Heredity 73 (1994) 142—154 Received 21 December 1993 Genetical Society of Great Britain

Restricted gene flow within the morphologically complex species moiis (): contrasting evidence from the mating system and pollen dispersal

SIEGFRIED L. KRAUSS Department of Biolo gical Sciences, University of Wo//ongong, Northfields Avenue, Wo/longong, NSW2522, Australia

Severelyrestricted gene flow may be a factor contributing to the remarkable amount of morpho- logical variation within the complex species Persoonia mollis, in which nine subspecies are recognized. The mating system and realized pollen dispersal were studied to assess their effect on gene flow. Mating system parameters were estimated in seven natural populations over two seasons using allozyme electrophoresis. Realized pollen dispersal was measured in two natural populations over two seasons by monitoring the dispersion of a rare allozyme from a known source in each population. Single- and multilocus estimates of outcrossing rate (t)wereconsistently equal to or greater than one (i.e. complete outcrossing). Realized pollen dispersal distances showed that 99 per cent of the pollen received by given females was donated by males on average within 33 m. However, 70 per cent of all pollen dispersal was on average to the paternal plant's immediate neighbour. Genetic neighbourhood sizes due to pollen dispersal alone ranged from one to five and paternity pool sizes ranged from four to 22 plants. These population sizes are small enough to allow genetic differentiation in the absence of selection. However, in contrast to the expectation that small population size leads to biparental inbreeding and reduced heterozygosity compared with Hardy—Weinberg expectations, the mean fixation index (F) of —0.035indicated a slight excess of heterozygotes in the seed cohort. This apparent paradox could be the result of selection for heterozygous seeds, disassortative mating or more likely because gene flow through seed dispersal substantially increases the neighbourhood sizes estimated here through pollen dispersal alone.

Keywords:geneflow, genetic neighbourhood, outcrossing rate, paternity pool, Persoonia mollis, realized pollen dispersal.

Introduction crossing rate). The contribution of pollen dispersal to gene flow is directly correlated with the amount of out- Morphologicaldiversity within species can be brought crossing, as selfed seeds have a pollen dispersal about by a restriction in gene flow (Ehrlich & Raven, distance of zero. 1969; Wright, 1969; Endler, 1977; Slatkin, 1985, The relationship between gene flow and the genetic 1987; Templeton, 1989). Gene flow occurs in flower- structure of a continuously distributed population was ing plants via pollen and seed, with pollen dispersal fre- inferred by Wright (1943, 1946, 1969). The important quently forming the larger component (Levin & parameter is the effective number of individuals in a Kerster, 1974; Beattie & Culver, 1979; Loveless & genetic neighbourhood (Ne). A neighbourhood is an Hamrick, 1984; Fenster, 1991; Waser, 1993). Realized area from which parents of central individuals may be pollen dispersal is a function of the movement of pollen treated as if drawn at random (Wright, 1946). The by vectors as well as post-pollination events that deter- neighbourhood area due to pollen dispersal is a func- mine systems of mating (Levin & Kerster, 1974; Levin, tion of pollen dispersal distances weighted by outcross- 1981; Loveless & Hamrick, 1984). The most funda- ing rate and when multiplied by the effective density of mental of these post-pollination events is the relative breeding organisms, provides a measurement of effec- success of outcross pollen over self pollen (i.e. the out- tive neighbourhood size (Wright, 1946; Crawford, MATING SYSTEM AND POLLEN DISPERSAL IN P. MOLLIS 143

198 4a,b). Related to the concept of neighbourhood Materials and methods size is that of the paternity pool, where the paternity pooi is the maximum number of potential mates for individual females (Levin, 1988). Reproductive biology All else being equal, species with small effective Persooniamollisusuallyflowers from late December population size parameters are more likely than those until March. Flowers are pollinated by a suite of native with larger populations to undergo genetic differentia- bee species as well as the introduced European honey- tion over a given distance. Theoretical models suggest bee Apis mellifera (personal observation). Each mature that random differentiation (genetic drift) can occur plant produces thousands of flowers, each flower when neighbourhoods contain fewer than 200 individ- remaining receptive for about 5 days after anthesis uals and may be considerable when they have 20 or (Krauss, 1994). The fruits (drupes) usually contain one fewer (Wright, 1946). Many published estimates of seed and are retained on the plants until maturation in neighbourhood size in plants have been low (e.g. November. Natural fruit-set is usually about 20 per Crawford, 1984b; Cahalan & Gliddon, 1985; Bos et cent of flowers (Krauss, 1994). Dispersal may be al., 1986; Fenster, 1991) indicating that random differ- affected by a range of large birds (e.g. the Currawong, entiation within populations may be considerable. Strepera spp. (Barker & Vestjens, 1990)). Although However, high estimates of neighbourhood size (i.e. self-compatible, there is a strong preference for out- more than 200 and even up to 50 000) in plants are cross pollen, in terms of both quantity and quality of also known (e.g. Crawford, 1984b; Govindaraju, 1988; seeds produced (Krauss, 1994). There is no vegetative Young, 1988; Epperson & Allard, 1989; Broyles & reproduction. Wyatt, 1991; Eguiarte et al., 1993). Paternity pool size estimates range from 20 to over 30000 for 25 herb Studypopulations and tree species (Levin, 1988). Persoonia mollis (Proteaceae) is a remarkably com- Plantmating systems can be extremely flexible and can plex species morphologically. Nine allopatric or para- vary both spatially and temporally (Schemske & patric subspecies are currently recognized (Krauss & Lande, 1985). Therefore, estimates of outcrossing Johnson, 1991). The variation is complex because it is were obtained for seven natural populations (repre- often clinal within the range of a subspecies, with senting six of the nine subspecies). These populations transition zones of various widths between subspecies. and subspecies were: These transition zones are in places extremely narrow 1 Lyrebird Gully (Lyr), ssp. maxima and measured in the tens of metres with striking 2 Woodburn State Forest (Woo), ssp. caleyi changes in morphQlogy. The variation appears to have 3 Penrose State Forest (Pen), ssp. ledifolia a genetic basis (as, Opposed to phenotypic plasticity), as 4 Murphy's Glen, Blue Mountains National Park reciprocal transplants of different forms do not remove (Mur), ssp. mo/us the diagnostic differences (personal observation). 5 Monga State Forest (Mon), ssp. budawangensis It is a long-lived (about 10—30 years) fire-sensitive 6 Belanglo State Forest (Bel), ssp. revoluta shrub occurring in a range of habitats in south-eastern 7 Long Acre Fire Trail, Belanglo State Forest (Lon), Australia for about 400 km S and 200 km W from Syd- ssp. ledifolia. ney(Krauss & Johnson, 1991). This is one of a series of studies addressing the fac- Thesepopulations cover much of the natural range of tors contributing to the morphological variation within habitats and population densities of P. mollis. For three P. mollis. In this study, I investigated the pollen disper- of these populations (Lyrebird Gully, Woodburn, and sal component of gene flow using isozyme markers. I Murphys Glen), estimates of mating system parameters firstly estimated mating system parameters in seven were derived for two consecutive seasons (1990 and natural populations over two seasons. Secondly, I 1991). Realized pollen dispersal was estimated in two measured realized pollen dispersal in two of these of these populations (Lyrebird Gully and Long Acre). natural populations. Estimates of paternity pool size The Lyrebird Gully population is relatively disjunct and genetic neighbourhood size were derived from and isolated, while all others are more or less part of a these measurements. The aim was to characterize the continuous distribution. mating system and pollen dispersal within P. mollis and to assess whether a restriction in gene flow can explain Electrophoresis the morphological complexity within this species. Upto 40 mature fruits per plant were collected from 10—28 randomly chosen plants per population per 144 S. L. KRAUSS

