International Association for Ecology

Variation in Selection Pressures on the Goldenrod Gall Fly and the CompetitiveInteractions of Its Natural EnemiesAuthor(s): Warren G. Abrahamson, Joan F. Sattler, Kenneth D. McCrea and Arthur E.WeisSource: Oecologia, Vol. 79, No. 1 (1989), pp. 15-22Published by: Springer in cooperation with International Association for EcologyStable URL: https://www.jstor.org/stable/4218915Accessed: 02-05-2019 15:56 UTC

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide

range of content in a trusted digital archive. We use information technology and tools to increase productivity and

facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at

https://about.jstor.org/terms

Springer, International Association for Ecology are collaborating with JSTOR to digitize,preserve and extend access to Oecologia

This content downloaded from 64.107.81.200 on Thu, 02 May 2019 15:56:48 UTCAll use subject to https://about.jstor.org/terms

Oecologia (1989) 79:15-22 ( J a ? Springer-Verlag 1989

Variation in selection pressures on the goldenrod gall fly

and the competitive interactions of its natural enemies

Warren G. Abrahamson, Joan F. Sattler, Kenneth D. McCrea, and Arthur E. Weis*

Department of Biology, Bucknell University, Lewisburg, PA 17837, USA

Summary. Larvae of the tephntid fly Eurosta solidaginis induce ball-shaped galls on the stem of tall goldenrod, Soli- dago altissima. Survival probability depends on gall size; in small galls the larva is vulnerable to parasitoid oviposi- tion, whereas larvae in large galls are more frequently eaten by avian predators. Fly populations from 20 natural old fields in central Pennsylvania were monitored in 1983 and 1984 to examine the distribution of the selection intensity imposed by natural enemies, the parasitoids Eurytoma gi- gantea and E. obtusiventris, the inquiline Mordellistena un- icolor, and the predatory birds Dendrocopus pubescens and Parus atricapillus. Mordellistena and E. obtusiventris are able to attack galls of all diameters while E. gigantea and the predatory birds preferentially assaulted small and large diameter galls, respectively. Eurosta in intermediate sized galls had the highest survivorship, hence selection had a stabilizing component. However, parasitoid attack was more frequent than bird attack, and the two did not exactly balance, thus there was also a directional component. The mean directional selection intensity on gall size was 0.21 standard deviations of the mean, indicating that larger gall size was favored. Interactions among the insect members of the Eurosta natural enemy guild are complex and fre- quent.

Key words: Bird predation - Eurosta solidaginis - Parasi- toids - Selection pressures - Solidago altissima

Herbivore populations typically sustain heavy attack from intricate arrays of natural enemies (Bergman and Tingey 1979; Price et al. 1980). Gallmakers, for example, fre- quently support complex guilds of parasitoid wasps (Varley 1947; Askew 1961: Price and Clancy 1986) and avian preda- tors (Spofford 1977; Schlichter 1978; Confer and Paicos 1985). Although many studies have described the interac- tions of natural enemy guilds with their gallmakers, we know of no studies, except those of Weis and Abrahamson (1986) and Price and Clancy (1986), that have measured the selection intensities on gallmakers that are created by these enemies. The study reported here provides such selec- tion information for 20 old fields in central Pennsylvania.

The natural enemies of a galimaker clearly influence the survival of individual gallmakers. If survivorship varies among gallmakers with different traits, natural enemy at- tack can potentially alter gallmaker traits in subsequent generations. The relationships of E. solidaginis with its nat- ural enemies have been described in a number of studies

(e.g., Fitch 1855; Harrington 1895; Beck 1947; Uhler 1951; Judd 1953; Stinner and Abrahamson 1979; Abrahamson et al. 1983), however, none of these efforts determined the selection pressures these enemies create. The study reported here expands on these earlier works by (1) measuring the selection pressures on gall size that are caused by natural enemy attack and (2) examining the potential for competi- tive interactions among the guild of insect natural enemies.

