmogamous flowers has no genetic advantage.
2. We measured change in expression of cleistogamy/chasmogamy in glasshouse-grown plants of different sizes and architecture. Size was manipulated by varying the length of time for which the plants were grown under short days. Architecture was manipulated by removing stems.
3. We found that increased size led to an increase in the percentage of chasmogamy in M. nasutus. This is unexpected given the lack of genetic advantage of outcrossing and suggests that this effect can occur as a direct result of increased resource availability without the necessity of selection for progeny fitness.
4. An increase in plant size was also found to lead to an increase in resource sinks (developing flowers and seed capsules on other parts of the plant). As increased plant size can therefore produce both increased total resources and increased number of resource sinks, the net resources available to a developing flower will be a balance of these two effects. Resource re-allocation is shown to occur between flowers on a stem, between stems and between flowers and seed set. In each case, this results in a net decrease in resources to new flowers and so results in there being only sufficient resources for cleistogamous flowering.
5. We discuss the difficulties of using species with varying growth forms to test the hypothesis that expression of allogamous chasmogamy increases with plant size because of the genetic advantage of outcrossing.
It is thought that the original angiosperm condition was one of allogamy and that autogamous systems are derived from these allogamous systems by modifications of existing floral structures (Stebbins 1970). A number of angiosperm species exhibit a mixed mating system where plants are dimorphic and produce some flowers adapted for allogamy and others adapted for autogamy. The autogamous flowers of dimorphic species are usually much reduced in size and are cleistogamous. Despite looking superficially like small buds which have failed to open, cleistogamous flowers set good yields of seed. Although a few dimorphic species have been reported as having very low levels of outcrossing (Schoen 1984; Cole & Biesboer 1992) the chasmogamous flowers are generally thought to be allogamous (barring possible geitonogamy and insect mediated selfing) (Schemske 1978; Waller 1979, 1984; Wilkin 1982; Schnee & Waller 1986; Knight & Waller 1987). The first review of the occurrence of cleistogamy (Darwin 1877) found it to be widely distributed among the angiosperms, occurring in 55 genera spread through 25 families. The phenomenon has attracted several further reviews (Uphof 1938; Maheshwari 1962; Lord 1981; Campbell et al. 1983) and cleistogamy is currently estimated to occur in 120 species from 65 genera and 25 families. Considering the angiosperms in general, however, it remains a rare method of effecting autogamy (Schoen & Lloyd 1984).
The long-term stability of such mixed mating systems has been questioned as partial autogamy does not allow for the reduction in inbreeding depression by the purging of deleterious recessive alleles (Lande & Schemske 1985; but see Stewart 1994). Cleistogamous seed will be autogamous and autogamous seed produced in a plant with a mixed mating system would, on average, be expected to suffer greater inbreeding depression than seed from a purely autogamous species. Decreased fitness of cleistogamous seed compared with chasmogamous seed from the same plant has been found in Impatiens spp. (Schemske 1978; Waller 1984; Knight & Waller 1987) and Triodanis perfoliata (Gara & Muenchow 1990), although not in Collomia grandiflora (Wilken 1982). In Danthonia spicata the relative fitness of cleistogamously and chasmogamously produced offspring is influenced by environmental conditions (Clay 1983). There is then considerable evidence in the literature that suggests that cleistogamy in an otherwise allogamous species is unlikely to be an evolutionary stable strategy unless cleistogamy has an adaptive value.
A commonly expressed advantage of cleistogamy is its cost efficiency in seed production (Schemske 1978; Lord 1980; Waller 1984). Because structures for attracting pollinators are reduced to a minimum, cleistogamous flowers require less resource allocation to attractive and male functions than chasmogamous ones (Cruden 1977; Waller 1979). Connected with an ability to produce cheaper flowers, is the advantage of an ability to combine this plasticity with an ability to detect environmental cues and respond by expressing the most effective breeding system for those conditions. An ability to produce cleistogamous flowers may therefore allow a plant to continue to flower and set seed in an environment too poor for chasmogamous flowering. There are many well-documented cases where external environmental influences, such as drought or shade, have influenced both the initiation of cleistogamous flower production and the proportion of a plant's flowers which are produced cleistogamously (Uphof 1938; Brown 1952; Evens 1956; Langer & Wilson 1964; Waller 1979; Richards 1986).