season. Individual plants were tagged and mapped and are unlinked (Ritland & Jam, 1981). To test for non- resampled for pollen when flowering (about 2 months random associations among alleles at two loci, the after fruiting). Fruits were stored at 4°C. Buds were digenic composite disequilibrium coefficient AB (Weir collected just prior to anthesis and immediately stored & Cockerham, 1989; Weir, 1990) was calculated for all in liquid nitrogen and later at —80°C. Seeds were pairwise combinations of loci using the program of removed from fruits and ground in chilled ceramic spot Weir (1990). AAB is equal to zero if the allelic state at plates in one drop of the Tris-HC1 pH 7.5 grinding one locus is not correlated with that at another. Devia- buffer of Soltis etat.(1983) but with mercaptoethanol tions from zero were detected using x2withone degree increased to 1 per cent. Whole anthers were removed of freedom, calculated as the squared estimates of dis- from buds and similarly ground. equilibrium divided by the sampling variances (calcu- Thirty enzyme systems were assayed using a range lated under the hypothesis that the true value of the of grinding buffers and running conditions. Three loci disequilibrium is zero) (Weir, 1990). Estimates of dis- were sufficiently resolvable and polymorphic in seed equilibrium were calculated from seed genotypes, for the following enzymes: alcohol dehydrogenase which violates the assumption of independence in the (ADH, E.C. 1.1.1.1), glucose-6-phosphate isomerase x2testdue to common maternal, and often paternal, (GPI, E.C. 5.3.1.9) and phosphoglucomutase (PGM, parents (Waller & Knight, 1989). However, as the error E.C. 5.4.2.2). For Gpi-1 and Pgm-1, maternal geno- about the estimate is reduced by using these data rela- types were determined by electrophoresis of pollen tive to independent data, a nonrejection of the null and Mendelian inheritance was confirmed from the hypothesis of random association (gametic equili- segregation patterns in their open-pollinated families brium) among alleles at two loci remains valid. Use of of seeds. Alcohol dehydrogenase (Adh-1) was not these data, however, increases the probability of a expressed in pollen, so maternal genotypes were type-I error. Therefore, a significant result needs to be inferred from the progeny arrays. Electrophoresis was interpreted with caution and could be the result solely conducted with 12 per cent w/v starch gels in a Tris- of violation of the assumption of independence. The maleate buffer solution at pH 7.4 (Selander et at., parental genotypes could not be used to test for 1971). Gels were run for 16 h at 90 mA at 4°C. Stain gametic disequilibrium because of small sample sizes. recipes were based on Wendel & Weeden(1989).

