Pollen Production and Plant Density Affect Pollination and Seed Production in Taxus Canadensis Author(S): Taber D

Home , Plant density

Pollen Production and Plant Density Affect Pollination and Seed Production in Taxus Canadensis Author(s): Taber D. Allison Reviewed work(s): Source: Ecology, Vol. 71, No. 2 (Apr., 1990), pp. 516-522 Published by: Ecological Society of America Stable URL: http://www.jstor.org/stable/1940305 . Accessed: 06/04/2012 10:55

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Ecological Society of America is collaborating with JSTOR to digitize, preserve and extend access to Ecology.

http://www.jstor.org Ecology, 71(2), 1990, pp. 516-522 ? 1990 by the Ecological Society of Amenca

POLLEN PRODUCTION AND PLANT DENSITY AFFECT POLLINATION AND SEED PRODUCTION IN TAXUS CANADENSIS'

TABER D. ALLISON2 Bell Museum of Natural Historyand Departmentof Ecology and Behavioral Biology, Universityof Minnesota, Minneapolis, Minnesota 55455 USA

Abstract. Mean pollen productionand mean nearestneighbor distance were recorded forseveral populations of Taxus canadensis and correlatedwith the proportionof ovules pollinated and seed set. Distance and pollen production togetherexplained 86% of the variation in pollination success, each variable significantlyadding to the regressionwhen adjustingfor the other. Seed set was correlatedsignificantly with pollen productionand nearestneighbor distance separately,but the multipleregression including the lattertwo variables was not significant.Seed set was correlatedmost stronglywith pollination success and mean ovule production(R2 = 0.71), suggestingthat variation in seed set among Taxus populations was a combination of differencesin pollen and resourceavailability. Key words: ApostleIslands; plant density; pollen availability; pollen production; pollination efficiency;seed production; Taxus canadensis; wind pollination.

INTRODUCTION Most theorieson sexual allocationstrategies of plants The efficiencyof wind pollination is generallyas- assume thatpollen availabilitytypically does not limit sumed to decrease as the concentrationof airborne seed productionin plant populations (Charnov 1982, pollen decreases (Regal 1982, Whitehead 1983). This Willson and Burley 1983). Supportingstudies usually low efficiencymay be due to large distances between involve a singlepopulation of a particularspecies. Pol- conspecifics,low pollen productionby individuals,and len limitationof seed productionhas been observed in poor pollen dispersal. These conditions are assumed several species (Bierzychudek1981), and pollen avail- to affectboth ovule fertilizationand seed production abilitymay vary among populations of a species (Al- negatively. lison 1987). Measurements of wind-borne pollen from point The difficultiesin describing the relationship be- sourcesshow highly leptokurtic dispersion (e.g., Wright tweenpollen availabilityand pollinationefficiency are: 1952, 1953, Faegri and van der Pijl 1979), and con- (1) findingplant populations coveringa wide range of sequently,as plant spacing increases, the probability plant abundance; (2) simultaneouslymeasuring pollen of pollinationis assumed to decrease. Studies of agri- productionand seed productionin severalpopulations; culturalspecies (Bateman 1947, Sarvas 1968) support and (3) long-distancepollen dispersal among popula- this assumption,but evidence fromnatural plant pop- tionsthat complicates the dynamicsof pollen dispersal ulations is limited and indirect(see, forexample, Le- withinpopulations (Lanner 1966, Koski 1970). Long- men 1980, Farris and Mitton 1984). No quantitative distance pollen dispersal may cause significantpolli- relationshipsbetween pollination efficiency and pollen nation in otherwisepresumably isolated populations availabilityhave been developed for natural popula- (Squillace 1967). tions of wind-pollinatedspecies. Here I reporton therelationship between plant spac- Determiningthe relationshipbetween pollen avail- ing, pollen production,and pollination efficiency(es- abilityand seed productionin wind-pollinatedspecies timated by both the proportionof ovules pollinated is importantbecause of the potential for gene flow and ovules matured as seeds) in Taxus canadensis among populations and the influenceof pollen avail- Marsh. All data discussed in this paper were collected abilityon sex expression.Models of gene flowbetween fromnaturally occurring T. canadensis populations. plant populations,for example, are typicallybased on pollen dispersion curves from point sources (Koski METHODS 1970), but the behavior of a pollen cloud frompoint sources vs. that originatingfrom diffusepopulation T. canadensis is an evergreen,wind-pollinated, co- sources may be different(Levin and Kerster 1974). niferousshrub found in the understoryof moist deciduous and mixed deciduous-coniferousforests in the 1 Manuscript received 13 October 1988; revised 20 June northeasternUnited States and southeasternCanada 1989; accepted 27 June 1989. 2 Presentaddress: Departmentof Botany, The Ohio State (Martell 1974). It is monoecious, producing separate Universityat Marion, 1465 Mt. Vernon Avenue, Marion, male and female strobili on the same plant (Cham- Ohio 43302 USA. berlain 1966, Allison 1987). Female strobiliare uni- April 1990 POLLINATION AND SEED PRODUCTION IN TAXUS 517

