Phenotypic and Genetic Correlations Among Floral Traits in Two Species of Thalictrum

Transcription

1 gen, non-protein nitrogen and nucleic acids during wheat grain development. Aust J Physiol 6: Evers AD, Development of the endosperm of wheat. Ann Bot 34: Kidd AD, Francis D, and Bennett MD Replicon size, mean rate of DNA replication and duration of cell cycle and its component phase in eight monocotyledonous species of contrasting DNA values. Ann Bot 59: Kowles RV and Phillips RL, DNA amplification patterns in maize endosperm nuclei during kernel development. Proc Natl Acad Sci USA 82: Kowles RV, Srienc F, and Phillips RL, Endoreduplication of nuclear DNA in the developing maize endosperm. Dev Genet 11: Nagl W, Polytene chromosomes of plants. Int Rev Cytol 73: Phillips RL, Wang AS, and Kowles RV, Molecular and developmental cytogenetics of gene multiplicity in maize. Stadler Genet Symp 15: Ramachandran C and Raghavan V, Changes in nuclear DNA content of endosperm cells during grain development in rice (Oriza sativa). Ann Bot 64: Corresponding Editor: Prem P. Jauhar Phenotypic and Genetic Correlations Among Floral Traits in Two Species of Thalictrum S. L. Davis The evolution of dioecy in plants is expected to be followed by sex-specific selection, leading to sexual dimorphism. The extent of the response to selection depends on the genetic covariance structure between traits both within and between the sexes. Here I describe an investigation to determine phenotypic and genetic correlations between reproductive traits within cryptically dioecious Thalictrum pubescens and within morphologically dioecious T. dioicum. Females of T. pubescens produce flowers having stamens and pistils, appearing hermaphroditic. Genetic correlations were estimated as family-mean correlations among paternal half-sib families. Positive phenotypic and genetic correlations between parts of the same reproductive organs, as the anther and filament of the stamen, indicate developmental associations between these traits in both species. Negative genetic correlations were detected between pistil number and size of reproductive organs in T. dioicum and showed the same direction, but not significance, in T. pubescens. There was a negative phenotypic correlation between the number of stamens and the number of pistils within female flowers of T. pubescens. Within T. pubescens, there was a positive genetic correlation between the number of stamens in males and the number of pistils in females, indicating that floral evolution in males and females may not be independent in this species. Avoidance of inbreeding depression and resource reallocation have long been considered primary factors driving the evolution of gender dimorphism and/or dioecy from cosexuality in plants, yet their relative contributions to this pathway are still argued ( Bawa 1980; Charlesworth and Charlesworth 1978; Charnov et al. 1976; Darwin 1877; Willson 1979). Whatever the nature of selection driving the evolution of separate sexes, the response to that selection will depend on the relationship, particularly the genetic correlations, between male and female reproductive traits ( Lande 1980). Theoretical predictions as to what the genetic relationship between reproductive traits might be and what empirical studies have shown have not always coincided. For example, a major assumption underlying theoretical models of sex allocation is the existence of a fixed amount of resources available for reproduction that can be allocated in various ways among the different sexual functions (Charlesworth and Morgan 1991). This leads to the expectation of negative genetic correlations between male and female traits within cosexuals. Such negative correlations between male and female traits should help augment selection toward sexual specialization among individuals, as individuals that allocate more to male function conversely allocate less to female function (Mazer et al. 1999). Yet studies have often found positive correlations in allocation to male and female functions (Ashman 1999; Campbell 1997; Elle 1998; Fenster and Carr 1997; O Neil and Schmitt 1993). Positive correlations between traits can be caused by greater genetic variation among individuals in resource acquisition than in allocation (Houle 1991), so individuals with more resources available would allocate more to both male and female traits. Selection can also shape genetic correlations between two traits that interact together to perform a given function. If functional integration between traits increases an individual s fitness, selection would act to increase the positive correlation between those traits (Conner 1997; Conner and Via 1993). Once dioecy has been established, further changes between the sexes in resource allocation patterns are expected to occur because of sex-specific selection (either sexual or natural selection (Geber 1995), leading to sexual dimorphism (Cox 1981; Lloyd and Webb 1977; Meagher 1994; Willson 1979). However, positive genetic correlations between traits in each sex would limit the independent evolution of those traits in males and females (Lande 1980; but see Cheverud