Estimationof pollen dispersal Estimationof mating system parameters Realizedpollen dispersal was measured in two natural Wright'sfixation index (F) was calculated from the populations (Lyrebird Gully in seasons 1990 and 1991 progeny in each population. This value gives the pro- and Long Acre in 1991) where paternity could be portional increase of homozygosity in excess of that assigned to one plant that possessed an allele unique in expected under panmixia (the Hardy—Weinberg expec- that population. Rare alleles in seed were detected tation). The standard error was calculated from formu- electrophoretically and the distance from the maternal lae in Brown et al. (1975). Based on the mixed-mating plant to the known source of the rare allele (paternal model (Fyfe & Bailey, 1951; Clegg, 1980; Brown etal., plant) measured. Estimates of pollen dispersal derived 1985), estimates of the proportion of seeds derived from genetic markers are more accurate than those from outcrossed ovules (t) and the frequency of alleles derived indirectly from observations of pollinator in the pollen pool (p) were calculated jointly for each movements or from the movement of coloured dyes as locus using the Maximum-Likelihood (ML) procedure they take into account both pre- and post-fertilization of Brown et at. (1975) and the Expectation-Maximi- components of pollen dispersal (Schaal, 1980; Levin, zation (EM) algorithm of Cheliak et at. (1983). Multi- 1981; Waser & Price, 1982; Handel, 1983; Hamrick & locus estimates of t were calculated where appropriate Schnabel, 1985; Campbell, 1991; Adams eta!., 1992). using the procedures of Green et a!. (1980), Shaw et at. In the Lyrebird Gully population in 1990, one plant (1981) and the joint Maximum-Likelihood methods of was homozygous for a rare allele at the Pgm-1 locus Ritland & Jam (1981). The variance for each Ritland & (DD). All the other nine plants sampled were homo- Jam (1981) multi- and average single-locus estimate of zygous for the common allele (AA) at this locus. t and the variance for the difference between these two Therefore any heterozygous AD progeny on the AA estimates of t for each population were calculated maternal plants must have been fathered by pollen using the bootstrap method with the unit of resampling from the DD plant, assuming that all the pollen donors within families. are located within the boundaries of the study popula- An assumption of the multi-locus models for the tion. In 1991, this procedure was repeated and estimation of mating system parameters is that the loci included all 16 reproductively mature plants in the MATING SYSTEM AND POLLEN DISPERSAL IN P. MOLLIS 145 study area. Two AA plants from the 1990 sampling Results had died in the intervening period, giving a total of 18 plants in the study area over the 2 years. One maternal heterozygote AD at the Pgm-1 locus was located at Mating system parameters one end of this population's distribution (and 70 m Allelefrequencies for the maternal parents, progeny from the DD maternal). All others were AA. The and pollen pool are given in Table 1. Allele frequencies paternity of heterozygous progeny at the Pgm-1 locus across the three cohorts (parents, progeny and pollen) for homozygous AA maternal plants could be assigned were in close agreement. These data suggest that on almost all occasions to either of the two plants maternal plants contribute equally to the pollen pool. possessing the rare D allele because these plants were No loci were variable in all populations. Indeed, in only homozygous for alternative alleles at the Gpi-1 locus. two of the seven populations were all three loci Only those cases where paternity could be assigned variable. with absolute confidence were used for measures of Estimates of Wright's fixation index (F) for each dispersal distance in both 1991 and 1990. Doubtful locus show a consistent trend towards a negative value cases, where the offspring were double heterozygotes across all populations and loci, although no individual at Pgm-1 and Gpi-1, were not included. Therefore, estimate of F was significantly different from zero estimates for 1990 also utilized genotypes at the Gpi-1 (Table 2). The frequency of negative values of F (14 of locus. At a second population (Long Acre), one of 19 19) was in excess of one-half (x= 4.26;P <0.05) flowering plants was heterozygous at the Gpi-1 locus; indicating an overall excess of heterozygotes in the all others were homozygous for the common allele. progeny compared with the Hardy—Weinberg expecta- The proportions of seed with known paternity were tionofF=0. multiplied by the square of the distance from the No single-locus estimate of outcrossing (t) was signifi- known source plant (p2) and summed to give an esti- cantly less than unity and indicates complete outcross- mate of the variance of dispersal distance of pollen ing (Table 3). The mean single-locus outcrossing rate was such that (7 absolute =p2/n(Crawford, 1984a,b), where 1.04. This level of outcrossing leads to an indirect n is the sum of the proportions of seed with known estimateofthefixationindex Fe=(1 —t)/(1+ t)=—0.02. paternity. This variance was halved to derive (7 axial Fe is close to the direct estimate of the mean fixation (Crawford, 1984a,b). An estimate of neighbourhood index of F —0.035 and indicates that the number of area was derived according to the general formula heterozygotes is as expected on the basis of the mating Na =r[(tko axiai/2) + k5o], where Na is the neigh- system(i.e. implies an inbreeding equilibrium) bourhood area, t =outcrossingrate, and k and k5 are (Hedrick, 1985). All single-locus estimates of outcross- correction factors for kurtosis (i.e. non-normality of the ing derived using the EM algorithm of Cheliak et al. pollen and seed dispersal distributions) (Crawford, (1983) gave the maximum possible value from this 1984b). The contribution of pollen dispersal to total procedure of t =1.00,except for Murphys Glen dispersal is halved as total dispersal is the sum of (1990) Pgm-1 (t=0.96) and Monga (1990) Pgm-1 paternal dispersal (via pollen and seed) and maternal (t =0.95).Estimates of pollen allele frequencies dispersal (via seed only) (Crawford, 1984a,b). That is, derived using the EM algorithm were in close agree- the maternal contribution to gamete dispersal via ment to those derived using the ML procedure. pollen is zero (Crawford, 1984a,b). Kurtosis is zero, Multi-locus estimates of outcrossing (t) were con- and k and k5 =4, when the dispersal distributions are sistently equal to or greater than unity (Table 4). There normal. When t =1,and pollen dispersal is normally is heterogeneity among the different multi-locus esti- distributed around a mean of zero, and seed dispersal mates of t within the Lyrebird Gully population. is not measured (i.e. a= 0), this reduces to Ritland's multi- and mean single-locus estimates of t N0 =2nuaxial (Crawford, 1 984b).Estimates of effect- failed to converge for two populations (Table 4). This ive neighbourhood size resulting from pollen dispersal can be caused by small population size and asymmetri- only were derived by multiplying Na by the density of cal allelic frequencies, leading to small numbers of breeding plants per m2. detectable outcrosses. These values were not included Assuming a normal distribution of pollen dispersion in the calculation of mean values for t. The multi-locus with zero mean, 99 per cent of the pollen received by estimate of t was significantly greater than the mean given females is donated by males within a circleof single-locus estimate of t only in the Lyrebird Gully radius 3a and area A= 9rru axial (Levin, 1988). The (1991)population(P <0.01). number of potential fathers for the progeny of a refer- ence female is Ad, where d is the density of pollen- producing plants. 146 S. L. KRAUSS

Table1Allelefrequencies for variable loci in parental, progeny and pollen pooi cohorts and sample sizes of parents and progeny, for each population and season

Pgm-I Gpi-1 Adh-1 A B C D E F n A B C n A B C n

Lyr (1990) Parents 0.91 — — 0.09 — — 11 0.35 — 0.65 10 Progeny 0.86 — — 0.14 — — 178 0.37 — 0.63 172 Pollen 0.87 0.64 Lyr(1991) Parents 0.91 — — 0.09 — — 16 0.50 — 0.50 16 Progeny 0.88 — — 0.12 — — 3540.44 — 0.56 350 Pollen 0.88 0.58 Woo (1990) Parents 0.95 —— — 0.05 — 11 0.91 — 0.09 11 0.86 0.14 — 11 Progeny 0.92 — — — 0.08 — 190 0.90 — 0.10 127 0.81 0.19 — 189 Pollen 0.87 0.92 0.84 Woo (1991) Parents 0.79 — — — 0.21 — 29 0.93 — 0.07 28 0.90 0.10 — 29 Progeny 0.79 — — — 0.21 — 229 0.95 — 0.05 212 0.82 0.18 — 229 Pollen 0.77 0.97 0.75 Pen (1991) — Parents 0.95 — — — 0.05 10 0.95 0.05 — 10 0.90 — 0.10 10 — Progeny 0.95 — — — 0.05 186 0.95 0.05 — 178 0.90 — 0.10 186 Pollen 0.98 0.98 0.92 Mur (1990) Parents 0.590.09 — 0.32 — — 11 Progeny 0.590.07 — 0.34 — — 203 Pollen 0.59 0.03 0.38 Mur (1991) Parents 0.62 0.05 — 0.33 — — 21 Progeny 0.58 0.01 — 0.41 — — 258 Pollen 0.57 0.01 0.42 Mon (1990) — Parents 0.67 — 0.04 0.29 — 12 0.90 — 0.10 10 — Progeny 0.62 — 0.08 0.30 — 137 0.92 — 0.08 126 Pollen 0.51 0.10 0.39 0.91 Bel (1990) Parents 0.90 — 0.10 10 Progeny 0.93 — 0,07 195 Pollen 0.98 Lon(1991) Parents 0.97 0.03 — 19 Progeny 0.940.06 179 Pollen 0.96

both 1990 and 1991 seasons (Table 5). Therefore Gameticdisequilibrium . . . thesetwo loci may not give independent estimates of The disequilibrium statistic was significantly greater mating system parameters in this population. However, than zero for the association between alleles at the due to violations of the assumption of independence Pgm-1 and Gpi-1 loci in the Lyrebird Gully population in the x2test,conclusions about the nonrandom asso- MATING SYSTEM AND POLLEN DISPERSAL IN P. MOLLIS 147