TABLE 1. Means of male strobilusand ovule productionper branch,nearest neighbor distance, pollination success, and seed set of studypopulations in 1985. Data were calculated as describedin the Results section.All sites are in zone 15.

Male No. strobilus Ovule Nearest plants produc- produc- neighbor Pollination Site sampled tion* tion* (m)* success Seed set Site coordinatest Apostle Islands Basswood Island 50 0.46 0.25 3.13 .231 .103 673700m E/5192500m N Rocky Island 97 1.90 0.28 3.13 .463 .134 676500 m E/5211000m N OtterIsland 57 4.20 2.24 0.20 .791 .541 675100m E/5206000m N Outer (closed) 92 22.17 3.06 0.11 .853 .393 695500m E/5216600m N StocktonIsland 45 23.48 11.77 1.58 .576 n.a.t 686700m E/5198100m N Outer (gap) 25 163.30 35.82 0.03 .932 .212 696100m E/5216500m N Cedar Creek 25 0.97 0.46 0.15 .473 .284 484300 m E/5028100m N North Grey Cloud 31 87.25 9.24 0.03 .913 .309 500400m E/4962200m N * Means of male strobilusproduction, ovule production,and nearest neighbordistance are back-transformedfrom logl0 transformationsof individual plant values. Transformedmeans were used in comparisons of means among populations (Allison 1987). t Universal TransverseMercator (UTM) Grid coordinatesof the studysites (describedin the Methods section). t Not available. ovulate; afterfertilization, a femalestrobilus produces ple plants 1 mo afterthe 1983 springcensus on the a singleseed surroundedby a red, fleshyaril. Apostle Islands and at all studysites in 1984 and 1985. The forestunderstory is typicallycooler than the Earlierobservations indicated that no mortalityof de- canopy,creating a temperatureinversion that restricts veloping seeds had occurredby this time. Ovules that verticalmovement of pollen (Levin and Kerster1974). did not develop were still presenton plants but were The decreased turbulence and wind velocity of the yellow-whitein color and small, easily distinguished understorywill also reduce pollen transport(Raynor fromthe green,swollen, developing ovules. Circum- 1967, Tauber 1967). As a result,the influenceof long- stantial support for the conclusion that developing distance pollen dispersal should be minimized, and ovules have been pollinated comes fromhand-polli- pollinationwithin a T. canadensis population should nation experimentsthat increase the number of de- be most stronglydetermined by the plant densityand veloping ovules (Allison 1987). Althoughfertilization pollen productionwithin that population. of ovules typicallyoccurs within 1 mo of pollination Most data were collectedfrom several T. canadensis in T. canadensis(Dupler 1917), in theabsence ofdirect, populationsat theApostle Islands National Lakeshore, cytologicalevidence I cannotbe certainthat developing an archipelago of 21 islands near the south shore of ovules were fertilized.Thus, I used this census to es- Lake Superior,where there is considerable variation timate pollination success (proportionof ovules pol- in plantdensity and pollen productionamong T. cana- linated) in the various populations. Ovules not devel- densispopulations (Allison 1987). In additiondata were oping at this time were assumed to have received no collected on sexual reproductionat two studysites in (or insufficient)pollen. southeasternMinnesota. Precise location of all study Seed set (proportionof ovules that become seeds) sites is listed in Table 1. was measuredin all populationsbeginning in mid-July At all sites I selected T. canadensis rametsor plants and continuinginto September.Ripe seeds werecount- along transectsthat were systematicallyplaced in each ed when I observed eithera brown seed surrounded of the studypopulations. I recordedthe distanceto the by a red, fleshyaril, or an expanded receptacleindi- nearest neighborof each sample plant to the nearest cating that a ripe aril with seed had been removed. centimetre.Mean nearestneighbor distance was then Considerable mortalityof developing ovules occurs calculated for each population. I measured the pro- duringthe summer,although the proximate mecha- duction of male and female strobiliand seeds in the nisms controllingovule abortionin T. canadensis are sample plantsin these populations in 1983, 1984, and unknown. 1985. Female strobiliare uniovulate;thus female strobilus production is equivalent to ovule production. RESULTS Regression analysis showed a strongcorrelation be- Mean male strobilusand ovule productionper ra- tweenthe size ofmale strobiliand thenumber of pollen met,nearest neighbor distance, pollination success, and grains per strobilus(r2 = 0.89). I recorded no signif- seed set forall populations in 1985 are listed in Table icantdifferences in themean size of male strobiliamong 1. This was the last year data were collected and is several populations (Allison 1987). Counts of male representativefor data fromother years that were in- strobili,therefore, are good estimates of plant pollen cluded in the regressions.Means of male strobiluspro- production. duction, ovule production,and nearest neighbordis- I countedthe numberof developingovules on sam- tance are back-transformedfrom log10 transformations 518 TABER D. ALLISON Ecology, Vol. 71, No. 2

