Ecology, 77(8), 1996, pp. 2382-2392 ? 1996 by the Ecological Society of America PHENOTYPIC VARIATION IN A ZOOPLANKTON EGG BANK' NELSON G. HAIRSTON, JR. AND COLLEEN M. KEARNS Section of Ecology and Systematics, Cornell University, Ithaca, New York 14853-2701 USA STEPHEN P. ELLNER Biomathematics Program, Department of Statistics, North Carolina State University, Raleigh, North Carolina 2 7695-8203 USA Abstract. Dormant propagule pools may store potentially significant genetic variation that can influence the rate and direction of microevolution via directional selection, tem- porally fluctuating selection, and evolution of trait covariance between timing of emergence from the propagule pool and fitness characters expressed in the active population. The third process can interact with either of the first two to produce distinct effects. Each process can lead to a different distribution of genotypes and phenotypes between active and dormant subpopulations. We compared the phenotypic distributions of an important fitness character for individuals collected from active and diapausing subpopulations of a freshwater co- pepod, Diaptomus sanguineus, with a long-lived egg bank. The character, seasonal timing of the switch from production of immediately hatching eggs to diapausing eggs, determines the relative representation of copepods with different switch dates in future generations and is subject to fluctuating selection due to year-to-year changes in the timing and intensity of the seasonal onset of fish predation. The mean timing of diapause is significantly later in the season for copepods reared from long-lived diapausing eggs than it is for copepods reared from individuals collected from the water column. Phenotypic variance for diapause timing does not differ between the two subpopulations. Within the sediment subpopulation, the distribution of diapause timing depends upon two features of the diapausing eggs: (1) individuals originating from eggs near the sediment surface exhibit a slightly earlier switch date with greater phenotypic variance than individuals from deep in the sediments, and (2) individuals from eggs that hatched shortly after they were collected from sediments have a later seasonal switch to diapause than those that hatched later in time. We hypothesize that our results are explained by adaptive covariance between traits that influences how long an egg spends in the sed- iments before hatching and traits that influence the seasonal timing of diapause. The co- variance may result from either phenotypic plasticity or from genetic covariance between diapause timing and hatching probability. Key words: copepod; diapause timing; diapausing eggs; Diaptomus sanguineus; dormant prop- agule pool; egg bank; evolution; fluctuating selection; photoperiod, diapause timing; temperature, diapause timing; trait covariance. INTRODUCTION to maintain genetic variation when selection fluctuates An extraordinary diversity of organisms, including through time by providing a mechanism of temporal representatives from all kingdoms, produce long-lived bet hedging distinct from maximization of geometric dormant stages such as cysts, spores, seeds, or eggs mean fitness (Chesson 1984, Seger and Brockmann capable of surviving for multiple years or decades 1987, Ellner and Hairston 1994). Third, the duration (Hairston et al. 1996). In a few instances, survival for of dormancy may itself evolve as a function of traits centuries or even millennia has been documented (Hair- expressed in the active stage, which may in turn alter ston et al. 1996). The effects of these dormant prop- the direction or intensity of selection on the active traits agule pools on evolutionary dynamics within popula- (Brown and Venable 1986, Rees 1994). tions are likely to be substantial, particularly when the Reciprocally, both directional and fluctuating selec- distribution of dormant genotypes is different from that tion should result in a gradual accumulation of geno- in the active population. Three distinct effects have types in the dormant propagule pool that differ from those expressed in the active population at any one been identified in the theoretical literature. First, emer- time. Directional selection should lead to the mean ge- gence of dormant stages may slow down response rate notype in the propagule pool being different from that to directional selection by increasing average genera- in the active population, whereas fluctuating selection, tion time (Templeton and Levin 1979, Brown and Ven- in contrast, should result in greater genotypic variance able 1986). Second, dormant propagule pools may act in the propagule pool than in the active population. In I Manuscript received 21 June 1995; revised 8 March 1996; the case of evolution of covariance between dormancy accepted 13 March 1996. duration with other traits, the effects on the distribution 2382 December 1996 VARIATION IN A ZOOPLANKTON EGG BANK 2383 of genotypes must depend upon the specifics of the ferences between a dormant propagule pool and active system. For example, Rees' (1994) suggestion that members of a population. McGraw and his colleagues adult life span should covary negatively with duration have provided the only investigations of this kind (Ben- of dormancy does not dictate any difference between nington et al. 1991, Vavrek et al. 1991, McGraw 1993). the genotype distributions of the adult life span trait Capitalizing on the presence of viable plant seeds bur- for active and dormant individuals. However, Evans ied in arctic tundra for up to 150 yr (McGraw et al. and Cabin's (1995) proposal that "adaptive syn- 1991), they compared various performance measures dromes" associate genes for terminating dormancy in (e.g., plant size, growth rate, competitive ability) for a particular type of year (i.e., environment) with genes three species. In each case, mean phenotype (and, pre- for performance in that year type, should lead to greater sumably genotype) differed between seed-bank and ac- genetic variation in the dormant pool than expressed tive individuals. This suggests directional selection has at any one time in the active population. taken place over the period from production of the Despite the potential evolutionary importance of oldest seeds to the present (Bennington et al. 1991, prolonged dormancy, only a few studies have compared Vavrek et al. 1991, McGraw 1993). At the same time, the genotypes or phenotypes of dormant propagules it is interesting to note that although these investigators with active individuals from single populations. Most uncovered substantial phenotypic variance within the of the research in this area has been either with plant different subpopulations (seed bank vs. active), the ab- seed banks or with freshwater zooplankton egg banks. solute level of variation did not appear to be greater Dormant seeds may promote genetic constancy in an- in the seed-bank-derived individuals than in the active nual plant populations (Epling et al. 1960, Gottlieb ones for any of the species (although the authors did 1974), or similarly, zooplankton diapausing eggs may not discuss this comparison explicitly). This result sug- slow response to directional selection (Hairston and De gests that neither fluctuating selection nor evolution of Stasio 1988). Bosbach et