Keywords: Functional traits, growth strategy, iteroparous, life history, maximum height, reproductive allocation

Diversity of RA schedules

The partitioning of energy between reproduction and other activitiesthroughout a plant's lifetime – such as growth, storage, and defense – is arguablythe most fundamental component of its life history (Harper and Ogden1970; Bazzaz et al.2000). Here we refer to the fractionof surplus energy that is allocated to reproduction in a given period as reproductiveallocation (RA), where surplus energy is that which remains after the costs ofrespiration and tissue turnover have been paid. As RA is expressed as a proportion ofenergy, it falls between 0 and 1. The change inRA with respect tosize or age will be termed anRA schedule. We use surplus energyinstead of net primary productivity as the energy pool to be subdivided, because formost perennial species, reproductive investment does not appear to come at theexpense of existing tissues. This assumption is evident in the allometry of mosttrees, in which all size dimensions tend to increase over time. Use of “surplusenergy” also aligns our study with many theoretical models, which invest inreproduction only after paying maintenance costs (e.g., early review by Kozlowski1992) and plant growth models(e.g., papers by Thornley1972;de Wit1978; Mäkelä1997). RA schedules then enact theoutcome of a single fundamental trade‐off: the allocation of surplus energy betweengrowth and reproduction. As such, they summarize essential elements of a plant's lifehistory strategy: At what age do plants begin reproducing, what proportion of energygoes to reproduction, and how do plants moderate the proportion of energy theyallocate to reproduction as they age? The follow‐on information is equally important,for energy not allocated to reproduction is used for growth, increasing the plantsheight and thereby its ability to outcompete neighbors for light (or otherresources), hence increasing survival. From the perspective of other organisms, theRA schedule determines how gross primary productivity is allocated amongfundamentally different tissue types, that is, leaves, woody tissues, flowers,fruits, and seeds, the eventual food stuffs at the base of terrestrial food webs.

The diversity of life history strategies observed across extant plantspecies suggests many different RA schedules might be expected (Fig. 1). The two most extreme RAschedules include a slow increase in RA across a plant's lifetime (a graded RAschedule) and an RA schedule where maximum RA is reached and vegetative growth ceasesas soon as reproduction commences (a big bang schedule, indicating a switch fromRA = 0 to RA≈1 across a single growing season) (Fig. 1). Big bang reproducers are also termed semelparous ormonocarpic, a group that includes some annuals, several succulent shrubs, and atleast a hundred trees (Young2010; Thomas2011)(Fig. 1, panel B). It ispossible for a big bang species to cease growth and continue reproducing for severalyears, but most species die following a single large reproductive event (Young2010). A graded RA schedule, alsotermed iteroparous or polycarpic, can be further divided into RA schedules we termpartial bang, asymptotic, gradual, and declining, depending on how RA changes withsize (Fig. 1C–G). Gradedstrategies are diverse, including RA schedules displaying early reproductive onsetand high reproductive investment at the expense of growth and survival, as well asones with a long period devoted entirely to growth followed by more modestreproductive output. Figure 2highlights, using a simple plant growth model from Falster et al.2011, how differences in RA schedulealone can drive differences in growth, seed production, and biomass allocation.

Theoretical treatments of RA schedules

Theorists long ago adopted RA schedules as an elegant way to connectenergy allocation with life history (e.g., Cole1954; Myers and Doyle1983; Kozłowski and Uchmanski1987; Kozlowski1992; Engen and Saether1994; Miller et al.2008). By incorporating the growth‐reproduction trade‐off, optimal energyallocation models identify the RA schedule that maximizes seed production across theplant's lifecycle under a given set of environmental conditions and for a given setof physiological traits (Kozlowski1992). For instance, researchers have developed models that indicate howRA schedules vary with shifts in a variety of biotic and abiotic factors includingtissue turnover (Pugliese and Kozlowski1990), seed set (Miller et al.2008), age‐specific mortality (Charnov and Schaffer1973; Reznick and Endler1982; Engen and Saether1994), and environmentalstochasticity (King and Roughgarden1982; Gurney and Middleton1996; Katsukawa et al.2002).

In a simple linear system, big bang is always optimal

The history of using optimal energy allocation to model RA schedulestraces back to a seminal paper by Cole (1954). In his model, and subsequent similar ones, surplus energy can onlygo two places: to reproductive investment or vegetative production increasing thesize of the plant. Moreover, there is a linear rate of energy conversion into thesestructures, so the trade‐offs between growth and reproduction are also linear.Optimal energy models that include only this direct linear trade‐off find that thecomplete cessation of growth with reproductive onset, a single reproductive episode,and subsequent death (i.e., the big bang strategy from Fig. 1, where RA switches from 0 to 1) isalways optimal, because delayed reproduction when small and correspondingly greatergrowth leads to greater final reproductive output (Cole1954; Kozlowski1992; Perrin and Sibly1993; Engen and Saether1994). In these models, individuals with an iteroparous reproductivestrategy (i.e., with an earlier start to reproduction, an RA <1, and multiplereproductive episodes) have a lower lifetime reproductive output than big bangreproducers. This is because with the iteroparous reproductive strategy, the onset ofreproduction leads to decreased growth rates and a smaller adult size, resulting inlower lifetime surplus energy. The models predict that the size (or age) atreproduction of big bang reproducers shifts with factors such as growth rate, howincreased size translates to increased reproductive output, and the probability ofsurvival (Kozłowski and Wiegert1987; Perrin and Sibly1993); changing these parameters never causes the optimal RA schedule toshift away from big bang to a graded schedule. Yet the list of perennial semelparousplant species displaying a big bang strategy is relatively short, encompassingapproximately 100 trees and some palms, yuccas, and giant rosette plants from alpineAfrica (e.g., see Thomas2011).This disconnect between theoretical prediction and observation has come to be knownas Cole's Paradox (Charnov and Schaffer1973) and has led researchers to search for mechanisms favoring a gradedreproduction schedule.

Nonlinear trade‐offs or environmental stochasticity promote graded allocationstrategies

Cole's paradox has largely been resolved, as it is now known that avariety of other factors can shift the optimal energy allocation from “big bang” to a“graded” schedule. Specifically, models need to include either: (i) stochasticenvironmental conditions (King and Roughgarden1982) or (ii) secondary functions influencing howefficiently energy allocated to different goals (growth, reproduction) is convertedinto different outcomes (increased vegetative size, seed production). It seems thatif these conversion functions are nonlinear with respect to plant size, a gradedallocation may be favored.

In one class of nonlinear trade‐offs, an auxiliary factor causes thecost of increased reproductive or vegetative investment to increase more (or less)steeply than is predicted from a linear relationship. As a first example, consider afunction that describes how efficiently resources allocated to reproduction areconverted into seeds. Studying cactus, Miller et al. (2008) showed that floral abortion rates due to insect attackincreased linearly with RA. In other words, as RA increases, the cost of creating aseed increases, such that the cacti are selected to have lower RA and earlierreproduction than would be expected from direct costs of reproduction alone. A secondexample, Iwasa and Cohen's model (1989) showed that declining photosynthetic rates with size, a trenddetected in several empirical studies (Niinemets2002; Thomas2010), led to a graded RA schedule. Third, many models, often backed upwith data from fish or marine invertebrates, have shown that if mortality decreaseswith age or size, it benefits an individual to grow for longer and then beginreproducing at a low level – a graded RA schedule (Murphy1968; Charnov and Schaffer1973; Reznick and Endler1982; Kozłowski and Uchmanski1987; Engen and Saether1994). Overall, optimal energymodels show that a great diversity of graded RA schedules is possible, and that assuggested, both fundamental life history traits (mortality, fecundity) and functionaltrait values (photosynthetic rate, leaf life span, growth rates) could affect theshape of the RA schedule.

