Ecology, 84(9), 2003, pp. 2462±2475 ᭧ 2003 by the Ecological Society of America

MULTIGENERATIONAL EFFECTS OF FLOWERING AND FRUITING PHENOLOGY IN PLANTAGO LANCEOLATA

ELIZABETH P. L ACEY,1,4 DEBORAH A. ROACH,2 DAVID HERR,3 SHANNON KINCAID,1 AND RACHAEL PERROTT2 1Department of Biology, 312 Eberhart Building, University of North Carolina, Greensboro, North Carolina 27402-6170 USA 2Department of Biology, University of Virginia, Charlottesville, Virginia 22904-4328 USA 3Department of Mathematics, University of North Carolina, Greensboro, North Carolina 27402 USA

Abstract. Phenological patterns of ¯owering and fruiting can be in¯uenced by the effects of reproductive time on seed production. We propose here that these patterns are also in¯uenced by phenological effects on offspring quality. Furthermore, we hypothesize that there are cross-generational trade-offs between parental and offspring components of parental ®tness in¯uencing the evolution of reproductive phenology. To test our hypothesis, we examined the multigenerational effects of ¯owering and fruiting phenology in Plantago lanceolata. Offspring of 30 families were transplanted into ®eld plots to measure the effects of onsets of ¯owering and fruiting, duration of fruiting, percentage fungal infection, and damage by grasshoppers on total seed production, our measure of the within-generational component of parental ®tness. To gather information about cross-generational contributions to parental ®tness, we assessed the quality of off- spring produced at different times in terms of seed mass and germination. Families signi®cantly differed in ¯owering and fruiting onsets. Larger plants began ¯owering earlier, and earlier ¯owering plants matured fruits earlier and produced fruits for a longer time. Signi®cant family-mean correlations among these traits suggest that selection on any one trait will change all three traits. A negative family-mean correlation between fruiting onset and seed production suggests that we can expect an antagonistic trade-off in response to selection on these two traits. Early fruiting signi®cantly reduced seed predation by grasshoppers and increased seed production. In contrast, late-maturing seeds were sig- ni®cantly heavier and germinated more rapidly than did early-maturing seeds produced by the same plants. The directions of the multigenerational effects support the hypothesis that there are cross-generational trade-offs between parental and offspring components of pa- rental ®tness. The experiments indicate that multigenerational ®tness effects should be considered in future studies addressing the evolution of ¯owering and fruiting phenology. Key words: ¯owering; fruiting; multigenerational ®tness; parental effects; path analysis; path- ogen infection; phenology; Plantago lanceolata; seed predation.