et al. 1985). For example, selection against a particular trait may be countered by selection favoring that same trait in the opposite sex. Positive genetic correlations between the sexes are predicted because when dioecy first arises, sexual dimorphism should be low and homologous characters in the two sexes should be under the control of the same or overlapping sets of genes in each sex (Meagher 1992, 1994). One possible example of this constraint on the evolution of sexual dimorphism may be found in dioecious species in which at least one of the sexes produces full size, yet nonfunctional, sexual organs of the opposite sex, making them appear hermaphroditic. This condition is known as cryptic dioecy (Mayer and Charlesworth 1991). The maintenance of cryptic dioecy could be explained by the existence of positive genetic correlations between the sexes so that, for example, selection for resource reallocation and reduction of anthers in females would be countered by selection favoring anthers in males. Hence the study of the genetic relationship between reproductive traits in cryptically dioecious species and their relatives offer a unique opportunity to better understand factors influencing breeding system evolution and sexual dimorphism. In this article I describe an investigation to determine the relationship between reproductive traits within the cryptically dioecious species Thalictrum pubescens and its dioecious relative T. dioicum. As with most cryptically dioecious species, T. pubescens is morphologically androdioecious but functionally dioecious. Females produce flowers that have both stamens and pistils, making them appear hermaphroditic. T. dioicum is functionally and morphologically dioecious. Females of T. pubescens appear to invest the same amount of resources per stamen as males. Furthermore, a possible trade-off between stamen and pistil production within female flowers indicates that females may be maintaining stamen production at the expense of pistil production, and hence potential seed production ( Davis SL, in preparation). Therefore there should be strong selection against producing stamens in females. Davis (1997) found that females of T. pubescens whose stamens Downloaded from at Pennsylvania State University on March 6, 2014 Brief Communications 361

2 had been removed did not suffer reduced seed set due to a lack of adequate pollination by insects, most likely because this plant is also pollinated by wind. Hence stamens in females of this species are not likely maintained as a reward for pollinators. Consequently I have investigated the genetic relationship among the reproductive traits in this species to help understand why females are morphologically hermaphroditic. Materials and Methods Study Species T. pubescens and T. dioicum are members of the Ranunculaceae. Sex determination in Thalictrum sp. appears to be under nuclear control, with males serving as the heterogametic sex (reviewed in Meagher 1988). Both species lack petals and nectaries ( Boivin 1944). Sepals are small and whitish and fall off quickly as the flower ages. T. pubescens is a summer flowering perennial that grows in rich woods, low thickets, swamps, wet meadows, and stream banks. Females of T. pubescens are functionally unisexual, but morphologically hermaphroditic: females produce pistillate flowers with an average of 12 pistils with uniovulate ovaries and 8 stamens containing inaperturate pollen. Males produce only staminate flowers with an average of 38 stamens and no vestigial pistils (Davis SL, in preparation). Both wind and insects have been reported to pollinate this species (Davis 1997; Kaplan and Mulcahy 1971; Keener 1976). T. dioicum is an early spring flowering perennial that grows in rich rocky woods, ravines, and alluvial terraces ( Keener 1976). Flowers of both sexes are both functionally and morphologically unisexual. This species is solely wind-pollinated and shows characteristics associated with this mode of pollination: stamens of males are pendulous and pistils of females are upright. Field sites for the populations to be used in the study are Cedar Bluffs State Nature Preserve located near Bloomington, Indiana, for T. dioicum and a small unused field located near Fort Hill State Historical Park near Hillsborough, Ohio, for T. pubescens. Phenotypic and Genetic Correlations When character traits to be correlated cannot be measured on the same individual (such as individuals growing in different environments), genetic correlations cannot be estimated directly ( Via 1984). Genetic correlations between the sexes are directly analogous to genetic correlations between environments in that the two sexes represent different physiological environments (Meagher 1992). An approximation to the genetic correlation is the correlation of family means. The estimation of covariation of family means (cov m ) contains both the covariance among families and a fraction of the within-family error variance, so cov m cov among (1/n)cov within, where n is the number of individuals per family. Hence the standard family mean correlation is expected to estimate the true value as the number