Fewer studies have addressed the issue of whether phenotypic plasticity in expression of cleistogamy can be an adaptive response to variation in the intrinsic plant environment. The most fundamental intrinsic factor is plant size. Previous studies in Collomia spp. (Wilkin 1982), Impatiens spp. (Waller 1980) and in several amphicarpic species (see review Cheplick 1994) have indicated that an increase in plant size leads to a proportional increase in allocation to chasmogamous flowers. In these species, the chasmogamous flowers are outcrossing in function. As a result, the resource allocation pattern found has been interpreted as cleistogamous flowering ensuring a basic seed output while larger plants are, in addition, able to afford the risks of outcrossing (hence potentially increased fitness) as they have a higher fecundity.
The Snout Nosed Monkey Flower, Mimulus nasutus Greene, is a dimorphic species which is unusual because its chasmogamous flowers are small and autogamous (Kiang 1973). Given this, the seed set from chasmogamous flowers in M. nasutus is usually genetically equivalent to that from the cleistogamous flowers. Under these conditions, the above hypothesis that relative expression of chasmogamy increases with plant size because of the genetic advantage of outcrossing would not be expected to apply. This paper compares the pattern of relative resource allocation to cleistogamy and autogamous chasmogamy in M. nasutus in an attempt to establish whether the breeding system of the autogamous flowers leads to different response in floral dimorphism to that reported for allogamous chasmogamy in other species.
The paper describes three experiments. The first experiment considers the effect of an increase in plant size on expression of cleistogamy. As an increase in plant size often causes a change in plant architecture, the second experiment investigates the effect of a change in plant architecture on expression of cleistogamy. Finally, as an increase in plant size results in the production of more flowers and hence fruit capsules, the third experiment considers the effect of a change in allocation of resources to fruit capsule production.
Experiment 1. The effect of an increase in plant size on the expression of cleistogamy
MATERIALS AND METHODS
This investigation used two M. nasutus populations with different growth forms, NAS1 from near Chico, Butte County, CA, and NAS2 from Skagg's Springs, Sonoma County, CA. The NAS2 population has a growth form very representative of most M. nasutus populations studied whilst the NAS1 population exhibits most extremely some of the features, such as early flowering, small flowers and little apical dominance, which distinguish M. nasutus from its progenitor species Mimulus guttatus. No phenotypic variation in growth form was detected within the M. nasutus populations.
Plants were grown in a glasshouse at a temperature of between 15 and 25 °C under natural daylight plus light supplied by overhead high-pressure sodium discharge lamps. A range of plant sizes was achieved by growing plants up from seedlings under an 8-h daylength (which is too short to initiate flowering in M. nasutus) for a varying number of weeks before transferring them to randomized positions under a 16-h daylength that stimulated flowering. The seedlings were grown in individual 7-cm diameter pots and re-potted as necessary.
Five plants of each population were grown in short days for each of the following number of weeks: 0, 2, 4, 6 and 8. As the flowering stems developed they were categorized into five stem classes according to relative size as follows: (1) stem class 1, apical stem, always solitary; (2) stem class 2, large side stem; (3) stem class 3, medium side stem; (4) stem class 4, small side stems (produced at base of plant); (5) stem class 5, tiny side stems (produced mostly as subsidiaries to other stems but also at the base of the plant).
All stems in this annual herb species become flowering stems under long daylength. The number of each stem class produced by a plant was recorded and the following information was recorded for each stem: (a) total number of flowering nodes produced; (b) percentage of flowers which are chasmogamous; (c) flower width of first flower produced on second node of main stem. Factor (c) was used as an estimate of corolla size for both chasmogamous and cleistogamous flowers: the second node was used because the first node sometimes produces atypically small flowers.
RESULTS AND INTERPRETATION
The growth pattern for plants of both populations is such that there is a gradual increase in the number of all stem types other than the primary stem as the plant ages. This is particularly the case for the smaller stem types (Fig. 1). This growth pattern results in a decrease in apical dominance with increase in plant size. The difference in apical dominance between the two populations is clear in the smaller plants but diminishes with increasing plant size.
A comparison of the number of nodes on each stem type shows a gradual increase in number as size increases (Fig. 2). This indicates that a proportion of extra resources can always be directed into an increase in stem size (as measured by number of flowering nodes produced) as well as into the production of more minor stems. This process of extra allocation to existing stems somewhat ameliorates the loss of apical dominance caused by branching that occurs with increase in plant age. The less branched NAS2 population maintains generally longer stems for its age than does the NAS1 population. This again suggests that there may be a resource trade-off between stem size and number, and that populations may vary in their relative resource allocation to these two growth parameters.
The percentage of chasmogamy was found to increase with plant size for all but the smallest stem types (Fig. 3). The overall rate of increase in percentage chasmogamy with plant size was greater for smaller plants than for larger ones (Table 1). The possibility that larger plants show a lower rate of increase of chasmogamy because of an increasing resource sink, either to other stems or to other nodes on the same stem, is investigated in the two experiments described below.