INTRODUCTION hamer 1991, English-Loeb and Karban 1992, Diggle The time of ¯owering and fruit ripening in plants 1995, Bishop and Schemske 1998, Ollerton and Diaz can evolve in response to many selective pressures that 1999, O'Neil 1999, Husband and Schemske 2000, Pico in¯uence seed production. Onset of ¯owering/fruiting and Retana 2000). For many species, these selection must occur early enough during a growing season to pressures may function in concert to modify reproduc- allow for successful pollination, fertilization, and fruit tive phenology (e.g., Schemske et al. 1978, Bawa 1983, maturation (e.g., see reviews by Rathcke and Lacey Evans et al. 1989, English-Loeb and Karban 1992, Bro- 1985, Primack 1987, Ollerton and Lack 1992). The dy 1997, PicoÂand Retana 2000, Pilson 2000). length of this growing season can be determined by In contrast to the many empirical studies focusing physical factors, such as temperature and water avail- on seed production, few studies have examined how ability. Within this window, seed production can be ¯owering and fruiting phenology may evolve in re- in¯uenced by the timing of pollinators, pathogens, seed sponse to selection pressures acting on the progeny. predators, resource availability, and the synchronous This lack of attention to offspring ®tness is also man- ¯owering of individuals having the same ploidy level ifest in theoretical models of ¯owering time (e.g., Pal- (e.g., Augspurger 1980, Bawa 1983, Schmitt 1983, tridge and Denholm 1974, Cohen 1976, Vincent and Schemske 1984, Campbell and Motten 1985, Rathcke Pulliam 1980, Chiariello and Roughgarden 1984, de and Lacey 1985, Primack 1987, de Jong and Klink- Jong et al. 1992). Individual seed mass can change over a reproductive season (e.g., Cavers and Steel 1984, Manuscript received 14 February 2002; revised 7 January Thompson and Pellmyr 1989, Galen and Stanton 1991, 2003; accepted 16 January 2003. Correponding Editor: T.-L. Ashman. Winn 1991, Wolfe 1992, Diggle 1995, Ollerton and 4 E-mail: [email protected] Diaz 1999, Simons and Johnston 2000), and ¯owering 2462 September 2003 MULTIGENERATIONAL EFFECTS OF TIMING 2463 time can in¯uence offspring germination, survival, their results, Lacey and Herr hypothesized that from and/or ®tness (Lacey and Pace 1983, Case et al. 1996, the perspective of offspring ®tness, natural selection Ollerton and Diaz 1999, PicoÂand Retana 2000, Wolf favors parents that ¯ower later in the ¯owering season and Burns 2001, Galloway 2002). Thus, offspring pro- when it is warmer in the piedmont, North Carolina. If duced at different times are not necessarily equally ®t. they are correct, then one must ask why P. lanceolata These studies suggest that offspring quality, as well as begins ¯owering so much earlier, when it is cooler. quantity, could in¯uence the evolution of ¯owering and Some genotypes begin ¯owering in late April. Our hy- fruiting phenology. pothesis is that there are cross-generational trade-offs The in¯uence of offspring quality on the evolution between parental and offspring components of parental of reproductive timing is also suggested by recent stud- ®tness. Thus, although early-¯owering parents may ies of parental environmental effects. Greenhouse and produce less ®t offspring, they may produce suf®ciently growth-chamber studies show that the environment more offspring to offset the reduction in offspring ®t- during fertilization and early embryonic development ness. of an offspring while still attached to the maternal par- To test our hypothesis we conducted two types of ent can in¯uence offspring seed mass and germination experiments: one to measure the within-generational (e.g., Riddell and Gries 1958, Koller 1962, Robertson component of parental ®tness, reproductive success, et al. 1962, Sawhney and Naylor 1979, Gutterman and one to measure the cross-generational component, 1980±1981, Siddique and Goodwin 1980, Wulff 1986, i.e., offspring ®tness, with Plantago lanceolata. In the Case et al. 1996, Lacey 1996). It can also in¯uence ®rst, we examined the phenological patterns of ¯ow- offspring ®tness in the ®eld (Lacey and Herr 2000). ering and fruiting and their effects on seed production, The studies cited here examined the speci®c effects of our within-generational ®tness measure. Also, we ex- postzygotic temperature and/or photoperiod on off- amined the phenological patterns of pathogen and seed spring traits. Because temperature and photoperiod, on predator attack. We hoped to answer the following average, change predictably throughout a growing sea- questions: (1) What are the phenological patterns of son, these studies suggest that by altering ¯owering/ ¯owering, fruiting, pathogen attack, and seed predation fruiting phenology, an individual plant has the potential in a North Carolina population of P. lanceolata? (2) to change its postzygotic temperature or photoperiod, How do ¯owering and fruiting times affect within-gen- and thereby also change the quality of its offspring. erational ®tness, as measured in terms of seed produc- Differences in offspring quality could favor certain tion? (3) Can ®tness differences associated with ¯ow- ¯owering/fruiting times over others. ering/fruiting phenology be explained by timing of The above studies suggest that if we wish to under- pathogen attack and seed predation? We collected data stand why plants ¯ower and fruit when they do, we on three phenological traits: ¯owering onset, fruiting should assess the multigenerational ®tness effects of onset, and fruiting duration. We assessed the impact of ¯owering/fruiting phenology. Ideally, one should as- family and maternal plant size on these traits and de- sess the effect of parental environment on parental ®t- termined their phenotypic and family-mean correla- ness in the parental habitat, the within-generational ®t- tions. Then we used structural equation modeling (e.g., ness component for a parent, and offspring ®tness in Mitchell 1992, Petraitis et al. 1996, Grace and Pugesek the offspring habitat, the cross-generational component 1998, Scheiner et al. 2000) to examine the causal re- of parental ®tness (Donohue and Schmitt 1998). Two lationships among plant size, ¯owering/fruiting phe- attempts to do this, in contexts unrelated to ¯owering/ nology, infection, predation, and seed production. fruiting phenology, suggest that there may be trade- In the second set of experiments, in the laboratory offs between these parental and offspring components we assessed the impact of fruiting phenology and fam- (Donohue and Schmitt 1998, Donohue 1999). ily on two offspring traits: seed mass and germination. To explore further the cross-generational interactions Lacey and Herr (2000) observed that high postzygotic between ®tness components in parental and offspring temperature reduced seed mass but increased germi- generations, which might arise because of parental en- nation (Lacey 1996, Lacey and Herr 2000). Percentage vironmental effects and which might in¯uence the evo- germination strongly in¯uenced overall offspring ®t- lution of ¯owering and fruiting phenology, we exam- ness. Therefore, we asked the following questions: (1) ined the multigenerational consequences of ¯owering How does fruiting time affect these ®tness compo- and fruiting phenology in Plantago lanceolata. Our nents? (2) Are the effects in the directions suggested study was motivated by Lacey and Herr's (2000) ob- by the previous experiments? In other words, does de- servation that postzygotic temperature can in¯uence layed fruiting reduce seed mass and increase germi- offspring ®tness in ®eld-grown P. lanceolata. Using nation? two parental temperature regimes that resemble mean METHODS monthly temperatures for May and July, during which time P. lanceolata ¯owers in North Carolina, Lacey Biology of experimental species and Herr observed that high postzygotic temperature Plantago lanceolata L. (Plantaginaceae), ribwort increased offspring ®tness by almost 50%. Based on plantain, is a short-lived perennial herb that grows in 2464 ELIZABETH P. LACEY ET AL. Ecology, Vol. 84, No. 9 disturbed sites, abandoned crop ®elds, and lawns in son (E. P. Lacey, unpublished data). Therefore, in this temperate North America and in its native Eurasia. The experiment, we sought to estimate the damage through- species was introduced into North America ϳ100±200 out the season and its impact on seed production. years ago (Cavers et al. 1980). Individuals typically live from 2 to 5 years (e.g., Cavers et al. 1980, An- Experimental designs tonovics and Primack 1982, Lacey and Herr 2000, Seeds for our experiment were collected from 30 Roach 2003). After germinating in the fall or spring, plants naturally growing along a transect in a ®eld on an individual grows vegetatively as a rosette. With the the campus of Duke University in Durham, NC. Be- onset of ¯owering, a plant produces reproductive spikes cause this species is self-incompatible and predomi- at the ends of long stalks arising from leaf axils. The nantly wind pollinated, we treated seeds collected from ¯owering season for P. lanceolata in the North Car- a particular individual as constituting a maternal half- olina (NC) piedmont extends from late April into Au- sib family. We germinated seeds in the Duke University gust. Growth chamber and ®eld experiments show that Phytotron to minimize juvenile mortality and trans- ¯owering phenology is both genetically and environ- planted a total of 1200 six-week-old seedlings back mentally determined (e.g., Antonovics and Primack into the ®eld where the maternal parents grew. Seed- 1982, Wolff 1987, Wolff and van Delden 1987, van der lings from each family were planted in a randomized Toorn and van Tienderen 1992). Mowing regime is block design within two 10.3 ϫ 4.4 m blocks. Within known to in¯uence the evolution of ¯owering time (van each block, plants were located every 15 cm with stag- der Toorn and van Tienderen 1992). gered rows 10 cm apart. Except for a 3-cm hole created Various pathogens and phytophagous insects can at- for each seedling, the surrounding plant community tack reproductive spikes during the ¯owering/fruiting was left undisturbed. The plants used in our experiment season (e.g., de Nooij and Mook 1992). We focused on constituted a subset of a much larger set of plants es- two pathogens that appeared in our study population. tablished to study age-speci®c demography (Roach The disease symptoms matched those produced by two 2003). fungal pathogens that are known to infect P. lanceo- Seedlings were planted in April 1997. In November, lata: Diaporthe adunca (Rob.) Niessl, syn. Phomopsis we counted the number of leaves on each individual subordinaria (Desm.) Trav. and Fusarium moniliforme and measured the length and width of the longest leaf. var. subglutinans (Booth 1971). Diaporthe adunca These data were used to estimate total leaf area, a good (hereafter referred to as Diaporthe) causes the stalk of estimate of plant biomass (see Lacey 1996). Total leaf a spike just below the ¯owers to turn brownish-black area was estimated as the product of leaf number, lon- and wither about three days after fungal entry (de Nooij gest leaf length, and longest leaf width. Plants ®rst and van der Aa 1986). Spike development is arrested, ¯owered in May 1998, their second year. We collected which reduces seed production, and the infected spike our phenological data in summer 1998. Data from 2562 may or may not turn downward. Uninfected spikes on mature spikes produced by 444 plants were used in our the same rosette develop normally. Diaporthe is trans- analyses. Because some offspring died before ¯owering mitted by weevils, which puncture spikes during feed- or did not ¯ower, offspring sample sizes differed among ing. These wounds provide doors for fungal entry (de families. There were from 4±10 offspring per block. Nooij and van der Aa 1986). Infection can also occur For 92% of the family by block combinations, we col- in high humidity in the absence of wounding (Linders lected data from six or more offspring per family per et al. 1996). Fusarium moniliforme var. subglutinans, block. Data were analyzed by week, starting with week (hereafter referred to as Fusarium) produces a pink 1, which began Monday, 4 May 1998, and ending with mycelium over a variable portion of a spike, and it also week 14, which began 3 August 1998. Each week, we reduces seed production (Alexander 1982). Infection collected phenological data for each plant. We de®ned occurs during ¯owering, and one to many spikes of an ``onset of ¯owering'' as the week that we ®rst observed individual plant may be infected. Data suggest that in- a developing spike on a plant. As soon as a spike began fection is not systemic (Alexander 1982). Both diseases to turn brown and before it had dispersed seeds, it was produce a distinctive phenotype on the spikes. Thus, collected. We de®ne ``onset of fruiting'' as the week we assume that the above species were the sources of when the ®rst mature spike was collected. Duration of infection for our study plants. fruiting, an estimate of the fruit maturation period, Insect herbivores that feed on reproductive spikes equaled the week of the last collection minus the week include weevils (de Nooij and van der Aa 1986) and of the ®rst collection. At collection time, evidence of grasshoppers (E. P. Lacey and D. Roach, personal ob- fungal presence was recorded. If a spike showed no servation). We focused on grasshoppers because grass- withering and darkening of its stalk and no mycelium, hopper damage could be easily detected. Grasshoppers we assumed that the spike was healthy, i.e., not in- chew the tips of maturing infructescences and/or create fected. In the laboratory, we inspected each spike for lateral indentations on an infructescence. A previous evidence of grasshopper damage and again looked for experiment suggested that grasshopper damage wors- evidence of a mycelium using a dissecting microscope. ens during the latter part of the ¯owering/fruiting sea- Each spike was weighed to the nearest 0.001 mg. September 2003 MULTIGENERATIONAL EFFECTS OF TIMING 2465