Table2 Estimates of Wright's fixation index (F), determined from seed for each locus in each population and season

Population Pgm-1 Gpi-1 Adh-1 Mean

Lyr(1990) —0.12(0.18) —0.05 (0.10) —0.09 Lyr(1991) —0,11 (0.17) 0.03 (0.06) —0.04 Woo (1990) 0.04 (0.33) —0.11 (0.23) 0.22 (0.21) 0.05 Woo (1991) 0.01 (0.11) —0.05 (0.26) 0.00(0.12) —0.01 Pen(1991) —0.05 (0.43) —0.05 (0.45) —0.11 (0.25) —0.07 Mur(1990) —0.16 (0.09) —0.16 Mur(1991) —0.03 (0.07) —0.03 Mon (1990) —0.07 (0.10) —0.02 (0.35) —0.05 Bel(1990) —0.01(0.35) —0.01 Lon(1991) —0.06 (0.27) —0.06 Mean —0.06 —0.04 0.03 —0.035

Values are standard errors and the arithmetic mean.

Table 3 Single-locus estimates of outcrossing (t)foreach variable locus in each population and season

Population Pgm-1 Gpi-1 Adh-1 Mean

Lyr(1990) 1.11 (0.05)* 1.07 (0.09) 1.09 Lyr(1991) 1.11 (0.03)** 0.99(0.06) 1.05 Woo(1990) 0.81 (0.32) 1.19(0.36) 1.18 (O,Q3)** 1.06 Woo(1991) 1,11 (0.09) 1.06 (0.22) 0.84 (0.21) 1.00 Pen(1991) 1.05 (0.32) 1.05 (0.32) 1.18 (0.24) 1.09 Mur(1990) 0.96 (0.21) 0.96 Mur(1991) 0.99(0.09) 0.99 Mon(1990) 1.07 (0.19) 1.04 (0.28) 1.06 Bel(1990) 0.91 (0.25) 0.91 Lon(1991) 1.08 (0.32) 1.08 Mean 1.03 1.07 1.03 1.04

Values are standard errors and mean. 1at the 5 per cent level; **>1at the 1 per cent level.

Table 4 Multi-locus estimates of outcrossing rate (t)andstandard errors derived from the procedures of Green eta!. (1980) (tm(G)), Shaw at a!. (1981) (tm(S)), and Ritland & Jam (1981) (tm(R)), for populations with more than one variable locus. Minimum variance mean single-locus estimates (ts(R)), and the difference between the multi- and single-locus estimates (tm(R) —ts(R))for each population and season are given

Population tm(G) tm(S) tm(R) ts(R) tm(R)—ts(R) DOC

Lyr(1990) 1.41 (0.07)** 1.14(0.08) 1.00(0.19) 1.01 (0.13) —0.01 (0.07) 94(172) Lyr(1991) 1.22 (0.07)** 1.03 (0.06) 1.26 (O,O4)** 1.07(0.06) 0.19 (0.06)1 170 (350) Woo (1990) 1.35 (0.15)* 1.20(0.13) # # — 51(127) Woo (1991) 1.01 (0.08) 1.06 (0.09) 1.10 (0.08) 1.13 (0.08) —0.03 (0.03) 80 (212) Pen(1991) 1.40(0.27) 1.38 (0.26) # # — 24(178) Mon (1990) 1.00 (0.10) 1.13 (0.13) 1.29 (0.16)* 1.08 (0.18) 0.21 (0.13) 48 (126) Mean 1.23 1.16 1.16 1.10 0.09

DOC indicates the number of discernible outcrosses detected in the seed, with the total number of seed in parentheses. *>1at the 5 per cent level; t>1at the 1 per cent level; 'indicates tm(R) —ts(R)>0 at the 1 per cent level; # indicates that the estimate of tdidnot converge. 148 S.L. KRAUSS

ciation of alleles at these two loci in this population persal was here treated as normally distributed around cannot be made. No other significant departures from a mean of zero. Ninety-nine per cent of the pollen equilibrium between loci were detected in the other received by given females was donated by males on populations (Table 5). Estimates of mating system average within 33 (±7.3)m. parameters at each of the three loci at all other popula- tions are therefore likely to be independent. Discussion Thecharacterization of the mating system and pollen Pollen dispersal dispersal within the complex species Persoonia mo//is Thespatial distributions of all plants within the indi- has revealed an apparent paradox. Pollen dispersal was cated boundaries of the two natural populations in shown to be extremely restricted. Theoretical models which pollen dispersal was estimated are shown in Fig. 1. and computer simulations show that inbreeding and Paternity could be assigned unequivocally to 61 seeds of a total of 620 seeds (9.8 per cent). The frequency of the seeds with known paternity on each maternal plant and the distances of these maternal plants to the pater- 11 12 nal plant are given in Table 6. On average, 70 per cent LyrebirdGully of all dispersal through pollen was to the paternal creek plant's immediate neighbour and 87 per cent was 10 restricted to the four nearest neighbours. The longest dispersal distance recorded was 72 m, which was the 1 13th closest plant to the paternal plant (Table 6). These 5 results assume no pollen contamination from outside 24 the study area. The close agreement for allele frequen- 6 3 N cies between maternal and pollen cohorts (Table 1) and 13 the extremely restricted nature of pollen dispersal 18 within these populations (Table 6) suggest that pollen contamination from outside the study area was negli- 17 2Dm gible. Estimates of neighbourhood area and effective size due to realized pollen dispersal and the paternity pool area and size are given in Table 7. In all populations, 18 is kurtosis was minor but due to small sample sizes no 16 17 test of significance was possible. Therefore, pollen dis- 14 15 LongAcre Table 5 Estimates of gametic disequilibrium ABforeach population in which multi-locus outcrossing rates were 1213 determined S 11

Population (year) loci AAB SD g 10 Lyr (1990) Pgm/Gpi 0.08730.0107 100.30*** 4 8 Lyr(1991)Pgm/Gpi 0.03720.0088 19.06*** 3 2 7 Pen(1991)Pgm/Adh —0.0023 0.0041 0.28 Pen (1991)Pgm/Gpi —0.0018 0.0027 0.30 Pen(1991) Gpi/Adh —0.0066 0.0032 2.28 I I Woo(1990)Pgm/Gpi —0.0075 0.0044 1.64 6 8m Woo(1990)Gpi/Adh 0.01160.0086 2.29 Woo (1990) Pgm/Adh — 0.0099 0.0059 2.19 Fig. I Spatial distribution of plants in the two undisturbed Woo (1991) Pgm/Gpi 0.00050.0050 0.01 natural populations in which realized pollen dispersal was Woo(1991)Gpi/Adh 0.00050.0048 0.01 measured (Lyrebird Gully (Lyr) and Long Acre (Lon)). Woo (1991) Pgrn/Adh 0.00930.0098 0.82 Plants possessing unique alleles are indicated. Note that in Mon (1990) Pgm/Gpi 0.00640.0 159 0.20 1991 in the Lyrebird Gully population, plants 4 and 7 had died and that not all plants were included in the 1990 esti- AAB significantly different from 0 (P <0.001). mates. See also Table 6. Table 6 Number of heterozygous seed (i.e. with known paternity) and the total number of seed genotyped for each homozygous (AA ) maternal plant