of individual plant values. Transformedmeans were 1.0 A used in comparisonsof means among populations(Al- lison 1987), and thereforewere also used in regressions. I plottedweighted means of pollinationsuccess and A seed set against mean male strobilusproduction per < 0.8- 0 plantand mean nearestneighbor distance for each pop- 0.8 ulation. Weighted means of pollination success and seed setwere used in theregressions to reducethe effect 07 * 1985 of plants that produced only one or two ovules on 0.6- population means and variances. These were calculat- 0 A 1984 ed by multiplyingeach plant's ovule productionby the C number of ovules produced by that plant divided by 0 *H s * z ~~~~~~~~1983 the mean ovule productionfor the population. No sig- 0~~~~~~~~ . nificantdifferences were observed among years in the 0.4 regressionof eitherpollination success or seed set ver- 0. sus pollen productionand nearest neighbordistance; thus the data fromdifferent years were combined into one regression(e.g., see Fig. 1). 0.2 , I 0 100 200 Pollination success is a rapidly saturatingpositive 300 functionof male strobilusproduction (Fig. 1A) and a negativefunction of nearestneighbor distance or plant X Male Strobilus Production per Plant spacing (Fig. 1B). Regressions performed on log10 transformationsof both independent variables produced significantstraight-line functions (male strobili: 1.0 B Fig. 2A, Table 3a; nearestneighbor distance: Fig. 2B). * 1985 Log10transformation of the independentvariables was used to improve the fitof linear functionsto the data, la _A 1984 althoughit is apparent fromFig. 2A that some non- co 0.8 linearityremains. Because percentage data are ob- .C* 0 1983 viously constrainedwhile pollen productiondata are CL less so, some nonlinearrelationship is inevitable. Nearest neighbordistance and mean male strobilus 7 production of a population were significantlynega- 0.6- tivelycorrelated (Table 2); high-densityT. canadensis C populations tended to have high male strobiluspro- .2* duction per plant (see Table 1). This reflectsthe influ- 0 ence of past deer historyon Apostle Islands popula- a 0.4 0 tions used in the regression(see Discussion). Enough variabilityin thedata set existed,however, so thatboth variables contributedsignificantly to the regressionaf- teradjusting for the other.Stepwise regression on pol- 0.2 . a lination success indicated that male strobilusproduc- 0 1 2 3 4 tion contributedmore to pollination success than did distance to nearestneighbor. Together male strobilus production and nearest neighbordistance accounted X Nearest Neighbor Distance (m) for86% of the variance observedin pollinationsuccess FIG. 1. Plot of pollinationsuccess (proportion of ovules among sites (Table 3b). fertilized)vs. (A) mean male strobilusproduction and (B) Seed set was not as well correlatedwith male stro- meannearest neighbor distance of different Taxus canadensis bilus production (Table 2 and Table 3c) or nearest populations. neighbordistance (Table 2), althoughboth correlations were significant.Stepwise regressionshowed that nei- is stronglycorrelated with both log10male strobilus thervariable significantlyadded to theregression when production and pollination success (Table 2). When adjustingfor the other. Seed setis also indirectlyrelated adjustingfor male strobilusproduction, however, ovule to these two variables by pollinationsuccess (Table 2; production does not significantlyexplain additional P < .005). variationin pollinationsuccess. The reverseis not true; Ovule productionis anotherfactor that might influ- male strobilus production adds significantlyto the ence pollinationsuccess and seed set. In fact,in a com- regressionwhen adjustingfor ovule production.Ovule parison among populations,the log10ovule production production is significantlynegatively correlated with April 1990 POLLINATION AND SEED PRODUCTION IN TAXUS 519 1.0 A tionin naturalpopulations of a wind-pollinatedspecies. As expected,pollination success in T. canadensis pop- *u * ulationsis positivelycorrelated with pollen production A and negativelycorrelated with plant spacing.The sim- ilarityin thepatterns of pollinationsuccess in different