al. (1982) found greater allo- adaptive syndromes was important over the period zyme diversity in annual plant populations germinating studied, or that seeds, once buried in the permafrost, from recently plowed soils than in those germinating contributed little or nothing to future population dy- on less disturbed sites and interpreted this as evidence namics. Evans and Cabin (1995), studying a desert an- for stored genetic variation in the seed bank. Compar- nual plant, found that there was significant genetic vari- isons of allozymic diversity between relatively short- ation for both tendency to germinate in different en- lived (3-6 or 7 yr) seed banks and growing plants yield vironments and for post-germination performance. conflicting results. Tonsor et al. (1993) showed differ- They speculated that selection acting on both sub- ences in allele frequencies but not in total genetic vari- populations could lead to the evolution of covariance ance between the two components of the population, between seed bank and post-germination traits. and concluded that the seed bank had little effect on An opportunity to explore variation stored in a dor- microevolutionary dynamics in their population. In mant propagule pool is presented by a population of contrast, Evans and Cabin (1995) found substantial ge- calanoid copepods, Diaptomus sanguineus, living in netic differences between the germinated and dormant Bullhead Pond, a small lake in southern Rhode Island. subpopulations, though their results were potentially This copepod population is active under the ice during confounded by spatial structure. Two investigations of winter when it makes eggs that hatch immediately. In zooplankton electrophoretic variation found no statis- spring each year, the copepods switch to making dia- tically significant differences in allozyme heterozy- pausing eggs (Hairston and Munns 1984). For this pop- gosity between water-column-derived and sediment- ulation, we know that the annual spring onset of fish derived subpopulations (Wolf and Carvalho 1989; M. predation acts as a selection force on the timing of G. Boileau and N. G. Hairston, Jr., unpublished manu- diapause initiation (Hairston and Munns 1984, Hairston script). It is not clear, however, what to conclude from 1987), that onset of fish predation (and hence selection these allozyme studies, whether plant or zooplankton. on optimal diapause timing) fluctuates between years Though expectation of genetic differences between (Hairston 1988), that diapause timing is heritable under dormant propagules and active individuals in any given both laboratory and field conditions (Hairston and Dil- population is based on the proposition that selection is lon 1990), and that the mean phenotype of the popu- acting, electrophoretically identified genotypes are fre- lation shifts between years in response to selection quently assumed to be selectively neutral (e.g., Kimura (Hairston and Dillon 1990). In addition, the diapausing 1983, Futuyma 1986). If they are neutral, year-to-year eggs produced by the copepods are present in high changes via genetic drift could still result in an accu- density (>105 eggs/M2) in the lake sediments (De Stasio mulation of genetic variation in the dormant stages, but 1989), and many of these eggs are quite old (mean age only when the population size of active, breeding in- 70.4 yr, maximum age 330 yr; Hairston et al. 1995). dividuals is small. This was distinctly not the case in We hypothesize that the pool of eggs of D. sanguineus any of the studies just cited. present in the sediments of Bullhead Pond possesses a Phenotypic characters of demonstrable fitness sig- distribution of genotypes distinct from that expressed nificance provide a better way to explore genetic dif- in the water column in any given year. 2384 NELSON G. HAIRSTON, JR. ET AL. Ecology, Vol. 77, No. 8 The character upon which selection acts is the sea- 12.00 x 6.00, 12.75:11.25 x 8.0?, and 13.25:10.75 x sonal timing of the switch from immediately hatching 10.50. Illumination was provided by daylight fluores- eggs to diapausing eggs. For this copepod species, dia- cent lamps (17-38 pu.mol.s-' m-2 depending upon lo- pause timing is cued by a combination of photoperiod cation within the growth chamber). and temperature (Hairston and Olds 1986, Hairston et Copepods active in the water column, and egg-con- al. 1990, Hairston and Kearns 1995). Here we compare taining sediments were collected on 2 March 1993 and the egg-type distributions produced in response to pho- 16 June 1993, respectively. Water-column animals toperiod and temperature cues by female D. sanguineus (fifth instar copepodids) were obtained by plankton net drawn from copepods either hatched from the sediment before the population had begun to produce diapausing egg bank, or from eggs produced by water-column in- eggs. They were returned to the laboratory and 144 dividuals. We show that the mean timing of diapause females were placed individually in 125-mL glass jars, differs significantly between these two subpopulations, with two males per jar for mating, under short-day (L: but that the variances of egg bank and water-column D, 8:16) and low temperature (4.8?C) environments to copepods are similar. In addition, there are differences ensure that all clutches produced were immediately in the distribution of diapause between copepods hatching (Hairston and Olds 1986, 1987). One hundred hatched from young eggs obtained from near the sed- twenty-one females became ovigerous over a 17-d pe- iment-water interface and copepods hatched from older riod. An additional 82 egg-bearing females were ob- eggs obtained from deeper in the sediment, as well as tained over the same period from the stock carboy in differences in diapause timing among copepods that which plankton were transported from Bullhead Pond. hatch readily in the laboratory and those that take lon- All ovigerous females were isolated in filtered lake ger to respond to favorable hatching conditions. Our water in 7-mL wells of 12-well plastic tissue culture data appear to show covariance between time spent in plates until their eggs had hatched. diapause and seasonal timing of onset of diapause. Lake sediments were obtained from the lake center by scuba using six benthic core tubes (16 cm diameter, MATERIALS AND METHODS 15 cm deep). In the laboratory, the top 5 cm of sediment We measured the distributions of diapause timing in was removed from each core in two layers (0-2 and the laboratory by rearing copepods under photoperiods 2-5 cm), with each layer analyzed separately. We chose and temperatures chosen to mimic natural conditions 5 cm as the maximum depth to analyze because viable near the time that the population of Diaptomus san- egg density is substantially lower below 5 cm compared guineus naturally switches from producing immediate- with that above (Hairston et al. 1995). We chose to ly hatching eggs to diapausing eggs in Bullhead Pond. divide the sediment at 2 cm because experiments using Our estimate of the distribution of switch dates is based polystyrene beads (similar in size and specific gravity on 9 yr of data (Hairston 1987, Hairston and De Stasio to diapausing eggs) show that the top 2 cm of the sed- 1988, Hairston and Kearns 1995). We determined the iment at the center of Bullhead Pond is mixed to a fraction of female copepods making either immediately modest extent by bioturbation (Kearns et al., in press, hatching eggs or diapausing eggs at four photoperiod N. G. Hairston and C. M. Kearns, unpublished data), and temperature conditions: ? 