Need for empirical data

While the outcomes of the many optimal energy models show that RAschedules shift depending on a plant's collection of life history and physiologicaltraits, there is little empirical data to test the outcomes of these models.Widespread collection of empirical data has been limited due to the effort requiredto accurately determine the many sinks for surplus energy, including growth, storage,defense, and reproduction. In particular, very few data on lifetime reproductiveallocation exist for long‐lived species, due to the impracticalities of assessingreproductive output across an individual tree's lifetime.

In this study, our first aim is to review the available empirical RAschedules in nonclonal, woody plants with bisexual flowers. We present a summary ofempirical data for the handful of studies quantifying complete RA schedules, as wellas some data sets that include only particular features of an RA schedule, such asthe shape of the curve. Despite several reviews about elements of plant reproduction(Bazzaz et al.2000; Obeso2002; Moles et al.2004; Weiner et al.2009; Thomas2011), none have explicitly focusedon RA schedules or the integration between empirical data and the outcome oftheoretical models. This review focuses on perennial species, for recent work hasestablished a framework for investigating reproductive output (RO) in annuals (Weineret al.2009). Studyingreproductive investment in perennial species is more challenging, but very relevant,as these species are the dominant contributors to woody plant biomass worldwide. Wepredict that species will display a diversity of RA schedules and that shorter livedspecies will have relatively high RA and reach their maximum RA more quickly than dolonger‐lived species. Second, we summarize studies that compared RA or RA schedulesacross individuals, populations, or species growing under different disturbanceregimes or with different resource availabilities, and hence give insight on whatenvironmental, life history, or functional traits might alter either RA at a givenage or size or the entire RA schedule. We expect 1) that individuals in poor resourceenvironments will postpone reproduction and have lower annual RA and 2) thatindividuals in disturbance‐prone environments will begin reproducing at younger agesand have higher annual RA. In the discussion, we compare the information gleaned fromour compilation of RA schedules with that provided by measures of RO and the researchquestions each method best address.

Defining and quantifying reproductive allocation schedules

A conceptual outline of the energy budget for a plant illustrates howRA is calculated (Fig. 3). Tocalculate the amount of energy allocated to growth, it is necessary to distinguishbetween growth that replaces lost tissues and growth that increases the size of theplant. Beginning at Figure 3A,consider that a plant of a given size and with a given collection of functionaltraits has a given gross primary production (GPP) and respiration costs. Subtractingrespiration from GPP yields net primary production (NPP). Some of this NPP will beused to replace lost or shed tissue (Fig. 3C), with the remainder designated as “surplus energy”(Fig. 3D). (Energy can also beallocated to storage or defense, but for simplicity these are not included. Ifsurplus energy is allocated to storage – and hence unmeasured – surplus energy willbe underestimated and RA will be an overestimate.) Note that total growth on theplant in a given year is not one of the boxes, because it represents a combination ofenergy used to replace lost tissues, that is, the portion of NPP a plant used tomaintain current size, and the portion of surplus energy allocated to growing to abigger size during the survey period.

To properly quantify an RA schedule, one must measure all the energyallocated to growth and reproduction over time. In principle, an RA schedule concernsthe instantaneous fraction of surplus energy allocated among growth and reproduction.In reality, RA should be measured over longer time periods, because growth andreproduction often occur at different points during the growing season. The energybudget is therefore typically tabulated on a per year basis. Some species haveinconsistent year‐to‐year reproductive output, termed masting. For these species, theenergy budget must be tabulated across a mast year and the number of nonmast yearsthat follow. The weight of dry biomass is the most commonly used proxy for “energy,”but the kilojoules energy contained in the biomass or the mass of a specific limitingelement are valid alternatives. It is important that the same energy units be usedfor both reproductive and vegetative material.

Reproductive investment should be measured over an entire reproductivecycle and include energy invested both in seed and accessory tissues, the lattertermed accessory costs. Accessory costs include the construction of prepollination(flower, nectar, and pollen) and postpollination (packaging, protective and dispersaltissues; aborted ovules) floral parts. Total accessory costs are highly variable andcan be as much as 99% or as little as 15% of reproductive energy investment(Table 1).

To calculate the investment in growth, one must determine how muchbigger the plant is, relative to a year earlier. Unless you are able to follow asingle plant through its life, you must find individuals of different sizes,preferably of known age, on which to measure RA. These individuals should be growingunder similar environmental conditions and in a similar community of species. Oneapproach to estimating a complete RA schedule for long‐lived species is to pick aknown chronosequence, as is available with plantation trees and in locations with aknown disturbance (and germination) history (Zammit and Zedler1993; Cleary et al.2008; Genet et al.2010). Combining RA measurementsfrom plants across a range of sizes yields an RA schedule; a curve showing how anindividual's relative investment in reproduction shifts with plant size or age(Fig. 1). We have focused onsize‐related patterns, as size has been shown to have a greater influence on RA thanage (Herrera1991; Pino et al.2002). In particular, size isthe primary factor determining the onset of reproduction in competitive environments(Pino et al.2002).

Literature

Here we review what can be learned about RA data from existing studieson 34 populations, representing 32 species. These are the only studies we found inthe literature that include data either on how RA changes with size (or age) or thatcompare RA across populations or closely related species. We searched widely in theliterature using both Web of Science and Google Scholar for studies that had measuredreproductive investment at multiple ages, across different resource environments orunder different disturbance regimes. Some studies used a known chronosequence, somefollowed the same individuals (or population) across multiple years, and yet othersused co‐occurring individuals of different sizes to construct a RA schedule.Additional studies report measures of RO, proxies for RA, such as flowering intensity(e.g., Herrera and Jovani2010)or number of reproductive modules (e.g., Miller et al.2008), but not actual biomass or energy allocation toreproduction. Ideally, RA values were available for individuals at multiple sizes (orages), such that a RA schedule could be plotted. Knowing RA at reproductive onset and2–3 later time points is sufficient to predict the shape of the RA schedule, but ofcourse more data points increased the precision with which the RA schedule could bedrawn. We included studies from which the shape of the RA schedule can be estimated,even if absolute RA values cannot be calculated. The categorization of RA scheduletypes (Fig. 1) is based on avisual assessment, as data are insufficient for a statistical classification. Studiessolely reporting plots of reproductive biomass against plant size have not beenincluded as they have been thoroughly reviewed recently (Weiner et al.2009; Thomas2011) and do not provide any meansof determining whether a plant with a large reproductive capacity has a high rate ofmass production or large allocation to reproduction. Most of the studies includedhave not themselves explicitly plotted RA schedules, but instead provide data thatcan be used to quantify RA schedules (see[Link] for details). The studies comparing RA in populationsor species subjected to different resource conditions or disturbance regimes do nothave data on different sized individuals; instead, these data indicate how thesevariables might shift certain parts of an RA schedule.

Based on published information, RA was calculated as the proportion oftotal surplus energy, on a per time basis, allocated to reproduction. One year (orone growing season) is the commonly used time interval. Energy units used are pergram dry mass or kilojoules (determined by burning the samples). Total surplus energyis calculated as the sum of RO, “growth beyond replacement,” as defined inFigure 3, energy storedunderground, and energy allocated to defense. RO is the sum total of all types ofreproductive investment: flowers, nectar, aborted fruit, mature fruit, and vegetativestructures associated only with flowering. It is noted in Table 1 when studies report totalnew growth, not growth beyond replacement; using total new growth instead of “growthbeyond replacement” overestimates surplus energy and underestimates RA. Very fewstudies consider energy stored underground and energy allocated to defense. Whenavailable, these are summed with growth, otherwise this pool is ignored (set tozero). If growth beyond replacement is not directly reported, it is estimated fromdata on increase in stem diameter and increase in leaf area. RA is then calculatedand plotted against plant size (or age) to determine the shape of the RA schedule.Unfortunately, most studies report data for only some reproductive components,usually ignoring shed accessory tissues. The missing reproductive costs are thus notincluded in our analysis, which will cause RA to be underestimated.

