Evolution of life-history timing in a leafmining moth:Phenotypic selection in patches with manipulateddevelopment time

ROBERT McGREGOR*

Behavioral Ecology Research Group, Department of Biological Sciences, Simon Fraser University, Burnaby, British Columbia

V5A 1S6, Canada

Summary

Phyllonorycter mespilella (HuÈ bner) is a leafmining moth with two stages of larval development: the initial sap-

feeding (SF) stage followed by the tissue-feeding (TF) stage. Phenotypic selection by parasitoids on theduration of the SF stage (SF duration) was measured in arti®cial patches of larvae placed in the ®eld during

the diapausing generation. Pretreatment of larvae with di�erent photoperiods allowed creation of patches thatvaried in the time-course of appearance of TF larvae. The shorter the photoperiod pretreatment, the sooner

TF larvae tended to appear. Some patches were left exposed and others were caged to exclude parasitoids.Positive directional selection on SF duration was detected in exposed patches, and no selection was detected in

caged patches. Directional selection in exposed patches was caused by both parasitoid oviposition and otherunidenti®ed sources of mortality. The other sources of mortality may have included host feeding by parasi-

toids on TF larvae. A larger proportion of parasitoid eggs were oviposited on TF larvae in patches where TFlarvae appeared the earliest, but this variation in parasitoid oviposition did not result in signi®cant di�erences

in directional selection intensity among patches with early, intermediate and late appearance of TF larvae.Although the general form of the ®tness function was very similar when compared among patch types, no

signi®cant directional selection could be detected in patches where TF larvae appeared late, and the causes ofdirectional selection appeared to vary between patches where TF larvae appeared at early and intermediate

dates.

Keywords: ®tness functions; leafminers; life-history timing; parasitoids; phenotypic manipulations; phenotypicselection; Phyllonorycter

Introduction

Insect parasitoids are assumed to act as selective in¯uences on a variety of traits in herbivorousinsects (Price et al., 1980; Je�ries and Lawton, 1984; Lawton, 1986; Bernays and Graham, 1988).The within-generation consequences of natural selection on quantitative traits can be determinedby making direct measurements of phenotypic selection in natural populations (Lande and Arnold,1983; Endler, 1986; Mitchell-Olds and Shaw, 1987; Schluter, 1988; Crespi, 1990; Wade and Kalisz,1990). However, attempts to measure selection by parasitoids on traits in herbivorous insects havebeen few (Rausher, 1992; Weis et al., 1992; McGregor, 1996). In this paper, I describe measure-ments of phenotypic selection by parasitoids on the timing of life-history events in a leafminingmoth. In particular, I report on measurements of phenotypic selection in a manipulative experi-ment where the phenotypic distribution of life-history timing is arti®cially altered. I attempt toaddress the in¯uence of variation in parasitoid behaviour on the form of selection.

*Address all correspondence to Robert McGregor, PARC Agassiz, Agriculture and Agri-Food Canada, PO Box 1000,

Agassiz, British Columbia V0M 1A0, Canada.

Evolutionary Ecology 1998, 12, 629±642

0269-7653 Ó 1998 Chapman & Hall

The ®tness function, derived from measurements of phenotypic selection, formalizes the rela-tionship between phenotype and a measure of ®tness (Schluter, 1988). Phenotypic manipulationscan be used to expand the range of phenotype over which the function is estimated in a naturalpopulation, or to increase the precision of the estimate in the tails of the phenotypic distributionwhere fewer data are available (Schluter, 1988; Anholt, 1991). The ®tness consequences of phe-notypic manipulations have been studied for traits such as clutch size, male tail length and tes-tosterone levels in birds (Andersson, 1982; Dijkstra et al., 1990; Ketterson and Nolan, 1992;Mùller, 1994), egg size in lizards (Sinervo et al., 1992) and body size in damsel¯ies (Anholt, 1991).Comparisons can also be made of phenotypic selection and ®tness functions among groups ofindividuals with manipulated phenotypic distributions. For example, detection of di�erences inphenotypic selection between treatment groups with di�ering phenotypic distributions can be usedto indicate the existence of frequency-dependent selection (Wade and Kalisz, 1990; Schluter, 1994).When predation or parasitism is the cause of selection, variation in phenotypic selection can

result from variation in the behaviour of the predators or parasites. Many animals change theirrates of acceptance for di�erent prey types based on the local distribution of prey (Krebs andMcCleery, 1984; Shettleworth, 1984; Stephens and Krebs, 1986). Such variation in prey selectionbehaviour may, in turn, result in variation in natural selection on characters that determine thedistribution of prey phenotypes. In animals with sequential life-history stages that vary in their riskof predation, natural selection should in¯uence the timing of transitions between stages to maxi-mize survival from predation (Rowe and Ludwig, 1990). If the predator's attack on di�erent life-history stages varies with the frequency distribution of those stages, then the form of selection onlife-history timing may also vary.Insect parasitoids often display oviposition preferences for particular developmental stages of

their hosts (Vinson, 1976; Vinson and Iwantsch, 1980; Strand, 1986; van Alphen and Vet, 1986;Godfray, 1994). This stage-speci®c oviposition can act as a selective in¯uence on the timing of life-history transitions (McGregor, 1996). However, oviposition behaviour in parasitic insects is oftenplastic, and can vary in response to the local distribution of host types (Mangel and Roitberg, 1989;Roitberg et al., 1992, 1993; Li et al., 1993). When oviposition preferences vary with the distributionof hosts available, the form of selection on the timing of life history may change. The in¯uence ofsuch variation in parasitoid behaviour on phenotypic selection can be determined by comparing®tness functions between treatment groups of host insects with manipulated life-history timing.Here, I present the results of such a manipulation for Phyllonorycter mespilella (HuÈ bner), a leaf-mining moth that feeds on the foliage of apple trees in the Okanagan Valley of British Columbia,Canada (Cossentine and Jensen, 1992).