1.5 and ?2.5 standard whereas deeper sediments are not. Moreover, radio- deviations from the field mean (= day of year 85, 26 isotope dating shows a steady decline in 210Pb activity March). These values were chosen to maximize the throughout the top 10 cm with only a slightly steeper statistical power of our analyses to detect differences decline in the top 2-3 cm than deeper in the sediment in the variance of switch date between treatments (pow- (Hairston et al. 1995), again indicating modest mixing er estimated by Monte Carlo simulations over the ex- only in the top few centimetres. Within the 0-2 cm pected range of variances based on previous studies of sediment layer there was a mean of 8.3 x 104 viable the diapause-timing distribution; Hairston and Dillon eggs/M2, with a mean age of 12.2 yr; in the 2-5 cm 1990). The standard deviation of switch date, though layer there was a mean of 5.6 x 104 viable eggs/M2, variable between years (Hairston and Dillon 1990), is with a mean age of 56.7 yr old (De Stasio 1989, Hair- -7 d (Hairston and Kearns 1995), thus the four lab- ston et al. 1995). Each sediment layer from each core oratory environments followed day length and lake was placed in a plastic tray (12 trays, each 38 x 15 X temperature conditions on day of year 67.5 (8 March), 5 cm) for a sediment depth of 1.5-2.0 cm and glass- 74.5 (15 March), 95.5 (5 April), and 102.5 (12 April). fiber filtered Bullhead Pond water was siphoned over The water temperatures were taken from averages of the top to bring total depth (sediment + water) in the 11 yr of data covering the chosen dates, using linear trays to 4 cm. The trays were incubated in a controlled extrapolation between measurement dates when nec- environment chamber at L:D, 13:11 and 150C. Each essary. These temperatures were then matched to pho- week for 52 wk, the water was siphoned out of each toperiods obtained from standard tables. Combinations tray and checked for copepod nauplii. The filtered water of photoperiod (to the nearest 15 min) and temperature was then returned to the trays and the sediment was (+0.5?C) were created in four controlled environment thoroughly stirred before being allowed to stand un- chambers at (L:D x ?C): 11.75:12.25 x 4.30, 12.00: disturbed until the next siphoning. December 1996 VARIATION IN A ZOOPLANKTON EGG BANK 2385 As they were obtained, nauplii from water column TABLE 1. Diapause response of Diaptomus sanguineus reared at four photoperiod x temperature combinations. and sediment subpopulations were distributed evenly among the four environmental treatments. Nauplii were Treatment: placed in 250-mL glass jars and fed laboratory-cultured day length (h) and temperature (?C) Chlamydomonas sp., with more of this alga added ev- 11.75 12.00 12.75 13.25 ery few days to keep culture jars slightly green. Eu- 4.30 6.00 8.00 10.50 glena gracilis was added to the diet when the animals Water column reached the copepodid stage. For the water-column sub- N (clutches) 265 121 243 203 population, where groups of nauplii were obtained from % immediate 91.3 74.4 12.8 2.0 individual clutches, siblings were reared one family per Sediment jar. Within each environmental chamber, the nauplii 0-5 cm (pooled) obtained from a particular sediment depth in a given N (clutches) 84 106 345 354 % immediate 96.4 98.1 34.5 8.8 week were reared together. Mature males and females 0-2 cm were mated within each environmental treatment as N (clutches) 49 50 205 170 previously described. Individual ovigerous females % immediate 95.9 96.0 32.7 15.3 from these laboratory-reared cultures were again iso- 2-5 cm N (clutches) 35 56 140 184 lated in tissue culture wells and monitored daily for % immediate 97.1 100.0 37.1 2.7 hatching. Eggs hatching within 2 wk (warm tempera- Note: Rows give numbers of viable clutches assayed for tures) or 4 wk (cold temperatures) of laying were scored females reared from copepods obtained from the water col- as immediately hatching eggs, those that had not umn and from two sediment depths, and the percentage of hatched by this time and had not decomposed were those clutches that were immediately hatching eggs (% im- mediate; as opposed to diapausing eggs). scored as diapausing eggs (Hairston and Munns 1984, Hairston and Olds 1984, 1986). Finally, in a preliminary study of diapause duration, reared in late season environments (GLIM; F,2 = a subset of the diapausing eggs produced by both water- 672.8, P < 0.005). column-derived and sediment-derived copepods were In the laboratory, diapausing eggs hatched from the maintained in continuous dark and in a temperature sediment trays for the entire period for which we mon- regime similar to lake conditions. The eggs were held itored them (363 d in 0-2 cm layer, 292 d in 2-5 cm in their respective treatment temperatures for the first layer). Although hatching was much reduced at the end 3 to 5 wk post-production, after which they were cycled of this period, it had not terminated (Fig. 1). A total through summer (13 wk at 15?C), fall (7 wk at 8?C), of 9290 copepod nauplii (7.70 x 104/m2) were obtained and winter (4?C) conditions, where they were held for from the 0-2 cm sediment trays, and 5768 nauplii (4.78 the remainder of the experiment. Eggs were scored for X 104/m2) from the 2-5 cm trays. These numbers are hatching at intervals of -2-4 mo. The conditions do strikingly close to estimates of mean egg densities (De not strictly mimic the natural environment to which the Stasio 1989, Hairston et al. 1995; see Methods), and eggs are exposed (except deep in the mud), but rather taken at face value imply -92 and 85% hatching of serve as an initial index by which the relative diapause the viable eggs from the 0-2 and 2-5 cm sediments, duration of eggs from the two subpopulations can be respectively. From a random sample of these nauplii, compared. Eggs that had hatched were recognizable as 889 females produced viable clutches. As with water- either nauplii (live or dead), or as empty egg cases split column-derived copepods, there is a significant effect along the equatorial seam that opens at the time of of environment (photoperiod x temperature) on egg hatching (Champeau 1970). type (Table 1; GLIM, F1,2 = 89.64, P < 0.02). Does the phenotypic distribution of diapause re- RESULTS sponse differ between copepods derived from the water column and those derived from the sediments? To an- Of the copepods collected from the water column, swer this question, we pooled the results from the two 832 females produced viable eggs. The proportions of sediment layers. Results from the 0-2 and the 2-5 cm females producing immediately hatching or diapausing layers will be compared separately below. The distri- eggs depended upon the photoperiod X temperature bution of clutch types differs significantly between the combination in which the copepods had been reared water-column-derived and the pooled sediment-derived (Table 1). We analyzed these data, and other similar copepods (GLIM; F1,5 = 24.82, P < 0.005). This sig- types of data discussed below, using Generalized Lin- nificant effect results from a difference in distribution ear Interactive Modeling (GLIM) and accounted for means rather than a difference in variances (Table 2, potential overdispersion by allowing the program to Fig. 2). Assigning the photoperiod X temperature treat- estimate the scale parameter (Crawley 1993). Those ments to the days of the year they mimic, maximum copepods raised in early-season environments (short likelihood estimates of the means and standard devia- day, low temperature) produced a significantly greater tions of switch-to-diapause dates were computed using fraction of immediately hatching eggs than did those an underlying Gaussian distribution for each population. 