Individual components of an RA schedule are presented in Table 2 and discussed below. Theyinclude the shape of the RA schedule, RA at maturation, maximum RA, and size atmaturation. For the following studies, the numbers presented in Table 2 were taken directly fromthe published articles: Pitelka1977; Pritts and Hancock1983; Oyama1990;Alvarez‐Buylla and Martinez‐Ramos1992; Comps et al.1994; Ehlers and Olesen2004; Poorter et al.2005; Read et al.2006,2008; Miller et al.2008. For the remainingstudies, we calculated RA schedules using published data (see[Link] for details).

Reproducibility

All analyses were conducted with R software (R Core Team2014). The code and data forproducing all figures in this study is available athttps://github.com/dfalster/Wenk_RA_review.

Lifetime reproductive allocation schedule

The species sampled exhibit an enormous variety of reproductivestrategies, from truly big bang species (Fig. 1B, Table 2) to a great diversity of graded reproductionschedules (Fig. 1C–G, Table 2). We included only twospecies with big bang RA schedules; all others exhibit one of the graded RAschedules. Three species, including most perennial herbaceous species studied, rampup to their maximum RA within a few years of reproductive onset (Pitelka1977; Ehlers and Olesen2004) and are classified as “partialbang” (Fig. 1B). Eight speciesshow a more gradual increase in RA, but still reach a definite plateau, the“asymptotic” type in Fig. 1D(Piñero et al.1982; Oyama1990; Alvarez‐Buylla andMartinez‐Ramos1992; Genet et al.2010). Five of the longestlived species, including both evergreen and deciduous temperate trees, continue toincrease RA throughout their lives, never reaching an obvious asymptote (Comps et al.1994; Hirayama et al.2004,2008), and are therefore labeled “gradual‐indeterminate”(Fig. 1E). No species had anRA schedule we visually categorized as “gradual‐determinate” (Fig. 1F). This collection of RA schedulesmatched our expectations that some species displayed few years of relatively high RAand others many years of mostly lower RA. Faster growth allowed a monocarpic speciesTachigali vasquezii to reach a large size and reproductivematurity more quickly than co‐occurring iteroparous species; that is, faster growthallowed the onset of reproduction to be advanced (Poorter et al.2005).

In most of the studies considered, the maximum RA achieved ismaintained until the end of life, in agreement with evolutionary theory predictingincreasing or stable RA until death (Roff2002; Thomas2011). However, there are three species,Vacciniumcorymbosum (Pritts and Hancock1985),Abies veitchii (Kohyama1982), and high elevationpopulations ofAbies mariesii (Sakai et al.2003), where RA decreases late inlife and thus exhibit a “declining” RA schedule (Fig. 1G, Table 2).

Reproductive allocation at maturation

Threshold reproductive allocation was reported for 15 species andpopulations. Long‐lived iteroparous species usually initially have very low RAvalues, such as 0.05 forRhopalostylis sapida (Nikau Palm) (Enright1985) and 0.08 for beech(Genet et al.2010) (Table 2). By contrast, shorterlived species can have quite high RA values the year they commence reproduction, suchas 0.25 forVaccinium corymbosum (Pritts and Hancock1985) and 0.18 forLupinusvariicolor (Pitelka1977) (Table 2). Two semelparous perennial species, ones with a big bang schedule wherethey instantaneously reach RA = 1, are included in Table 2. Several hundred additionalspecies are known to have this life history (Young1984,2010; Klinkhamer et al.1997; Thomas2011).

Maximum reproductive allocation

Thirteen of the studies reported maximum RA. For semelparous species,such asTachigali vasquezii andCerberiopsiscandelabra, it is always close to 1 (Poorter et al.2005; Read et al.2006). Iteroparous species usuallyhave a maximum RA between 0.4 and 0.7 (Table 2), although values as low as 0.1 have been recordedin an alpine community (Hemborg and Karlsson1998). Long‐lived iteroparous species are expected to havelower maximum RA than shorter lived species, as they are diverting more resources tosurvival, both in the form of more decay and herbivore resistant leaves and stems andother defense measures. These species compensate for a lower RA by having moreseasons of reproductive output. However, no clear trend in longevity versus maximumRA is noted among the studies in Table 2, with the highest RA, 0.70, recorded in a temperatepalm that lives for more than 250 years.

Shifts in reproductive allocation with disturbance frequency or resourceavailability

Comparisons across species or populations that are subject to differentenvironmental conditions have identified certain RA schedule components thatrecurrently co‐vary, suggesting convergent adaptation. In each case, the twopopulations (or species) grow either in locations that differ in resourceavailability or in disturbance frequency (effecting mortality), with resultant shiftsin RA schedule components. Species or populations with smaller threshold size orearlier maturation, generally have higher RA, supporting traditional life historytheory that weedy species have higher fecundity (Stearns1992; Table 3). Higher mortality is alsocorrelated with this fast‐growth strategy, suggesting that the aforementioned traitscompensate for having fewer years to reproduce. Lower resource availability isrecurrently correlated with lower RA and delayed maturation. Of these studies, onlySakai et al. (2003) havesufficient data to plot complete RA schedules (see Table 3), with the other studiesonly providing data on portions of the RA schedules such as size at reproductiveonset, initial RA, or maximum RA.

Alternative measures of reproductive function

Much research has focused on components of reproductive function,including measures of reproductive output (RO; Henery and Westoby2001; Niklas and Enquist2003; Weiner et al.2009), relationships betweenreproductive output versus vegetative mass (RV curves; Weiner et al.2009), a species' maximum height(Wright et al.2010; Cornwellet al.2014), and relative sizeat onset of maturity (RSOM; Wright et al.2005; Falster and Westoby2005; Thomas2011). We now consider the value of these metrics, versus RA, inquantifying reproductive patterns and their relative benefits for addressingdifferent research questions.

Reproductive output is the measure of seed production per unit time(either in numbers or units mass). To first order, plants increase reproductiveoutput by growing larger as the productive capacity of a plant increases along withits total leaf area (Müller et al.2000; Niklas and Enquist2003; Weiner et al.2009; Fig. 4). Therelationship between plant size and RO can be examined by constructing a log–logregression of cumulative lifetime RO against vegetative size – an “RV curve” (Samsonand Werk1986; Klinkhamer et al.1992; Bonser and Aarssen2009; Weiner et al.2009). An RV curve allows one toestimate the lifetime RO of an individual of a given size, an important metric for adiversity of plant population biology, agricultural, and conservation biologyresearch questions. In contrast, an RA schedule only informs us of the amount ofenergy invested in reproduction, and thus, how many offspring are produced, if growthrates are also known, leading to criticism that using allocation ratios to measurechanges in reproductive output across a plant's lifetime is limiting (Jasienski andBazzaz1999; Müller et al.2000; Weiner2004). If the RV curve is known fora species, the size of all individuals in a population can rapidly be estimated andthe total RO calculated. A RV curve is equally applicable for high and low resourceenvironments and different population densities, because differences in plant sizelead to corresponding shifts in RO.

For other research questions however, RA schedules add information:they frame reproductive investment as a trade‐off to growth and separate the effectsof large plant size and large reproductive investment on RO. RA schedules embody howincreased allocation to reproduction impacts growth in a given year (or growingseason) and therefore affects both the competitive interactions between species in acommunity and individual survival. One species could grow fast and have early RO,while another could have slower growth and delayed RO; both could have similar RVcurves, but very different life spans, for the species diverting resources toreproduction at a smaller size is likely to be outcompeted for light (or water ornutrients) by co‐occurring species and be shorter lived.