X0.8* yearsand the significanceof these patternseven when data are pooled fromgeographically separated populations is striking.If, for example, pollen production decreasedin a populationin one year,this was reflected > 0.6- in reducedpollination success in thatpopulation. Vari- C ables such as springweather conditions apparently had 0 U 1985 littleinfluence on pollen dispersaland pollination(see 0 also Sarvas 1962, Matthews 1963). Othershave noted o 0.4-0 A 1984 thatthe morphology of wind-pollinated flowers ensures IL A 0 thatpollen release generallyoccurs under optimal con- 1983 ditions forpollen transport(e.g., Bianchi et al. 1959). A The use of spacing as a measure of abundance em- 0.2- phasizes the leptokurticcharacter of pollen dispersal -1 0 1 2 3 and implies that the pollination dynamics of a wind- Male Strobilus Productionper Plant pollinated plant is dependenton its nearestneighbor. Log1 0 Expressingabundance as plant densitymay be a more appropriatemeasure, as it emphasizes that the pollen 1.0 B reachinga plant has an areal or diffuseorigin rather than a point origin (Levin and Kerster 1974). In the %A 0 1985 currentstudy, I do not have the data to resolve this issue adequately. A A 1984 Mean nearestneighbor distance can be convertedto @ 0.8- C1 * 1983 0:1 density(Southwood 1978). Using the lattervariable in

CL the regressionwill not affectthe magnitude of the regressioncoefficient, but the sign of the coefficient 0.6 (0 obviously will be reversed.Nevertheless, the average 0. distancebetween plants does accuratelydescribe pollen E availabilityfor the average plant in a population,par- ticularlywhen combined with mean pollen production. so In particular,the two variables interactto influence ? 0~~~~~~~~0.4- pollination efficiency.For example, plant spacing on L_ A Basswood and Rocky Island are identical (Table 1), but the higherpollen productionof Rocky Island cor- 0. A~~~~~~~ respondsto higherpollination success for T. canadensis in In 0.2- individuals that population. contrast,Stockton -2 -1 0 Island and Outer Island have similar levels of pollen

Log 1 oNearest NeighborDistance TABLE 2. Matrixof regressioncoefficients for log,0 male and FIG. 2. Plotof pollination success vs. (A) log,(male stro- log10female strobilus production, logo nearest neighbor bilusproduction and (B) log,(nearest neighbor distance. Near- distance,pollination success, and seed set. All regressions est neighbordistance is measuredin metres.The regression are significantat P < .05 with n = 18. equationfor part A is givenin Table 3a. Variable seed set when the latteris adjusted forpollination suc- Polli- cess (Table 3d). In populations with equivalent pollination Loglo Loglo Loglo suc- Seed nationsuccess, those populations that, on average,have Variable male female nearest cess set fewerovules per plant will have higherseed set. Log10male Log10female .945 Loglo nearest -.702 -.631 DISCUSSION Pollination This is the firststudy that quantitatively relates pol- success .872 .789 -.810 Seed set .558 .569 -.499 .713 linationefficiency to plant spacing and pollen produc- 520 TABER D. ALLISON Ecology, Vol. 71, No. 2