The probability of E. solidaginis survival depends on gall size since at least some of the gallmaker's natural ene- mies differentially attack galls of various sizes (Milne 1940; Miller 1959; Uhler 1961; Cane and Kurczewski 1976; Weis and Abrahamson 1986). Gallmakers in small galls have been shown to be more vulnerable to oviposition by the parasitoid wasp E. gigantea (Weis and Abrahamson 1985; Weis et al. 1985) and large galls are more frequently at- tacked by downy woodpeckers and black-capped chicka- dees (Confer and Paicos 1985; Weis and Abrahamson 1986). While Abrahamson et al. (1983) studied the insect natural enemy guild as a whole and reported its impact on gallmaker survival, the present study examines the im- pacts of the individual species comprising the guild (both insect and avian) and measures the selection pressure each exerts on the gallmaker.

Although the gall is plant tissue, aspects of the gall phe- notype are likely influenced by the insect. A genetically coded stimulus from the insect induces the development of a gall from normal tissue (Weis and Abrahamson 1986; Abrahamson and Weis 1987; Carango et al. 1988). Since some of the gallmaker's natural enemies differentially at- tack galls of various diameters, the natural enemies should exert selection pressure on the gallmaker that could influ- ence the evolution of the gallmaker's contribution to gall diameter (Weis and Abrahamson 1985). We have deter- mined Eurosta mortality in relation to gall size and natural enemy attack, and thus we are able to measure selection intensities. Specifically, we asked the questions: (1) What are the abundances of the gallmaker and its insect natural enemies? (2) To what extent is Eurosta mortality caused by insect and bird enemies? (3) Which natural enemies ex- hibit gall size preferences, what are these preferences, and

* Current address: Department of Biological Sciences, Northern Illinois University, DeKalb, IL 60115, USA

Offprint requests to: W.G. Abrahamson

This content downloaded from 64.107.81.200 on Thu, 02 May 2019 15:56:48 UTCAll use subject to https://about.jstor.org/terms

16

what selection pressures result?, and (4) To what extent

do the insect natural enemies compete for the gallmaker?

Natural history of the system

The univoltine gallmaking fly, Eurosta solidaginis (Fitch; Diptera: Tephritidae) causes the tall goldenrod (Solidago altissima L., Compositae) to develop a globose, tumor-like stem growth. Oviposition occurs in mid- to late May in central Pennsylvania. The gall appears about 3 wk after oviposition and reaches full size in another 3 to 4 wk (Weis and Abrahamson 1985). The larva enters diapause in late September and pupates within the gall during the following March or April (Uhler 1951).

Eurytoma obtusiventris (Gahan; Hymenoptera: Euryto- midae) is an internal parasite of Eurosia (Uhler 1951). Para- sitism most likely occurs while the galimaker eggs are hatch- ing and the larvae are boring through the stem (Abraham- son unpubl. data). When attacked, the gailmaker forms a premature puparium in late summer after which the para- sitoid consumes the host. E. obtusiventris larvae remain in- side the host's puparium throughout winter and pupate the following spring (Weis and Abrahamson 1985).

Eurytoma gigantea (Walsh; Hymenoptera: Eurytomi- dae) is an external parasite that attacks Eurosta after maxi- mum gall size has been reached (Weis and Abrahamson 1985). It consumes the gallmaker by the end of August and remains in the central chamber of the gall to feed on plant tissues (Uhler 1951). E. gigantea females probe galls of all sizes but can only inject eggs when the gall wall is thinner than the length of their ovipositor. This parasitoid is therefore limited to smaller galls (Weis and Abrahamson 1985; Weis et al. 1985).

Mordellistena unicolor (Lec; Coleoptera: Mordellidae) feeds on gall tissues by chewing narrow channels through gall parenchyma and vascular regions. Larvae, which hatch from eggs oviposited on the gall surface in early July (Weis and Abrahamson 1985), bore into the gall tissue by early August. Although Mordellistena is considered an inquiline (Uhler 1951), it usually eats the gallmaker by the end of the growing season.

Two species of birds attack Eurosta during winter months, the downy woodpecker, Dendrocopus pubescens, and the black-capped chickadee, Parus atricapillus (Schlichter 1978). The degree of mortality caused by these birds is highly variable from site to site and year to year (Weis et al. unpublished data). Milne (1940), for example, found that bird predation accounted for 44.8% of total Eurosta mortality whereas Miller (1959) and Uhler (1961) showed bird predation to be much less at 7% and 2.3% of total gallmaker population, respectively.