To evaluate the quality of seeds produced at different ing duration, and the phenotypic correlation between times during the fruiting season, we compared the seed fruiting onset and skewness of the fruit maturation mass and the percentage of germination of seeds ma- curve. For the last analysis, we computed the skewness turing during two weeks: week 8 (19±26 June) and of the fruit maturation (duration) curve for each fruiting week 11 (13±17 July). These two weeks represented onset. Only weeks 7±11 had large enough sample sizes early and mid-season fruiting. To control for maternal of individual plants to be used in this analysis. Then and offspring variation, seeds from these maturation we examined the correlation between skewness and times were compared for the same set of maternal fam- fruiting onset. A correlation between these two traits ilies and offspring. Thus, the experimental design for could indicate that fruiting onset and the duration of seed mass was 12 half-sib maternal families ϫ 3±6 fruit maturation undergo correlated evolution. offspring nested within family ϫ 10 seeds per offspring ANOVA (SAS 2000) was used to examine the effect ϫ 2 seed maturation times, for a total of 900 seeds. of maternal family and block on fall leaf area. Before Seeds were weighed individually to the nearest 0.001 the analysis, we examined every family by block com- mg. For the germination experiment, we used 30 seeds bination to check for normality. Thirteen outliers, per offspring for a total of 2700 seeds. which were greater than three interquartile ranges from To examine germination, we sowed seeds on ®lter the median leaf area value, were deleted to normalize paper in covered petri plates. Filter paper was moist- the data for the analysis (N ϭ 431). Family and its ened daily. Each plate was marked with a grid to keep interaction were treated as random variables. track of individual seeds. Twenty-®ve seeds were as- To estimate seed production for each spike and to signed to each plate. The seeds from each offspring compare the effects of infection and grasshopper dam- were randomly assigned across all plates, with the con- age on seed production, we counted all healthy seeds straint that no more than one seed/offspring/maturation (i.e., all seeds that were brown and plump) on four time could be assigned to any particular plate. Plates subsets of spikes. We did not count aborted seeds, were kept in a growth chamber at 85% relative hu- which are black and ¯at. The four subsets were: (1) midity and a 16-h light:8-h dark cycle. Temperatures undamaged spikes, which showed no signs of pathogen resembled those of Durham, NC, in May: night; 15ЊC infection or predation (N ϭ 101), (2) spikes showing and day; 20ЊC (Teramura et al. 1981). Day of germi- evidence of grasshopper damage only (N ϭ 29), (3) nation was recorded for each seed over 43 days, the spikes showing evidence of infection only by Fusarium last seven of which showed no new germination. Seeds (N ϭ 28), and (4) spikes showing evidence of infection had been stored dry in envelopes in the lab for one only by Diaporthe (N ϭ 20). In our total data set, only year prior to this test. These storage conditions max- a handful of plants were both infected and eaten, and imize seed viability (Steinbauer and Grigsby 1975), and none showed signs of infection by both fungi. The the storage time allowed for completion of most, if not spikes whose seeds were counted were chosen so that all, after-ripening of seeds (Blom 1992). a range of families and spike masses would be repre- sented. These data were used in linear regression anal- Statistical analyses yses (SAS 2000) to determine how seed number Because the phenological data could not be nor- changed with spike mass. We also tested quadratic and malized, we used nonparametric tests to examine ¯ow- cubic regression models, but for all subsets the rela- ering/fruiting onset and duration. We used the multi- tionship between seed number and spike mass was variate G test (log-likelihood ratio) (Bishop et al. 1975, strongly linear. Therefore, the linear regression equa- Sokal and Rohlf 1995) to examine the effects of ma- tions were used to estimate seed number for spikes ternal family and block on onset of ¯owering, onset of whose seeds were not counted. The t test was used to fruiting, and fruiting duration (G test program, M. compare the slopes of regression lines for three of the Rausher, personal communication). The G test can be four subsets of spikes. The requirement that the resid- used to test the independence of three or more variables uals be normally distributed was satis®ed in those sub- when each variable corresponds to a dimension of a sets. No statistical test was needed to evaluate the effect multidimensional contingency table. In our case, fam- of the fourth subset. ily and block were two dimensions of the table, and To explore the interactions among ¯owering/fruiting onset of ¯owering, onset of fruiting, or fruiting duration phenology, predation, infection, and seed production, was the third dimension. To reduce the number of emp- we used a structural equation modeling (SEM) pro- ty cells in the table, we lumped some weeks together cedure (PROC CALIS, SAS 2000) to perform a path for ¯owering/fruiting onset and duration (¯owering on- analysis of the data (e.g., Mitchell 1992, Petraitis et al. set levels used: week 2, weeks 3±4, weeks 5±8; fruiting 1996, Grace and Pugesek 1998, Scheiner et al. 2000). onset levels: weeks 7±8, weeks 9±10, weeks 11±14; SEM allows one to perform path analyses of variables duration levels: Ͻ1 wk, 1±3 wk, 4±6 wk). of interest. Given an a priori path model that de®nes The Spearman rank correlation test (SAS 2000) was the causal relationships between multiple dependent used to examine phenotypic and family-mean corre- variables, one uses path analysis to measure the lations among ¯owering onset, fruiting onset, and fruit- strength of the causal relationships among the variables 2466 ELIZABETH P. LACEY ET AL. Ecology, Vol. 84, No. 9

FIG. 1. Alternative path models tested for goodness-of-®t to our data for Plantago lanceolata. In Model 1, paths go from fruiting onset and duration to predation (``1'' in parentheses). In Model 2, these paths are reversed (``2'' in parentheses). Model 2 is used to measure the causal relationships among size, phenological traits, infection, seed predation, and seed production. Model 2 was used to calculate path coef®cients. The strength of each coef®cient is indicated by the width of the arrow. Solid lines indicate positive direct effects; dotted lines indicate negative effects.

(Wright 1934, Li 1975). One can also use SEM to test ward. Leaf area directly in¯uences ¯owering onset, whether or not a path model chosen for the analysis ¯owering onset in¯uences onset of fruiting, fruiting ®ts the data. This allows the testing of alternative mod- onset in¯uences duration of fruiting, and fruiting du- els to see which model provides the best ®t. We tested ration, predation, and infection all directly in¯uence two models, ®rst with a goodness-of-®t chi-square (␹2) seed production (Seeds). Because grasshopper damage test. A signi®cant ␹2 value indicates that a model does showed seasonal change, we also included paths be- not ®t the data. We also report Bentler's and Bonett's tween the three phenological traits and predation. The Normed Fit Index (NFI), which is based on the model two models differed in the directions of these paths. ␹2 relative to that of a model that assumes independence We did not include paths between the phenological of all variables. NFI ranges between 0 and 1, with NFI traits and infection because probability of infection did Ͼ 0.90 indicating a good ®t (Bentler 1989). Because not show seasonal change. Because leaf area is a good our variables were not normally distributed, we used indicator of resource accumulation by a plant, we in- a weighted least squares procedure to test for goodness cluded paths from leaf area to infection, predation, of ®t of our models to the data. This procedure assumes fruiting duration, and seed production. Larger plants no particular distribution for the variables in the mod- might be more susceptible to infection and predation els. because larger plants produce more targets (spikes) for Our path models (N ϭ 444) included leaf area, the infection and predation. Dudycha and Roach (in press) three ¯owering/fruiting traits, percentage spikes in- observed that the frequency of Fusarium increases with fected by either fungus (Infection), percentage of spike number. Also, larger plants have more resources spikes damaged by grasshoppers (Predation), and seed available for lengthening the fruiting period and pro- production (Fig. 1). Family was not included because ducing more seeds. We included a path from ¯owering it is a categorical variable and has no quantitative value onset directly to seed production because onset might that is biologically meaningful. Because of the tem- in¯uence seed production independently of fruiting if, poral nature of the data, the causal relationships, i.e., for example, it affects pollinator activity. Plantago lan- direct paths, between most variables are straightfor- ceolata is pollinated by both wind and insects. We also September 2003 MULTIGENERATIONAL EFFECTS OF TIMING 2467 included a path from infection to predation because we (SAS 2000) of germination day on seed mass for each thought it possible that grasshoppers may discriminate maturation time. The t test was used to compare the between infected and uninfected spikes. Discrimination slopes of regression lines for each time. Germination would explain the negligible number of spikes that day was logit-transformed to improve the normality of were both damaged and infected. residuals. Our two models differed in the direction of the paths between predation and fruiting onset and duration (Fig. RESULTS 1). One could argue that time of onset of fruit matu- ration and length of the fruit maturation period in¯u- Flowering and fruiting patterns ence the probability of predation. Because grasshopper abundance likely increases throughout the reproductive Flowering onset occurred throughout May and June season, the later ¯owering and fruiting plants are likely (from week 2 to 8) across all plants. However, because to suffer more predation. Alternatively, one could argue most plants began ¯owering in May, family means for that predation might alter a plant's rate of reproductive onset ranged from 2.4 to 4.6 wk (Fig. 2). Fruits matured development, thereby affecting fruit ripening and du- from mid-June to early August (week 7 to 14). Family ration. Therefore, we tested both. In Model 1 (Fruiting means for onset of fruiting ranged from 8.5 to 10.6 wk. to Predation), paths go from both fruiting onset and Spike maturation for the whole population climbed duration to predation (paths marked with a ``1'' in Fig. from 53 spikes maturing in week 7 to a peak of 1111 1). In Model 2 (Predation to Fruiting), the direction of spikes in week 11, and then dropped to 19 in week 14. the paths is reversed (marked with a ``2''). Estimated seed number per spike for all spikes col- Using our better path model, we estimated the path lected, regardless of attack by pathogens or predators, coef®cients, which measure the strength of the causal declined from 99 seeds per spike in week 7 to 13 seeds relationships among variables. A path coef®cient es- in week 14 (Fig. 3a). timates the direct effect between pairs of traits, for Flowering onset, fruiting onset, and fruiting duration example, the effect of infection directly on seeds. Path were all signi®cantly correlated with each other (Table coef®cients can also be used to estimate an indirect 1). Earlier ¯owering plants began fruiting earlier and effect between two traits as mediated by one or more produced fruits for a longer time. Family-mean and other variables, for example, the path from infection phenotypic correlations were highly signi®cant. Ad- to seeds via predation. The indirect effect is the product ditionally, the family-mean correlation between fruit- of the path coef®cient from infection to predation and ing onset and seed production was signi®cant. The the coef®cient from predation to seeds. shape of the spike maturation curve for an individual To evaluate the phenological effects on seed mass, plant, on average, changed as onset of fruiting was we used a three-way ®xed-model ANOVA to test the delayed. Most plants that began maturation in week 7 effects of maturation time, family, and offspring nested matured their spikes over 4±6 weeks. Few plants ma- within family on seed mass, that is, the mass of seeds tured spikes more quickly. Most plants that began mat- produced by the offspring. All independent variables uration in week 8 matured spikes over 2±4 weeks. were treated as ®xed. The variable offspring was treated Plants that began fruit maturation latest (in week 12, as ®xed because we deliberately chose only offspring 13, or 14) all matured spikes within one week. Skew- whose fruiting duration extended over at least four ness of the fruit maturation curve was signi®cantly cor- weeks (weeks 8±11). Choosing only these offspring related with fruiting onset (N ϭ 5, Spearman rank cor- and ensuring that we had replicate offspring values relation coef®cient ϭ 1.00, P Ͻ 0 .0001; Kendall's tau within family also restricted the families used in the ϭ 1.00, P Ͻ 0.014). Skewness changed from negative experiment. These restrictions limit the breadth of our to positive as onset of fruiting was delayed. conclusions to the particular families used in the ex- Onset and duration of ¯owering differed among fam- periment. However, the restrictions allow us to examine ilies. Family, but not block, signi®cantly in¯uenced the effects of maturation time independently of mater- nal genetic effects, which would have been a confound- onsets of ¯owering (G ϭ 101.73, df ϭ 58, P Ͻ 0.001) ing factor if we had ignored the origin of the seeds. and fruiting (G ϭ 82.16, df ϭ 58, P Ͻ 0.05). The No transformation of the seed mass data was required. analysis of duration indicated that there was a signif- We used the ␹2 test to examine the effects of seed icant block-by-family interaction (G ϭ 102.59, df ϭ maturation time on total germination and percentage 58, P Ͻ 0.001), which is explained by the fact that the germination by day 2. This day was chosen because by families contributing most to these differences differed day 2, ϳ40% of all seeds germinating throughout the between blocks. In the two-way ANOVA of fall total test had germinated. By day 3, 70% of the seeds had leaf area, the block-by-family interaction was signi®- germinated. Thus, using day 2 gave us the greatest cant (P Ͻ 0.043). When each block was analyzed sep- ability to detect whether or not maturation time in¯u- arately, families did not differ signi®cantly in fall leaf ences early germination. To examine the effect of seed area (Block 1, F ϭ 1.43, df ϭ 29, 197, P Ͻ 0.084; mass on germination, we performed a linear regression Block 2, F ϭ 1.02, df ϭ 29, 174, P Ͼ 0.4). 2468 ELIZABETH P. LACEY ET AL. Ecology, Vol. 84, No. 9