RA schedules are also important for dissecting the contribution ofyearly growth versus preexisting size to RO; RV curves and plots of the ratio of ROto plant biomass versus plant size provide no data on how much a plant grows in agiven year, just how large it is. Consider Figure 4 that presents data on annual RO in relation to size for 47coexisting plant species. It shows that for most species, RO increases with size, butthat species differ by at least two orders of magnitude in the amount of productionat any given size. Do such differences reflect different levels of photosyntheticproductivity? Or do they indicate different levels of allocation to seed production?If one knew both the plant's RA schedule and its growth rates, one could separate theeffects of RA and productive capacity on RO. Two plants of a given size could haveidentical RO, but one would have higher productive capacity and a lower RA and asecond plant could have the reverse. As plants age their pool of surplus energy maybegin to plateau or even decrease, both through declining photosynthetic capacity(Niinemets2002; Thomas2010) and increasing tissuereplacement costs. Plots of RO against plant size indicate RE approaches anasymptote. Yet from the limited empirical data (Table 2) and optimal energy theory we know that RA may notbe constant as a plant increases in size. Indeed, unlike RE, RA often continues toincrease across an individual's life and the rate of increase in RA with size varieswith life history.

Maximum height and RSOM, the ratio of threshold size (size atreproductive onset) to maximum size, are two other metrics used to assess thetrade‐off between growth and reproduction. Like RA, they are based on the assertionthat allocation to reproduction impacts growth (Thomas1996; Davies and Ashton1999). RSOM is used to summarize the trade‐off betweencontinued faster growth rates and greater maximum height versus earlier reproduction,curtailed growth, and lower maximum height (Thomas2011). The premise for using maximum height is that aspecies with a greater maximum height has delayed diverting energy to reproductionfor longer and hence maintained a greater growth rate for longer during development(Turner2001; Westoby et al.2002). The tallest species ina community are predicted to be the long‐lived, later reproducing species thatallocate less of their yearly energy to reproduction. Greater maximum height wascorrelated with higher potential growth rate in adults in tropical forests (Wrightet al.2010), but this study doesnot include any data on reproductive output. The advantage of using maximum height asa proxy for reproductive allocation is that it is easy to measure: Data now exist forover 20,000 species (Cornwell et al.2014). The main problem with maximum height is that it quantifies theoutcome of both demographic luck and a whole host of individual trade‐offs, not justthe RA trade‐off. Moreover, the nature of all these trade‐offs may shift with ageand/or across its geographic range. As is shown in Figure 2, different RA schedules can yieldthe same final maximum height, but with different growth rates along the way, leadingto different competitive interactions. Thus, both RSOM and maximum height might bemore usefully seen as outcomes of an RA schedule rather than predictors of it.

While the above‐mentioned measures of reproductive function may beeasier to quantify across large numbers of species, they cannot substitute for acomplete RA schedule. In particular, none of those measures directly captures theseasonal or yearly decision faced by the plant of where to allocate surplus energy,making them difficult to incorporate into process‐based models of vegetation dynamics(e.g., Fisher et al.2010;Falster et al.2011; Scheiteret al.2013). Neither RV curvesnor current season RO can be incorporated into such models, because both only capturethe output of energy allocation, rather than the process itself. In contrast, an RAschedule has a direct process‐based definition: it specifies the proportion of energyallocated to reproduction as a fraction of the total energy available, at each sizeor age.

Considerations when measuring reproductive allocation schedules

Overall, we advocate for greater measurement of RA schedules. Given RAschedules have been called the measure of greatest interest for life historycomparisons (Harper and Ogden1970; Bazzaz et al.2000), we are surprised by just how little data exist. As described above,we are aware of the variety of challenges that exist to accurately collect this data,including accounting for shed tissue, all reproductive costs, and the yearly increasein size across multiple sizes and/or ages. In addition to these methodologicaldifficulties, we will briefly introduce some other intricacies.

There has been debate as to the appropriate currency for measuringenergy allocation. Almost all studies use dry weight or calorie content (joules) astheir currency. Ashman (1994),whose study had one of the most complete point measures of RA, showed that carboncontent is an inferior predictor of underlying trade‐offs compared to nitrogen andphosphorus content, although the general patterns of allocation did not shift withcurrency. Other studies have found all currencies equally good (Reekie and Bazzaz1987; Hemborg and Karlsson1998), supporting the theorythat a plant is simultaneously limited by many resources (Chapin et al.1987).

A complicating factor in determining RA schedules (or any plot showingyearly reproductive investment), is that many species do not have consistentyear‐to‐year reproductive output (Kelly and Sork2002; Smith and Samach2013). Indeed, many species, including ones represented in 3of the studies included in Table 2, mast, indicating they have years with far‐above average reproductiveinvestment, following by one or more years with near‐zero reproduction. For thesespecies, reproductive investment must be the average of a mast year and the relativenumber of nonmast years observed in that species.

A topic we have not seen discussed in the RA allocation literature ishow to account for the transition of sapwood to heartwood. If functionally deadheartwood were considered part of the shed tissue pool, far more of a plant's annualenergy production would be spent replacing this lost tissue, decreasing surplusenergy and greatly increasing estimates of apparent RA for all plants, especially asthey approach the end of life. It may even result in more iteroparous speciesactually approaching RA = 1 in old age, as is predicted in many models.

A recent model, however, suggests that reproductive restraint can bebeneficial late in life, if it allows an individual to survive for an additionalseason and have even a few additional offspring (McNamara et al.2009). An alternative hypothesis putforward is that species that can be long‐lived may none‐the‐less benefit from high RAearly in life, because the patch environment will be most favorable to the species'recruitment closer to the time the individual itself germinated (Kohyama1982; Nakashizuka et al.1997; Ehlers and Olesen2004). Under this scenario, thespecies may quickly reach a high RA and later as the patch environment degradesdisplay reproductive restraint if there is a small probability individuals cansurvive until the patch environment is again ideal for recruitment. This argumentmost obviously applies to understory species increasingly shaded by a canopy (Prittsand Hancock1985; Ehlers andOlesen2004), but was alsoproposed by Kohyama (1982) toexplain decreasing RA with stand age in a canopy tree. Alternatively, these patternsmay result from incomplete measurements, such as underestimating tissue turnoverrates (Fig. 3). At this point,there is just too little data to draw many general conclusions, or assess whethermethods of data collection are influencing our results.

Utility of reproductive allocation schedules and future directions

Over 40 years ago, Harper and Ogden (1970) recognized the intrinsic value for RA in understandingplant function, stating that “Ideally a measure of reproductive effort would involvethe determination of starting capital, gross production, and that fraction which isoutput in the form of propagules.” Energy invested in reproduction reduces the poolof energy available for plant growth – either growth in height, maintaining access tolight or growth in leaf area, and hence photosynthetic gain. As such, we and othershave argued that RA schedules elegantly describe a core life history trade‐off forplants. A focus on the allocation of energy by the plant at a given age or sizeallows RA schedules to be easily incorporated into a variety of process‐based plantgrowth and ecosystem models (e.g., Fisher et al.2010; Falster et al.2011; Scheiter et al.2013). The division of energy between growth andreproduction is also the foundation of optimal energy models (Myers and Doyle1983; Kozlowski1992; Perrin and Sibly1993; Reekie and Avila‐Sakar2005; Miller et al.2008).