2386 NELSON G. HAIRSTON, JR. ET AL. Ecology, Vol. 77, No. 8 120- 0-2 cm 100- 6 80- FIG. 1. Hatching rate of Diaptomus sangui- neus nauplii (larval copepods) from diapausing 0) eggs obtained from two layers of sediment (0- ~C60- 2 cm and 2-5 cm depth) collected from Bullhead Pond, Rhode Island. The arrow shows the point 40 of the division at 8 October (day 111 of sedi- Z 20 m ~ ~~25c ment incubation) between "early" and "late" hatching nauplii. See Results for further expla- 40 nation and for methodological details. 0 50 100 150 200 250 300 350 400 No. Days Incubated Confidence intervals (95% ci) were computed by boot- copepods than in either the 0-2 cm copepods or the strapping from the data (n = 104 replicates). By this water-column-derived copepods (Table 2; 95% cis do method, the mean date on which D. sanguineus from not overlap). the water column switched to producing diapausing We noticed while collecting our data that sediment- eggs is day 81.9 (23 March) whereas that for sediment derived copepods from the 0-2 cm layer showed a copepods is day 90.4 (31 March). The means differ distinct temporal pattern in diapause response that was significantly (Table 2; 95% cis do not overlap), whereas absent in copepods from the 2-5 cm layer. Those that the variances do not (standard deviation 95% cis over- hatched early in the experiment were more likely to lap). exhibit a later switch to diapause than were those that There are subtle differences in the diapause response hatched late in the experiment (Table 2). This pattern within the sediment-derived copepods. Maximum like- can be seen most easily in the data for the laboratory lihood estimates of mean diapause timing did not differ environment simulating day of year 95.5 (see "sedi- significantly between copepods from the 0-2 cm layer ment [pooled]" curve in Fig. 2) because in this data (day 90.1) and copepods from the 2-5 cm layer (day set an intermediate fraction of individuals made im- 91.3) (Table 2; 95% c's overlap). The phenotypic vari- mediately hatching eggs. There is a distinct drop in the ance, however, is significantly lower in the 2-5 cm fraction of copepods that made immediately hatching TABLE 2. Maximum likelihood estimates of the mean and standard deviation of diapause timing of Diaptomus san- 0 100- guineus computed using an underlying Gaussian distribu- tion for populations from the water column and sediments o 80- Sediment of Bullhead Pond, Rhode Island. (pooled) O60- Switch date (day of year) Sample Mean 95% Ci SD 95% ci 40- Water column 81.9 80.6-83.1 10.6 9.6-11.6 Cu Water Column I >% 20- Sediment Pooled (0-5 cm) 90.4 89.1-91.8 9.4 7.8-10.7 0-2 cm 90.1 88.4-91.8 11.4 9.3-13.5 0) 2-5 cm 91.3 89.2-93.8 6.9 4.0-8.8 E 50 75 100 125 E 0-2 cm Day of Year early hatching 92.3 90.5-94.1 10.0 7.5-12.3 late hatching 83.1 79.4-87.5 9.7 6.3-13.2 FIG. 2. Maximum likelihood Gaussian distributions fit to 2-5 cm the diapause responses, expressed as fraction of females mak- early hatching 92.5 90.9-93.9 5.2 3.7-6.6 ing egg clutches that hatch immediately (as opposed to dia- late hatching 91.1 85.8-95.3 12.2 1.8-20.6 pause), of Diaptomus sanguineus from Bullhead Pond, Rhode Notes: The copepods were reared at four photoperiod x Island. Response was determined for copepods reared from temperature combinations that mimicked four dates during parents obtained directly from the water column of the lake the spring when the copepods naturally switch from produc- or from diapausing eggs hatched from lake sediments. Four tion of immediately hatching eggs to diapausing eggs. Con- combinations of photoperiod and temperature that mimic the fidence intervals (95% Ci) were computed by bootstrapping seasonal transition from winter to spring in the lake were from the data (n = 104 replicates). used. January 1 was day 1. December 1996 VARIATION IN A ZOOPLANKTON EGG BANK 2387 60 mo (N. G. Hairston, Jr. and C. M. Kearns, unpublished data, B. T. De Stasio, Jr., personal communication). 40- Six months after the end of March is the beginning of October, which is close to the time when D. sanguineus diapausing eggs begin to hatch in Bullhead Pond each *A 20- year (Hairston and Munns 1984). Thus our early group consists only of copepods that hatched from eggs laid E E in years prior to 1993; our late group includes copepods 22 40 76 110 194 251 that hatched from eggs laid after the population switched to making diapausing eggs in March 1993. No. Days Incubated For the 2-5 cm data, early-hatching and late-hatch- FIG. 3. The fraction of copepods (N = 205 inds.) hatched ing copepods did not differ significantly in diapause from Bullhead Pond sediment (0-2 cm depth) that made im- response (GLIM allowing for overdispersion, F,5 = mediately hatching eggs in a 12.75:11.25 (L:D) and 8.0?C environment (simulates day 95.5 of year; January I was day 0.05, P > 0.75). The mean switch dates for the 2-5 I) as a function of the length of time that eggs in the exper- cm copepods are day 92.5 early-hatching copepods and iment were exposed to hatching conditions. Two presentations day 91.1 for the late-hatching copepods; the 95% con- of the same data are given. The continuous line shows a fidence intervals overlap markedly (Table 2). locally weighted regression scatterplot smoothing with a dis- The diapausing eggs collected from the water-col- tinct drop at 8 October. The histogram bars divide the data into five time intervals for statistical analysis: the first three umn-derived and the sediment-derived copepods began intervals differ significantly from the last two (see Results). to hatch after different periods in diapause (Fig. 4), even though the two groups were maintained under identical conditions. Diapausing eggs from the water- eggs in this environment among those that hatched be- column-derived copepods began hatching within 7 mo fore early October and those that hatched later (Fig. after they were laid. In contrast, diapausing eggs from 3). "Lowess" (locally weighted regression scatterplot the sediment-derived copepods did not begin to hatch smoothing) analysis (Velleman 1995) of the data shows in significant numbers until >12 mo after they were a distinct drop in the fraction of copepods making im- laid. Only :45% of the eggs had hatched 20 mo after mediately hatching eggs in early October. For further the experiment started, with a low but steady hatching analysis, we divided the data into five intervals of rate in both subpopulations. Again, we do not believe roughly equal sample size (interval duration varied that the quantitative details of the hatching phenology [Fig. 3] because more copepods hatched early