Natural history

Two sequential stages occur during larval development in Phyllonorycter species: the sap-feeding(SF) and tissue-feeding (TF) stages (Pottinger and Leroux, 1971). During the SF stage, larvaeenlarge a blotch-shaped mine by shearing mesophyll cells and feeding on plant sap (Pottinger andLeroux, 1971). Growth in body size occurs during the SF stage, but, to a large extent, feeding in theSF stage functions to enlarge the area of the leafmine in which all subsequent feeding occurs anddevelopment to the pupal stage is completed. Most larval growth in Phyllonorycter species occursduring the TF stage (Pottinger and Leroux, 1971). Tissue-feeeding larvae spin silk threads acrossthe bottom surface of the leafmine that cause the leaf surface to fold into a tentiform structure thatwill accommodate the increased body size of the late-stage larva and pupa. Tissue-feeding larvaefeed on parenchyma cells on the upper surface of this feeding chamber, and pupate inside the mine(Pottinger and Leroux, 1971).

630 McGregor

Pnigalio ¯avipes (Ashmead) and a Sympiesis species (Hymenoptera: Eulophidae) are the mostcommon parasitoids attacking P. mespilella in the Okanagan Valley (Cossentine and Jensen, 1992).Parasitoids of these genera oviposit predominantly on TF larvae and host feed mainly on SF larvae(Askew and Shaw, 1979; Maier, 1982; van Driesche and Taub, 1983; Laing et al., 1986; Casas,1989; Barrett and Brunner, 1990; Varela and Welter, 1992). The high potential for both growth andparasitoid oviposition attack during the TF stage de®nes a growth±mortality trade-o� that shouldin¯uence the evolution of the timing of the transition between the SF and TF stages.Prolonged extensions of the duration of the SF stage have been observed in a number of

Phyllonorycter species in the diapausing generation (Pottinger and Leroux, 1971; Maier, 1984;Laing et al., 1986; Barrett and Brunner, 1990; Varela and Welter, 1992; McGregor, 1997). Al-though the length of the SF stage is extended, no corresponding increase in the size of SF larvaeoccurs in the diapausing generation compared to summer generations (Pottinger and Leroux,1971). Phenotypic selection by parasitoids may, in part, explain this seasonal extension of theduration of SF development, which occurs without an apparent pay-o� in larval growth. Ovipo-sition attack by parasitoids on TF larvae caused positive directional selection on the date oftransition to the TF stage in one population of P. mespilella during the diapausing generation(McGregor, 1996). Individuals that began TF feeding at a later date had a higher probability ofsurvival from parasitism because they delayed TF development until after the most intense periodof parasitoid oviposition.Parasitoid oviposition on SF larvae is more frequent in the fall (autumn) generation than in

summer generations in P. mespilella (McGregor, 1996). Seasonal increases in oviposition on SFlarvae by Eulophid parasitoids have also been observed in other Phyllonorycter species (Barrett andBrunner, 1990; Varela and Welter, 1992). Host acceptance decisions by parasitoids attackingPhyllonorycter larvae may depend on the local frequencies of SF and TF larvae. Sap-feeding larvaemay be accepted more often for oviposition when TF larvae are rare, as is the case early in the fallgeneration. The survival advantage of delaying TF development to avoid parasitoid attack will bereduced if the probability of oviposition attack in the SF stage increases. Phenotypic selection on TFdate, and the form of the ®tness function, may thus depend on the frequencies of SF and TF larvae.In this study, phenotypic selection by parasitoids on SF duration was measured in a single

population of P. mespilella in the ®eld. Patches of larvae were created with manipulated phenotypicdistributions of SF duration. The resulting patch types contained di�erent relative frequencies ofSF and TF larvae during the period of parasitoid attack. The following questions were addressed inthe study: (1) Does phenotypic selection on SF duration occur in this population of P. mespilella?(2) Is selection on SF duration caused by parasitoid oviposition, or by other sources of mortality?(3) Does the pattern of parasitoid oviposition on SF and TF larvae vary among patch types withdi�erent life-history timing? (4) Does phenotypic selection on SF duration vary among these patchtypes as a consequence of changes in the pattern of parasitoid behaviour?