FIG. 2. Mean onsets of ¯owering and fruiting and mean end of fruiting for offspring of the 30 experimental families of Plantago lanceolata. The dotted line shows the time between initiation of ¯owering and onset of fruit maturation. The solid line shows the duration of fruiting.

Temporal patterns of grasshopper damage and NFI exceeded 0.90 for both models, indicating that both fungal infection provided an excellent ®t when compared to a null mod- Damage by grasshoppers increased throughout the el that assumes independence among all variables summer and was always at a higher frequency than the (Model 1, NFI ϭ 0.996; Model 2, NFI ϭ 0.997). Be- cause the P value associated with the 2 test for Model incidence of fungal infection (Fig. 3b). Infection by ␹ 2 was more than twice the P value for Model 1, we Diaporthe never exceeded 4.2% of spikes collected in used Model 2 to explore further the interactions be- any one week. Infection by Fusarium never exceeded tween the phenological variables, predation, and seed 6.9% for any week. In contrast, the percentage of spikes production. damaged by grasshoppers rose from 15% in week 7 to Relationships between variables in the model varied 84% in week 14. greatly in strength, which can be seen by looking at Damage by fungi or grasshoppers reduced seed pro- the path diagram (Fig. 1) and the causal effects (Table duction per spike, as evidenced by the regressions of 2). The absolute values for the direct effects ranged seed production on spike mass for the four subsets of from 0.03 (¯owering onset to predation) to 0.73 (fruit- spikes (Fig. 4). The t tests showed that grasshopper ing onset to fruiting duration). The stronger relation- damage and Fusarium infection reduced seed produc- ships among variables (path coef®cients Ͼ0.21) were tion signi®cantly and by a similar amount when com- contained in three paths. Two run from leaf area to seed pared to healthy spikes, even when one lowers the crit- production: (1) the direct path, and (2) the indirect path ical P value to account for multiple comparisons through ¯owering and fruiting (Fig. 1). The direct ef- (healthy vs. eaten spikes, t ϭ 3.79, P Ͻ 0.001; healthy fect, however, explained most (71%) of the total causal vs. Fusarium-infected spikes, t ϭ 3.13, P Ͻ 0.002; effect of leaf area on seed production. Of the total eaten vs. Fusarium-infected spikes, t ϭ 0.45, P Ͼ 0.6). causal effect of ¯owering onset on seed production, the Infection by Diaporthe was most deleterious. Diapor- indirect effect through fruiting was greater. It explained the-infected spikes produced no seeds. 56% of the total effect. The third path is the direct effect of predation on seed production. This direct ef- Phenological effects on seed production fect explained almost all (96%) of the total effect of Both path models provided a good overall ®t to the predation on seed production. Finally, considering the data. The ␹2 values of both were nonsigni®cant, indi- paths between predation and the three phenological cating that the covariance structure speci®ed by both traits, the coef®cient for the path from predation to models could not be rejected given the covariance fruiting onset is at least double the coef®cients for the structure of the data (Fruiting to Predation Model 1, ␹2 other paths (Table 2). Moreover, this is the only phe- ϭ 10.41, df ϭ 6, P Ͼ 0.1; Predation to Fruiting Model nological trait that has a signi®cant family-mean cor- 2, ␹2 ϭ 7.17, df ϭ 6, P Ͼ 0.3). The Bentler and Bonett relation with seed production (Table 1). September 2003 MULTIGENERATIONAL EFFECTS OF TIMING 2469

0.015, P Ͻ 0.014; late maturation, day ϭ 0.006[mass] ϩ 0.07, R2 ϭ 0.019, P Ͻ 0.005). This relationship was similar for both maturation times, as evidenced by the lack of difference between the slopes of the regression lines for the maturation groups (t ϭϪ1.03, P Ͼ 0.30).

DISCUSSION Even though we recognize that our data are limited to one season and one population, we feel that the patterns observed here provide useful information about general patterns concerning ¯owering/fruiting phenology. Our data show that ¯owering and fruiting times can produce multigenerational effects in P. lan- ceolata. Within the parental generation, plants that ¯owered and matured fruit earlier produced more seeds. When considering the seeds produced by an in- dividual over its reproductive season, late-maturing seeds were heavier and germinated faster than did ear- ly-maturing seeds. When combined with the results of the experiment by Lacey and Herr (2000), these data provide evidence that reproductive phenology affects both offspring quality and quantity in P. lanceolata. The fact that several other studies (Lacey and Pace 1983, Case et al. 1996, Ollerton and Diaz 1999, Pico and Retana 2000, Wolf and Burns 2001, Galloway 2002) have also detected phenological effects on off- spring phenotype suggests that such multigenerational effects may be widespread in plants. Effects on seed size and germination likely affect offspring ®tness. Van