Plant Dist.1 No. of seed(total) Plant Dist.2 No. of seed(total) Plant Dist.3 No. of seed(total) Plant Dist.4 No. of seed(total)

10 7 19(28) 15 4 4(24) 10 7 17(17) 2 3.5 3(24) 11 18 1(7) 17 5 1(24) 7 13 0(7) 6 10 0(7) 12 30 2(20) 13 7 0(33) 12 30 1(14) 7 10 0(8) 9 32 2(36) 14 8 1(30) 5 37 0(17) 3 14 2(7) 5 37 1(20) 3 15 0(20) 4 41 0(15) 4 14 0(4) 1 45 0(20) 2 18 2(20) 1 45 0(16) 5 14 1(8) 2 51 0(20) 6 18 1(14) 2 51 0(20) 8 20 0(8) 3 52 0(20) 18 22 0(24) 3 52 0(20) 9 25 0(8) 18 52 0(24) 1 28 0(20) 6 72 0(20) 10 25 0(8) 13 60 0(33) 5 31 0(20) Totall8/146 11 25 0(8) 14 61 0(30) 9 46 2(36) 14 25 0(8) 15 71 0(24) 10 74 0(28) 15 27 0(8) 6 72 1(14) 11 80 0(7) 12 28 0(8) 17 72 0(24) 12 97 0(20) 16 28 0(8) Tota126/320 Total=11/320 17 29 0(8) 13 30 0(8) 18 35 0(8) 19 40 0(8) Total 6/154 z Dist.1 and Dist.2 are the distances of AA maternal plants from the DD (plant 8) and AD (plant 16) paternal plants, respectively, in the Lyrebird Gully (1991) C) Dist.3 is the distance of AA maternals from the DD 8) paternal in the Lyrebird Gully (1990) population. Dist.4 is the distance of AA maternals C', population. (plant -< from the AB in the Acre (1991) C') (plant 1)patemal Long population. -1 All distances are in meters. See also 1. Fig. z> 0 0-o I- 2m 0 C', m

C,) I- 2

0 N - CD 150 S. L. KRAUSS

Table 7Neighbourhood and paternity pooi area and size estimates derived from realized pollen dispersal distances

Population/year/sourcenArea d u Na Ne 'a P, Lyr 1990 (DD) 1840000.003 52.9 3321.01496 4.5 Lyr 1991 (DD) 2665000.002299.818843.8847716.9 Lyr1991(AD) 1165000,002190.911992.4539810.8 Lon 1991 (AB) 620000.01 77.3 4864.9218621.9 Mean 9753.0438913.5

'Source' refers to the genotype of the paternal plant; n is the number of seeds that were known to be fathered by the source plant; area is the total area of the study population in m2, with density (d) in plants/rn2; a is the axial variance of pollen dispersal; Na and Ne are the neighbourhood area and effective size estimates respectively assuming normally distributed dispersal; 'a and P., are the paternity pooi area (in m2) and size, respectively.

excess homozygosity are expected from such severely directly correlated to the amount of outcrossing. restricted pollen dispersal (Brown, 1979, 1990; Turner Therefore, complete outcrossing within P. molhis etal.,1982). In contrast to theoretical expectations means that the mating system promotes gene flow. though, these P. mollis populations were found to be Some of the single- and multi-locus estimates of t completely outcrossing. Indeed, an excess of hetero- weresignificantly in excess of unity (Tables 3 and 4) and zygotes was found in the progeny cohort. Three are therefore 'biologically unreasonable'. Violations of possible explanations, namely (i) disassortative mating, one or more of the assumptions of the mixed-mating (ii) heterozygote advantage, and (iii) seed dispersal model (Clegg, 1980; Brown et at., 1985) will cause esti- distances in excess of pollen dispersal, are discussed in mates of ttodiffer from their true values (Brown, 1979, detail below. The latter appears the most likely 1990). One assumption is that the pollen involved in explanation of the apparent paradox and suggests that each outcrossing event is drawn at random from a restricted gene flow is not a factor contributing to the uniform population of pollen (i.e. a randomly mating morphological complexity within P. moths. population) (Clegg, 1980; Brown et at., 1985). Esti- mates of realized pollen dispersal show that this assumption is violated in P. mo//is populations where Matingsystem parameters 70 per cent of all pollen dispersal from one plant is on Ratesof outcrossing (t) in Persoonia mo/his are among average to its immediate neighbour (Table 6). the highest recorded (Schemske & Lande, 1985; Deviation from the mixed-mating model assumption Brown, 1979, 1990). P. mo/his is clearly completely of uniform pollen frequency over maternal genotypes outcrossing in all populations investigated. Elsewhere, can also be caused by subpopulation structure (Shaw I have used experimental self- and cross-pollinations to et at., 1981; Ennos & Clegg, 1982; Hamrick, 1982; show that these outcrossing rates are high because of a Ellstrand & Foster, 1983). Some contrived types of 'pseudo' self-incompatibility mechanism preventing subpopulation structure, such as over-dispersed most but not all self-pollen tube growth in the style plantations of genotypes, have led to upward biases in (Krauss, 1994). Even following successful pollination, estimates of t(e.g.Ellstrand & Foster, 1983). However, there is a strong preference for outcross pollen over an over-dispersed population structure is unlikely to be self pollen both in terms of quantity and quality of the detected with the present data set due to small popula- seeds produced (Krauss, 1994). Fruit-set following self- tion size, few loci and the generally asymmetric allele ing was 5 per cent that of outcrossed flowers, while frequencies. Furthermore, it is unlikely that disassorta- fruit weight following selfing was 70 per cent that of tive mating and over-dispersion could be the case here outcrossed flowers (Krauss, 1994). The current study for all loci in all populations. indicates that its pollinators are effectively providing P. In one population (Lyrebird Gully 1991), the multi- mo//is plants with the pollen they prefer, in the sense locus estimate of t(1.26 was significantly that all seeds are the product of fertilization between higher than the mean single-locus estimate of t the maternal ovule and a male gamete of different (1.070.06), which suggests consanguineous mating genotype. The amount of gene flow through pollen is (i.e. mating among close relatives) (Ritland & Jam, MATING SYSTEM AND POLLEN DISPERSAL IN P. MOLLIS 151