Since the natural enemies attack the same resource but at different times, there are predatory relationships among them. Although E. obtusiventris attack occurs first, E. gi- gantea and M. unicolor larvae incidentally consume E. obtu- siventris when they attack the gallmaker. Similarly, Mordell- istena larvae typically consume E. gigantea larvae if the inquilines find their way to the gall's central chamber. Pre- dation of the parasitoids by birds has also been document- ed. Cane and Kurczewski (1976) found that the larvae of F. gigantea were taken, but suggested that E. obtusiventris might be distasteful as it was consistently avoided. Schlichter (1978) found the Mordellistena larvae were like- wise not taken by birds.

Methods

Patterns of gallmaker and enemy abundance were examined in 20 natural old fields of varying stages of succession with- in a 15 km radius of Lewisburg, PA USA (40057' N, 760 53' W). Analysis of gall abundance was done by system- atically placing 50 (0.5 x 1 m) quadrats in an approximately

equidistant grid across each field. Fields were censused by counting all ball galls in each quadrat between 17 August and 29 September 1983. During the periods of 12 No- vember-I 1 December 1983 and 7-28 April 1984, a mini- mum of 97 galls were systematically collected from each of the 20 fields. Galls from both autumn and spring collec-

tions were measured to the nearest mm by passing them through circle stencils and then galls were dissected to deter- mine their contents (i.e., gallmaker, parasitoid, or inqui- line). Successful bird predation was recorded when beak chisel holes penetrated to the central chamber of an empty gall. Chisel holes to the central chamber with evidence of Mordellistena presence were not recorded as bird success since the Mordellistena could have consumed the gallmaker prior to the bird's attack (unknown bird success). Galls with chisel holes not reaching the central chamber and still containing Eurosta or a parasitoid were recorded as unsuc- cessful bird predation. Evidence of an exit hole of adult E. gigantea, Mordellistena tunnels, gallmaker death, or empty galls was also recorded.

To determine the extent of mortality caused by each natural enemy, percentages of surviving Eurosta, parasi- toids, inquiline, and galls successfully attacked by birds were calculated. This was done for both November and April data based on the percentages found at dissection after excluding galls containing dead Eurosta and empty galls. The November data were used to determine gallmaker and enemy occurrences before predation by birds and the April data were compared to November data to assess natu- ral mortality over winter and to determine bird predation levels. Galls are primarily a winter food resource for these birds. April data were used to assess the survival at the end of Eurosta's annual life cycle.

The gall diameters for each content class were approxi- mately normally distributed, but their variances were not equal. This prohibited the use of a parametric one-way an- ova or multiple comparison test such as the Student-New- man-Keuls procedure. We therefore used the Kruskal-Wal- lis non-parametric one-way anova, and multiple Student's t-tests using separate variance estimates, correcting to give an experiment-wise error rate of P <0.05 (Sokal and Rohlf 1981).

Selection intensities on gall size exerted by natural en- emy attack were calculated as the difference between the mean gall diameter of the selected individuals and the mean gall diameter of the entire population, divided by the popu- lation standard deviation (Falconer 1981). This index is a measure of the magnitude and direction of a selective pressure but not of the evolutionary response to that pres- sure. We will separately present the results of a five-year study in which we measured the evolutionary responses to selection pressures (Weis et al., unpublished work).

The standard statistical tests used to determine if a sin- gle selection intensity is different from zero, or if several selection intensities significantly differ from one another, are flawed when assessing mortality selection. A case in point is the regression test based on the discovery by Price

This content downloaded from 64.107.81.200 on Thu, 02 May 2019 15:56:48 UTCAll use subject to https://about.jstor.org/terms

17

Table 1. Density of ball galls based on quadrat sampling and percentages of galls (excluding empty galls and those in which no living occupant was found, D/E) containing living Eurosta solidaginis (Eur.), Mordellistena unicolor (Mord.), Eurytoma gigantea (E. gig.), Eurytoma obtusiventris (E. obt.) or which were successfully attacked by birds (bird). N is the number of galls dissected