as op- are relevant to field conditions: we wish here only to posed to late in the experiment, Fig. 1). Among the emphasize that the timing of initiation of hatching dif- first three intervals, copepods did not differ signifi- fers between the subpopulations (Gehan's test, z = cantly in proportion producing immediately hatching 7.08, P < 0.001, n = 1279; Gross and Clark 1975). eggs (GLIM likelihood ratio test; G2 = 2.50, df = 2, DISCUSSION NS). Likewise, among the last two intervals, copepods did not differ significantly in proportion producing im- The timing of the switch to diapause in March or mediately hatching eggs (GLIM likelihood ratio test; April is an important fitness character for Diaptomus G2 = 0.00, df = 2, NS). There is, however, significant sanguineus in Bullhead Pond. It dictates the vulnera- variation among all groups (GLIM likelihood ratio test; bility of a female's progeny to annual springtime in- G = 24.66, df = 4, P < 0.005). When we divide all of the data from all four photoperiod X temperature X 0.5- environments (cf. Table 1) into "early-hatching" co- pepods and "late-hatching" copepods using early Oc- i 0.4- tober (8 October, Julian day 281) to divide between the two groups there is a significant difference in diapause . 0.3- response (GLIM allowing for overdispersion; F,1 = Water 18.75, P < 0.01). The mean switch dates for the 0-2 Lu 0.2- column cm copepods are day 92.3 for the copepods that hatched co 0.1 _ Sediments early in the experiment and day 83.1 for the copepods E that hatched late, and the 95% confidence intervals do n 0.0T not overlap (Table 2). 0 0 2 4 6 8 10 12 14 16 18 20 An early October division of the data makes sense Egg Age (mo) biologically. The D. sanguineus in Bullhead Pond switch each year to making diapausing eggs in late FIG. 4. The fraction of viable diapausing eggs laid by the water-column-derived and the sediment-derived copepods March (Hairston and Munns 1984). In the laboratory, hatching over a period of 20 mo. Eggs from the water-column under ideal hatching conditions, newly produced dia- subpopulation tend to hatch after a shorter period of diapause pausing eggs began to hatch only after a period of 6 than do eggs from the sediment subpopulation. 2388 NELSON G. HAIRSTON, JR. ET AL. Ecology, Vol. 77, No. 8 creases in fish predation (Hairston and Munns 1984, pausing eggs ranging in age (Hairston et al. 1995) from Hairston 1987), and hence directly impacts her repre- 0 to 76 yr (i.e., 0-5 cm) on average switch to diapause sentation in future generations. Copepods that switch 9-10 d later than copepods derived from animals swim- from making immediately hatching eggs to diapausing ming in the water column. eggs too late in the season lose fitness because some Is there a plausible explanation for this difference in of their offspring get eaten before they can mature. diapause timing between the water-column and sedi- Those that switch too early in the season lose repre- ment subpopulations? One obvious possibility is that sentation in future generations because their offspring there has been recent directional natural selection fa- lose time to mature and to make diapausing eggs of voring an earlier switch to diapause. Because diapause their own. We have shown that there are distinct dif- timing in Bullhead Pond is known to be an adaptation ferences in the phenotypic expression of diapause tim- to avoid the seasonal onset of fish predation (see In- ing depending upon whether the copepods originated troduction for details; Hairston 1987, Hairston and Dil- from the active subpopulation in the water column or lon 1990), a change to an earlier seasonal diapause from the diapausing subpopulation in the sediment. would have occurred if the fish predation had increased Within the sediment subpopulation, there are further markedly in the recent past. We have, in fact, observed differences depending both upon whether the copepods the opposite of such a response: 2 yr after fish were were hatched from diapausing eggs obtained from near completely eliminated from Little Bullhead Pond (200 the sediment surface (0-2 cm) or from deeper sedi- m from Bullhead Pond) by a single-year drought, the ments (2-5 cm), and upon whether the copepods came mean timing of diapause shifted to later in the spring from sediment diapausing eggs that readily hatched by 26 d (Hairston and Walton 1986). In Bullhead Pond, (i.e., early in the egg hatching experiment) or from eggs however, we know that changes in selection of this that were slower to hatch. In addition, the water-column magnitude have not occurred in at least the past decade. and sediment subpopulations differ in the time between Selection on diapause timing of D. sanguineus has been when eggs are laid and when they begin to hatch. quantified in Bullhead Pond in nine separate years be- Do these phenotypic patterns have an important ge- tween 1979 and 1989 (Hairston and Dillon 1990, S. netic component? We suspect that the answer is yes, Ellner and N. G. Hairston, Jr., unpublished manuscript). but we cannot rule out an important plasticity com- This is approximately the same period covered by the ponent. In laboratory breeding experiments, diapause top centimetre of sediment (Hairston et al. 1995). Be- photoperiod response in D. sanguineus had a significant tween 1979 and 1989, selection fluctuated but showed heritability (h2 = 0.48, Hairston and Dillon 1990). In no consistent directionality. The average magnitude of the field, the mean timing of diapause for this species the change in diapause timing between consecutive shifted between years in a pattern consistent with in- years was 2.47 d, with a total change over the entire terannual variation in the timing and intensity of the seasonal onset of fish predation in both the Bullhead period (1979-1989) of 2.17 d. Since 1989, there have Pond population (Hairston and Dillon 1990) and in a been no obvious directional changes in fish abundance nearby population in Little Bullhead Pond (Hairston or in the volume of water in Bullhead Pond (decrease and Walton 1986, Hairston and De Stasio 1988), dem- in volume translates into an increase in fish density, onstrating that diapause timing is also heritable in the Hairston 1988) that would have caused an alteration in field. The copepods in our experiment were all reared selection pressure sufficient to produce the 9- or 10-d from eggs in the laboratory using a "common garden" difference in diapause timing between the sediment and experimental design, and the differences in diapause water-column subpopulations (N. G. Hairston, Jr., per- response we have documented describe consistent pat- sonal observation). The top two centimetres of sedi- terns across photoperiod x temperature treatments and ment do contain eggs laid up to 28 yr ago, 17 yr before among subpopulations. These arguments provide cir- the period for which we have direct evidence of co- cumstantial evidence for heritable differences in dia- pepod and fish dynamics. However, changes in fish pause timing between the sediment and water-column predation in Bullhead Pond are related directly to subpopulations of D. sanguineus in Bullhead Pond. changes in pond volume (Hairston 1988), which are in Our hypothesis at the start of this research was that turn a function of changes in water table