Yet, our ability to systematically study the life history strategies ofreal plants and relate these to basic theory seems limited by the paucity ofcurrently available data. We expect further integration of RA schedules into plantgrowth models will help clarify several empirical patterns. For example, growth ratesamong larger plants show only weak relationship to leaf traits (Wright et al.2010) – this could be becausesubstantial variation in RA among species veils the underlying effects of traitsinfluencing mass production and deployment (Thomas2010). Better empirical data on RA would also allow thewealth of predictions made by optimal energy models to be tested. For example, dophysiological traits affecting growth and mortality rates have consequences for RAschedules, as theory would suggest (Pugliese and Kozlowski1990) (Iwasa and Cohen1989)? Miller et al. (2008) provides a rare exception,where empirical data was incorporated into an optimal energy model, convincinglyshowing that plant seed set, and hence RA, is strongly affected by insect attack.More data on RA schedules could also greatly improve our ability to modelbiogeochemical cycles and ecosystem food webs. The processes controlling allocationof carbon between different plant tissues has been identified as one of the mostuncertain features of current biosphere models (De Kauwe et al.2014). Whether carbon is allocatedto building leaf, stem, or reproductive material has potentially large implicationsfor predicted carbon fluxes and plant growth rates (Thomas2011). For example, in a widely usedmodel of regional carbon uptake and population dynamics, the ecosystem demographymodel (Moorcroft et al.2001), afixed fraction (0.3) of surplus energy is allocated to reproduction. Our resultssuggest this amount is lower than the maximum achieved by most species, but also thatallocation varies substantially through ontogeny.

To address these key questions, make better comparisons and determinemore generalities, data for RA schedules must be collected across many species usingsimilar if not identical methods. Life history and functional traits must be measuredfor each species in order to determine how variation in these traits correlates withRA schedules. For decades, theoreticians have been using RA schedules as afundamental evolvable trait (Myers and Doyle1983; Iwasa and Cohen1989; Kozlowski1992). It's time we empiricists collected some data.

Enright1985

The manuscript provides data on net annual production across alltree sizes, dividing a plant into leaves, stem, underground tissues, andreproductive tissues. Total leaf number is constant among reproductively matureplants, and therefore, new leaf production is simply replacing shed tissues. RAis therefore calculated as yearly reproductive production divided by the sum ofnet annual production of stems, underground tissues, and reproduction. True RAvalues will be somewhat higher than those reported, especially for the oldestplants, as root turnover rates are not known.

Genet2010

The manuscript provides data on net annual increase in stembiomass, annual non‐structural carbohydrate production, and seed production.Reproductive biomass includes seed and cupule mass, but not other accessorycosts. RA is calculated as seed production divided by the sum of increase instem biomass, annual non‐structural carbohydrate production, and seedproduction. Data were collected in a significant mast year forFagussylvatica, such that RA averaged over several years would likely belower. ForFagus, the accessory costs ignored should be smallcompared with seed and cupule mass, but their inclusion would increase RA.

Hirayama2004

The manuscript provides data on increase in “woody organ”biomass, total leaf biomass, and reproductive costs. Reproductive costs includeaccessory costs, including the number of flowers, aborted fruit, and maturefruit. The allometric equation they used to determine increase in “woody organ”biomass and leaf production are based on averages for an entire community andare not specific to the species they studied. They do not have data on plantage or yearly increase in leaf biomass, such that it is difficult to estimateproportion of total leaf biomass that replaces shed tissue and proportion thatrepresents an increase in leaf biomass. As a rough approximate, we used averageannual increase in dbh across all tree sizes (as there was no clear trend withtree size) to estimate difference in age between their 5 trees. We then dividedthe difference in leaf weight between trees by the difference in age todetermine the annual increase in leaf mass. This rough calculation indicatedthat the majority of leaf biomass replaces the previous year's shed leaves,while a small fraction represents an increase in leaf biomass. RA is thencalculated as reproductive production divided by a sum of increased woodbiomass, increased leaf biomass, and reproductive production. This yields RAvalues up to 3 times greater than presented in their manuscript, although thepattern, a continued increase in RA with height is unchanged.

Hirayama2008

The manuscript provides data on increase in woody material,increase in leaf biomass, and reproductive investment. Reproductive investmentincludes mature fruit (including dispersal and protective material) as well asaborted flowers and fruits. RA is calculated as reproductive investment dividedby the sum of increase in woody material, increase in leaf biomass, andreproductive investment. These species show a distinct pattern of biennialmasting, so we averaged RA during a high and low mast year to determine thetrees' actual allocation patterns. Note that the manuscript uses total leafbiomass, not incremental increase, in their calculation of RA, such that theirestimates of RA are much lower than the ones we have used. The manuscript doesnot provide a threshold size or age for these species, only indicating theyused a range of dbh values above 20 cm.Quercus salicina has aconstant RA across this range, leading us to believe the smallest treesincluded are larger than trees at the “threshold size.” We therefore do notinclude a threshold size or RA for this species.

Kohyama1982

The data are from a figure in the manuscript. The manuscript doesnot indicate how “net reproductive effort” is calculated so we are uncertainwhether it accounts for shed tissues. The reproductive investment componentincludes accessory tissues associated with the cone and seed formation. Mastingoccurs every 4 years, and numbers presented in the study are averaged acrossyears to determine the long‐term RA.

Pinero 1982

The manuscript includes data on net annual production of roots,trunk, live leaves, dead leaves (interpreted to mean shed leaves), seeds, andreproductive accessory tissues. RA is calculated as the sum of reproductiveproduction divided by the sum of net annual production of roots, trunk, liveleaves, and reproductive materials.

Pritts1985

The manuscript includes data on annual biomass production ofvegetative and reproductive structures for a range of plant ages (theirFig. 3). Vegetativestructures are divided into stems, roots, and leaves. Total new vegetativeproduction is the sum of new stem production, new root production, and theincrease in leaf biomass over the previous year. We use the increase in leafproduction asVaccinium corymbosum is a winter deciduousspecies and the other portion of leaf production offsets previously shedtissue. RA is calculated as reproductive production divided by the sum ofreproductive and new vegetative production.

Sakai2003

The manuscript provides equations for increment increase of wood(R²H increment) and annual reproductive biomass for threepopulations ofAbies mariesii. Neither annual leaf and rootproduction, nor turnover of these structures, is known. RA could therefore notbe calculated, although the possible shape of the RA curve could be. Using theequations provided and estimating a wood density of 0.5, annual wood productionand reproductive production are determined. RA is then calculated asreproductive production divided by the sum of wood production and reproductionproduction. If a hypothetical increase in leaf area is included, shifting fromdouble wood production in young plants to a small fraction of wood productionin mature plants, the shapes of the curves of the middle and low populationplants are unchanged, while the initial RA at the high site is much reduced andthe RA across plant size is fairly unchanged.