FIG. 3. (a) Weekly reproduction as mean seeds per spike (horizontal bars, with vertical bars representing Ϯ1 SE) and (b) disease and grasshopper damage at the ®eld site as fre- TABLE 1. Phenotypic and family-mean Spearmann rank cor- quency of mature spikes either infected by Diaporthe adunca relation coef®cients are shown for pairs of variables in (open bars), Fusarium moniliforme (striped bars), or grass- Plantago lanceolata. hopper damaged (solid bars). The number of spikes infected by both fungi or that showed evidence of grasshopper damage Coef®cient and infection was negligible. Pairs of variables Phenotypic Family-mean LA±FL Ϫ0.32*** Ϫ0.13 ns LA±FR Ϫ0.19*** Ϫ0.04 ns Effects on seed mass and germination LA±DUR 0.18*** 0.04 ns LA±PRED Ϫ0.13 ns Ϫ0.12 ns Late-maturing seeds were signi®cantly heavier than LA±INF 0.04** 0.37* early-maturing seeds when averaged over families LA±SEEDS 0.39*** 0.25 ns (1.21 Ϯ 0.01 mg for week 8, 1.26 Ϯ 0.01 mg for week FL±FR 0.50*** 0.64*** FL±DUR Ϫ0.39*** Ϫ0.49** 11 [means Ϯ 1 SE]; Table 3). Also, there was a sig- FL±PRED Ϫ0.02 ns 0.28 ns ni®cant family-by-time interaction, indicating that the FL±INF Ϫ0.10*** 0.07 ns effect of maturation time on seed mass had a genetic FL±SEEDS Ϫ0.34*** Ϫ0.20 ns component. FR±DUR Ϫ0.73*** Ϫ0.82*** FR±PRED 0.16* 0.23 ns Total percentage germination, i.e., over 43 days, was FR±INF Ϫ0.06** Ϫ0.03 ns very high for both seed maturation times (88.03% and FR±SEEDS Ϫ0.44*** Ϫ0.36* 89.47% for weeks 8 and 11, respectively), and early- DUR±PRED Ϫ0.04 ns Ϫ0.003 ns DUR±INF 0.02** 0.03 ns and late-maturing seeds did not differ in ®nal germi- DUR±SEEDS 0.48*** 0.25 ns nation (␹2 ϭ 1.36, df ϭ 1, P Ͼ 0.2). However, per- PRED±INF Ϫ0.18** Ϫ0.16 ns centage germination by day 2 did differ signi®cantly. PRED±SEEDS Ϫ0.27*** 0.04 ns Later maturing seeds had a higher percentage germi- INF±SEEDS Ϫ0.04* Ϫ0.06 ns nation (32.7% for week 8, 38.2% for week 11; ␹2 ϭ Notes: Abbreviations are: LA, leaf area; FL, ¯owering on- 7.92, df ϭ 1, P Ͻ 0.005). Seed mass signi®cantly in- set; FR, fruiting onset; DUR, fruiting duration; INF,infection; PRED, predation; SEEDS, seed production. Total offspring ¯uenced germination day for both maturation times for all 30 families ϭ 444. (early maturation, day ϭ 0.005[mass] ϩ 0.07, R2 ϭ * P Յ 0.05; ** P Յ 0.01; *** P Յ 0.001; ns, P Ͼ 0.05. 2470 ELIZABETH P. LACEY ET AL. Ecology, Vol. 84, No. 9

FIG. 4. Relationship between spike mass (mass) and seed number (seeds) with associated regression lines for three groups of spikes of Plantago lanceolata: (1) undamaged and uninfected (gray circles; regression equation: seeds ϭϪ1.310 ϩ 0.4830[mass] [solid line], R2 ϭ 0.79, P Ͻ 0.0001), (2) Fusarium-infected (solid squares; regression equation: seeds ϭ 4.225 ϩ 0.0843[mass] [solid line], R2 ϭ 0.14, P Ͻ 0.05), and (3) grasshopper-damaged (open triangles; regression equation: seeds ϭ 0.170 ϩ 0.1486[mass] [dashed line], R2 ϭ 0.67, P Ͻ 0.0001). Diaporthe-infected spikes are not shown because they never produced seeds.

der Toorn and Pons (1988) observed that early-ger- set was the only phenological trait that showed a sig- minating plants have a competitive advantage in P. ni®cant family-mean correlation with seed production. lanceolata. Seed mass and germination can in¯uence This suggests that selection for earlier fruit maturation lifetime ®tness or its components in other species (e.g., should increase seed production, our within-genera- Schaal 1980, Stanton 1984, Roach 1986, Winn 1988, tional ®tness component. BieÁre 1991, Galen and Stanton 1991, Donohue and Multigenerational effects could act synergistically or Schmitt 1998, Simons and Johnston 2000). antagonistically to determine the net direction of evo- Multigenerational effects are likely to in¯uence the lutionary change in reproductive phenologies (Lacey evolution of reproductive timing in P. lanceolata. Our and Herr 2000). Our data support the hypothesis that data, as well as those of others (Primack and Anto- the multigenerational effects are antagonistic, i.e., that novics 1981, Wolff 1987, Wolff and van Delden 1987, there are cross-generational trade-offs between parental van Tienderen and van der Toorn 1991, Lacey 1996, and offspring components of parental ®tness in P. lan- Lacey and Herr 2000; D. Roach, unpublished data) ceolata. From the point of view of reproductive output show that genetic variation underlies the phenotypic (e.g., seed production), which is how ®tness is typically variation in reproductive phenology in natural popu- measured in ecological and evolutionary studies, se- lations and that the genetic control is partially inde- lection appears to favor plants that begin to ¯ower and pendent of plant size, which is also under partial ge- fruit early in the season. In contrast, from the point of netic control. Therefore, selection pressures acting in view of offspring ®tness, selection appears to favor the parental and progeny generations can potentially parents that ¯ower and fruit later. This latter point is interact to in¯uence the evolution of ¯owering and supported by the Lacey and Herr (2000) study. If the fruiting phenology. The amount of genetic variation relative contributions of parental and offspring com- underlying the phenotypic variation in phenological ponents of parental ®tness change little over years, then traits determines the degree of response to selection. one would predict that ¯owering/fruiting time would The signi®cant family-mean correlations between ¯ow- contract around the period that maximizes the joint ering and fruiting onsets and duration suggest that all ®tness effects. However, if yearly ¯uctuations in weath- three phenological traits are genetically correlated with er cause ¯uctuations in the relative balance between each other, and therefore should coevolve. Fruiting on- parental and offspring components, then a prolonged September 2003 MULTIGENERATIONAL EFFECTS OF TIMING 2471