of individuals per family increases (Via 1984). It is safe to assume that any significant negative correlations within or among taxa represent real genetic constraints, because any bias due to environmental effects in the greenhouse would likely skew values in a positive direction (Fry 1993; Houle 1991; Mazer and Hultgård 1993). Seeds from 30 open-pollinated maternal plants of each species were collected from the field sites described above. Seeds of T. dioicum were collected in April 1994 and seeds of T. pubescens were collected in August Seeds of both species were planted in late September 1994 in trays of MetroMix soil and placed in a 4 C cold room for 6 weeks. Germination rate is low ( 30%), so large numbers of seeds by the handful per maternal plant were sown. Seeds from 28 maternal lines of T. dioicum and 21 maternal lines from T. pubescens germinated. Trays were then transferred to a mist room in a greenhouse at Indiana University. After the seeds germinated and seedlings reached a reasonable size (after 7 8 weeks), 2 15 seedlings per maternal line, depending on how many had germinated, were transplanted into 4 in. clay pots filled with a 3:1 mixture of soil and peat. Plants were placed in a greenhouse at the Botanical Experimental Field Station at Indiana University. After another 8 10 weeks of growth, the plants were transplanted into 8 in. pots, filled with a 3:1 mixture of soil and peat. When these plants flowered they were used to create paternal half-sib families in both species as described below. For T. dioicum, plants flowered in January March A total of 26 males, each from a different maternal line, were each crossed with three to five females from maternal lines different from each other and from the male being crossed. Seeds from each cross were collected and coldtreated, and germinated as described above. A maximum of 15 seedlings from each father mother combination was transplanted into 4 in. pots. Plants began to flower in February 1996, before they were transplanted, so they were kept in the 4 in. pots. On the first or second day of flowering, two flowers per plant were collected in 70% ethyl alcohol (ETOH) for measurements to be taken in the laboratory. Two flowers per plant seems an adequate number, as the within-plant variation among floral traits is much lower than the among-plant variation in floral traits. There were a total of 26 paternal families with an average of 12.4 individuals per family. Flowering rate was low in T. pubescens. Less than 20% of T. pubescens plants flowered in 1995, and another round of flowering started in August Although not ideal, by using plants in both rounds of flowering, I was able to use 17 males to pollinate two to three females each. Seeds from each paternal half-sib family were planted and cold-treated in September 1997 as described above. A maximum of 15 seedlings from each father mother combination of T. pubescens were transplanted into 4 in. pots in October To promote greater flowering rates, these plants were exposed to another round of cold-stratification by leaving the greenhouse unheated unless the temperature dropped below 40 F starting in early January In mid-march 1998, as the temperature increased, the plants were transplanted into 8 in. pots and moved into a greenhouse with the minimum temperature control turned off. Plants started flowering in early May. On the first or second day of flowering for each plant, two flowers were collected in 70% ETOH for measurements to be taken in the laboratory. One family failed to have any seeds germinate and one other failed to have any plants flower, giving a total of 15 families with an average of 16.2 individuals per family. Measurements on all flowers were taken using a micrometer on a dissecting microscope under 8 magnification. Traits measured were number of stamens, number of pistils, stigma length, ovary length, anther length, filament length, and sepal length. Phenotypic correlations were calculated by using the entire dataset for each species. Pearson product-moment correlation coefficients were calculated for each pair of variables within each sex. Because of the small number of plants that flowered at one time precluded a complete crossing design, the power of this experiment was limited. Variance components for maternal and paternal ef- 362 The Journal of Heredity 2001:92(4)

3 Table 1. Phenotypic and genetic correlations between floral traits in female T. dioicum Ovary length Number of pistils Sepal length Stigma length x 1 SD Ovary length (185) ** (184) ** (187) Number of pistils * (26) (182) (185) Sepal length (26) (26) ** (184) Stigma length (26) (26) (26) Phenotypic correlations are shown above the diagonal and family-mean correlations are shown below the diagonal. N is indicated in parentheses. Those correlations that are significant after correcting for multiple tests by using a sequential Bonferroni technique are shown in bold. fects could not be estimated and heritabilities for individual traits were not calculated. For estimates of genetic correlations, the mean of each trait was calculated for each paternal half-sib family. Pearson product-moment