1981; Shaw et a!., 1981). Over all populations, the observation provides support for the hypothesis of mean multi-locus estimate of t was approximately 1.2, extremely restricted pollen dispersal within P. mo/us. while the overall mean single-locus estimate of t was The estimates of paternity pool size for P. mo//is are 1.04. However, there is no unambiguous genetic among the smallest recorded (Levin, 1988). These evidence to suggest the presence of mating among small paternity pools, and the high incidence of near- close relatives within these populations. Although the est-neighbour pollination (on average 70 per cent of all relative longevity of P. mollis plants provides the pollinations), indicate clearly that each of the sampled opportunity for parents and their offspring to co-exist populations in this study does not constitute a panmic- and therefore mate, this is unlikely as fire, which kills tic unit. These populations are effectively fractured adult plants, appears to promote seed germination into smaller, overlapping paternity pools. Small pater- (Abbott & van Heurck, 1988). Seed dormancy nity pools and nearest-neighbour pollination result in mechanisms within Persoonia (Abbott & van Heurck, differences in allelic frequencies in the effective pollen 1988) that delay germination for up to 3 years (Fox et pooi of each plant within each population (i.e. the al., 1987), combined with extremely low to negligible Wahlund effect), as well as partial inbreeding through seed germination rates (Abbott & van Heurck, 1988; mating among inbred relatives (i.e. consanguineous personal observation), substantially decrease the matings) (Hedrick, 1985). These two factors (Wahlund numbers of siblings germinating together. However, effect and consanguineous matings) should rapidly pro- assuming a normal distribution of seeds, some sib- duce low levels of heterozygosity, high fixation indices mating would be expected. More precise estimates of (F) and low outcrossing rates (t) (Brown, 1979, 1990; mating system parameters are required to determine (i) Turner et a!., 1982; Hedrick, 1985). Contrary to the extent of apparent selfing due to consanguineous expectations, Persoonia mo//is exhibits complete out- matings (e.g. Ritland, 1984) in contrast to (ii) disassor- crossing and excess heterozygosity compared with tative mating as an explanation of the estimates of t in Hardy—Weinberg expectations in the seed cohort excess of unity, as discussed above. (F =—0.035).Two explanations for the difference in homozygosity for the model and for P. mo//is are (i) there is selection for heterozygous seeds over homo- Paternitypool zygous seeds during seed maturation, or (ii) seed dis- Asonly one or two individuals in these populations persal is much greater than pollen dispersal. carry rare markers, generalizations to patterns of effec- In P. mollis it may be that the mating system is being tive pollen dispersal may be difficult to make (Hamrick maintained at, and in excess of, complete outcrossing & Schnabel, 1985; Adams eta!., 1992). This is particu- by the evolutionary pressure of heterozygote advan- larly so if the rarity of an allele is due to selection tage. Increased fitness associated with increased against it. However, in both the Lyrebird Gully and heterozygosity has been demonstrated in a number of Long Acre populations, the close agreement of the rare plant species (e.g. Ledig, 1986; James & Kennington, allele frequencies in the maternal and progeny cohorts 1993), although other studies have also shown no (Table 1) suggests that selection against the rare allele, relationship (Mitton & Grant, 1984). The classical at least prior to germination of seed, is insignificant. expectation is that high levels of outbreeding should be Combined with outcrossing rates at unity, these data favoured when heterozygosity enjoys an advantage further suggest no loss of pollen viability of paternal (Maynard Smith, 1977). However, in many outbreed- plants relative to other plants in these populations and ing species a general deficit of heterozygotes compared therefore that the paternal plants are contributing with panmictic expectations has been found (Brown, equally to the pollen pool. Although the use of rare 1979). This has been termed the heterozygosity para- alleles is an accurate means of estimating realized dox (Brown, 1979). Conversely, a general excess of pollen dispersal, the method is restricted to special cir- heterozygotes has been found in many inbreeding cumstances in natural populations. As such, generaliza- species. The most favoured explanation for excess tions may be difficult to support. However, it was the heterozygosity among inbreeders is heterozygote most appropriate technique to measure realized pollen advantage for chromosomal segments that include the dispersal here, as a lack of many polymorphic loci marker loci (Brown, 1979). The extent of heterozygote ruled out the use of a paternity analysis approach (e.g. advantage in the seed cohort in P. mo//is needs to be Adams et a!., 1992). Even though only a small number more fully investigated, particularly as a study of of paternal plants from different subspecies in popula- genetic structure within and among 18 populations of tions of different densities were used to estimate effec- this species suggests no excess of heterozygotes tive population size parameters in this study, only a (F =0.04)in the adults (S. L. Krauss, unpublished small range was found in these estimates (Table 7). This data). 152 S. L. KRAUSS