TABLE 2. Total causal, direct, and indirect effects for pairs er been observed to exceed 10% in this population of variables in Path Model 2. (Alexander 1982, Dudycha and Roach, in press). Be- Effect cause both path models ®t our data well, the relation- Variables ship between predation and fruiting is unclear. Earlier (from±to) Total Direct Indirect² maturing fruits suffered less grasshopper damage, but LA±FL Ϫ0.33 Ϫ0.33 0 whether or not damage in¯uences maturation rate is LA±FR Ϫ0.19 0 Ϫ0.19 LA±DUR 0.17 0.04 0.13 unknown. These relationships need further study. Also, LA±PRED Ϫ0.14 Ϫ0.13 Ϫ0.0005 as P. lanceolata individuals become older/larger, their LA±INF 0.06 0.06 0 fruiting duration is extended later and later into the LA±SEEDS 0.42 0.30 0.12 season (D. Roach, unpublished data). One can spec- FL±FR 0.50 0.50 Ϫ0.005 FL±DUR Ϫ0.37 0 Ϫ0.37 ulate that with increasing size, a plant is better able to FL±PRED Ϫ0.03 Ϫ0.03 0 tolerate seed loss from grasshoppers. FL±SEEDS Ϫ0.24 Ϫ0.10 Ϫ0.13 Our observation that late-maturing seeds were heavi- FR±DUR Ϫ0.73 Ϫ0.73 0 FR±SEEDS Ϫ0.27 0 Ϫ0.27 er than were early-maturing seeds did not match our DUR±SEEDS 0.37 0.37 0 predicted results based on Lacey's experiment (1996). INF±PRED Ϫ0.19 Ϫ0.19 0 In a growth chamber experiment, she observed that INF±FR Ϫ0.03 0 Ϫ0.03 INF±DUR 0.01 0 0.01 low, rather than high, temperature increased seed mass. INF±SEEDS Ϫ0.06 Ϫ0.11 0.05 This observation is consistent with results of other PRED±FR 0.16 0.16 0 growth chamber experiments examining the effects of PRED±DUR Ϫ0.03 0.08 Ϫ0.12 temperature in other species (e.g., Robertson et al. PRED±SEEDS Ϫ0.24 Ϫ0.23 Ϫ0.01 1962, Sawhney and Naylor 1979, Gutterman 1980± Notes: Total causal effects ϭ direct ϩ indirect effects. See 1981, Siddique and Goodwin 1980, Wulff 1986). Our Table 1 for abbreviations. ²0ϭ no effect. results suggest that a seasonal change in some variable other than temperature more than offsets the effect of high temperature on seed mass. Most studies docu- menting seasonal change in seed mass have reported a ¯owering/fruiting season would be expected. Such ¯uc- decline in mass over time, attributable to declining re- tuations could help to explain why P. lanceolata has sources (e.g., Cavers and Steel 1984, Thompson and such a long ¯owering season. While our experiments Pellmyr 1989, Galen and Stanton 1991, Winn 1991, do not allow us to quantify the relative contributions Wolfe 1992, Diggle 1995, Simons and Johnston 2000). of parental and offspring components of parental ®t- Clearly, declining resources do not explain our results. ness, nor quantify their annual ¯uctuations, they do One possible explanation is that plants may alter their indicate that both components should be considered in resource allocation within a spike throughout the re- studies addressing the evolution of reproductive phe- productive season such that later spikes produce fewer nology and that, ultimately, such quanti®cation is need- but larger seeds. Alternatively, maturation time effects ed. might be caused by differences in pollen donors. These Seed predation intensity changes over the reproduc- paternal effects could be genetically or environmen- tive season in many species. For some, seed predation tally based. At this point, the cause of the observed is highest early in the season (e.g., Evans et al. 1989, increase in mass is unknown. In spite of this, our results BieÁre and Honders 1996, Pilson 2000), for others, it is are in the predicted direction, if the hypothesis that highest late (e.g., Schemske 1984, BieÁre and Honders delayed ¯owering/fruiting enhances offspring ®tness is 1996, Bishop and Schemske 1998), and for a few spe- true. Large seeds often germinate more quickly and cies, it is highest during the time of peak reproduction produce larger seedlings than do small seeds (e.g., (e.g., Marquis 1988, Evans et al. 1989, English-Loeb Black 1958, Harper 1977, Schaal 1980, Stanton 1984, and Karban 1992). In our study, seed predation in- Roach 1986, Winn 1988, Galen and Stanton 1991, Si- creased throughout the season. Spike damage by grass- hoppers exceeded 80% at season's end. As evidenced by the path analysis results, predation negatively af- fected seed production per plant. Grasshopper repro- TABLE 3. Fixed-effects model ANOVA results for the ef- fects of maturation week, family, and offspring nested with- duction and survival generally increase as weather be- in family on seed mass. comes warm, dry, and sunny (Dempster 1963), which typi®es the changing weather conditions from April to Source df FP August, the reproductive season for P. lanceolata,in Maturation week 1 8.76 0.003 piedmont, North Carolina. Family 11 1.35 0.2 The data suggest that seed predation by grasshop- Offspring (Family) 34 8.92 Ͻ0.0001 Week Family 11 pers, unlike fungal infection, favors earlier ¯owering ϫ 4.24 Ͻ0.0001 and fruiting in P. lanceolata. In contrast to predation, Error 859 the percentage of spike infection by Fusarium has nev- Note: F values were calculated using Type III SS. 2472 ELIZABETH P. LACEY ET AL. Ecology, Vol. 84, No. 9 mons and Johnston 2000). Larger seeds typically are 1980±1981, Schaal 1984, Wulff 1986, Miao et al. more strongly favored in competitive, shaded, or stress- 1991a, b, Schmitt et al. 1992, Platenkamp and Shaw ful habitats, although there are some exceptions (re- 1993, Schmid and Dolt 1994, Lacey 1996, Sultan 1996, viewed in Donohue and Schmitt 1998). Mazer and Wolfe 1998), as well as offspring ®tness or Our germination results are also in the predicted di- its components in nature (Schmitt et al. 1992, Galloway rection, although they do not exactly match expecta- 1995, Donohue and Schmitt 1998, Lacey and Herr tions based on the experiment by Lacey and Herr 2000). There is presently, however, little evidence that (2000). On average, temperature increases from April these effects are evolutionarily important (e.g., see Do- to August in piedmont, NC (Teramura 1978), and in nohue and Schmitt 1998, Mazer and Wolfe 1998). One our experiment, the daily temperature averaged over reason may be that most attention has been directed the two weeks prior to our seed collections in weeks toward measuring the phenotypic effect of the parental 8 and 11 were 20.1 Ϯ 1.2ЊC (mean Ϯ 1 SE) and 27.2 environment, and little attention has been directed to- Ϯ 0.4ЊC, respectively (Roach 2003). Lacey and Herr ward traits that produce a parental environmental ef- observed that high postzygotic temperature increased fect. Traits that produce parental environmental effects total germination under ®eld conditions in spring, but should evolve in response to the direction and mag- we found no difference in total germination between nitude of the effects on offspring phenotype and to the early- and late-maturing seeds. However, we did ®nd selective pressures faced by offspring. Our data suggest that late-maturing seeds germinated more quickly. This that reproductive phenology evolves partially in re- difference is not likely explained by a difference in sponse to the parental environmental effects that it in- dormancy because seeds in both groups showed very duces. high germination overall, and they had ample time to ACKNOWLEDGMENTS complete after-ripening prior to the germination test. Also, because early and late seeds were produced by Many students, at both Duke University and the University of Virginia, assisted us in the ®eld, and we thank them all. the same maternal parents, the difference is not ex- We also thank T.-L. Ashman for her constructive comments plained by maternal genetic effects. Our experimental on a previous draft of this manuscript, the Duke Phytotron design did not allow us to test for possible paternal staff for their assistance, Mark Rausher for use of his G test effects associated with maturation time. program, and the following institutions for ®nancial support: NIH (PO1-AG08761 to D. A. Roach) and NSF (DEB-94- The biological signi®cance of a one-day difference 15541 to the Duke Phytotron). in germination is presently unclear. One could argue that it is likely to be important in highly competitive LITERATURE CITED situations, where P. lanceolata can and does thrive, Alexander, H. M. 1982. Demography of an intraspeci®c var- e.g., lawns. Rapidly germinating individuals have a iation in Plantago lanceolata in relation to infection by the competitive advantage over slowly germinating indi- fungus Fusarium moniliforme var. subglutinans. Disserta- tion. Duke University, Durham, North Carolina, USA. viduals in P. lanceolata (van der Toorn and Pons 1988) Antonovics, J., and R. B. Primack. 1982. Experimental eco- and in other species (e.g., Black 1958, Harper 1977, logical genetics in Plantago VI. The demography of seed- Schaal 1980, Roach 1986, Winn 1988, BieÁre 1991, Ga- ling transplants of P. lanceolata. Journal of Ecology 70: len and Stanton 1991, Simons and Johnston 2000). Ear- 55±75. Augspurger, C. K. 1980. Mass ¯owering of a tropical shrub lier germination may be more advantageous among (Hybanthus prunifolius): in¯uence on pollinator attraction fall-germinating seeds if it speeds establishment rate and movement. Evolution 34:475±488. and, thereby, reduces winter mortality. Further studies Bawa, K. S. 1983. Patterns of ¯owering in tropical plants. will have to be conducted to assess the effect of re- Pages 394±410 in C. E. Jones and R. J. Little, editors. A productive timing on fall vs. spring germination. In handbook of experimental pollination ecology. Van Nos- trand Reinhold, New York, New York, USA. Campanula americana (Galloway 2002) and in Daucus Bentler, P. M. 1989. EQS structural equations program man- carota (Lacey and Pace 1983), earlier ¯owering plants ual. BMDP Statistical Software, Los Angeles, California, produce seeds that are more likely to germinate in the USA. fall because earlier ¯owering plants disperse their seeds BieÁre, A. 1991. Parental effects in Lychnis ¯os-cuculi. II. Selection on time of emergence and seedling performance earlier, in time for fall germination. This situation does in the ®eld. Journal of Evolutionary Biology 3:467±486. not apply to NC piedmont populations of P. lanceolata BieÁre, A., and S. J. Honders. 1996. Impact of ¯owering phe- because even later ¯owering plants disperse their seeds nology of Silene alba and S. dioica on susceptibility to well before the fall germination season. fungal infection and seed predation. Oikos 77:467±480. Finally, many studies have documented parental en- Bishop, Y. M., S. E. Fienberg, and P. W. Holland. 1975. Dis- crete multivariate analysis: theory and practice. Massachu- vironmental effects in plants (e.g., see reviews: Roach setts Institute of Technology Press, Cambridge, Massachu- and Wulff 1987, Gutterman 1992, Wulff 1995, Byers setts, USA. and Shaw 1998, Donohue and Schmitt 1998). Studies Bishop, J. G., and D. W. Schemske. 1998. Variation in ¯ow- show that environmental factors such as parental nu- ering phenology and its consequences for lupine colonizing Mount St. Helens. Ecology 79:534±546. trient level, shading, substrate material, drought stress, Black, J. N. 1958. Competition between plants of different temperature, photoperiod, and density can affect off- initial seed sizes in swards of subterranean clover (Trifo- spring phenotype (e.g., Durrant 1962, Gutterman lium subterraneum L.) with particular reference to leaf area September 2003 MULTIGENERATIONAL EFFECTS OF TIMING 2473