correlation coefficients were then calculated for each pair of traits among half-sib families. The number of data points could have been increased by collapsing the design and using full-sib families. However, these data points would not be independent of one another, since in the breeding design each mother was crossed with multiple fathers. The correlation between the number of stamens in males and the number of pistils in females in both species was calculated, as, in a sense, these characters represent homologous structures in the two sexes. In addition, for T. pubescens, correlations were calculated between the pairs of traits shared by males and females: number of stamens, stamen length, anther length. Because the correlations were calculated among paternal sibships, variation due to maternal effects should be minimal. Log-transformation of nonnormally distributed variables did not change the substance of the results, so the analyses on untransformed data are presented for ease of interpretation. For each group of correlations within each sex (i.e., within-female phenotypic correlations, within-male genetic correlations), significance levels were corrected for multiple tests using a sequential Bonferroni procedure (Rice Table ). The table-wise error rate was set at 0.05, which was then divided by the number of tests within the table. If the lowest P value within the table was less than this new adjusted value, that correlation was considered significant. The next lowest P value was then compared to a new value of 0.05 divided by the number of tests minus 1, and so on, until a nonsignificant test was reached. Results Phenotypic and Genetic Correlations between floral traits in male T. dioicum Number of stamen Anther length Filament length Sepal length x 1 SD Number of stamen (122) (122) (122) Anther length (26) ** (123) ** (123) Filament length (26) (26) ** (123) Sepal length (26) (26) (26) Phenotypic correlations are shown above the diagonal and family-mean correlations are shown below the diagonal. N is indicated in parentheses. Those correlations that are significant after correcting for multiple tests by using a sequential Bonferroni technique are shown in bold. Figure 1. Correlations among paternal family means between the number of stamens per flower produced in males and the number of pistils produced per flower in females: (A) Thalictrum dioicum; (B) Thalictrum pubescens. Thalictrum dioicum Two classes of phenotypic correlations were significant in both sexes. First, structures within organs were positively correlated: within the pistil, stigma length was positively correlated with ovary length (Table 1), and within the stamen, anther length was strongly correlated with filament length ( Table 2). Second, positive phenotypic correlations exist in both females and males between the size of sexual organs and the size of the sepals (Tables 1 and 2). No significant phenotypic correlations were associated with either the number of pistils or the number of stamens. There were no significant genetic correlations within male floral traits ( Table 2). A significant negative genetic correlation was present between ovary size and pistil number within female flowers ( Table 1). As ovary size increases, pistil number decreases. The genetic correlation between stamen number in males and pistil number in females was not significant (r 0.061, N 26, P.767; Figure 1A). Thalictrum pubescens Patterns of phenotypic correlations in T. pubescens were similar to those found in T. dioicum (Tables 3 and 4). Within the pistils of females, stigma length was positively correlated with ovary length ( Table 3), and within the stamens of both males and females, anther length was strongly correlated with filament length (Tables 3 and 4). Sepal length was positively correlated with the size of stamens and pistils in females and with stamen size in males. The number of pistils in females was negatively correlated with the size of the ovary and with the size of the stamen parts: flowers with large numbers of pistils had smaller stamens and pistils. In addition, there was a negative correlation between the number of stamens and the number of pistils within female flowers ( Table 3). Most of the genotypic correlations within females were in the same direction as the phenotypic correlations. However, after correcting for multiple tests, none of these correlations were significant ( Table 3). Within male flowers, anther length was again significantly correlated with filament Downloaded from at Pennsylvania State University on March 6, 2014 Brief Communications 363