Field Gall N % of galls by content in November N % of galls by content in April density

per m2 % Excluding dead or empty galls % Excluding dead or empty galls D/E D/E

Eur. Mord. E. gig. E. obt. Bird. Eur. Mord. E. gig. E. obt. Bird

BVSA < 0.04 99 39 23 27 28 22 0 100 64 25 28 22 25 0 147R 0.08 100 56 50 36 9 5 0 100 57 56 26 9 2 7 Furn 0.20 100 57 49 30 21 0 0 100 53 36 30 32 0 2 Pit 0.28 100 49 71 20 4 6 0 100 51 37 14 18 0 31 147E 0.36 100 40 70 18 12 0 0 97 58 66 15 20 0 0 Marsh 0.36 105 44 51 14 24 0 12 100 32 34 9 7 0 50 Hess 0.40 99 40 75 10 1 0 0 5 101 49 1 3 6 13 4 63 Beag 0.44 100 50 44 42 12 2 0 100 52 38 33 21 8 0 Violet 0.52 145 30 36 39 20 6 0 109 59 49 16 24 9 2 MP 0.72 100 36 47 8 16 2 28 100 29 38 21 7 0 34 Ind park 0.84 118 42 48 20 17 13 1 114 54 27 25 15 12 21 Pott 1.16 100 34 77 9 9 5 0 100 51 65 20 12 0 2 Owenl 1.16 100 38 63 15 18 5 0 100 41 71 12 15 2 0 254 1.24 100 47 53 21 11 2 13 100 50 46 22 8 2 22 147S 1.68 100 49 57 20 22 2 0 100 43 40 14 5 0 40 Aikley 1.80 100 53 34 34 13 19 0 287 48 49 23 15 13 1 Stein 2.04 252 33 15 55 15 14 0 147 58 27 31 23 18 2 Allen 2.08 100 25 53 33 13 0 0 99 39 62 20 8 3 7 Owen 2 3.88 284 40 63 17 13 7 1 100 56 80 9 7 5 0 KF 4.00 100 48 71 15 13 0 0 99 46 77 11 8 0 4

All fields 2402 42 50 26 15 6 3 2253 50 47 19 14 6 15 combined

(1970) that the regression coefficient or relative fitness over phenotypic score was algebraically equivalent to the selec- tion intensity. However, when a fitness component with binary values (i.e., survived vs. died) is used, the basic as- sumption of a normally distributed dependent variable is violated and test results invalid. Endler (1987) offered a modified two sample t-test to determine if the difference between the mean phenotype of the entire population dif- fered from the mean phenotype of the selected subset of the population. When applied to longitudinal studies such as our's, it violates the assumption of independence of the two samples because the selected individuals are included in both samples. This gives the test a conservative bias (i.e., real selection will too often go undetected). For these rea- sons we estimated confidence limits on the selection intensi- ties measured in each of the 20 populations by a bootstrap- ping routine.

Bootstrapping is a technique used to derive the error distribution of a statistic by intensive resampling of an em- pirical data set (Efron 1981). A computer program is used to draw a "sample" of observations, with replacement, from the data set and then compute the desired statistic. When repeated many times, this procedure yields a distribu- tion of values for the statistic, from which confidence inter- vals can be derived. In this study a single data set consisted of the trivariate distribution of gall diameter (in mm), sur- vival from parasite attack (valued as 0 or 1) and survival from bird attack (also valued as 0 or 1) for one of the fields. One thousand "samples" were drawn from this dis- tribution, and the selection intensity for the parasite and bird episodes of selection, as well as total directional selec-

tion, were calculated for each " sample." We considered the 95% confidence interval to extend from the 2.5th to the 97.5th percentile of the distribution of the intensities from the 1000 "samples." The whole procedure was re- peated for each of the 20 populations. An observed selec- tion intensity was considered significant if its lower confi- dence limit was greater than zero, and two selection intensi- ties were considered significantly different if their 95% con- fidence intervals did not overlap.

Results

Gallmaker mortality sources

The density of Eurosta galls and the attack rate of each natural enemy were highly variable from field to field for both the November and April collections (Table 1). In No- vember, 29.3% of all galls (i.e., not excluding dead or empty galls) contained surviving Eurosta as compared to only 23.6% in April. This difference was significant (X2 = 19.30, P<0.001) and was primarily due to winter exploitation of galls by birds. After exclusion of dead or empty galls, galls containing Eurosta larvae accounted for as few as 15% to as many as 77% of the remaining galls in the November sample and as few as 13% to as many as 80% in the April sample. Eurosta mortality resulting from successful bird at- tacks was only 1.5% of all galls in November but increased to 7.2% by April (X2=91.74, P<0.001). Excluding dead or empty galls, successful bird attacks accounted for 3% of the mortality in November and 15% by April.