TABLE 3. Summaryof regressionequations for pollination The regressionequations can be used to estimate successand proportionseed set. Eq. 3a correspondsto the pollinationsuccess and seed set in other T. canadensis plotin Fig. 2B. populations. To testthe regressionmodel, I used data a) Y= 0.42 + 0.23X,;r2 = 0.75 on male strobilusproduction, nearest neighbor dis- Y = pollinationsuccess (percent ovules pollinated) tance, and pollination success collected in 1985 from X, = logOmean male strobilusproduction per ramet Oak Island of the Apostle Islands. These data were not b) Y= 0.41 + 0.16X, - 0.12X2;r2 = 0.86 used in buildingthe originalregressions. T. canadensis Y= pollinationsuccess plants on Oak Island are small and, on average, pro- X, = log10mean male strobilusproduction per ramet duce fewmale strobili.I substitutedthe followingval- X2= log1omean nearest neighbor distance (measured no. of male in metres) ues into the equation of Table 3b: log10 strobiliproduced per ramet= 0.47; log10of the nearest c) Y= 0.22 + 0.08X; r2 = 0.31 neighbordistance (measured in metres)= -0.07. The Y = proportionovules setting seed X, = log1omean male strobilusproduction per ramet predictedpollination success forOak Island based on theseparameters is 0.494; theobserved value was 0.485. d) Y= -0.03 + 0.57X, - 0.008X2;r2 = 0.71 Y = proportionovules setting seed The extremelyhigh predictability (98%) ofthe equation X, = pollination success in this particularinstance may be fortuitous.The re- X2= meanovule production per ramet sults suggests,however, that the relationshipbetween pollination success, plant spacing, and male strobilus productionin T. canadensis is strongand may be use- production,but the greaterspacing betweenplants on fullyapplied to estimatingthe importance of pollen Stocktonresults in lower pollination success. availability as a regulatorof pollination success and Seed set (in comparisonto pollinationsuccess) is not seed set in T. canadensis populations throughoutthe as well correlatedwith pollen productionor plantspac- species' range. ing, possibly reflectingthe greatervariety of factors The regulationof seed set by pollen availability in influencingseed set. For example, inbreedingdepres- plant populations is widely debated, although most sion resultingfrom increased self-fertilizationin low- models of plant sexual reproductionassume that re- densitypopulations of wind-pollinatedspecies (e.g., sourcesand not pollen limitseed set (Willson and Bur- Farrisand Mitton 1984) could account forthe reduced ley 1983, Lloyd 1984, but see Bierzychudek1981). In seed set in low-densityT. canadensis populations. T. fact,the relationships described in thispaper werepos- canadensis, however,appears to be highlyself-fertile sible due to the wide naturalvariation in pollen avail- (Allison 1987). ability,a resultof the differencesin plant densityand Rodent predation on developing seeds may reduce pollen productionrecorded in the various populations seed set in T. canadensis. I frequentlyobserved empty of T. canadensis. The wide variation in plant density husksof developing seeds on theforest floor, and small and pollen productionthat I measured is due princi- seed receptaclesbearing tooth marks were oftenleft on pallyto the deer historyof the Apostle Islands. Several the plant, indicatingthat developing seeds had been islands currentlyhave deer or had deer in the past. removed. The intensityof seed predationon T. cana- Otherislands neverhad deer. The yew populations on densis varies annually and among populations (T. D. islands that have deer now or had deer in the past Allison,personal observation). consistof widelyspaced plants thatproduce littlepol- Differencesin resource availability may also influ- len and have low levels of pollinationsuccess and seed ence patternsof seed set observed among the different set. Hand-pollination significantlyincreased seed set sites.The interactionbetween resource availability and in these populations, indicatingthat seed production pollen availability as an influenceon seed set is sug- was pollen limited (Allison 1987). gested by the equation in Table 3d, which combines All of these relationshipsare based on population pollinationsuccess and ovule productionas variables means; mean pollen production per plant and mean determiningseed set. When pollinationsuccess (influ- nearestneighbor distance estimate the average amount enced by pollen availability) in two populations is of pollen produced at an average distance from the equivalent,populations with high ovule productionper average plant. There is considerable variation in pol- plant will have lower seed set than populations with lination efficiencyamong plants within populations, low average ovule productionper plant. This suggests however,indicating that pollen availabilityis not the resource limitation of seed production (Stephenson same forall plants. Greatestvariation among plants is 1980). Within Taxus populations, however,the rela- recordedin populationswith low mean pollinationsuc- tionshipbetween ovule numberand seed set was weak cess. In part, this is a statisticalartifact related to the and frequentlynot statisticallysignificant. Although nature of percentagedata, but in these populations resourcesmay influenceseed set in T. canadensis, the some individuals have high levels of pollination and majorityof variationobserved in seed set among sites seed set. (51%) can be explained by differencesin mean polli- Althoughthe pollen cloud is derived fromthe entire nationsuccess. The latter,in turn,is stronglycorrelated population, a varietyof factorscould result in local with pollen productionand plant spacing. variationin the densityof the pollen cloud reachinga April 1990 POLLINATION AND SEED PRODUCTION IN TAXUS 521