Materials and methods

Experimental design and data collection

Phyllonorycter mespilella pupae were collected during the second generation in August 1993 in aresearch orchard located at Kelowna, British Columbia. Three generations of P. mespilella arecompleted per year in the Okanagan Valley (Cossentine and Jensen, 1992). Leaves containingpupae were collected and returned to the laboratory. Pupae were dissected from leafmines, sexedand then placed individually in gelatin capsules. Each pupa was randomly assigned to a growthchamber (set at a temperature of 25oC) with either a 16-h or a 13-h photoperiod.

Life-history timing in a leafmining moth 631

Photoperiod pretreatments were chosen to simulate di�erent points in the growing season.Seasonal changes in photoperiod cause changes in the duration of development and mediate theinduction of diapause in P. mespilella (McGregor, 1997). A 13-h photoperiod occurs at Kelowna inearly September. It was predicted that larvae treated with a 13-h photoperiod would respond as ifless growing time was available before winter, resulting in shorter SF durations. The larvae treatedwith 16-h photoperiods were predicted to have longer SF durations.Upon emergence, adult male and female P. mespilella were placed in pairs in gelatin capsules for

24 h to induce mating. Mated female leafminers were transferred individually into ovipositioncages that contained single apple seedlings 30±40 cm in height. After 24 h, the females were re-moved from the cages and the seedlings, with resulting full-sib families of leafminers, were held inthe same growth chamber for 6 days. Ovipositions occurred over a 7-day period (17±23 August). Atthe end of the 6-day pretreatment, seedlings were transferred to a greenhouse, and held at ambientphotoperiod until all other ovipositions and photoperiod pretreatments were complete.On 29 August, all seedlings were returned to the orchard in Kelowna. Twelve patches, each

consisting of 10 seedlings, were positioned at the base of orchard trees on platforms at a height of1 m. On the platforms, the leaves on the seedlings were at the same height as the lowest foliage inthe orchard trees. Three types of patch were created: four patches consisting of 10 seedlingspretreated with the 16-h photoperiod; four patches consisting of 5 seedlings from each of the twophotoperiod pretreatments; and four patches consisting of 10 seedlings pretreated with the 13-hphotoperiod. These will be referred to as the 16-h, mixed and 13-h patch types, respectively. Thepatches were arranged in a rectangular grid with 15 m between each position. The type of patchplaced at each position was randomly assigned. In addition, two patches were created that con-tained 10 seedlings from each photoperiod pretreatment. These patches were caged to excludeparasitoids, and located 15 m from the edge of the treatment grid. All leafmines on the seedlingswere individually marked and numbered.Leafmines were checked daily for the transition between the SF and TF stages of development.

The date on which the ®rst longitudinal wrinkle appeared on the bottom surface of the leafminewas recorded as the date of transition to the TF stage (TF date). Longitudinal wrinkles appear onthe lower surface of Phyllonorycter leafmines when TF larvae spin silk threads early in the fourthinstar (Pottinger and Leroux, 1971). This is the ®rst external evidence that the moult to the TFstage has occurred. The duration of the SF stage was calculated and recorded for each larva as thenumber of days between the date of oviposition and the TF date.After the TF dates were recorded, the seedlings were left in the orchard until 27 October. At this

time, they were returned to the laboratory, and each marked leafmine was dissected. Each indi-vidual was assigned to one of three categories: survived, parasitized or other. Individuals that hadsurvived were TF larvae or pupae that were alive at the time of dissection. Parasitized individualshad an associated parasitoid egg, larva or pupa. Individuals in the `other' category had died ofunknown causes.

Analysis of phenotypic manipulations

Analysis of variance (ANOVA) in SF duration was conducted by patch type for data from the 12exposed patches (all statistical analyses were conducted using SYSTAT; Wilkinson et al., 1992).Mean SF duration was compared among patch types using Tukey tests. The proportion of SF andTF larvae present in each patch type was calculated for each Julian date during the experiment. Thecumulative proportion of TF larvae appearing during the experiment in the di�erent patch typeswas compared by pairwise Kolmogorov-Smirnov two-sample tests.

632 McGregor

Analysis of parasitoid oviposition

The number and proportion of individuals parasitized in the SF and TF stages was recorded foreach patch type. v2 analysis was used to test whether the distribution of parasitoid oviposition onthe two larval stages was independent of patch type.

Analysis of survival from parasitism

Standard univariate selection statistics were calculated for SF duration for those individuals thatsurvived the SF stage. Directional selection intensity was calculated as:

i � xa ÿ xbs

where xb is the mean SF duration before selection, xa is the mean SF duration after selection, and sis the standard deviation of SF duration before selection (Endler, 1986). Selection on the varianceof the phenotypic distribution was quanti®ed by calculating the statistic j from the formula:

j � ma ÿ mbmb

where mb is the phenotypic variance of SF duration before selection and ma is the variance afterselection (Endler, 1986). Selection statistics were tested for signi®cant di�erences from zero using atwo-tailed t-test for values of i and a two-tailed F-test for values of j. These are standard tests usedto detect the existence of signi®cant directional, stabilizing or disruptive selection (Endler, 1986).Values of selection statistics were ®rst calculated (and tested for signi®cant di�erences from zero)

for all of the data collected from all 12 exposed patches. This analysis was done to determinewhether signi®cant phenotypic selection on SF duration occurred in this population of P. me-spilella. Values of i and j were also calculated for subsets of the data from which individuals of oneof the two mortality categories had been removed. Selection statistics were calculated from thesereduced data sets to examine the contribution of individual mortality sources to phenotypic se-lection. Selection statistics were also calculated for individuals caged to exclude parasitoids. Valuesof i and j calculated for exposed larvae were compared with values calculated for larvae caged toexclude parasitoids to examine the contribution of parasitoids (and other sources of mortalityexcluded by cages) to phenotypic selection.Values of selection statistics were calculated for each replicate patch in the experiment, and