Table 1. . A comparison of the two populations of Mimulus nasutus showing how the number of chasmogamous (CH) nodes increases steadily with an increase in plant size. The mean percentage of chasmogamy (%CH), however, increases much more slowly. (Size increase was achieved by growing the plants in short days for greater number of weeks)
Average flower width per stem was always largest on the main stem and decreased with stem class (Fig. 4). Larger plants produced larger flowers on all but the smallest stem classes (class 5) where the flowers remained cleistogamous. Some of the increase in size for flowers of stem class 4 as plant size increased was caused by stems of this class in larger plants being able to produce chasmogamous flowers. As was found for the percentage of chasmogamy, the relative increase in floral width decreases with plant size, suggesting a greater rate of accumulation of resource sinks than of new resources.
Experiment 2. The effect of change in plant architecture on expression of cleistogamy
MATERIALS AND METHODS
Twenty seeds of the NAS2 population were germinated and grown in short-day conditions (described above) for 2 weeks. They were then transferred to a 16-h daylength and allowed to come into flower. Five plants were then assigned at random to each of the following treatments: (1) all stems removed as they developed except for stem class 1, the main stem (i.e. stem classes 2, 3 and 4 removed); (2) all stems removed as they developed except for stem classes 1 and 2 (i.e. stem classes 3 and 4 removed); (3) all stems removed as they developed except for stem classes 1, 2 and 3 (i.e. stem class 4 removed); (4) control, no stems removed.
The main stem of each plant was then scored for the total number of nodes produced and the percentage of these which were chasmogamous.
RESULTS AND INTERPRETATION
As more stem classes are left on the plant, fewer nodes are produced on the main stem (F = 18·57, P < 0·001) (Table 2). This suggests that some resources obtainable by the main stem are being relocated to other stems despite these stems also having their own photosynthesizing leaf bracts and calyces. The presence of other stems also slightly increased the percentage of cleistogamy expressed by the main stem (Table 2). This decrease was not significant (F = 2·39, P = 0·107).
Table 2. . The effect of different stem-removal treatments on the mean number of nodes and on the percentage of cleistogamous flowers produced by the remaining apical stem (stem class 1): sample size per treatment = 5
Experiment 3. The effect of a reduction in allocation of resources to seed formation
MATERIALS AND METHODS
Thirty seedlings of the NAS2 population were allowed to grow in short days for 2 weeks as described above except that any developing side shoots were removed, leaving only a central rosette. They were then transferred to long days and allowed to come into flower, but only to produce one main stem. All side stems continued to be removed as soon as they started to develop so that resources were not diverted into them. Ten plants were then each assigned, at random, to one of the following treatments: (a) each flower removed the day after anthesis (i.e. before any resource allocation to seed set begins); (b) one flower per node removed the day after anthesis, leaving the other flower to set seed normally; (c) control, no flowers removed.
The total number of nodes produced and the percentage of these which were cleistogamous were scored for each plant until flowering ceased.
RESULTS AND INTERPRETATION
The number of nodes produced increased significantly when lower nodes are removed (F = 47·48, P < 0·001) (Table 3). This suggests that the removal of a resource sink into seed set may release resources for further flower production and agrees with results found for M. guttatus (Macnair & Cumbes 1990; Mossop et al. 1994). Flower removal also results in a higher percentage chasmogamy (F = 5·10, P = 0·013) suggesting greater resources promote chasmogamy and, thus, that resource deprivation plays a part in the induction of cleistogamy (Table 3).
Table 3. . The effect of different flower-removal treatments on the mean number of nodes and on the percentage of cleistogamous flowers produced by the remaining apical stem (stem class 1): sample size per treatment = 10; flowers were removed on the day of anthesis
This study has shown that, although plant size is a major determining force acting on the expression of cleistogamy in M. nasutus, the effect of larger plants having larger total resources is balanced by the effects of larger plants having a larger number of resource sinks in the form of maturing seed and of larger plants having decreased apical dominance. Decreased apical dominance results in the generation of numerous small resource sinks (other flowers on the same or other stems). These flowers are often cleistogamous but also exert a drain, reducing the size and ultimately the morphology of flowers on the more major stems. The limiting resource controlling the switch from chasmogamy to cleistogamy in M. nasutus appears to be carbohydrate rather than an inorganic nutrient (Diaz 1994).