plant. Characteristicsof a plant's nearest neighbors, commentsgreatly improved this manuscript.Field workwas e.g.,their distance and pollenproduction, or theamount supportedby grantsfrom Sigma Xi, the Universityof Min- of pollen produced by the plant,if self-fertile,could be nesota, and the Minnesota Zoological Society. important.For example, T. canadensis is self-fertile, LITERATURE CITED and I observed a weak, but significantrelationship between a plant's pollination success and its pollen pro- Allison, T. D. 1987. The reproductivebiology of Canada in yew (Taxus canadensis Marsh.) and its modificationby duction some but not all, Apostle Islands popula- herbivory.Dissertation. University of Minnesota, Minne- tions (maximum r = 0.25). On Rocky Island, however, apolis, Minnesota, USA. pollination success of individual plants is not signifi- Bateman,A. J. 1947. Contaminationof seed crops.II.Wind- cantlycorrelated with nearest neighbor distance or the pollination.Heredity 1:235-246. pollen productionof the plant's nearestneighbor. This Bianchi, D. E., D. J. Schwemmin,and W. H. Wagner, Jr. 1959. Pollen release in the common ragweed (Ambrosia indicates that determinationof the factors influencing artemisiffolia).Botanical Gazette 120:235-243. pollen availabilityfor individual wind-pollinated plants Bierzychudek,P. 1981. Pollinatorlimitation of plant repro- is not simple. Chance, in particular,may play a large ductive effort.American Naturalist117:838-840. role. Chamberlain,C. J. 1966. Gymnosperms,structure and evolution. Dover, New York, New York, USA. The variationin thepollen availabilitywithin a pop- Charnov,E. L. 1982. The theoryof sex allocation. Princeton ulation has importantconsequences for our attempts UniversityPress, Princeton,New Jersey,USA. to model sexual allocationin plants.For example,wide Dupler,A. W. 1917. The gametophytesof Taxus canadensis variationin genderhas been observed among individ- Marsh. Botanical Gazette 64:115-136. uals ofplant populations (e.g., Primack and Lloyd 1980). Faegri, K., and L. van der Pijl. 1979. The principles of pollinationecology. Pergamon,Oxford, England. Typically,such variation is assumed to reflectthe in- Farris,M. A., and J. B. Mitton. 1984. Population density, fluenceof plant status (i.e., the size, age, and overall outcrossingrate, and heterozygotesuperiority in ponderosa resourcebudget of the plant; Lloyd and Bawa [1984]) pine. Evolution 38:1151-1154. and not pollen availabilityon plant reproductivesuc- Koski, V. 1970. A studyof pollen dispersalas a mechanism in cess. There is growingevidence, that of gene flow conifers.Communicationes Instituti Fores- however, pollen talis Fenniae 70:1-78. availabilityvaries among populationsof a species (this Lanner, R. M. 1966. Needed: a new approach to the study study; Worthen and Stiles 1988). Pollen availability of pollen dispersion.Silvae Genetica 15:50-52. may also varyamong individuals withina population, Lemen, C. 1980. Allocation of reproductiveeffort to the limitingseed set in some plants while resourcesmay male and femalestrategies in wind-pollinatedplants. Oeco- logia (Berlin) 45:156-159. limitseed set in others.Variations in reproductivesuc- Levin, D. A., and H. W. Kerster. 1974. Gene flowin seed cess and sex expressionamong plantscould resultfrom plants. EvolutionaryBiology 7:139-220. variationsin resourceand pollen availability. Lloyd, D. G. 1984. Genderallocations in outcrossingcosex- The equations describedhere (Table 3) may provide ual plants. Pages 277-300 in R. Dirzo and Jos6 Sarukhan, editors. on plant population ecology. Sinauer, a quantitativebasis forstudying in sex Perspectives changes expres- Sunderland,Massachusetts, USA. sion in response to changes in pollen availability.For Lloyd, D. G., and K. S. Bawa. 1984. Modificationof the example, froma plant's perspective,the equation es- genderof seed plants in varyingconditions. Evolutionary timnatingpollination success (Table 3b) also estimates Biology 17:255-338. the probabilityof one of the plant's ovules being pol- Martell,A. M. 1974. Canada yew. Pages 158-160 in J. D. Gill and W. M. Healy, editors.Shrubs and vines forNorth- linated (e.g., for Oak Island, 0.494, see above). This easternwildlife. United StatesForest Service General Tech- probabilityincreases as pollen availability increases, nical Report NE-9. indicatingthat a plant's gain in femalefitness is partly Matthews,J. D. 1963. Factors affectingthe productionof influencedby pollen availability. Female flowersor seed by foresttrees. Forestry Abstracts 24. R. and D. G. Lloyd. 1980. Sexual strategiesin structures(i.e., ovules) functionas collectorsin Primack, B., pollen plants.IV. The distributionsof gender in two monomorphic wind-pollinatedplants. Increasing the number of ovules shrubpopulations. New Zealand Journalof Botany 18:109- when pollen availability and the probabilityof polli- 114. nationis low ensuresthat a minimumnumber of ovules Raynor, G. S. 1967. Effectsof a foreston particulatedis- will be pollinated. At low levels of pollen availability, persion. Pages 581-588 in C. A. Mawson, editor. USAEC MeteorologicalInformation Proceedings. Chalk River Nu- greaterfitness gains for a wind-pollinatedplant may clear Laboratories,Ontario, Canada. be achieved by increased allocation to ovules (female Regal, P. J. 1982. Pollinationby wind and animals: ecology function)rather than to pollen (male function).Further of geographicpatterns. Annual Review of Ecologyand Sys- work in wind-pollinatedspecies is required to deter- tematics13:497-524. R. 1962. on the and seed mine the importanceof pollen availabilityto seed pro- Sarvas, Investigations flowering crop of Pinus silvestris.Communicationes Instituti Fores- duction and plant sex expression. talis Fenniae 53:1-198. 1968. Investigationson the floweringand seed crop of Picea abies. CommunicationesInstituti Forestalis Fen- ACKNOWLEDGMENTS niae 67: 1-84. I thankD. Thiede, S. Householder, P. Regal, and the Park Southwood, T. R. E. 1978. Ecological methods: with par- Rangers of the Apostle Islands Natural Lakeshore for their ticularreference to the studyof insectpopulations. Second assistancein conductingthis research. I thankE. Cushingand edition. Methuen, London, England. P. Morrow,J. Thomson, and two anonymousreviewers whose Squillace, A. E. 1967. Effectivenessof 400-foot isolation 522 TABER D. ALLISON Ecology,Vol. 71, No. 2