ANOVA by patch type was conducted on these values of i and j. This analysis was done todetermine if there were signi®cant di�erences among the three patch types in the values of selectionstatistics. Power analysis was conducted on these ANOVA using the program GPOWER (Erd-felder et al., 1996). The data were then combined from the four replicates of each patch type tocreate pooled data sets for the 16-h, mixed and 13-h patches. As for the entire data set above,values of i and j were calculated for all of the data for each patch type, and for subsets of the datafrom which individuals of one of the two mortality categories had been removed. Selection sta-tistics for the pooled data for the 16-h, mixed and 13-h patch types were tested for signi®cantdi�erences from zero using the tests described above.Survival functions were generated for data from all 12 exposed patches, and pooled data for each

of the three exposed patch types, using a program that ®ts cubic spline regressions to survival data(Schluter, 1988). The data were searched within the range )5 to +10 for the value of the smoothingparameter that minimized the `generalized cross-validation' score (Schluter, 1988). Standard errorsaround survival functions were estimated from the predicted values of cubic spline regressions ®t to200 bootstrap samples of the data (E�ron, 1981; Schluter, 1988).

Life-history timing in a leafmining moth 633

All measurements of selection, in this experiment, were made on individuals that survived the SFstage. Because the phenotype (SF duration) of individuals that die during the SF stage is neverexpressed, no information can be derived regarding selection during that stage. Phenotypic se-lection on SF duration may be caused by mortality sources that act during the SF stage, but itcannot be directly quanti®ed. Here, I quantify phenotypic selection on the variation in SF durationthat remains after mortality during the SF stage has occurred.

Results

Phenotypic manipulations

Mean SF duration was shorter for 13-h patches (40.9 � 0.4 days) than for 16-h patches(45.1 � 0.4 days) or mixed patches (44.8 � 0.5 days; F� 27.6, d.f.� 2, P<0.001). However, therewas no signi®cant di�erence in mean SF duration between 16-h patches and mixed patches. Alarger proportion of larvae reached the TF stage at an earlier date in 13-h patches than in mixedpatches (Kolmogorov-Smirnov two-sample test, P<0.05) or 16-h patches (P<0.01) (Fig. 1). Al-though there was no di�erence in mean SF duration between 16-h and mixed patches, the timecourse of appearance of TF larvae varied between these patch types (Kolmogorov-Smirnov two-sample test, P<0.01). A larger proportion of larvae reach the TF stage at an earlier date in mixedpatches than in 16-h patches.

Figure 1. Time course of transition to the TF stage for Phyllonorycter mespilella in arti®cial patches withmanipulated development times. The ®gure shows the proportion of individuals within 13-h, mixed and 16-h

patches that had completed the SF stage at each Julian date.

634 McGregor

Pattern of parasitoid oviposition

More eggs were oviposited in TF mines, than in SF mines, in 13-h patches (Fig. 2). Approximatelyequal numbers of parasitoid eggs were oviposited in SF and TF mines in both mixed and 16-hpatches (Fig. 2). The proportion of eggs oviposited in SF and TF mines was independent of patchtype when mixed patches and 16-h patches were compared (v2� 0.001, d.f.� 1, P� 0.98). When theproportion of eggs oviposited in SF and TF mines was compared between 13-h patches and 16-hpatches (v2� 14.7, d.f.� 1, P<0.001), or between 13-h and mixed patches (v2� 15.4, d.f.� 1,P<0.001), the pattern of oviposition was not independent of patch type.

Phenotypic selection on SF duration

Signi®cant positive directional selection occurred on SF duration in larvae developing in the 12exposed patches (Table 1). The ®tness function for combined data from exposed patches shows amonotonic increase in survival probability across the range of SF duration (Fig. 3). Signi®cantnegative selection on the variance of SF duration also occurred (Table 1). Signi®cantly negativevalues of j can indicate the existence of stabilizing selection (Lande and Arnold, 1983; Endler, 1986).However, because negative values of j can also be caused by the reduction in variance associatedwith directional selection (Endler, 1986), and because no internal survivorship peak occurs in the®tness function (Fig. 3), it is unlikely that stabilizing selection on SF duration occurred.The values of i and j calculated for individuals in caged patches are extremely low in magnitude,

and not signi®cantly di�erent from zero (Table 1). This indicates that no change in the mean or

Figure 2. Proportion of parasitoid eggs oviposited in SF and TF mines in 13-h (n� 114), mixed (n� 99) and16-h (n� 87) patches. Hatched bars show the proportion of parasitoid eggs oviposited in SF mines, and

un®lled bars show the proportion of parasitoid eggs oviposited in TF mines, in each patch type.