The stage in a plant's flowering cycle during which cleistogamous flowers are produced varies with the species involved and is likely to be under at least some ontogenetic constraint. In M. nasutus chasmogamous and cleistogamous flowering can occur simultaneously on different stems. The simultaneous production of cleistogamous and chasmogamous flowers also occurs in Linaria canadensis, Impatiens spp. (Uphof 1938) and Danthonia spp. (Clay 1982). However, cleistogamous flowers can be produced before chasmogamous ones as in Drosera spp., Antirrhinum spp. (Darwin 1877; Uphof 1938), Ononis parviflora (Darwin 1877) and Specularia perfoliaya (Trent 1940), or after chasmogamous flowering as in Viola spp. (Beattie 1976) and Oxalis spp. (Darwin 1877; Uphof 1938). The extent to which the occurrence of cleistogamy is ontogenetically fixed to occur at a particular developmental stage or environmentally determined is not known in most cases.
A switch from chasmogamy to cleistogamy, even in a species such as M. nasutus that produces relatively small chasmogamous flowers, results in a major reduction in resources to attractive and male functions. Cleistogamous flowers have a more complete transfer of pollen to the stigma and will have no opposing selection maintaining higher pollen levels as a closed flower automatically results in complete pollen discounting. Pollen/ovule ratios have been used as an indicator of breeding system (Cruden 1977) and under this criterion, cleistogamy appears as the most energetically efficient sexual breeding system. Reduced resource allocation to pollen allocation has, in other species, been shown to allow a phenotypic re-allocation of resources into maternal expenditure (Schoen 1984; Atlan et al. 1992). Theoretical models (Schoen & Lloyd 1984) also indicate that the evolution of cleistogamy from allogamy is favoured in those species which have an ability to reduce a large male allocation and re-allocate resources to female reproductive effort. Such re-allocation is generally considered to occur within the same flower, to ovule or pollen production (Charlesworth & Charlesworth 1987; Brunet 1992). In addition, however, shifts in resource allocation to seed production have been suggested, especially for cleistogamous flowers (Schoen & Lloyd 1984). Finally, resource re-allocation has been found in M. guttatus, a close relative of M. nasutus, between earlier and later produced flowers (Mossop et al. 1994) and between seed set and the production of further flowers as flowering and seed set overlap (Macnair & Cumbes 1990). The present study of M. nasutus indicates that resource re-allocation can occur not only within a flower, but also between flowers on a stem, between stems and between flowers and seed set.
The overall result of the balance of forces of increasing size and decreasing apical dominance in M. nasutus is that an increase in plant size results in an increase in total number of chasmogamous flowers. However, the proportion of flowering which is chasmogamous increases much more slowly with increasing plant size. The pattern of increasing chasmogamy with an increase in plant size agrees with that found in Impatiens spp. (Waller 1980), Collomia spp. (Wilkin 1982) and Amphicarpaea bracteata (Schnee & Waller 1986) where an increase in size diverted proportionally more resources into outcrossing chasmogamous flowers. At face value, this would suggest that it is irrelevant that the chasmogamous flowers of M. nasutus carry little outbreeding potential, as they are still supported preferentially to cleistogamous flowers when resources are abundant. However, this may be owing more to ontogenetic constraints on the process of enlargement of a M. nasutus plant than to any selection for diversion of resources into its autogamous chasmogamous flowers. A pattern of an increase in size causing an increase in percentage cleistogamy has been documented for at least two species; the grass, Danthonia spicata (Clay 1982) and the herb Oxalis montana (Jasieniuk & Lechowicz 1987). Clay (1982) does not comment on this pattern further, but the main explanations proposed by Jasieniuk & Lechowicz (1987) for the occurrence in O. montana of an increase in cleistogamy in favourable conditions are the hypotheses that (1) poor environments may have a greater negative effect on inbred progeny than outbred progeny (Schemske 1983) and that (2) heterosis from outbreeding is expressed more strongly in harsher environments (Lloyd 1980). The authors also propose that, as O. montana is a perennial, producing individual unbranched stems from fleshy bases, one branch is not so strongly affected by the development of the rest of the plant. It seems possible that developmental constraints alone may cause an increase in chasmogamy with size in one genus (in particular one with species which exhibit strong apical dominance), while differences in ontogeny can, alone, cause the opposite effect in another genus. If this is the case then a rigorous testing of the hypothesis that expression of outcrossing chasmogamy increases with size because of the genetic advantage of outcrossing is all but impossible, as it would require pairs of dimorphic taxa which are similar in every way (including their percentage cleistogamy) but where the chasmogamous flowers of one member are allogamous and those of the other are autogamous.
This research was supported by a SERC studentship, receipt of which is gratefully acknowledged. We thank an anonymous reviewer for their useful comments on an earlier draft of this paper.
Present address: Farmland Ecology Unit, Department of Land Resources, Scottish Agricultural College, Aberdeen AB21 9TS, UK.