1. Alvarez‐Buylla, E. R., and Martinez‐Ramos M.. 1992. Demography andallometry ofCecropia obtusifolia, a neotropical pioneer tree ‐an evaluation of the climax ‐ pioneer paradigm for tropical rainforests. J. Ecol.80:275–290. [Google Scholar]
2. Ashman, T.‐L.1994. A dynamic perspective on thephysiological cost of reproduction in plants. Am. Nat.144:300–316. [Google Scholar]
3. Barot, S., Mitja D., Miranda I., Meija G., and Grimaldi M.. 2005. Reproductiveplasticity in an Amazonian palm. Evol. Ecol.Res.7:1051–1065. [Google Scholar]
4. Bazzaz, F. A., Ackerly D. D., and Reekie E. G.2000. Reproductive allocation in plantsPp. 1–30inFenner M., ed. Seeds: the ecology of regeneration inplant communities, 2nd ednCABI,Oxford. [Google Scholar]
5. Bender, M. H., Baskin J. M., and Baskin C. C.. 2000. Age of maturity andlife span in herbaceous, polycarpic perennials. Bot. Rev.66:311–349. [Google Scholar]
6. Bonser, S., and Aarssen L.. 2009. Interpretingreproductive allometry: individual strategies of allocation explain size‐dependentreproduction in plant populations. Perspect.Plant Ecol. Evol. Syst.11:31–40. [Google Scholar]
7. Burns, J. H., Blomberg S. P., Crone E. E., Ehrlén J., Knight T. M., Pichancourt J.‐B., et al. 2010. Empiricaltests of life‐history evolution theory using phylogenetic analysis of plantdemography. J. Ecol.98:334–344. [Google Scholar]
8. Chapin, F. S., Bloom A. J., Field C. B., and Waring R. H.. 1987. Plant responses tomultiple environmental factors. Bioscience37:49–57. [Google Scholar]
9. Charnov, E. L., and Schaffer W. M.. 1973. Life historyconsequences of natural selection Coles result revisited. Am. Nat.107:791–793. [Google Scholar]
10. Chave, J., Coomes D., Jansen S., Lewis S. L., Swenson N. G., and Zanne A. E.. 2009. Towards a worldwidewood economics spectrum. Ecol.Lett.12:351–366. [DOI] [PubMed] [Google Scholar]
11. Chen, H., Felker S., and Sun S.. 2010. Allometry ofwithin‐fruit reproductive allocation in subtropical dicot woodyspecies. Am. J. Bot.97:611–619. [DOI] [PubMed] [Google Scholar]
12. Cleary, M. B., Pendall E., and Ewers B. E.. 2008. Testing sagebrushallometric relationships across three fire chronosequences in Wyoming,USA. J. Arid Environ.72:285–301. [Google Scholar]
13. Cole, L. C.1954. The population consequences of lifehistory phenomena. Q. Rev. Biol.29:103–137. [DOI] [PubMed] [Google Scholar]
14. Comps, B., Thiébaut B., Barrière G., and Letouzey J.. 1994. Répartition de labiomasse entre organes végétatifs et reproducteurs chez le hêtre européen(Fagus sylvatica L), selon le secteur de la couronne et l'âgedes arbres. Annales des SciencesForestières51:11–26. [Google Scholar]
15. Cornelissen, J. H. C., Lavorel S., Garnier E., Díaz S., Buchmann N., Gurvich D. E., et al. 2003. A handbook ofprotocols for standardised and easy measurement of plant functional traitsworldwide. Aust. J. Bot.51:335. [Google Scholar]
16. Cornwell, W. K., Westoby M., Falster D. S., FitzJohn R. G., O'Meara B. C., Pennell M. W., et al. 2014. Functionaldistinctiveness of major plant lineages. J.Ecol.102:345–356. [Google Scholar]
17. Davies, S. J., and Ashton P. S.. 1999. Phenology andfecundity in 11 sympatric pioneer species ofMacaranga(Euphorbiaceae) in Borneo. Am. J.Bot.86:1786–1795. [PubMed] [Google Scholar]
18. De Kauwe, M. G., Medlyn B. E., Zaehle S., Walker A. P., Dietze M. C., Wang Y.‐P., et al. 2014. Where does thecarbon go? A model–data intercomparison of vegetation carbon allocation andturnover processes at two temperate forest free‐air CO2 enrichmentsites. New Phytol.203:883–899. [DOI] [PMC free article] [PubMed] [Google Scholar]
19. Ehlers, B. K., and Olesen J. M.. 2004. Flower production inrelation to individual plant age and leaf production among different patches ofCorydalis intermedia. Plant Ecol. formerly 'Vegetatio',174:71–78. [Google Scholar]
20. Engen, S., and Saether B. E.. 1994. Optimal allocation ofresources to growth and reproduction. Theor.Popul. Biol.46:232–248. [Google Scholar]
21. Enright, N. J.1985. Age, reproduction and biomassallocation inRhopalostylis sapida (Nikau Palm).Austral Ecol.10:461–467. [Google Scholar]
22. Falster, D. S., and Westoby M.. 2005. Tradeoffs betweenheight growth rate, stem persistence and maximum height among plant species in apost‐fire succession. Oikos111:57–66. [Google Scholar]
23. Falster, D. S., Brännström Å., Dieckmann U., and Westoby M.. 2011. Influence of fourmajor plant traits on average height, leaf‐area cover, net primary productivity,and biomass density in single‐species forests: a theoreticalinvestigation. J. Ecol.99:148–164. [Google Scholar]
24. Fisher, R., McDowell N., Purves D., Moorcroft P., Sitch S., Cox P., et al. 2010. Assessinguncertainties in a second‐generation dynamic vegetation model caused by ecologicalscale limitations. New Phytol.187:666–681. [DOI] [PubMed] [Google Scholar]
25. Forbis, T. A., and Doak D. F.. 2004. Seedlingestablishment and life history trade‐offs in alpine plants.Am. J. Bot.91:1147–1153. [DOI] [PubMed] [Google Scholar]
26. Franco, M., and Silvertown J.. 2004. A comparativedemography of plants based upon elasticities of vital rates.Ecology85:531–538. [Google Scholar]
27. Garcia, M. B., Pico F. X., and Ehrlen J.. 2008. Life span correlateswith population dynamics in perennial herbaceous plants. Am. J. Bot.95:258–262. [DOI] [PubMed] [Google Scholar]
28. Genet, H., Bréda N., and Dufrêne E.. 2010. Age‐related variationin carbon allocation at tree and stand scales in beech (Fagussylvatica L.) and sessile oak (Quercus petraea(Matt.) Liebl.) using a chronosequence approach. Tree Physiol.30:177–192. [DOI] [PubMed] [Google Scholar]
29. Greene, D. F., and Johnson E. A.. 1994. Estimating the meanannual seed production of trees. Ecology75:642–647. [Google Scholar]
30. Gurney, W. S. C., and Middleton D. A. J.. 1996. Optimal resourceallocation in a randomly varying environment. Funct. Ecol.10:602–612. [Google Scholar]
31. Harper, J. L., and Ogden J.. 1970. The reproductivestrategy of higher plants: I. The concept of strategy with special reference toSenecio vulgaris L. J.Ecol.58:681. [Google Scholar]
32. Hemborg, Å. M., and Karlsson P. S.. 1998. Somatic costs ofreproduction in eight subarctic plant species. Oikos82:149–157. [Google Scholar]
33. Henery, M., and Westoby M.. 2001. Seed mass and seednutrient content as predictors of seed output variation betweenspecies. Oikos92:479–490. [Google Scholar]
34. Herrera, C. M.1991. Dissecting factors responsible forindividual variation in plant fecundity. Ecology72:1436–1448. [Google Scholar]
35. Herrera, C. M., and Jovani R.. 2010. Lognormaldistribution of individual lifetime fecundity: insights from a 23‐yearstudy. Ecology91:422–430. [DOI] [PubMed] [Google Scholar]