height, which fluctuating selection on the timing of diapause com- has shown no trend since 1948 when the U.S. Geolog- bined with prolonged egg diapause would lead to the ical Survey first began collecting data near Bullhead storage of higher levels of trait variation among dia- Pond (Hairston 1988). In addition, if the nature of fish pausing eggs in the sediment than would be expressed predation had changed prior to 1979, we would have at any one time by copepods active in the water-col- predicted an impact on the mean and variance of dia- umn. This is clearly not the case. Phenotypic variances pause timing for copepods drawn from the 0-2 cm are remarkably similar between these groups (Table 2, sediment layer relative to either the water-column co- water column vs. pooled sediment). Instead, the really pepods or to those from the deeper sediment layer be- striking difference between the subpopulations is in t hcaeuse the time period encompassed by the 0-2 cm sed- mean timing of diapause. Copepods derived from dia- iment layer would span the period of directional se- December 1996 VARIATION IN A ZOOPLANKTON EGG BANK 2389 lection and response. We observed no such effect (Ta- eggs that receive the cue within the first year (i.e., ble 2). offspring collected directly from the water column, or With recent selection response ruled out, we hy- young eggs from the surface of the 0-2 cm layer) tend pothesize that the differences in diapause timing be- to switch to diapause early in the season, whereas eggs tween sediment and water-column subpopulations are that have been in the sediment for multiple years (i.e., due to a covariance between diapause traits expressed from deeper in the sediment) tend to switch to diapause in eggs and adults. Specifically, there is covariance, later in the season. One piece of our data is not, how- possibly genetic, between adult "switch date" (the day ever, consistent with this interpretation. Water-column- of the year when an adult switches from making im- derived copepods that switched to diapause early in the mediately hatching eggs to making diapausing eggs), season made diapausing eggs that tended to begin and the propensity of an egg to remain dormant for an hatching in the laboratory after only 6 mo; sediment- extended time in the sediment. derived copepods that switched to diapause late in the Two lines of evidence support this covariance hy- season made diapausing eggs that tended to begin pothesis. First, diapausing eggs produced by water-col- hatching in the laboratory after 12 mo (Fig. 4). Thus umn-derived adults with an early switch date tend to it was not egg age that determined when the copepods hatch sooner than eggs produced by sediment-derived switched to diapause, but rather when the copepods adults with a late switch date (cf. Figs. 2 and 4). Sec- switched to diapause that determined the timing of ond, early switch date is observed in two groups of hatching. One would have to add another layer of com- copepods: those in the water column, and those eggs plexity (i.e., a maternal effect of timing of switch to from the surface sediments (0-2 cm) that take a long diapause on duration of egg diapause) in order to main- time to hatch (Table 2). We know that there is a 6-mo tain this phenotypic plasticity hypothesis. period before new eggs can hatch (Fig. 4), hence the An alternative explanation for the trait association late-hatching subpopulation includes some eggs pro- is that there is genetic determination of diapause timing duced in the previous year, while the early-hatching (Hairston and Dillon 1990) and duration of diapause subpopulation consists entirely of eggs that were pro- in the sediment (which has not yet been shown, and is duced in prior years and remained in diapause. The only hypothesized here). If copepods that are geneti- latter eggs grow up to have the later mean switch date cally predisposed to hatch within the 1 st yr after being (Table 2). Thus the subset of eggs that are known to produced also tend to have genes that dictate an early have remained in diapause for > 1 yr have a later-than- seasonal switch to diapause, while those that are ge- average switch date, which again indicates a covariance netically predisposed to hatch after a longer period of between egg diapause and adult switch date. We give time also tend to have genes that dictate a late switch a possible explanation for this trait association below, to diapause, we could account for all of our current but first we suggest two ways in which this trait as- data. Genetic covariance of this sort would need to be sociation might be maintained: phenotypic plasticity tested in controlled breeding experiments in the labo- and genetic covariance. ratory. Under phenotypic plasticity, either the timing of the We conclude our discussion with an hypothesis that switch to diapause or the timing of egg hatching, or would explain the trait association we observe whether both, must be affected in a repeatable way by some it is maintained by phenotypic plasticity or genetic co- environmental factor. For example, if there is genetic variance. We hypothesize that there may be two alter- variation for the timing of the onset of diapause (as native adaptive strategies for a copepod to persist in shown by Hairston and Dillon 1990) and diapause du- the temporally variable environment of Bullhead Pond. ration is influenced by the time of year the diapausing In both, diapause is an important short-term adaptation egg is produced (postulated here for the sake of ar- for avoiding seasonally intense predation, but its value gument), then the trait association we observe could as a long-term persistence mechanism differs. One be produced. Another possibility is that the age at strategy is for nearly all diapausing eggs to hatch the which a diapausing egg hatches affects diapause timing year after they are produced. This is a potentially risky of the resulting copepod. The environmental cue to long-term strategy because any year with an early and hatch is a function of light, oxygen, and probably tem- intense onset of fish predation could lead to little or perature (N. G. Hairston, Jr. and C. M. Kearns, un- no recruitment of these short-diapause-duration cope- published data). Those eggs near the sediment-water pods. The risk can be countered, however, if individuals interface are likely to receive the hatching cue whereas also possess an early seasonal switch to diapause; that those buried deeper in the sediment almost certainly is, a safe diapause timing strategy. We call this first do not receive it (hence they hatched in our study only approach the "risky-in-the-sediment/safe-in-the-water- when we gave them the appropriate cue in the labo- column" strategy. The alternative strategy is for only ratory; Fig. 1). If an egg becomes buried in the sedi- a small fraction of the diapausing eggs produced in any ment before receiving the hatch cue, aging might affect given year to hatch in any single subsequent year. This how it responds to the water-column environment once is a relatively safe long-term strategy in that years of it does hatch. Most of our data would be explained if poor or zero recruitment are survived by the diapausing 2390 NELSON G. HAIRSTON, JR. ET AL. Ecology, Vol. 77, No. 8 eggs that did not hatch. Eventually, when