This content downloaded from 64.107.81.200 on Thu, 02 May 2019 15:56:48 UTCAll use subject to https://about.jstor.org/terms

18

Table 2. Results of a Kruskal-Wallis non-parametric one-way anova for mean gall diameter by field for all galls and each content; and results of a one-way anova on gall diameter by content (twenty fields combined) at two sampling times. All analyses exclude empty galls and those in which no living occupant was found. Those means followed by the same letter are not significantly different from one another at P = 0.05 (see methods for details of analysis)

November April

d.f. H Sig. of H d.f. H Sig. of H

Kruskal-Wallis one-way anova - gall diameter by field for:

all galls 19 146.0 <0.001 19 178.9 <0.001 Eurosta solidaginis 19 121.4 <0.001 19 93.7 <0.001 Mordellistena unicolor 19 63.8 <0.001 19 32.9 0.024 Eurytoma gigantea 19 31.5 0.036 19 44.3 <0.001 Eurytoma obtusiventris 13 23.3 0.039 11 25.3 0.008 Bird attacked galls 5 6.1 NS 14 56.5 <0.001

Kruskal-Wallis one-way anova - diameter by:

Content 4 303.0 < 0.001 4 235.7 < 0.001

Mean gall diameters (mm + std. dev.)

Eurytoma gigantea 17.81 + 2.78 a 17.25 + 2.73 a Mordellistena unicolor 19.81 + 3.52 b 19.97+ 3.32 b Eurosta solidaginis 21.63 + 2.32 c 21.07 + 2.39 c Eurytoma obtusiventris 21.80 + 2.40 c 21.52 + 1.99 c, d Bird attacked galls 22.62 + 1.88 c 21.83 + 2.31 d

Gallmaker mortality as influenced by gall size

Eurytoma gigantea was restricted to galls of a particular size range while the other insect enemies were not. This conclusion results from a one-way anova, performed on the November data set for gall diameters by gall content, that showed that the diameters of galls with surviving Eur- osta, M. unicolor, and E. obtusiventris varied significantly among the fields while the diameters of E. gigantea attacked galls did not.

In order to determine the relationships between natural enemies and gall sizes, we used the multiple comparison procedure on the November data to separate galls by con- tents into three significantly different subsets of gall diame- ter. Subset 1 consisted of galls containing E. gigantea (X= 17.81 mm); subset 2 consisted of galls containing Mordell- istena (X= 19.81 mm); subset 3 consisted of galls including both surviving Eurosta (X= 21.63 mm) and E. obtusiventris (X== 21.80 mm). These results indicate that attack by some natural enemies is gall size specific. Student's t-tests com- paring the mean gall diameters of the surviving gallmakers, parasitoids, and the inquilines with the mean gall diameter of the remainder of the population (all 20 fields combined) showed that the diameters of galls containing surviving Eur- osta and E. obtusiventris were significantly larger than the galls not containing the respective insect (t = 12.99, P < 0.001; t=4.66, P<0.001, respectively); even though E. ob- tusiventris can attack galls of all diameters. Galls containing E. gigantea and Mordellistena were significantly smaller than galls not containing the respective insect (I = 15.48, P<0.001; t=5.36, P<0.001, respectively); even though Mordellistena can attack galls of all diameters.

Birds preferentially attacked larger galls where the prob- ability of finding the gallmaker larva was highest. The den- sity of galls successfully attacked by birds was positively correlated with the population mean gall diameter (r= 0.448, P = 0.025) indicating that levels of bird attack were

higher in fields with more large galls. Total bird-attacked galls (galls categorized as successfully attacked, unsuccess- fully attacked, and unknown success of bird-attack) did not correlate with gall density. A Kruskal-Wallis nonpara- metric one-way anova performed on the April data set showed that the mean diameter of galls successfully at- tacked by birds varied significantly among the fields (Ta- ble 2). A multiple Student's t-test procedure found that suc- cessful bird attacked galls belonged to a mean diameter subset significantly larger than the galls containing E. gigan- tea, Mordellistena, surviving Eurosta, and the mean diame- ter of galls not attacked by birds (latter tested with Stu- dent's t-test, t=7.88, P= 0.01).