36. Hirayama, D., Itoh A., and Yamakura T.. 2004. Implications fromseed traps for reproductive success, allocation and cost in a tall tree speciesLindera erythrocarpa. Plant Species Biol.19:185–196. [Google Scholar]
37. Hirayama, D., Nanami S., Itoh A., and Yamakura T.. 2008. Individual resourceallocation to vegetative growth and reproduction in subgenusCyclobalanopsis (Quercus, Fagaceae) trees.Ecol. Res.23:451–458. [Google Scholar]
38. Iwasa, Y., and Cohen D.. 1989. Optimal growthschedule of a perennial plant. Am.Nat.133:480–505. [Google Scholar]
39. Jasienski, M., and Bazzaz F. A.. 1999. The fallacy of ratiosand testability of models in biology. Oikos84:321–326. [Google Scholar]
40. Karlsson, P. S., Svensson B. M., Carlsson B. Å., and Nordell K. O.. 1990. Resource investmentin reproduction and its consequences in threePinguiculaspecies. Oikos59:393–398. [Google Scholar]
41. Katsukawa, Y., Katsukawa T., and Matsuda H.. 2002. Indeterminate growthis selected by a trade‐off between high fecundity and risk avoidance in stochasticenvironments. Popul. Ecol.44:265–272. [Google Scholar]
42. Kelly, D., and Sork V. L.. 2002. Mast seeding inperennial plants: why, How, Where?Annu. Rev. Ecol. Syst.33:427–447. [Google Scholar]
43. King, D., and Roughgarden J.. 1982. Graded allocationbetween vegetative and reproductive growth for annual plants in growing seasons ofrandom length. Theor. Popul. Biol.22:1–16. [Google Scholar]
44. Klinkhamer, P. G. L., Meelis E., de Jong T. J., and Weiner J.. 1992. On the analysis ofsize‐dependent reproductive output in plants. Funct. Ecol.6:308–316. [Google Scholar]
45. Klinkhamer, P. G. L., Kubo T., and Iwasa Y.. 1997. Herbivores and theevolution of the semelparous perennial life‐history of plants.J. Evol. Biol.10:529–550. [Google Scholar]
46. Kohyama, T.1982. Studies on theAbiespopulation of Mt. Shimagare II. Reproductive and life historytraits. Bot. Mag. Tokyo95:167–181. [Google Scholar]
47. Kozlowski, J.1992. Optimal allocation of resources togrowth and reproduction: implications for age and size atmaturity. Trends Ecol. Evol.7:15–19. [DOI] [PubMed] [Google Scholar]
48. Kozłowski, J., and Uchmanski J.. 1987. Optimal individualgrowth and reproduction in perennial species with indeterminategrowth. Evol. Ecol.1:214–230. [Google Scholar]
49. Kozłowski, J., and Wiegert R. G.. 1987. Optimal age and sizeat maturity in annuals and perennials with determinate growth.Evol. Ecol.1:231–244. [Google Scholar]
50. Lord, J. M., and Westoby M.. 2006. Accessory costs ofseed production. Oecologia150:310–317. [DOI] [PubMed] [Google Scholar]
51. Mäkelä, A.1997. A carbon balance model of growth andself‐pruning in trees based on structural relationships. For. Sci.43:7–24. [Google Scholar]
52. McGill, B., Enquist B., Weiher E., and Westoby M.. 2006. Rebuilding communityecology from functional traits. Trends Ecol.Evol.21:178–185. [DOI] [PubMed] [Google Scholar]
53. McNamara, J. M., Houston A. I., Barta Z., Scheuerlein A., and Fromhage L.. 2009. Deterioration, deathand the evolution of reproductive restraint in late life. Proc. Biol. Sci.276:4061–4066. [DOI] [PMC free article] [PubMed] [Google Scholar]
54. Miller, T. E. X., Tenhumberg B., and Louda S. M.. 2008. Herbivore‐mediatedecological costs of reproduction shape the life history of an iteroparousplant. Am. Nat.171:141–149. [DOI] [PubMed] [Google Scholar]
55. Moles, A. T., Falster D. S., Leishman M. R., and Westoby M.. 2004. Small‐seeded speciesproduce more seeds per square metre of canopy per year, but not per individual perlifetime. J. Ecol.92:384–396. [Google Scholar]
56. Moorcroft, P. R., Hurtt G. C., and Pacala S. W.. 2001. A method for scalingvegetation dynamics: the ecosystem demography model (ED). Ecol. Monogr.71:557–585. [Google Scholar]
57. Müller, I., Schmid B., and Weiner J.. 2000. The effect ofnutrient availability on biomass allocation patterns in 27 species of herbaceousplants. Perspect. Plant Ecol. Evol.Syst.3:115–127. [Google Scholar]
58. Murphy, G. I.1968. Pattern in life history and theenvironment. Am. Nat.102:391–403. [Google Scholar]
59. Myers, R. A., and Doyle R. W.. 1983. Predicting naturalmortality rates and reproduction–mortality trade‐offs from fish life historydata. Can. J. Fish Aquat. Sci.40:612–620. [Google Scholar]
60. Nakashizuka, T., Takahashi Y., and Kawaguchi H.. 1997. Production‐dependentreproductive allocation of a tall tree speciesQuercus serrata. J. Plant. Res.110:7–13. [DOI] [PubMed] [Google Scholar]
61. Niinemets, U.2002. Stomatal conductance alone does notexplain the decline in foliar photosynthetic rates with increasing tree age andsize inPicea abies andPinus sylvestris. Tree Physiol.22:515–535. [DOI] [PubMed] [Google Scholar]
62. Niklas, K., and Enquist B.. 2003. An allometric modelfor seed plant reproduction. Evol. Ecol.Res.5:79–88. [Google Scholar]
63. Obeso, J. R.2002. The costs of reproduction inplants. New Phytol.155:321–348. [DOI] [PubMed] [Google Scholar]
64. Oyama, K.1990. Variation in growth and reproductionin the neotropical dioecious palmChamaedorea tepejilote. J. Ecol.78:648–663. [Google Scholar]
65. Perrin, N., and Sibly R. M.. 1993. Dynamic models ofenergy allocation and investment. Annu. Rev.Ecol. Syst.24:379–410. [Google Scholar]
66. Piñero, D., Sarukhan J., and Alberdi P.. 1982. The costs ofreproduction in a tropical palm,Astrocaryum mexicanum. J. Ecol.70:473. [Google Scholar]
67. Pino, J., Sans F. X., and Masalles R. M.. 2002. Size‐dependentreproductive pattern and short‐term reproductive cost inRumexobtusifolius L. ActaOecol.23:321–328. [Google Scholar]
68. Pitelka, L. F.1977. Energy allocations in annual andperennial lupines (Lupinus: Leguminosae). Ecology58:1055–1065. [Google Scholar]
69. Poorter, L., Zuidema P. A., Peña‐Claros M., and Boot R. G. A.. 2005. A monocarpic treespecies in a polycarpic world: how canTachigali vasqueziimaintain itself so successfully in a tropical rain forestcommunity?J. Ecol.93:268–278. [Google Scholar]
70. Poorter, L., McDonald I., Alarcon A., Fichtler E., Licona J.‐C., Pena‐Claros M., et al. 2010. The importanceof wood traits and hydraulic conductance for the performance and life historystrategies of 42 rainforest tree species. New Phytol.185:481–492. [DOI] [PubMed] [Google Scholar]
71. Pritts, M. P., and Hancock J. F.. 1983. Seasonal and lifetimeallocation patterns of the woody goldenrod,Solidagopauciflosculosa Michaux. (Compositae). Am. J. Bot.70:216–221. [Google Scholar]
72. Pritts, M. P., and Hancock J. F.. 1985. Lifetime biomasspartitioning and yield component relationships in the highbush blueberry,Vaccinium corymbosum L. (Ericaceae). Am. J. Bot.72:446–452. [Google Scholar]
73. Pugliese, A., and Kozlowski J.. 1990. Optimal patterns ofgrowth and reproduction for perennial plants with persisting or not persistingvegetative parts. Evol. Ecol.4:75–89. [Google Scholar]