Commission of New South Wales permit nos 3895 and Neighbourhood size 4565. This work was funded by an Australian Post- Theestimates of genetic neighbourhood size from graduate Research Award. Assistance from the Royal pollen flow for P. mo//is are among the smallest Botanic Gardens, Sydney, and support from the recorded (Crawford, 1984b; Cahalan & Gliddon, Linnean Society of New South Wales via a Linnean 1985; Bos et a!., 1986; Govindaraju, 1988; Levin, Macleay Fellowship and two Joyce W. Vickery 1988; Palmer et at., 1988; Young, 1988; Epperson & Research Fund grants and the Ecological Society of Allard, 1989; Broyles & Wyatt, 1991; Fenster, 1991; Australia via a Student Travel Grant are gratefully Eguiarte et at., 1993). These population sizes, on the acknowledged. David Ayre, Tony Brown, Libby basis of pollen dispersal, are small enough for genetic Howitt, Rod Peakall, David Skelly, Nick Waser, Peter differentiation in the absence of selection to occur Weston, Rob Whelan and an anonymous reviewer (Wright, 1943, 1946; Endler, 1977) and for random commented on and substantially improved earlier fluctuations in gene frequency to be important in drafts of this manuscript and I thank David Ayre, Peter evolution. However, a formal estimation of the genetic Weston and Rob Whelan for their advice and super- neighbourhood size in P. moltis must also include the vision. This is publication no. 121 from the Ecology contribution of seed dispersal. and Genetics Group of the University of Wollongong. Pollen dispersal has been shown to be much more important than seed dispersal in many bee-pollinated plants (e.g. Levin & Kerster, 1974; Beattie & Culver, 1979; Fenster, 1991; Waser, 1993). However, in these References other studies seed dispersal has been relatively passive, or with short-range explosive mechanisms, or with ABBOTF,I. AND VAN HEURCK, t'. 1988. Widespread regeneration short-range dispersers. P. moths fruits are fleshy failure of Persoonia elliptica (Proteaceae) in the northern drupes that are primarily gravity dispersed and secon- of , f. R. Soc. W. Aust., 71, darily eaten and dispersed by large birds that have 15—22. often large ranges. The potential, then, is for long- ADAMS, W. T., GRIFFIN, A. R. AND MORAN, G. F. 1992. Using pater- distance dispersal of fruits and consequently dispersal nity analysis to measure effective pollen dispersal in plant distances substantially greater than those of pollen. populations. Am. Nat., 140, 762—780. BARKER, R. D. AND VESTJENS, W. J. M. 1990. The Food of Austral- Consequently, the estimates of neighbourhood area ian Birds, vol. 2. Passerines. CSIRO, Canberra. and size may be severely underestimated and seed dis- BEA'VFIE, A. I. AND CULVER, 0. c. 1979. Neighbourhood size in persal may be a substantially larger component of gene Viola. Evolution, 33, 1226—1229. flow than pollen dispersal. DOS, M., HARMENS, H. AND VREILING, K. 1986. Gene flow in Therefore, pollen dispersal creates paternity pools Plantago. I. Gene flow and neighbourhood size in P. of extremely small size that severely restrict gene flow lanceolata. Heredity, 56, 43—54. among P. moths individuals. However, seed dispersal BROwN,A. H. ID. 1979. Enzyme polymorphism in plant popula- and the mechanisms affecting germination, coupled tions. Theor. Pop. Biol., 15, 1—42. with outcrossing rates at unity, appear to promote gene BRowN, A. H. 0. 1990. Genetic characterization of plant mating flow and ultimately a more homogeneous population systems. In: Brown, A. H. D., Clegg, M. T., Kahler, A. L. structure. One interpretation is that gene flow is and Weir, B. S. (eds) Plant Population Genetics, Breeding and Genetic Resources, pp. 145—162. Sinauer Associates, sporadic, with dispersal by pollen and seed over Sunderland, MA. extremely short distances occurring most of the time BROWN, A. H. D., BARRETF, S. C. H. AND MORAN, G. F, 1985. Mating but with occasional dispersal by seed over much longer system estimation in forest trees: models, methods and distances occurring frequently enough to produce meanings. In: Gregorius, H. R. (ed.) Population Genetics in genetic homogeneity. Indirect estimates of gene flow Forestry, pp. 32—49. Springer, Berlin. from data on population structure would be expected BROWN, A. H. D., MATHESON, A. C. AND ELIDRIDGE, K. ci. 1975. to show estimates of gene flow in excess of these direct Estimation of the mating system of Eucalyptus obliqua estimates derived from data on realized pollen disper- L'Hérit. by using allozyme polymorphisms. Aust. J. Bot., sal alone (Slatkin, 1985). Preliminary results of indirect 23, 93 1—949. estimates of gene flow within P. mo//is from isozyme BROYLES, S. B. AND WYA'flT, R. 1991. Effective pollen dispersal in data appear to support this interpretation. a natural population of Asclepias exaltata: the influence of pollinator behaviour, genetic similarity and mating success. Am. Nat., 138, 1239—1249. Acknowledgements CAHALAN, C. M. AND GLIDDON, C. 1985. Genetic neighbourhood sizes in Primula vulgaris. Heredity, 54, 65—70. Fieldwork was undertaken under N.S.W. National CAMPBELL, D. R. 1991. Comparing pollen dispersal and gene Parks and Wildlife permit no. B/403 and Forestry flow in a natural population. Evolution, 45, 1965—1968. MATING SYSTEM AND POLLEN DISPERSAL IN P. MOLLIS 153