Selection intensities

The selection intensities created by Eurosta mortality due to the various natural enemies varied considerably from field to field (Table 3). For instance, in the Route 254 field, we found very small directional selection intensities because the upward selection by E. gigantea was balanced by the downward selection by birds. Stabilizing selection predo- minated in this field. However, the Furnace Road field had fairly strong upward, directional selection as a result of weak downward bird selection. The Marsh Road field had comparatively weak selection by parasitoids but relatively strong selection by birds. The net result in this latter case was significant negative, directional selection. In other words, the bird attack was so strong in the Marsh Road field that smaller gall size was favored. In this latter case there is no stabilizing selection because the variance of the survivors is the same as the variance of the entire field population that starts the generation.

Considered over all fields, the mean diameter of galls with surviving Eurosta was larger than the mean of all galls due to mortality caused by the parasitoid, E. gigantea (Fig. 1). This mortality in small galls caused an upward

This content downloaded from 64.107.81.200 on Thu, 02 May 2019 15:56:48 UTCAll use subject to https://about.jstor.org/terms

19

Table 3. Population mean gall diameters (excluding empty galls and those in which no living occupant was found), surviving Eurosta mean gall diameters, and selection intensities (S.I., in units of standard deviation of mean) exerted on gall diameter by insect natural enemies (Mordellistena unicolor, Eurytoma gigantea, Eurytoma obtusiventris) and birds. Significance of selection intensities determined from bootstrapped 95% confidence intervals (95% C.l.)

Field Mean gall diameters (mm + s.d.) Selection intensities from mortality due to:

Original Surviving Insects Birds All combined population Eurosta

S.I. 95% C.I. S.I. 95% C.l. S.!. 95% C.I.

November

BVSA 21.78+2.97 23.86+1.83 0.70* (0.41, 1.03) 0 (0, 0) 0.70* (0.41, 1.03) 147R 19.45+ 2.65 20.18+2.42 0.28 (-0.002, 0.55) 0 (0, 0) 0.28 (-0.002, 0.55) Furn 18.64+3.53 20.38+3.07 0.49* (0.24, 0.75) 0 (0, 0) 0.49* (0.24, 0.75) Pit 21.00+2.51 20.97+2.38 -0.01 (-0.20, 0.17) 0 (0, 0) -0.01 (-0.20, 0.17) 147E 20.62+2.65 21.50+1.71 0.33* (0.16,0.50) 0 (0,0) 0.33* (0.16,0.50) Marsh 19.69+3.45 20.33+2.54 0.30* (0.11, 0.51) -0.10* (-0.21, -0.03) 0.19 (-0.06, 0.40) Hess 20.07+ 2.20 20.14+1.70 0.23* (0.08, 0.38) -0.07 (-0.16, 0) 0.15 (-0.04, 0.33) Beag 19.44+2.94 21 .23+2.05 0.61 * (0.37, 0.84) 0 (0, 0) 0.61 * (0.37, 0.84) Violet 21.28+2.90 22.22+1.31 0.32* (0.12, 0.54) 0 (0, 0) 0.32* (0.12, 0.54) MP 21.08+ 2.81 21.03+2.43 0.26* (0.10, 0.43) -0.28* (-0.46, -0.14) -0.02 (-0.30, 0.25) Ind Park 19.25+2.83 19.88+2.47 0.24 (-0.003, 0.46) -0.01 (-0.05, 0) 0.22 (-0.03, 0.45) Pott 21.94+ 2.42 22.51 + 1.97 0.24 * (0.09, 0.38) 0 (0, 0) 0.24 * (0.09, 0.38) Owen 1 20.36+2.60 21.23+1.68 0.33* (0.16, 0.49) 0 (0, 0) 0.33* (0.16, 0.49) 254 21.51 + 2.61 22.21 + 2.41 0.28* (0.07, 0.48) -0.05 (-0.14, 0.02) 0.27* (0.03, 0.53) 147S 20.86+ 2.41 21.90+ 2.01 0.43* (0.22, 0.65) 0 (0, 0) 0.43* (0.22, 0.65) Aikley 21.06+3.47 22.69+ 3.11 0.47* (0.15, 0.80) 0 (0, 0) 0.47* (0.15, 0.80) Stein 19.17+3.86 21.73+2.75 0.66* (0.41, 0.92) 0 (0, 0) 0.66* (0.41, 0.92) Allen 21.69+2.80 22.67+ 1.77 0.35* (0.17, 0.56) 0 (0, 0) 0.35* (0.17, 0.56) Owen 2 21.73+2.79 22.60+ 2.17 0.29* (0.18, 0.41) 0.01 (0, 0.03) 0.31 * (0.20, 0.42) KF 20.88+2.18 21.46+1.80 0.27* (0.07, 0.46) 0 (0, 0) 0.27* (0.07, 0.46)