Life-history timing in a leafmining moth 635

variance of SF duration occurred due to mortality during larval development in the caged patches.The signi®cant directional selection detected in exposed patches was thus caused by mortalityfactors that were excluded by the cages. Values of i calculated from subsets of the data with onlyparasitism or other mortality included were signi®cantly greater than zero (Table 1). This indicatesthat both parasitism mortality, and mortality from other unidenti®ed sources, contribute to di-rectional selection on SF duration in exposed patches. These other sources of mortality mayinclude host feeding by parasitoids on TF larvae.

Table 1. Selection statistics for within-generation survival versus SF duration in Phyllonorycter mespilella for

exposed patches and patches caged to exclude parasitoidsa

Patch type Data i j

Exposed (n=860) All 0.28*** )0.21*Parasitism 0.24*** )0.09Other 0.18** )0.23*

Caged (n=588) All 0.02 )0.01

aDirectional selection intensity, i, and variance selection intensity, j, were calculated for all data for each patch type, and for

data with only parasitism mortality, or other mortality, included. Statistics were tested for signi®cant di�erences from zero

using a two-tailed t-test for i and a two-tailed F-test for j: *P<0.05, **P<0.01, ***P<0.001.

Figure 3. Fitness function for within-generation survival versus SF duration of Phyllonorycter mespilella in allpatches exposed to parasitoids (n� 860). The solid curve is the ®tted cubic spline functions. The dashed lines

represent �1 S.E. of predicted values of the function from 200 bootstrap samples of the data.

636 McGregor

No signi®cant di�erences in directional selection intensity (i) could be detected among the patchtypes when values were compared among replicate patches (F� 0.78, d.f.� 2, P� 0.49). Similarly,no signi®cant in¯uence of patch type was detected on values of the variance selection statistic (j)when compared among patches (F� 3.18, d.f.� 2, P� 0.09). However, because of low sample size(i.e. only four replicates of each patch type), the power of these tests was low, and the chance ofType II error was high (b� 0.84 for i; b� 0.51 for j ). Further analysis will be presented for pooleddata for each patch type.Values of directional selection intensity (i) were positive and signi®cantly di�erent from zero

when calculated for pooled data for both the 13-h and mixed patches (Table 2). When i wascalculated for data with only parasitism mortality included, the value was signi®cantly greater thanzero for mixed patches, but not for 13-h patches (Table 2). When only other mortality was in-cluded, the value of i was signi®cantly greater than zero for 13-h patches, but not for mixed patches(Table 2). This indicates that the directional selection measured in mixed patches was caused byparasitism mortality, and that the selection in 13-h patches was caused by sources of mortalityother than parasitism. Values of directional selection intensity calculated from pooled data from16-h patches were not signi®cantly di�erent from zero (Table 2).Fitness functions for SF duration in the 13-h, mixed and 16-h patches show an increase in

survival probability with increasing values of SF duration (Fig. 4) similar to that observed in the®tness function for combined data for exposed patches (Fig. 3). The increase in survival withhigher SF duration is consistent with the detection of positive directional selection on SF durationin the 13-h and mixed patches. Although signi®cant directional selection on SF duration was notdetected in the 16-h patches, survival probability was still higher for individuals with longer SFdurations.Negative selection on the variance of SF duration was detected in mixed patches. Values of the

variance selection statistic, j, were negative and signi®cantly di�erent from zero when calculated forall of the data, or for data with only parasitism mortality included (Table 2). In 13-h patches, thevalue of j was only signi®cantly di�erent from zero when calculated from data with only othermortality included (Table 2). Because a signi®cant value of j was obtained only when calculatedfrom a reduced data set, this result will not be considered as evidence of selection on phenotypic

Table 2. Selection statistics for within-generation survival versus SF duration in Phyllonorycter mespilella for

exposed patches with manipulated development timesa

Patch type Data i j

13-h (n=399) All 0.28** )0.21Parasitism 0.18 0.003Other mortality 0.22* )0.27*

Mixed (n=241) All 0.33** )0.35*Parasitism 0.34** )0.37*Other mortality 0.16 )0.19

16-h (n=220) All 0.20 0.14

Parasitism 0.22 0.07Other mortality 0.06 0.10

aDirectional selection intensity, i, and variance selection intensity, j, were calculated for all data for each patch type, and for

data with only parasitism mortality, or other mortality, included. Statistics were tested for signi®cant di�erences from zero

using a two-tailed t-test for i and a two-tailed F-test for j: *P<0.05, **P<0.01, ***P<0.001.

Life-history timing in a leafmining moth 637

variance in 13-h patches. Values of j were not signi®cantly di�erent from zero for 16-h patches(Table 2).

Discussion

Positive directional selection on SF duration occurred on P. mespilella larvae in the Kelownapopulation during the diapausing generation. This result is consistent with a previous measurementof selection on SF duration from a population of P. mespilella in Summerland, British Columbia(McGregor, 1996). Selection on SF duration in this experiment was caused by mortality factors

Figure 4. Fitness functions for within-generation survival versus SF duration of Phyllonorycter mespilella in(a) 13-h patches (n� 399), (b) mixed patches (n� 241) and (c) 16-h patches (n� 220). The solid curves are the

®tted cubic spline functions. The dashed lines represent �1 S.E. of predicted values of the function from 200bootstrap samples of the data.