and the light microclimate. Australian Journal of Agricul- Evans, E. W., C. C. Smith, and R. P. Gendron. 1989. Timing tural Research 9:299±318. of reproduction in a prairie legume: seasonal impacts of Blom, C. W. P. M. 1992. Ecology of Plantago populations: insects consuming ¯owers and seeds. Oecologia 78:220± germination and establishment. Pages 88±97 in P. J. C. 230. Kuiper and M. Bos, editors. Plantago: a multidisciplinary Galen, C., and M. L. Stanton. 1991. Consequences of phe- study. Springer-Verlag, Heidelberg, Germany. nology for reproductive success in Ranunculus adoneus Booth, C. 1971. The genus Fusarium. Commonwealth My- (Ranunculaceae). American Journal of Botany 78:978±988. cological Institute, Kew, Surrey, UK. Galloway, L. F. 1995. Response to natural environmental Brody, A. K. 1997. Effects of pollinators, herbivores, and heterogeneity: maternal effects and selection on life-history seed predators on ¯owering phenology. Ecology 78:1624± characters and plasticities in Mimulus guttatus. Evolution 1631. 49:1095±1107. Byers, D. L., and R. G. Shaw. 1998. Genetics of maternal Galloway, L. F. 2002. The effect of maternal phenology on and paternal effects. Pages 97±111 in T. A. Mousseau and offspring characters in the herbaceous plant Campunula C. Fox, editors. Maternal effects as adaptations. Oxford americana. Journal of Ecology 90:851±858. University Press, Oxford, UK. Grace, J. B., and B. H. Pugesek. 1998. On the use of path Campbell, D. R., and A. F. Motten. 1985. The mechanism of analysis and related procedures for the investigation of eco- competition for pollination between two forest herbs. Ecol- logical problems. American Naturalist 152:151±159. ogy 66:554±563. Gutterman, Y. 1980±1981. Review: in¯uence on seed ger- Case, A. L., E. P. Lacey, and R. G. Hopkins. 1996. Parental minability: phenotypic maternal effects during seed mat- effects in Plantago lanceolata L. II. Manipulation of grand- uration. Israel Journal of Botany 29:105±117. parental temperature and parental ¯owering time. Heredity Gutterman, Y. 1992. Maternal effects on seeds during de- 76:287±295. velopment. Pages 27±59 in M. Fenner, editor. Seeds: the Cavers, P. B., I. J. Bassett, and C. W. Crompton. 1980. The ecology of regeneration in plant communities. CAB Inter- biology of Canadian weeds. 47. Plantago lanceolata L. national, Wallingford, UK. Canadian Journal of Plant Science 60:1269±1282. Harper, J. L. 1977. Population biology of plants. Academic Cavers, P. B., and M. G. Steel. 1984. Patterns of change in Press, London, UK. seed weights over time on individual plants. American Nat- Husband, B. C., and D. W. Schemske. 2000. Ecological uralist 124:324±335. mechanisms of reproductive isolation between diploid and Chiariello, N., and J. D. Roughgarden. 1984. Storage allo- tetraploid Chamerion angustifolium. Journal of Ecology cation in seasonal races of an annual plant: optimal versus 88:689±701. actual allocation. Ecology 65:1290±1301. Koller, D. 1962. Preconditioning of germination in lettuce at Cohen, D. 1976. The optimal timing of reproduction. Amer- time of fruit ripening. American Journal of Botany 49:841± ican Naturalist 110:801±807. 844. de Jong, T. J., and P. G. L. Klinkhamer. 1991. Early ¯owering Lacey, E. P. 1996. Parental effects in Plantago lanceolata. in Cynoglossum of®cinale L. constraint or adaptation. Func- L. I. A growth chamber experiment to examine pre- and tional Ecology 5:750±756. post-zygotic temperature effects. Evolution 50:865±878. de Jong, T. J., P. G. L. Klinkhamer, and M. J. van Staalduinen. Lacey, E. P., and D. Herr. 2000. Parental effects in Plantago 1992. The consequences of pollination biology for selec- lanceolata. L. III. Measuring parental temperature effects tion of mass or extended blooming. Functional Ecology 6: in the ®eld. Evolution 54:1207±1217. 606±615. Lacey, E. P., and R. Pace. 1983. Effect of parental ¯owering Dempster, J. L. 1963. The population dynamics of grasshop- and dispersal times on offspring fate in Daucus carota. per and locusts. Biological Reviews 38:490±529. Oecologia 60:274±278. de Nooij, M. P., and H. J. Mook. 1992. Interactions with Li, C. C. 1975. Path analysis: a primer. Boxwood Press, Pa- organisms other than plants. Pages 52±68 in P. J. C. Kuiper and M. Bos, editors. Plantago: a multidisciplinary study. ci®c Grove, California, USA. Springer-Verlag, Heidelberg, Germany. Linders, E. G. A., H. Turin, and O. F. R. van Tongeren. 1996. de Nooij, M. P., and H. A. van der Aa. 1986. Phomopsis The role of weevils and of mowing in epidemics of Dia- subordinaria and associated stalk disease in natural pop- porthe adunca in naturally occurring populations of Plan- ulations of Plantago lanceolata. Canadian Journal of Bot- tago lanceolata. Acta Oecologica 17:195±210. any 65:2318±2325. Marquis, R. J. 1988. Phenological variation in the neotropical Diggle, P. K. 1995. Architectural effects and the interpreta- understory shrub Piper arielnum: causes and consequences. tion of patterns of fruit and seed development. Annual Re- Ecology 69:1552±1565. view of Ecology and Systematics 26:531±552. Mazer, S. J., and L. M. Wolfe. 1998. Density-mediated ma- Donohue, K. 1999. Seed dispersal as a maternally in¯uenced ternal effects on seed size in wild radish: genetic variation character: mechanistic bias of maternal effects and selec- and its evolutionary implications. Pages 323±343 in T. A . tion on maternal characters in an annual plant. American Mousseau and C. Fox, editors. Maternal effects as adap- Naturalist 154:674±689. tations. Oxford University Press, Oxford, UK. Donohue, K., and J. Schmitt. 1998. Maternal environmental Miao, S. L., F. A. Bazzaz, and R. B. Primack. 1991a. Effects effects in plants: adaptive plasticity? Pages 137±158 in T. of maternal nutrient pulse on reproduction of two coloniz- A. Mousseau and C. Fox, editors. Maternal effects as ad- ing Plantago species. Ecology 72:586±596. aptations. Oxford University Press, Oxford, UK. Miao, S. L., F. A. Bazzaz, and R. B. Primack. 1991b. Per- Dudycha, J. L., and D. A. Roach. In press. Pathogen fre- sistence of maternal nutrient effects in Plantago major: the quency in an age-structured population of Plantago lan- third generation. Ecology 72:1634±1642. ceolata. Oecologia. Mitchell, R. J. 1992. Testing evolutionary and ecological Durrant, A. 1962. The environmental induction of heritable hypotheses using path analysis and structural equation change in Linum. Heredity 17:27±61. modelling. Functional Ecology 6:123±129. English-Loeb, G. M., and R. Karban. 1992. Consequences Ollerton, J., and A. Diaz. 1999. Evidence for stabilising se- of variation in ¯owering phenology for seed head herbivory lection acting on ¯owering time in Arum maculatum (Ar- and reproductive success in Erigeron glaucus (Composi- aceae): the in¯uence of phylogeny on adaptation. Oecol- tae). Oecologia 89:588±595. ogia 119:340±348. 2474 ELIZABETH P. LACEY ET AL. Ecology, Vol. 84, No. 9