All fields combined 20.62+ 3.10 21.63+ 2.32 0.34* (0.30, 0.39) -0.02* (-0.02, -0.01) 0.33* (0.28, 0.37)

April

BVSA 21.11 + 2.97 22.33+2.40 0.41 (-0.10, 0.87) 0 (0, 0) 0.41 (-0.10, 0.87) 147R 21.28+2.35 21.79+ 1.86 0.34* (0.14, 0.56) -0.10 (-0.22, 0) 0.22 (-0.04, -0.47) Furn 18.68+3.02 20.29+2.76 0.57* (0.28, 0.86) -0.02 (-0.08, 0) 0.53* (0.22, 0.84) Pit 19.71 +2.80 20.22+2.02 0.29* (0.07, 0.50) -0.18* (-0.35, -0.02) 0.18 (-0.13, 0.49) 147E 18.61 + 2.54 19.59+ 2.54 0.39* (0.20, 0.60) 0 (0, 0) 0.39* (0.20, 0.60) Marsh 21.35+2.82 20.30+1.94 0.21 * (0.09, 0.34) -0.60* (-0.83, -0.41) -0.37* (-0.72, -0.07) Hess 19.67+ 3.04 20.00+2.08 0.16 (-0.05, 0.34) -0.29 (-0.17, 0.11) 0.11 (-0.42, 0.60) Beag 19.08+ 3.02 20.61 + 2.45 0.51 * (0.22, 0.83) 0 (0, 0) 0.51 * (0.22, 0.83) Violet 18.73+2.83 19.36+ 2.50 0.30* (0.01, 0.61) -0.04 (-0.14, 0) 0.22 (-0.06, 0.56) MP 20.54+3.09 20.81+2.40 0.23* (0.06, 0.37) -0.19* (-0.35, -0.06) 0.09 (-0.18, 0.35) Ind Park 19.13 +2.39 19.36+ 2.13 0.33* (0.09, 0.61) -0.17* (-0.33, -0.05) 0.10 (-0.29, 0.52) Pott 20.73+2.76 21.62+ 1.95 0.33* (0.12, 0.52) -0.01 (-0.04, 0) 0.32* (0.11, 0.51) Owen 1 20.02+ 2.07 20.50+ 1.76 0.23* (0.07, 0.41) 0 (0, 0) 0.23* (0.07, 0.41) 254 20.38+ 2.86 19.91 + 2.41 -0.01 (-0.25, 0.20) -0.08 (-0.23, 0.02) -0.16 (-0.47, 0.11) 147S 22.11 + 2.95 21.83+2.95 0.13 (-0.04, 0.29) -0.23* (-0.45, -0.04) -0.09 (-0.47, 0.19) Aikley 21.57+2.47 21.10+2.04 0.22* (0.06, 0.38) -0.01 (-0.01, 0) 0.21 * (0.05, 0.37) Stein 19.34+ 3.17 20.94+ 1.89 0.58* (0.32, 0.84) -0.03 (-0.10, 0) 0.50* (0.23, 0.78) Allen 21.83+2.71 22.22+ 2.32 0.21 * (0.02, 0.37) -0.06* (-0.12, -0.01) 0.14 (-0.08, 0.33) Owen 2 22.25+2.47 22.46+2.19 0.09 (-0.11, 0.29) 0 (0, 0) 0.09 (-0.11, 0.29) KF 10.15+ 3.05 20.85+ 2.67 0.23* (0.05, 0.38) 0.01 (-0.01, 0.02) 0.23* (0.06, 0.40)

All fields combined 20.45+2.97 21.07+2.39 0.27* (0.22, 0.31) -0.08* (-0.10, -0.06) 0.21 * (0.15, 0.27)

* Significantly different from zero at P?0.05

shift in the mean diameter of galls containing surviving Eurosta which was partially counter-balanced by bird in- duced selection favoring Eurosta in smaller diameter galls. Eurosta in intermediate sized galls had the highest survivor- ship yielding a stabilizing component to selection. However, because E. gigantea attack was typically more frequent than bird attack and since the two mortality sources did not exactly balance, there was also a directional selection com-

ponent in most fields (Table 3). The net selection pressure from all natural enemies on the gallmaker resulted in an upward selection differential on mean diameter for galls containing surviving Eurosta of 0.21 standard deviations of the mean for the 1983-84 season.