638 McGregor

that were excluded by cages, and which included both parasitoid oviposition and other unidenti®edsources of mortality. Oviposition by parasitoids was clearly one cause of selection on SF durationin exposed patches. Although phenotypic selection has been analysed for organisms of many taxa,the ecological factors that cause selection have rarely been clearly identi®ed (Wade and Kalisz,1990). This study is one of very few that show parasitoids to be causes of selection on traits in theirhosts (Rausher, 1992; Weis et al., 1992; McGregor, 1996).It is likely that parasitoids also contributed to the other sources of mortality that caused selection

on SF duration, through host feeding on TF larvae. Eulophid parasitoids of Phyllonorycter specieshost-feed predominantly on SF larvae (Askew and Shaw, 1979; Maier, 1982; van Driesche andTaub, 1983; Laing et al., 1986; Casas, 1989; Barrett and Brunner, 1990; Varela and Welter, 1992).However, in patches where TF larvae appear earlier, parasitoids may host-feed more frequently onTF larvae. Unfortunately, it was not possible to identify host feeding as a mortality source whenleafmines were dissected. One primary function of host feeding in parasitoids is to gather nutrientsrequired for egg maturation (Jervis and Kidd, 1986; Heimpel and Collier, 1996). Newly emergedfemale parasitoids may search for larvae primarily for host feeding and not oviposition. In thatcase, individuals that enter the TF stage at the earliest date (short SF duration; 13-h patches) mayexperience an increased risk of host feeding. Host feeding on the TF larvae that appear ®rst wouldcause mortality of individuals with the shortest SF durations, resulting in positive directionalselection.Di�erences in the pattern of parasitoid oviposition were observed among the patch types in this

experiment. The patch types varied in the time course of appearance of TF larvae, and thus in therelative frequencies of SF and TF larvae that occurred at particular times. Parasitoids searchingconcurrently in the three patch types presumably had di�erent encounter rates with SF and TFlarvae. Oviposition occurred on TF larvae more frequently when more TF larvae appeared earlierin the season (as in 13-h patches). When fewer TF larvae were present early in the season (as inmixed or 16-h patches), a larger proportion of parasitoid oviposition occurred on SF larvae.Increases in the rate of oviposition on SF larvae by Pnigalio and Sympiesis species have previ-

ously been observed during the diapausing generation of Phyllonorycter species (Barrett andBrunner, 1990; Varela and Welter, 1992; McGregor, 1996). Because TF larvae appear later duringthe fall generation (Maier, 1984; Laing et al., 1986; Barrett and Brunner, 1990; Varela and Welter,1992), a seasonal increase in acceptance of SF larvae for oviposition may occur in response to alower encounter rate with TF larvae. Seasonal changes in acceptance of SF larvae for ovipositionmay re¯ect an adaptive shift in larval-stage preference that is mediated by encounter rates with SFand TF larvae. Many parasitic insects alter oviposition preferences based on the availability ofdi�erent host types (Mangel and Roitberg, 1989; Roitberg et al., 1992, 1993; Li et al., 1993).Alternatively, parasitoid females may accept SF larvae more often in the fall simply because theyare the most abundant host type during the period of search.Although di�erences in the pattern of parasitoid oviposition occurred among patch types, no

di�erences in directional selection intensity could be detected. This result should be interpreted withsome caution considering the low statistical power of the ANOVA. However, the ®tness functionsfor data from all exposed patches, and for data pooled for each patch type, show the same basicpattern. Survival probability increases monotonically with increasing SF duration. When selectionwas analysed within patch types, signi®cant positive directional selection on SF duration wasdetected in both the 13-h and mixed patches, but not in 16-h patches. Evidently, di�erential survivaloccurred as a function of SF duration in the 16-h patches, but the change in mean SF duration thatresulted was insu�cient to allow detection of signi®cant directional selection.The causes of directional selection appeared to vary between the 13-h and mixed patch types.

Analysis of data sets with only parasitism or other mortality included indicated that selection in

Life-history timing in a leafmining moth 639

13-h patches was caused by sources of mortality other than parasitism, and selection in mixedpatches was caused by parasitism mortality. More TF larvae appeared at earlier dates in 13-hpatches. If host feeding by parasitoids occurs predominantly on the ®rst TF larvae to appear, asdiscussed above, then we would expect host feeding to be heaviest on TF larvae in 13-h patches,and to cause directional selection on SF duration. However, as no data on host feeding mortalityare available, this argument is entirely speculative. In contrast, selection in mixed patches appearsto be caused by heavier parasitoid oviposition on individuals with shorter SF durations. This isconsistent with the earlier appearance of TF larvae in mixed patches compared to 16-h patches,where no selection could be detected.