74. Read, J., Sanson G. D., Jaffré T., and Burd M.. 2006. Does tree sizeinfluence timing of flowering inCerberiopsis candelabra(Apocynaceae), a long‐lived monocarpic rain‐forest tree?J. Trop. Ecol.22:621–629. [Google Scholar]
75. Read, J., Sanson G. D., Burd M., and Jaffré T.. 2008. Mass flowering andparental death in the regeneration ofCerberiopsis candelabra(Apocynaceae), a long‐lived monocarpic tree in New Caledonia.Am. J. Bot.95:558–567. [DOI] [PubMed] [Google Scholar]
76. Reekie, E. G., and Avila‐Sakar G.2005. The shape of the trade‐off function betweenreproduction and growth p. 247inReekie E. G., Bazzaz F. A., eds. Reproductive allocation inplants. Elsevier,Amsterdam. [Google Scholar]
77. Reekie, E. G., and Bazzaz F. A.. 1987. Reproductive effortin plants. 2. Does carbon reflect the allocation of otherresources?Am. Nat.129:897–906. [Google Scholar]
78. Reich, P. B., Walters M. B., and Ellsworth D. S.. 1992. Leaf life‐span inrelation to leaf, plant, and stand characteristics among diverseecosystems. Ecol. Monogr.62:365. [Google Scholar]
79. Reinartz, J. A.1984. Life history variation of commonmullein (Verbascum thapsus). I Latitudinal differences inpopulation dynamics and timing of reproduction. J. Ecol.72:897. [Google Scholar]
80. Reznick, D., and Endler J. A.. 1982. The impact ofpredation on life history evolution in Trinidadian guppies (Poeciliareticulata). Evolution36:160–177. [DOI] [PubMed] [Google Scholar]
81. de Ridder, F., and Dhondt A. A.. 1992a. The reproductivebehaviour of a clonal herbaceous plant, the longleaved sundewDroseraintermedia, in different heathland habitats. Ecography15:144–153. [Google Scholar]
82. de Ridder, F., and Dhondt A. A.. 1992b. The demography of aclonal herbaceous perennial plant, the longleaved sundewDroseraintermedia, in different heathland habitats. Ecography15:129–143. [Google Scholar]
83. R Core Team .2014. R: A language and environment forstatistical computing. R Foundation for StatisticalComputing, Vienna, Austria: URLhttp://www.R-project.org/. [Google Scholar]
84. Roff, D. A.2002. Life history evolution.Sinauer Associates, Incorporated,Sunderland, MA. [Google Scholar]
85. Sakai, A., Matsui K., Kabeya D., and Sakai S.. 2003. Altitudinal variationin lifetime growth trajectory and reproductive schedule of a sub‐alpine conifer,Abies mariesii. Evol. Ecol. Res.5:671–689. [Google Scholar]
86. Samson, D. A., and Werk K. S.. 1986. Size‐dependenteffects in the analysis of reproductive effort in plants. Am. Nat.127:667–680. [Google Scholar]
87. Santos‐del‐Blanco, L., Zas R., Notivol E., Chambel M. R., Majada J., and Climent J.. 2010. Variation of earlyreproductive allocation in multi‐site genetic trials of Maritime pine and Aleppopine. For. Syst.19:381–392. [Google Scholar]
88. Santos‐del‐Blanco, L., Climent J., Gonzalez‐Martinez S. C., and Pannell J. R.. 2012. Geneticdifferentiation for size at first reproduction through male versus femalefunctions in the widespread Mediterranean treePinus pinaster. Ann. Bot.110:1449–1460. [DOI] [PMC free article] [PubMed] [Google Scholar]
89. Scheiter, S., Langan L., and Higgins S. I.. 2013. Next‐generationdynamic global vegetation models: learning from community ecology.New Phytol.198:957–969. [DOI] [PubMed] [Google Scholar]
90. Sibly, R., Calow P., and Nichols N.. 1985. Are patterns ofgrowth adaptive?J. Theor. Biol.112:553–574. [DOI] [PubMed] [Google Scholar]
91. Smith, H. M., and Samach A.. 2013. Constraints toobtaining consistent annual yields in perennial tree crops. I: heavy fruit loaddominates over vegetative growth. PlantSci.207:158–167. [DOI] [PubMed] [Google Scholar]
92. Stearns, S. C.1992. The evolution of life histories.Oxford University Press,Oxford. [Google Scholar]
93. Svensson, B. M., Carlsson B. A., Karlsson P. S., and Nordell K. O.. 1993. Comparative long‐termdemography of three species ofPinguicula. J. Ecol.81:635. [Google Scholar]
94. Thomas, S. C.1996. Relative size at onset of maturity inrain forest trees: a comparative analysis of 37 Malaysian species.Oikos76:145–154. [Google Scholar]
95. Thomas, S. C.2010. Photosynthetic capacity peaks atintermediate size in temperate deciduous trees. Tree Physiol.30:555–573. [DOI] [PubMed] [Google Scholar]
96. Thomas, S. C.2011. Age‐related changes in tree growth and functionalbiology: the role of reproduction Pp.33–64inMeinzer F. C., Lachenbruch B. and Dawson T. E., eds. Size‐ and age‐related changes in treestructure and function. Springer,Netherlands,Dordrecht. [Google Scholar]
97. Thornley, J. H. M.1972. A model to describe the partitioningof photosynthate during vegetative plant growth. Ann. Bot.36:419–430. [Google Scholar]
98. Turner, I. M.2001. The ecology of trees in the tropical rainforest. Cambridge University Press,UK. [Google Scholar]
99. Weiner, J.2004. Allocation, plasticity and allometryin plants. Perspect. Plant Ecol. Evol.Syst.6:207–215. [Google Scholar]
100. Weiner, J., Campbell L. G., Pino J., and Echarte L.. 2009. The allometry ofreproduction within plant populations. J.Ecol.97:1220–1233. [Google Scholar]
101. Wesselingh, R. A., Klinkhamer P. G. L., de Jong T. J., and Boorman L. A.. 1997. Threshold size forflowering in different habitats: effects of size‐dependent growth andsurvival. Ecology78:2118–2132. [Google Scholar]
102. Westoby, M., Falster D., Moles A., Vesk P., and Wright I.. 2002. Plant ecologicalstrategies: some leading dimensions of variation between species.Annu. Rev. Ecol. Syst.33:125–159. [Google Scholar]
103. Wilson, A. M., and Thompson K.. 1989. A comparative studyof reproductive allocation in 40 British grasses. Funct. Ecol.3:297–302. [Google Scholar]
104. de Wit, C. T.1978. Simulation of assimilation.Centre for Agricultural Publishing andDocumentation, Respiration and Transpiration ofCrops. [Google Scholar]
105. Wright, I. J., Reich P. B., Westoby M., Ackerly D. D., Baruch Z., Bongers F., et al. 2004. The worldwideleaf economics spectrum. Nature428:821–827. [DOI] [PubMed] [Google Scholar]
106. Wright, S. J., Jaramillo M. A., Pavon J., Condit R., Hubbell S. P., and Foster R. B.. 2005. Reproductive sizethresholds in tropical trees: variation among individuals, species andforests. J. Trop. Ecol.21:307–315. [Google Scholar]
107. Wright, S. J., Kitajima K., Kraft N. J. B., Reich P. B., Wright I. J., Bunker D. E., et al. 2010. Functionaltraits and the growth–mortality trade‐off in tropical trees.Ecology91:3664–3674. [DOI] [PubMed] [Google Scholar]
108. Young, T. P.1984. The comparative demography ofsemelparousLobelia telekii and iteroparousLobeliakeniensis on Mount Kenya. J.Ecol.72:637. [Google Scholar]
109. Young, T. P.2010. Semelparity anditeroparity. Nat. Educ. Knowl.3:2. [Google Scholar]
110. Zammit, C. A., and Zedler P. H.. 1993. Size structure andseed production in even‐aged populations ofCeanothus greggii inmixed chaparral. J. Ecol.81:499–511. [Google Scholar]