a good-re- A. cruitment year comes, the egg bank is restocked. Co- 0) 100 pepods that adopt this "safe-in-the-sediment" strategy C Field should wait until relatively late in the season to switch 80- to diapause (i.e., a risky diapause timing strategy) be- I 60 Water column cause they rely on occasional good-recruitment years. Sediment The later the switch date, the higher the recruitment 40 (because the population grows larger before producing 20- diapausing eggs) and the higher the fitness of the co- E pepods until the seasonal increase in fish predation ter- E 0 F minates the reproductive season. We call the second 10 11 12 13 14 15 16 approach a "safe-in-the-sediment/risky-in-the-water- Day Length column" strategy. Our hypothesis, then, proposes the evolution of trait B. Water covariance between duration of egg diapause in the lake column Sediment sediments and diapause timing by copepods active in the water column. That association could be maintained either by phenotypic plasticity or genetic covariance. -A*,** El * The proposal differs qualitatively from what Evans and 80 82 84 86 88 90 92 94 Cabin (1995) suggest for plant seed banks in that they Day of Year postulate the evolution of particular genotypes that ger- FIG. 5. The timing of diapause of Diaptomus sanguineus minate in years with particular types of environments in Bullhead Pond, Rhode Island, lies intermediate between (which then evolve covariance with genotypes for post- subpopulations obtained from the lake water column and the germination performance in those environments). In lake sediments. (A) The fraction of immediately hatching contrast, our hypothesis proposes that some phenotypes eggs produced in the field (closed squares) as day length spend a short time in diapause whereas others spend a increases in spring (mean and 95% ci, n = 9 yr), and the diapause responses of copepods reared in the laboratory from long and variable time in diapause regardless of what parents obtained directly from the water column of the lake type of environment is present in any given year (but (open circles) or from diapausing eggs hatched from lake which also leads to evolution of covariation with post- sediments (open diamonds). Laboratory results are for four hatch performance in particular types of years). The combinations of photoperiod and temperature that mimic the postulated risky-safe/safe-risky corresponds to Levins' seasonal transition from winter to spring in the lake. (B) The mean timing of diapause in each of nine years in the lake in (1979) model of coexistence between strategies spe- relation to the mean timing observed for the "water-column" cializing on the mean and variance in a fluctuating and "sediment" (2-5 cm depth) subpopulations. environment. The risky-in-the-sediments individuals mostly have a switch date near to the optimal one in the mean environment, while later and riskier switch the optimal switch date is late. Other outcomes are dates predominate among the safe-in-the-sediments in- possible under different assumptions about the selec- dividuals. tion regime. A simple genetic model based on Ellner and Hairston If two strategies (i.e., safe-risky and risky-safe) co- (1994) suggests that the risky-safe/safe-risky strategy exist in the Bullhead Pond population of D. sanguineus, pair are a natural result of selection in this system. then the average diapause phenology expressed by the Details of this model will be presented elsewhere, and copepods in the lake should fall somewhere between we present a brief overview here only to point out that these two extremes. This is the case (Fig. 5A). The a model can produce the trait associations we observe. distribution of mean diapause timing over the 9 yr for The essential ingredients of the model are generation which we have data (Hairston and Dillon 1990, S. Ell- overlap, fluctuating selection on switch date, and her- ner and N. G. Hairston, Jr., unpublished manuscript) itable variation in both switch date and the annual is distinctly skewed towards early switch dates, but hatching fraction of diapausing eggs. The model pre- with a tail extending to later dates (Fig. SB). To illus- dicts that when the variance in the optimal switch date trate where these mean switch dates lie relative to the is low, all individuals are risky-safe: a high hatching dates we report here for the water-column and sediment fraction and a safe, early switch date. At higher vari- subpopulations, we use the means and 95% confidence ance, a dimorphism for hatching fraction develops: intervals from Table 2. The mean switch dates for all some individuals become safe-risky (lower hatching 9 yr lie between the extremes of the water-column and fraction associated with a later switch date). The safe- sediment (2-5 cm) subpopulations, consistent with the risky and risky-safe strategies cannot out-compete each interpretation that each year contains a mixture of the other, but they collectively out-compete any other strat- phenotypes present in the two subpopulations. Much egy. This specific outcome depends on there being a of the year-to-year variation in mean switch date can substantial "pay-off" for late switching in years when be ascribed to selection response (Hairston and Dillon December 1996 VARIATION IN A ZOOPLANKTON EGG BANK 2391 1990, S. Ellner and N. G. Hairston, Jr., unpublished cism of the manuscript. This research was supported by Na- manuscript), but some may also result from expression tional Science Foundation grant DEB-9119984. of varying proportions of the "water-column" and LITERATURE CITED "sediment" phenotypes. In most years, the switch to Bennington, C. C., J. B. McGraw, and M. C. Vavrek. 1991. diapause came relatively early, close to, but in only Ecological genetic variation in seed banks. II. Phenotypic and genetic differences between young and old subpopu- one case overlapping with, the 95% confidence interval lations of Luzula parviflora. Journal of Ecology 79:627- of the diapause timing in the water-column subpopu- 643. lation. Only in 1981 did the mean timing of diapause Bosbach, K., H. Hurka, and R. Haase. 1982. The soil seed fall close to the sediment diapause timing. In that year, bank of Capsella bursapastoris (Cruciferae): its influence Bullhead Pond was extremely shallow (Hairston 1988) on population variability. Flora 172:47-56. Brown, J. S., and D. L. Venable. 1986. Evolutionary ecology due to the same drought that dried up Little Bullhead of seed-bank annuals in temporally varying environments. Pond. Most of the D. sanguineus egg bank in Bullhead American Naturalist 127:31-47. Pond is in sediments near the center of the lake rather Champeau, A. 1970. Etude de la vie latente chez les Cal- than in near-shore sediment (De Stasio 1989). It seems anoides (Copepodes) caract6ristiques des eaux temporaires de Basse-Provence. Annales de la Facult6 des Sciences de likely, then, that when water level dropped during the Marseille 44:155-189. 