4 Table 3. Correlations between floral traits in female T. pubescens Number of stamen Anther length Filament length Ovary length Number of pistils Sepal length Stigma length x 1 SD Number of stamen 0.241* (145) 0.386** (145) (147) 0.281** (147) 0.406** (148) (147) Anther length (15) 0.436** (145) 0.279** (144) 0.380** (144) 0.603** (145) 0.364** (144) Filament length (15) (15) 0.285** (144) 0.335** (144) 0.442** (145) 0.300** (144) Ovary length (15) (15) (15) 0.259* (147) 0.266** (147) 0.321** (147) Number of pistils (15) (15) (15) (15) (147) (147) Sepal length (15) (15) (15) (15) (15) 0.364** (147) Stigma length (15) (15) (15) (15) (15) (15) Phenotypic correlations are shown above the diagonal and family-mean correlations are shown below the diagonal. N is indicated in parentheses. Correlations that are significant after correcting for multiple tests by using a sequential Bonferroni technique are shown in bold. length, and again there was a trend for these traits to be positively correlated with sepal size (Table 4). In T. pubescens, there were no significant correlations between stamen characters in males and stamen characters in females as were expected. However, there was a positive correlation among families between the number of stamens and the number of pistils: sibships with females that have many pistils also have males that have many stamens (Figure 1B). Discussion The repeated detection of correlations between specific traits among closely related species can provide insights into selection patterns and the joint evolution of these traits (Mazer and Hultgård 1993). Despite the lack of power in the breeding design, the phenotypic and genetic correlations found in this study were similar in both species. These patterns of covariation are discussed below in terms of how they may contribute to evidence of genetic constraints on the independent evolution of floral traits within the genus. Correlations Within Flowers Positive genetic correlations between parts of the same structure, such as those found here between anther and filament length in the stamen and ovary and stigma Table 4. Phenotypic and genetic correlations between floral traits in male T. pubescens Number of stamen Anther length Filament length Sepal length x 1 SD Number of stamen (188) (188) (187) Anther length (15) ** (188) ** (187) Filament length (15) * (15) ** (187) Sepal length (15) (15) (15) Phenotypic correlations are shown above the diagonal and family-mean correlations are shown below the diagonal. N is indicated in parentheses. Those correlations that are significant after correcting for multiple tests by using a sequential Bonferroni technique are shown in bold. length in the pistil, are not surprising and most likely reflect strong developmental associations between these traits. These correlations might impose constraints on selection responses if, for example, males are selected to have longer filaments in order to aid in pollen dispersal in a windpollinated plant, but are under conflicting selection pressure to maintain a certain anther size. A second prevalent positive correlation was that of the phenotypic correlation between the size of the sexual organs and that of the sepals in both sexes of both species. As both species are wind pollinated and sepals fall off shortly after the flower matures, the perianth (in this case just sepals) most likely does not function in pollinator attraction, and hence does not play a factor in determining their size in males versus females. Two other possible hypotheses, however, might lead to the observed correlations. First, developmental associations between the corolla and stamens (androecium) may cause a positive correlation between these structures. Second, the role of the perianth in enclosing the reproductive structures in the bud may result in a correlation between the size of enclosed structures and the sepals or petals (Delph et al. 1996). Sepal size was phenotypically correlated with stamen size in both males and females of T. pubescens, but showed no correlation with pistil size in females, so it appears that in T. pubescens the data follows the prediction that developmental associations with stamens contribute to the determination of sepal size. In the morphologically dioecious T. dioicum, however, sepal size was phenotypically correlated with stamen size in males and pistil size in females, so a developmental relationship with anthers alone cannot account for sepal size and the protective function of the sepals may be playing a more prominent role. There were no significant negative correlations, either phenotypic or genetic, within male flowers of either species. However, there was a negative genetic correlation between ovary size and pistil number in females of T. dioicum and a negative phenotypic correlation between the number of pistils and the size of both stamens and pistils in females of T. pubescens. If negative correlations indicate the presence of trade-offs in allocation between traits, this agrees with Bateman s principle (1948), which asserts that female reproduction is more likely to be limited by resources, whereas male reproduction is more often limited by access to mates. There was also a negative phenotypic correlation between the number of pistils and the number of stamens within female flowers of T. pubescens. In other words, plants appear to be limited in the total number of reproductive organs that they can make within the flower: plants that make more pistils per flower also produce fewer stamens. This limitation in flower size may be due to either the amount of space or the number of primordial meristems within the flower and not just the amount of resources available. Pistils have three times the biomass as stamens (Davis SL, in preparation). However, despite their relatively small cost compared to pistils in terms of resources, producing more stamens reduces pistil production. 364 The Journal of Heredity 2001:92(4)