Conclusions

Directional selection on the timing of life-history transitions in P. mespilella was detected in thisstudy, and selection was shown, at least in part, to be caused by parasitoid oviposition. Se-lection was measured in patch types with manipulated distributions of SF duration, in anattempt to characterize the in¯uence of the phenotypic distribution of life-history timing on thepattern of parasitoid behaviour and the form of phenotypic selection. The pattern of parasitoidoviposition varied among patch types, but this variation in the pattern of behaviour did notresult in variation in the general form or intensity of selection. However, the causes of selectionappeared to vary among the two patch types where TF larvae appeared earliest (13-h andmixed patches), and signi®cant directional selection could not be detected in patches where TFlarvae appeared the latest (16-h patches). Further resolution of the relationships between life-history timing in P. mespilella, variation in parasitoid behaviour and the form of phenotypicselection will require additional empirical work using the experimental approach describedabove.

Acknowledgements

Thanks to B.J. Crespi and B.D. Roitberg for helpful comments on previous versions of themanuscript. Many thanks to Wally Rendell for expert technical assistance. Financial support forthis work came from the Science Council of British Columbia and the Natural Sciences andEngineering Research Council of Canada. Facilities and technical support were provided by theAgriculture Canada Research Station at Summerland, British Columbia.

References

Andersson, M. (1982) Female choice selects for extreme tail length in a widowbird. Nature 299, 818±820.Anholt, B.R. (1991) Measuring selection on a population of damsel¯ies with a manipulated phenotype.

Evolution 45, 1091±1106.Askew, R.R. and Shaw, M.R. (1979) Mortality factors a�ecting the leafmining stages of Phyllonorycter

(Lepidoptera: Gracillaridae) on oak and birch. II. Biology of the parasite species. Zool. J. Linn. Soc. 67,51±64.

Barrett, B.A. and Brunner, J.F. (1990) Types of parasitoid-induced host preferences, and sex ratios exhibitedby Pnigalio ¯avipes (Hymenoptera: Eulophidae) using Phyllonorycter elmaella (Lepidoptera: Gracillar-

idae) as a host. Environ. Entomol. 19, 803±807.Bernays, E. and Graham, M. (1988) On the evolution of host speci®city in phytophagous arthropods. Ecology

69, 886±892.Casas, J. (1989) Foraging behavior of a leafminer parasitoid in the ®eld. Ecol. Entomol. 14, 257±265.

640 McGregor

Cossentine, J.E. and Jensen, L.B. (1992) Establishment of Phyllonorycter mespilella (HuÈ bner) (Lepidoptera:Gracillaridae) and its parasitoid Pnigalio ¯avipes (Hymenoptera: Eulophidae) in fruit orchards in the

Okanagan and Similkameen Valleys of British Columbia. J. Entomol. Soc. British Columbia 89, 18±24.Crespi, B.J. (1990) Measuring the e�ect of natural selection on phenotypic interaction systems. Am. Nat. 135,

32±47.Dijkstra, C., Bijlsma, A., Daan, S., Meijer, T. and Zijltra, M. (1990) Brood size manipulations in the kestrel

(Falco tinnunculus): E�ects on o�spring and parent survival. J Anim. Ecol. 59, 269±285.E�ron, B. (1981) Nonparametric estimates of standard error: The jacknife, the bootstrap and other methods.

Biometrika 68, 589±599.Endler, J.A. (1986) Natural Selection in the Wild. Princeton University Press, Princeton, NJ.

Erdfelder, E., Faul, F. and Buchner, A. (1996) GPOWER: A general power analysis program. Behav. Res.Meth. Instruments Comput. 28, 1±11.

Godfray, H.C.J. (1994) Parasitoids: Behavioral and Evolutionary Ecology. Princeton University Press,Princeton, NJ.

Heimpel, G.E. and Collier, T.R. (1996) The evolution of host-feeding behaviour in insect parasitoids. Biol.Rev. 71, 373±400.

Je�ries, M.J. and Lawton, J.H. (1984) Enemy free space and the structure of ecological communities. Biol. J.Linn. Soc. 23, 269±286.

Jervis, M.A. and Kidd, N.A.C. (1986) Host-feeding strategies in Hymenopteran parasitoids. Biol. Rev. 61,395±434.

Ketterson, E.D. and Nolan, V. (1992) Hormones and life histories: An integrative approach. Am. Nat. 140,S33±S62.

Krebs, J.R. and McCleery, R.H. (1984) Optimization in behavioural ecology. In Behavioral Ecology: AnEvolutionary Approach, 2nd edn (J.R. Krebs and N.B. Davies, eds), pp. 91±121. Blackwell Scienti®c,

Oxford.Laing, J.E., Heraty, J.M. and Corrigan, J.E. (1986) Summer diapause in Phyllonorycter blancardella (Fabr.)

(Lepidoptera: Gracillaridae) in Ontario. Can. Entomol. 118, 17±28.Lande, R. and Arnold, S.J. (1983) The measurement of selection on correlated characters. Evolution 37, 1210±

1226.Lawton, J.H. (1986) The e�ects of parasitoids on phytophagous insect communities. In Insect Parasitoids (J.

Waage and D. Greathead, eds), pp. 265±289. Academic Press, New York.Li, C., Roitberg, B. and Mackauer, M. (1993) Patch residence time and parasitism of Aphelinus asychis: A

simulation model. Ecol. Model. 69, 227±241.Maier, C.T. (1982) Parasitism of the apple blotch leafminer Phyllonorycter crataegella, on sprayed and un-

sprayed apple trees in Connecticut. Environ. Entomol. 11, 603±610.Maier, C.T. (1984) Seasonal development and ¯ight activity of the spotted tentiform leafminer, Phyllonorycter

blancardella (Lepidoptera: Gracillaridae), in Connecticut. Can. Entomol. 116, 435±441.Mangel, M. and Roitberg, B.D. (1989) Dynamic information and host acceptance by a tephritid ¯y. Ecol.