CHELIAK, W. M., MORGAN,K.,STROBECK, C., YEH, F. C. H. AND DANCIK, of Eucalyptus camaldulensis Dehnh. Aust. .1. Bot., 41, B. i'. 1983. Estimation of mating system parameters in 381—391. plant populations using the EM algorithm. Theor. Appi. KRAUSS, s. L. 1994. Preferential outcrossing in the complex Genet.,65, 157—161. species Persoonia mollis R. Br. (Proteaceae). Oecologia, CLEGG, M. T. 1980. Measuring plant mating systems. Bio- 97,256—264. Science, 30,814—818. KRAUSS, 5. L. AND JOHNSON, L. A. 5. 1991. A revision of the CRAwFoRD, T. ..1984a.The estimation of neighbourhood complex species Persoonia mollis (Proteaceae). Telopea, parameters for plant populations. Heredity, 52, 273—283. 4,185—199. CRAWFORD, T. i. 1984b. What is a population? In: Shorrocks, LEDIG, F. T. 1986. Heterozygosity, heterosis and fitness in out- B. (ed.) Evolutionary Ecology, pp. 135—173. Blackwell breeding plants. In: Soulé, M. E. (ed.) Conservation Scientific Publications, Oxford. Biology, the Science of Scarcity and Diversity, pp. 77—104. EGUIARTE, L. E., BURQUEX, A., RODRIGUEZ, J., MARTINEZ-RAMOS, M., Sinauer Associates, Sunderland, MA. SARUKHAN, J. AND PINERO, D. 1993. Direct and indirect esti- LEVIN, D. A. 1981. Dispersal versus gene flow in plants. Ann. mates of neighbourhood and effective population size in a Mo. Bot. Gard., 68, 233—25 3. tropical palm, Astrocayum mexicanum. Evolution, 47, LEVTN, D. A. 1988. The paternity pool of plants. Am. Nat., 75—87. 132, 309—317. EHRLICH, P. R. AND RAVEN, P. H, 1969. Differentiation of popula- LEVIN, D. A. AND KERSTER, Fl. W. 1974. Gene flow in seed plants. tions. Science, 165, 1228—1232. Evolutionary Biology, 7, 139—220. ELLSTRAND, N. C. AND FOSTER, K. w. 1983. Impact of population LOVELESS, M. D. AND HAMRICK, J. L. 1984. Ecological deter- structure on the apparent outcrossing rate of grain minants of genetic structure in plant populations. Ann. sorghum (Sorghum bicolor). Theor. Appi. Genet., 66, Rev. Ecol. Syst., 15, 65—95. 323—327. MAYNARD SMITH, J. 1977. The sex habit in plants and animals. ENDLER, J. A. 1977. Geographic Variation, Speciation and In: Christiansen, F. B. and Fenchel, T. M. (eds) Measuring Clines. Princeton University Press, Princeton. Selection in Natural Populations, pp. 315—331. Springer, ENNOS, R. A. AND CLEGG, M. T. 1982. Effect of population sub- Berlin. structuring on estimates of outcrossing rate in plant popu- MIrFON, J. B. AND GRANT, M. C. 1984. Associations among lations. Heredity, 48, 283—292. protein heterozygosity, growth rate and developmental EPPERSON, B. K. AND ALLARD, R. w. 1989. Spatial autocorrelation homeostasis. Ann. Rev. Ecol. Syst., 15,479—499. analysis of the distribution of genotypes within popula- PALMER, M., T1AVIS, J. AND ANTONOVICS, i. 1988. Seasonal pollen tions of Lodgepole Pine. Genetics, 121, 369—377. flow and progeny diversity in Amianthium muscaetoxi- FENSTER, C. B. 1991. Gene flow in Chamaecrista fasciculata cum: ecological potential for multiple mating in a self- (Leguminosae).I. Gene dispersal. Evolution, 45, incompatible, hermaphroditic perennial. Oecologia, 77, 398—409. 19—24. FOX, J., DIXON, B. AND MONK, 0. 1987. Germination in other RITLAND, K. 1984. The effective proportion of self-fertilization plant families. In: Langkamp, P. (ed.) Germination of with consanguineous matings in inbred populations. Australian Native Plant Seed, pp. 83—97. Inkata Press, Genetics, 106,139—152. Melbourne. RITLAND, K. AND JAIN, S. K. 1981. A model for the estimation of FYFE, J. L. AND BAILEY, N. T. i. 1951. Plant breeding studies in outcrossing rate and gene frequencies using n indepen- leguminous forage crops. 1. Natural cross-breeding in dent loci. Heredity, 47, 35—52. winter beans, J. Agric. Sci., 41,371—378. SCHAAL, B. 1980. Measurement of gene flow in Lupinus texen- GOVINDARAJU, D. R. 1988. Life histories, neighbourhood sizes sis. Nature, 284, 450—45 1. and variance structure in some North American conifers. SCHEMSKE, U. W. AND LANDE, R. 1985. The evolution of self- Biol. J. Linn. Soc., 35, 69—78. fertilization and inbreeding depression in plants. II. GREEN,A. G., BROWN, A.H. D. AND ORAM, R. N. 1980. Determina- Empirical observations. Evolution, 39, 4 1—52. tion of outcrOssing rate in a breeding population of SELANDER, R. K., SMITH, M. H., YANG, S. Y., JOHNSON, W. E. AND Lupinus albus L. (White Lupin). Z. Pflanzenzüchtg., 84, GENTRY, .i. B. 1971. Biochemical polymorphism and system- 18 1—19 1. atics in the genus Peromyscus. I. Variation in the old-field HAMRICK, i. L. 1982. Plant population genetics and evolution. mouse (Peromyscus polionotus). In: Studies in Genetics 6, Am. J. Bot., 69,1685—1693. pp.49-60. Univ. Texas PubI. 7103. HAMRICK, J. L. AND SCHNABEL, A. 1985. Understanding the SHAW, P. V., KAHLER, A. L. AND ALLARD, It w. 1981. A multilocus genetic structure of plant populations: some old problems estimator of mating system parameters in plant popula- and a new approach. In: Gregorius, H. R. (ed.) Population tions. Proc. Nail. Acad. Sci. USA., 78, 1298—1302. Genetics in Forestry, pp. 50—70. Springer, Berlin. SLATKIN, M. 1985. Gene flow in natural populations. Ann. Rev. HANDEL, 5. 1983. Pollination ecology, plant population struc- Ecol. Syst., 16,393—430. ture and gene flow. In: Real, L. (ed.) Pollination Biology, SLATKIN, M. 1987. Gene flow and the geographic structure of pp. 163—211. Academic Press, Orlando, FL. natural populations. Science, 236,787—792. HEDRICK, p, w. 1985. Genetics of Populations. Jones and Bart- SOLTI5, D. E., HAUFLER, C. H., DARROW, D. C. AND GASTONY, G. J. lett Publishers, Boston. 1983. Starch gel electrophoresis of ferns: a compilation of JAMES, S. H. AND KENNINGTON, w. j.1993.Selection against grinding buffers, gel and electrode buffers and staining homozygotes and resource allocation in the mating system schedules. Am. Fern J., 73, 9—27. 154 S.L.KRAUSS

TEMPLETON, A. R. 1989. The meaning of species and specia- WEIR, B. s. 1990. Genetic Data Analysis. Sinauer Associates, tion: a genetic perspective. In: Otte, D. and Endler, J. A. Sunderland, MA. (eds) Speciation and its Consequences, pp. 3—27.Sinauer WEIR, B. S. AND COCKERHAM, C. C. 1989. Complete charac- Associates, Sunderland, MA. terization of disequilibrium at two loci. In: Feldman, M. E. TURNER, M. E., STEPHENS, J. C. AND ANDERSON, W. w. 1982. (ed.) Mathematical Evolutionary Theory, pp. 86—110. Homozygosity and patch structure in plant populations as Princeton University Press, Princeton. a result of nearest-neighbour pollination. Proc. NatI. A cad. WENDEL, J. F. AND WEEDEN, N. F. 1989. Visualization and inter- Sci. U.S.A., 79, 203—207. pretation of plant isozymes. In: Soltis, D. E. and Soltis, WALLER, D. M. AND KNIGHT, S. E. 1989. Genetic consequences of P. S. (eds) Isozymes in Plant Biology, pp. 5—44. Dioscorides outcrossing in the cleistogainous annual, Impatiens Press, Portland, OR. capensis. III. Interlocus associations. Heredity, 63, 1—9. WRIGHT, 5. 1943. Isolation by distance. Genetics, 28, WASER, N. M. 1993. Population structure, optimal outbreeding 114—138. and assortative mating in angiosperms. In: Thornbill, N. WRIGHT, 5. 1946. Isolation by distance under diverse systems W. (ed.) The Natural Histoiy of Inbreeding and Outbreed- of mating. Genetics, 31, 39—59. ing: Theoretical and Empirical Perspectives, pp. 173—199. WRIGHT, 5. 1969. Evolution and the Genetics of Populations. University of Chicago Press, Chicago, IL. vol. 2, The Theory of Gene Frequencies. University of WASER, N. M. AND PRICE, M. v. 1982. A comparison of pollen and Chicago Press, Chicago. fluorescent dye carry-over by natural pollinators of YOUNG, H. ..1988.Neighbourhood size in a beetle pollinated Ipomopsis aggregata. Ecology, 63, 1168—1172. tropical aroid: effects of low density and asynchronous flowering. Oecologia, 76, 46 1—466.