Ollerton, J., and A. Lack. 1992. Flowering phenology: an Schemske, D. W., M. F. Willson, M. N. Melampy, L. J. Miller, example of relaxation of natural selection. Trends in Ecol- L. Verner, K. M. Schemske, and L. B. Best. 1978. Flow- ogy and Evolution 7:274±276. ering ecology of some spring woodland herbs. Ecology 59: O'Neil, P. 1999. Selection on ¯owering time: an adaptive 351±366. ®tness surface for nonexistent character combinations. Schmid, B., and C. Dolt. 1994. Effects of maternal and pa- Ecology 80:806±820. ternal environment and genotype on offspring phenotype Paltridge, G. W., and J. V. Denholm. 1974. Plant yield and in Solidago altissima L. Evolution 48:1525±1549. the switch from vegetative to reproductive growth. Journal Schmitt, J. 1983. Density-dependent pollinator foraging, of Theoretical Biology 44:23±34. ¯owering phenology, and temporal pollen dispersal pat- Petraitis, P. S., A. E. Dunham, and P. H. Niewiarowski. 1996. terns in Linanthus bicolor. Evolution 37:1247±1257. Inferring multiple causality: the limitations of path anal- Schmitt, J., J. Niles, and R. D. Wulff. 1992. Norms of reaction ysis. Functional Ecology 10:421±431. of seed traits to maternal environments in Plantago lan- PicoÂ, F. X., and J. Retana. 2000. Temporal variation in the ceolata. American Naturalist 139:451±466. female components of reproductive success over the ex- Siddique, M. A., and P. B. Goodwin. 1980. Seed vigour in tended ¯owering season of a Mediterranean perennial herb. bean (Phaseolus vulgaris L. cv. Appolo) as in¯uenced by Oikos 89:485±429. temperature and water regime during development and mat- Pilson, D. 2000. Herbivory and natural selection on ¯owering uration. Journal of Experimental Botany 31:313±323. phenology in wild sun¯ower, Helianthus annuus. Oecologia Simons, A. M., and M. O. Johnston. 2000. Variation in seed 122:72±82. traits of Lobelia in¯ata (Campanulaceae): sources and ®t- Platenkamp, G. A. J., and R. G. Shaw. 1993. Environmental ness consequences. American Journal of Botany 87:124± and genetic maternal effects on seed characters in Ne- 132. mophila menziesii. Evolution 47:540±555. Sokal, R. R., and F. J. Rohlf. 1995. Biometry. Third edition. Primack, R. B. 1987. Relationships among ¯owers, fruits, W. H. Freeman, New York, New York, USA. and seeds. Annual Review of Ecology and Systematics 18: Stanton, M. L. 1984. Seed variation in wild radish: effect of 409±430. seed size on components of seedling and adult ®tness. Ecol- Primack, R. B., and J. Antonovics. 1981. Experimental eco- ogy 65:1105±1112. logical genetics in Plantago V. components of seed yield Steinbauer, G. P., and B. Grigsby. 1975. Dormancy and ger- in the ribwort plantain Plantago lanceolata L. Evolution mination characteristics of the seeds of four species of 35:1069±1079. Plantago. Proceedings of the Association of Of®cial Seed Rathcke, B., and E. P. Lacey. 1985. Phenological patterns of Analysis 47:158±164. terrestrial plants. Annual Review of Ecology and System- Sultan, S. E. 1996. Phenotypic plasticity for offspring traits atics 16:179±214. in Polygonum perisicaria. Ecology 77:1791±1807. Riddell, J. A., and G. A. Gries. 1958. Development of spring Teramura, A. H. 1978. Localized ecotypic differentiation in wheat. III. Temperature maturation and age of seeds as three contrasting populations of Plantago lanceolata L. factors in¯uencing their response to vernalization. Agron- Dissertation. Duke University, Durham, North Carolina, omy Journal 50:743±46. USA. Roach, D. A. 1986. Variation in seed and seedling size in Teramura, A. H., J. Antonovics, and B. R. Strain. 1981. Ex- Anthoxanthum odoratum. American Midland Naturalist perimental ecological genetics in Plantago. IV. Effects of temperature on growth rates and reproduction in three pop- 117:258±264. ulations of Plantago lanceolata L. (Plantaginaceae). Amer- Roach, D. A. 2003. Age-speci®c demography in Plantago: ican Journal of Botany 68:425±434. variation among cohorts in a natural plant population. Ecol- Thompson, J. N., and O. Pellmyr. 1989. Origins of variance ogy 84:749±756. in seed number and mass: interaction of sex expression and Roach, D. A., and R. D. Wulff. 1987. Maternal effects in herbivory in Lomatium salmoni¯orum. Oecologia 79:395± plants: evidence and ecological and evolutionary signi®- 402. cance. Annual Review of Ecology and Systematics 18:209± van der Toorn, J., and T. I. Pons. 1988. Establishment of 235. Plantago lanceolata L. and Plantago major L. among grass. Robertson, R. N., H. R. Highkin, J. Smydzuk, and F.W. Went. 2. Shade tolerance of seedlings and selection on time of 1962. The effects of environmental conditions on the de- germination. Oecologia 76:341±347. velopment of pea seeds. Australian Journal of Biological van der Toorn, J., and P. H. van Tienderen. 1992. Ecotypic Science 15:1±15. differentiation in Plantago lanceolata. Pages 269±288 in SAS Institute. 2000. SAS 8.1: users' guide. SAS Institute, P. J. C. Kuiper and M. Bos, editors. Plantago: a multidis- Cary, North Carolina, USA. ciplinary study. Springer-Verlag, Heidelberg, Germany. Sawhney, R., and J. M. Naylor. 1979. Dormancy studies in van Tienderen, P. H., and J. van der Toorn. 1991. Genetic seed of Avena fatua. 9. Demonstration of genetic variability differentiation between populations of Plantago lanceolata. affecting the response to temperature during seed devel- I. Local adaptation in three contrasting habitats. Journal of opment. Canadian Journal of Botany 57:59±63. Ecology 79:27±42. Schaal, B. A. 1980. Reproductive capacity and seed size in Vincent, T. L., and H. R. Pulliam. 1980. Evolution of life Lupinus texensis. American Journal of Botany 67:703±709. history strategies for an asexual annual plant model. The- Schaal, B. A. 1984. Life-history variation, natural selection, oretical Population Biology 17:215±231. and maternal effects in plant populations. Pages 188±206 Winn, A. A. 1988. Ecological and evolutionary consequences in R. Dirzo and J. SarukhaÂn, editors. Perspectives on plant of seed size in Prunella vulgaris. Ecology 69:1537±1544. population ecology. Sinauer, Sunderland, Massachusetts, Winn, A. A. 1991. Proximate and ultimate sources of within- USA. individual variation in seed mass in Prunella vulgaris Scheiner, S. M., J. Mitchell, and H. S. Callahan. 2000. Using (Lamiaceae). American Journal of Botany 78:838±844. path analysis to measure natural selection. Journal of Evo- Wolfe, L. M. 1992. Why does the size of reproductive struc- lutionary Biology 13:423±433. tures decline through time in Hydrophyllum appendicula- Schemske, D. W. 1984. Population structure and local se- tum (Hydrophyllaceae)? Developmental constraints vs. re- lection in Impatiens pallida (Balsaminaceae), a sel®ng an- source limitation. American Journal of Botany 79:1286± nual. Evolution 38:817±832. 1290. September 2003 MULTIGENERATIONAL EFFECTS OF TIMING 2475

Wolfe, L. M., and J. L. Burns. 2001. A rare continual ¯ow- lanceolata L. II. Population characteristics. Heredity 58: ering strategy and its in¯uence on offspring quality in a 183±192. gynodioecious plant. American Journal of Botany 88: Wright, S. 1934. The method of path coef®cients. Annals of 1419±1423. Mathematics and Statistics 5:161±215. Wolff, K. 1987. Genetic analysis of ecologically relevant Wulff, R. D. 1986. Seed size variation in Desmodium pani- culatum. 2. Effects on seedling growth and physiological morphological variability in Plantago lanceolata L. II. Lo- performance. Journal of Ecology 74:99±114. calization and organization of quantitative trait loci. The- Wulff, R. D. 1995. Environmental maternal effects on seed oretical Applied Genetics 73:903±914. quality and germination. Pages 491±505 in J. Kigel and G. Wolff, K., and W. van Delden. 1987. Genetic analysis of Galili, editors. Seed development and germination. Marcel ecologically relevant morphological variability in Plantago Dekker, New York, New York, USA.