Entomol. 14, 181±189.McGregor, R. (1996) Phenotypic selection by parasitoids on the timing of life history in a leafmining moth.

Evolution 50, 1579±1584.McGregor, R. (1997) The in¯uence of photoperiod on larval development in the leafmining moth Phyllo-

norycter mespilella (Lepidoptera: Gracillaridae). Ann. Entomol. Soc. Am. 90, 333±336.Mitchell-Olds, T. and Shaw, R.G. (1987) Regression analysis of natural selection: Inference and biological

interpretation. Evolution 41, 1149±1161.Mùller, A.P. (1994) Sexual Selection and the Barn Swallow. Oxford University Press, Oxford.

Murdoch, W.W., Avery, S. and Smyth, M.E.B. (1975) Switching in predatory ®sh. Ecology 56, 1094±1105.Pottinger, R.P. and Leroux, E.J. (1971) The biology and dynamics of Lithocolletis (Phyllonorycter) blancar-

della (Lepidoptera: Gracillaridae) on apple in Quebec. Memoirs Entomol. Soc. Can. 77, 1±437.Price, P.W., Bouton, C.E., Gross, P., McPherson, B.A., Thompson, J.N. and Weis, A.E. (1980) Interactions

among three trophic levels: In¯uence of plants on interactions between insect herbivores and naturalenemies. Ann. Rev. Ecol. Syst. 11, 41±65.

Life-history timing in a leafmining moth 641

Rausher, M.D. (1992) Natural selection and the evolution of plant±insect interactions. In Insect ChemicalEcology: An Evolutionary Approach (B.D. Roitberg and M.B. Isman, eds), pp. 20±88. Chapman & Hall,

New York.Roitberg, B.D., Mangel, M., Lalonde, R., Roitberg, C., van Alphen, J. and Vet, L. (1992) Seasonal dynamic

shifts in patch exploitation by parasitic wasps. Behav. Ecol. 3, 156±165.Roitberg, B.D., Sircom, J., Roitberg, C., van Alphen, J. and Mangel, M. (1993) Life expectancy and repro-

duction. Nature 364, 108.Rowe, L. and Ludwig, D. (1990) Size and timing of metamorphosis in complex life cycles: Time constraints

and variation. Ecology 72, 413±427.Schluter, D. (1988) Estimating the form of natural selection on a quantitative trait. Evolution 42, 849±861.

Schluter, D. (1994) Experimental evidence that competition promotes divergence in adaptive radiation. Sci-ence 266, 798±800.

Shettleworth, S. (1984) Learning and behavioral ecology. In Behavioural Ecology: An Evolutionary Approach,2nd edn (J.R. Krebs and N.B. Davies, eds), pp. 170±194. Blackwell Scienti®c, Oxford.

Sinervo, B., Doughty, P., Huey, R.B. and Zamudio, K. (1992) Allometric engineering: A causal analysis ofnatural selection on o�spring size. Science 258, 1927±1929.

Stephens, D.W. and Krebs, J.R. (1986) Foraging Theory. Princeton University Press, Princeton, NJ.Strand, M.R. (1986) The physiological interactions of parasitoids with their hosts and their in¯uence on

reproductive strategies. In Insect Parasitoids (J. Waage and D. Greathead, eds), pp. 97±136. AcademicPress, New York.

van Alphen, J.J.M. and Vet, L.E.M (1986) An evolutionary approach to host ®nding and selection. In InsectParasitoids (J. Waage and D. Greathead, eds), pp. 23±62. Academic Press, New York.

van Driesche, R.G. and Taub, G. (1983) Impact of parasitoids on Phyllonorycter leafminers infesting apple inMassachusetts, U.S.A. Protect. Ecol. 5, 303±317.

Varela, L.G. and Welter, S.C. (1992) Parasitoids of the leafminer Phyllonorycter nr. elmaella (Lepidoptera:Gracillaridae) on apple in California: Abundance, impact on leafminer, and insecticide-induced mor-

tality. Biological Control 2, 124±130.Vinson, S.B. (1976) Host selection by insect parasitoids. Ann. Rev. Entomol. 21, 109±133.

Vinson, S.B. and Iwantsch, G.F. (1980) Host suitability for insect parasitoids. Ann. Rev. Entomol. 25, 397±419.Wade, M.J. and Kalisz, S. (1990) The causes of natural selection. Evolution 44, 1947±1955.

Weis, A.E., Abrahamson, W.G. and Andersen, M.C. (1992) Variable selection on Eurosta's gall size I: Theextent and nature of variation in phenotypic selection. Evolution 46, 1674±1697.

Wilkinson, L., Hill, M., Welna, J.P. and Birkenbeuel, G.K (1992) SYSTAT for Windows, Version 5. SYSTAT,Evanston, IL.

642 McGregor