1981 drought, deep sediments were disturbed by phys- Chesson, P. L. 1984. The storage effect in stochastic pop- ical (i.e., wave action) and biological (i.e., fish nesting) ulation models. Lecture Notes in Biomathematics 54:76- processes usually restricted to near the shore. These 89. deep sediments contain the store of risky-in-the-water- Crawley, M. J. 1993. GLIM for ecologists. Blackwell Sci- entific, New York, New York, USA. column/safe-in-the-sediment phenotypes, which when De Stasio, B. T., Jr. 1989. The seed bank of a freshwater introduced into the water column in large numbers crustacean: copepodology for the plant ecologist. Ecology cause the unusual shift in the mean timing of diapause 70:1377-1389. to very late in the season that we observed in 1981. Ellner, S., and N. G. Hairston, Jr. 1994. Role of overlapping generations in maintaining genetic variation in a fluctuating Long-lived dormant propagule pools such as the environment. American Naturalist 143:403-417. crustacean egg bank that we have investigated have Epling, C., H. Lewis, and F M. Ball. 1960. The breeding long been suspected of containing stored genetic vari- group and seed storage: a study in population dynamics. ation that might have significant impacts on ecological Evolution 14:238-255. and evolutionary processes (e.g., Epling et al. 1960, Evans, A. S., and R. J. Cabin. 1995. Can dormancy affect the evolution of post-germination traits? The case of Les- Templeton and Levin 1979, Brown and Venable 1986). querella fendleri. Ecology 76:344-356. Our data provide clear evidence that phenotypic (and Futuyma, D. J. 1986. Evolutionary biology. Sinauer, Sun- presumably genotypic) differences exist for an impor- derland, Massachusetts, USA. tant fitness character (timing of diapause) between sub- Gottlieb, L. D. 1974. Genetic stability in a peripheral isolate populations of the copepod D. sanguineus active in the of Stephanomeria exigua ssp. coronaria that fluctuates in population size. Genetics 76:551-556. water column and those residing in diapause in the Gross, A. J., and V. A. Clark. 1975. Survival distributions: sediments of Bullhead Pond. This means that pheno- reliability applications in the biomedical sciences. John Wi- typic variance for this trait in the total population (ac- ley and Sons, New York, New York, USA. tive and diapausing individuals) is substantially greater Hairston, N. G., Jr. 1987. Diapause as a predator avoidance adaptation. Pages 281-290 in W. C. Kerfoot and A. Sih, than that expressed in either of the subpopulations. editors. Predation: direct and indirect impacts on aquatic Furthermore, both the overall population mean and communities. New England University Press, Hanover, variance are different from what they would be in the New Hampshire, USA. absence of an egg bank. A major portion of the variance . 1988. Interannual variation in seasonal predation: its origin and ecological importance. Limnology and we see apparently results from covariance between du- Oceanography 33:1245-1253. ration of diapause in the egg bank and the fitness char- Hairston, N. G., Jr., and B. T. De Stasio, Jr. 1988. Rate of acter under investigation. This complicates in inter- evolution slowed by a dormant propagule pool. Nature 336: esting ways the role that dormant propagule pools may 239-242. play in natural populations. Not only may they act as Hairston, N. G., Jr., and T. A. Dillon. 1990. Fluctuating se- lection and response in a population of freshwater cope- a reservoir of genotypes produced in past environ- pods. Evolution 44:1796-1805. ments, but they may interact as a part of the selection- Hairston, N. G., Jr., T. A. Dillon, and B. T. De Stasio, Jr. response dynamics of current environment in ways that 1990. A field test for the cues of diapause in a freshwater can slow down response to selection, speed it up, or copepod. Ecology 71:2218-2223. alter its direction. Hairston, N. G., Jr., S. Ellner, and C. M. Kearns. 1996. Over- lapping generations: the storage effect and the maintenance of biotic diversity. Pages 109-145 in 0. E. Rhodes, Jr., R. ACKNOWLEDGMENTS K. Chesser, and M. H. Smith, editors. Spatial and temporal We thank Keith 0. Batson and Katherine M. Dowell for aspects of population processes. University of Chicago technical assistance in the laboratory and Charles McCulloch Press, Chicago, Illinois, USA. for statistical advice. Saran Twombly and her lab group mon- Hairston, N. G., Jr., and C. M. Kearns. 1995. The interaction itored copepod population development in Bullhead Pond and of photoperiod and temperature in diapause timing: a co- advised us when to make our collecting trips to Rhode Island. pepod example. Biological Bulletin (Woods Hole) 189:42- Steven A. Juliano and two referees provided valuable criti- 48. 2392 NELSON G. HAIRSTON, JR. ET AL. Ecology, Vol. 77, No. 8 Hairston, N. G., Jr., and W. R. Munns, Jr. 1984. The timing populations of Eriophorum vaginatum. Arctic and Alpine of copepod diapause as an evolutionarily stable strategy. Research 25:45-49. American Naturalist 123:733-751. McGraw, J. B., M. C. Vavrek, and C. C. Bennington. 1991. Hairston, N. G., Jr., and E. J. Olds. 1984. Population dif- Ecological genetic variation in seed banks: I. Establishment ferences in the timing of diapause: adaptation in a spatially of a time transect. Journal of Ecology 79:617-625. heterogeneous environment. Oecologia (Berlin) 61:42-48. Rees, M. 1994. Delayed germination of seeds: a look at the Hairston, N. G., Jr., and E. J. Olds. 1986. Partial photoperiod effects of adult longevity, the timing of reproduction, and control of diapause in three populations of the freshwater population age/stage structure. American Naturalist 144: copepod Diaptomus sanguineus. Biological Bulletin 43-64. (Woods Hole) 171:135-142. Seger, J., and H. J. Brockmann. 1987. What is bet-hedging? Hairston, N. G., Jr., and E. J. Olds. 1987. Population dif- Oxford Surveys in Evolutionary Biology 4:182-211. ferences in the timing of diapause: a test of hypotheses. Templeton, A. R., and D. A. Levin. 1979. Evolutionary con- Oecologia (Berlin) 71:339-344. sequences of seed pools. American Naturalist 114:232- Hairston, N. G., Jr., R. A. Van Brunt, C. M. Kearns, and D. 249. Tonsor, S. J., S. Kalisz, J. Fisher, and T. P. Holtsford. 1993. R. Engstrom. 1995. Age and survivorship of diapausing A life-history based study of population genetic structure: eggs in a sediment egg bank. Ecology 76:1706-1711. seed bank to adults in Plantago lanceolata. Evolution 47: Hairston, N. G., Jr., and W. E. Walton. 1986. Rapid evolution 833-843. of a life history trait. Proceedings of the National Academy Vavrek, M. C., J. B. McGraw, and C. C. Bennington. 1991. of Sciences (USA) 83:4831-4833. Ecological genetic variation in seed banks. III. Phenotypic Kearns, C. M., N. G. Hairston, Jr., and D. H. Kesler. 1996. and genetic differences between young and old seed pop- Particle transport by benthic invertebrates: its role in egg ulations of Carex bigelowii. Journal of Ecology 79:645- bank dynamics. Hydrobiologia, in press. 662. Kimura, M. 1983. The neutral theory of molecular evolution. Velleman, P. F 1995. Data desk, version 5.0, statistics guide. Cambridge University Press, Cambridge, England. Data Description, Ithaca, New York, USA. Levins, R. 1979. Coexistence in a variable environment. Wolf, H. G., and G. R. Carvalho. 1989. Resting eggs of lake- American Naturalist 114:765-783. Daphnia II. In situ observations on the hatching of eggs McGraw, J. B. 1993. Ecological genetic variation in seed and their contribution to population and community struc- banks: IV. Differentiation of extant and seed bank-derived ture. Freshwater Biology 22:471-478.