around a slash pine seed orchard.Journal of Forestry65: Willson,M. F., and N. Burley. 1983. Mate choice in plants. 823-824. PrincetonUniversity Press, Princeton,New Jersey,USA. Stephenson,A. G. 1980. Fruit set, herbivory,fruit reduc- Worthen,W. B., and E. W. Stiles. 1988. Pollen-limitedfruit tion,and the fruitingstrategy of Catalpa speciosa (Bignoni- set in isolated patches of Maianthemumcanadense Desf. aceae). Ecology 61:57-64. in New Jersey.Bulletin of the TorreyBotanical Club 115: Tauber, H. 1967. Differentialpollen dispersion and filtra- 299-305. tion. Pages 131-141 in E. J. Cushingand H. E. Wright,Jr., Wright,J. W. 1952. Pollen dispersionof some foresttrees. editors. Quaternarypaleoecology. Yale UniversityPress, United States Forest Service NortheasternForest Experi- New Haven, Connecticut,USA. ment Station Research Paper 46. Whitehead,D. R. 1983. Wind pollination:some ecological 1953. Pollen-dispersionstudies: some practicalap- and evolutionaryperspectives. Pages 97-108 in L. Real, plications.Journal of Forestry51:114-118. editor. Pollination biology. Academic Press, New York, New York, USA.