5 The genetic correlation between these traits mirrored the phenotypic correlation in sign, arguing that this may be a real trade-off, but was not statistically significant. The existence of negative genetic correlations may not be easily detected: for example, Mazer et al. (1999) found evidence of a negative genetic correlation between anther number and ovule number in Spergularia marina only after determining correlated responses to artificial selection. Correlations Between the Sexes Since the perianth in the two species studied here are relatively small compared to the reproductive structures within the flower, the size of the flowers in these species is most easily measured by the number of reproductive organs per flower. There was no genetic correlation between the number of pistils per flower in females and the number of stamens per flower in males in T. dioicum. Therefore the size of flowers in males and females should be able to respond independently to selection in this species. If the same genes control stamen production in males and females of T. pubescens, a positive genetic correlation should exist between these traits. Instead, the family-mean correlation between these traits was nonsignificant. On the other hand, the positive genetic correlation between stamen number in males and pistil number in females of T. pubescens still indicates that flower development in the two sexes is not independent, so selection acting differentially in each sex may not lead to separate evolutionary optima. Since pistils and stamens are not directly homologous, the correlation between these traits could be explained if it is flower size that is being genetically controlled, which would then determine the number of reproductive organs per flower that can be made. Furthermore, although the number of stamens in males is not directly correlated with the number of stamens in females, they are not independent. For example, families with large flowers would be comprised of males that have many stamens per flower and females that have many pistils per flower. The within-flower phenotypic correlations showed that females that produce many pistils per flower also produce fewer stamens. One possible explanation for the maintenance of the positive correlation between stamen number in males and pistil number in females would be if flower size was negatively correlated with flower number, so that families that have fewer flowers per plant also have bigger flowers in both males and females. Meagher (1994) found that artificial selection on flower size caused an indirect and inverse effect on flower number in Silene latifolia, suggesting a negative genetic correlation between these traits. A similar negative correlation between flower size and flower number may be maintaining the positive correlation between flower size in males and in females of T. pubescens. Since each pistil in females of T. pubescens is uniovulate, the number of pistils a female produces directly determines the number of seeds that female can potentially produce. If selection acts against stamen production in favor of more pistils, and hence more possible seeds, the negative correlation between stamen and pistil production within female flowers should be able to augment this selection and hasten a response toward further sexual dimorphism in T. pubescens. Two factors may help account for the maintenance of stamens. First, environmental variation may mask the trade-off between stamens and pistils in natural populations and reduce its effects ( Fry 1993; Houle 1991). Davis (in preparation) was able to detect a negative phenotypic correlation between these traits in only 1 of 4 years in the same natural population from which the seeds for this study were originally collected. This also happened to be a year when overall flower size was significantly smaller, suggesting reduced environmental variance in that year. So although there is a trade-off between stamen and pistil production, its effects on evolution may be limited. Second, conflicting selection pressures on flower size in males and flower size in females, along with the positive genetic correlation between these traits may prevent the loss of stamens in females. Males, whose reproduction is theoretically limited by access to mates, may be selected to have larger flowers with more stamens. Selection on increased flower size (more stamens) in males would result in a correlated response in females for more pistils. For females, whose reproduction may be limited by resources, producing more pistils leads to smaller pistils (as indicated by the negative phenotypic correlation), and hence possibly smaller, less viable seeds, so females may be under selection to produce flowers with fewer, larger pistils and more stamens. In this way, conflicting selection pressure in males and females, in conjunction with positive genetic correlations between males and females, may be maintaining genetic variation for stamen production in females. From the Department of Biology, Indiana University, Bloomington, Indiana. I wish to thank Lynda Delph, Edmund Brodie III, Harold Lindman, and Loren Rieseberg for guidance and comments on this work. I also wish to thank two anonymous reviewers for their comments. I especially want to thank D. J. McClellan, E. Levri, M. Levri, D. Dudle, and J. Walgenbach for help planting and transplanting Thalictrum. The greenhouse crew at Indiana University were important for their help in maintaining the plants. This work was supported through funds from the Indiana Academy of Sciences and the Department of Biology at Indiana University. Address correspondence to Sandra L. Davis, Department of Biology, University of Louisiana at Monroe, Monroe, LA 71203, or bidavis@alpha.nlu.edu The American Genetic Association References Ashman TL, Quantitative genetics of floral traits in a gynodioecious wild strawberry, Fragaria virginiana; implications for the independent evolution of female and hermaphrodite floral phenotypes. Heredity 83: Bateman AJ, Intrasexual selection in Drosophila. Heredity 2: Bawa KS, Evolution of dioecy in flowering plants. Annu Rev Ecol Syst 11: Boivin B, American Thalictra and their Old World allies. Rhodora 46:377, , Campbell DR, Genetic correlation between biomass allocation to male and female functions in a natural population of Ipomopsis aggregata. Heredity 79: Charlesworth D and Charlesworth B, A model for the evolution of dioecy and gynodioecy. Am Nat 112: Charlesworth D and Morgan MT, Allocation of resources to sex functions in flowering plants. Philos Trans R Soc Lond 332: Charnov E, Maynard Smith J, and Bull J, Why be a hermaphrodite? Nature 263: Cheverud JM, Dow MM, and Leutenegger W, The quantitative assessment of phylogenetic constraints in comparative analysis: sexual dimorphism in body weight among primates. Evolution 39: Conner JK, Floral evolution in wild radish: the roles of pollinators, natural selection, and genetic correlations among traits. Int J Plant Sci 158:S108 S120. Conner JK and Via S, Patterns of phenotypic and genetic correlations among morphological and life-history traits in wild radish, Raphunus raphanistrum. Evolution 47: Cox PA, Niche partitioning between sexes of dioecious plants. Am Nat 117: Darwin C, On the different forms of flowers on plants of the same species. London: John Murray. Davis SL, Stamens are not maintained as attractants to pollinators in females of cryptically dioecious Thalictrum pubescens Pursch (Ranunculaceae). Sex Plant Reprod 10: Delph LF, Galloway LF, and Stanton ML, Sexual dimorphism in flower size. Am Nat 148: Elle E, Quantitative genetics of sex allocation in the andromonoecious perennial, Solanum carolinense (L.). Heredity 80: Fenster CB and Carr DE, Genetics of sex allocation in Mimulus (Scrophulariaceae). J Evol Biol 10: Downloaded from at Pennsylvania State University on March 6, 2014 Brief Communications 365

6 Fry J, The general vigor problem: can antagonistic pleiotropy be detected when genetic covariances are positive? Evolution 47: Geber MA, Fitness effects of sexual dimorphism in plants. Trends Ecol Evol 10: Houle D, Genetic covariance of fitness correlates: what genetic correlations are made of and why it matters. Evolution 45: Kaplan S and Mulcahy D, Mode of pollination and floral sexuality in Thalictrum. Evolution 25: Keener C, Studies in the Ranunculaceae of the southeastern United States. II. Thalictrum L. Rhodora 78: Lande R, Sexual dimorphism, sexual selection and adaptation in polygenic characters. Evolution 34: Lloyd DG and Webb CJ, Secondary sex characters in plants. Bot Rev 43: Mayer S and Charlesworth D, Cryptic dioecy in flowering plants. Trends Ecol Evol 6: Mazer SJ, Delesalle VA, and Neal PR, Responses of floral traits to selection on primary sexual investment in Spergularia marina: the battle between the sexes. Evolution 53: Mazer SJ and Hultgård UM, Variation and covariation among floral traits within and among four species of northern European Primula (Primulaceae). Am J Bot 80: Meagher TR, Sex determination in plants. In: Plant reproductive ecology: patterns and strategies (Lovett-Doust J and Lovett-Doust L, eds). New York: Oxford University Press; Meagher TR, The quantitative genetics of sexual dimorphism in Silene latifolia (Caryophyllaceae). I. Genetic variation. Evolution 46: Meagher TR, The quantitative genetics of sexual dimorphism in Silene latifolia (Caryophyllaceae). II. Response to sex-specific selection. Evolution 48: O Neil P and Schmitt P, Genetic constraints on the independent evolution of male and female reproductive characters in the tristylous plant Lythrum salicaria. Evolution 47: Rice WR, Analyzing tables of statistical tests. Evolution 43: Via S, The quantitative genetics of polyphagy in an insect herbivore. II. Genetic correlations in larval performance within and among host plants. Evolution 38: Willson MF, Sexual selection in plants. Am Nat 113: Received May 18, 2000 Accepted February 14, 2001 Corresponding Editor: William F. Tracy 366 The Journal of Heredity 2001:92(4)