Figure 2 illustrates the percentage of galls attacked by E. obtusiventris is shifted to the right because some of the smaller galls that were occupied by E. obtusiventris were

This content downloaded from 64.107.81.200 on Thu, 02 May 2019 15:56:48 UTCAll use subject to https://about.jstor.org/terms

20

30 All Galls 100 Eurosta solidaginis

25 5 X 85.36 5 8-80 p (0.001

0 2_

a 0- 5 60 15

40

5 20

0~~~~~~~~~~~2

20 | | | f c20 , ,_

100 Mordellistena unicolor 100 Eurytoma obtusiventris

O5 X = 29.38 X x = 16.02 X 80- p (0.001 80 p < 0.003

- 60 - 60-

20 20

+U0 40 _

obuienrs Euryttaiomatggatean gal Birckdsb ids h

a2 20 - C20 U

40-

u)E0 u.yoaggne 400id 0 a-

(1 4161 0 22 >5(4 41 1 0222 2

Fi.1 aldaetr(m ysz clasfralgls;glscnan ing suvvnCuot oiaii,Mrelseauioo,Ertm -obtsvnti,Euyoa iate;ad al atcedb6irs0h

X2 value shown ompare agiven sbset to ll gall

Mordellistena unicolor

,, 60- v)

60

C) 50 M November N April

o 40

-~30

e 20 N I, 0Z --I 1 1 1 Q.- 0

(14 14 16 18 20 22 24 >25

Gall Diameter (mm)

Fig. 3. Percentage of galls by diameter size class (mm) attacked

by the inquiline Mordellistena unicolor at the November and April censuses

26.1%

BIRDS

14.5

Mordellistena 26.0% Eurosta 2 Eurytoma

unicolor solidaginis gigantea

286% t10.5% 21.9%

Eurytoma

obtusiventris

Fig. 4. Approximate frequencies of interactions among Eurosta so- lidaginis natural enemies that result in the death of one enemy by another

Eurytoma gigantea - November Eurytoma gigantea - April

100 - 100 . Observed attack

C)

5- . 0 s | t 5 | *11 l levels and e~~~~~___ stimated total attack lvl r hw o Q4 - d0 20 0 ~~0 i

1 20 - 20 <GEurytoma obtusiventris - November Eurytoma obtusiventris - April

0 15 - - ~~~~~15 -- Fig. 2. Percentage of galls attacked by Eurytoma o ~~~~~~~~~~~~~~~~~~~~species according to gall diameter size class (mm) for

n 10-- ~~~~~~10-- November and April censuses. Since one natural ~~ Li ~ enemy can consume another, both observed attack

5 - - 5 - - ~~~~~~ M ~levels and estimated total attack levels are shown for ? ? ?~ ~ ~ ~~~~~~ ? ?I each Eurytoma species at the November and April o ~0-- censuses (14 1 4 1 6 18 20 22 24 >25 114 14 1 6 18 20 22 24 125

Gall Diameter (mm)

also attacked by E. gigantea. Further, we found that Mor- dellistena experience higher winter mortality in smaller galls

(Fig. 3).

Insect natural enemy competition

A reasonably large fraction of the galls were attacked by more than one enemy but because the initial attacker was

consumed, only the ultimate survivor in the gall could be recorded. The approximate frequencies of interactions among the gallmaker's insect enemies were calculated and are shown as percentages in Fig. 4. The high levels of inter- action among the insect natural enemies prohibits a detailed analysis of individual species influence on the gallmaker.

Mordellistena is estimated to have consumed E. gigantea in 26.1% of the galls originally containing E. gigantea. To-

This content downloaded from 64.107.81.200 on Thu, 02 May 2019 15:56:48 UTCAll use subject to https://about.jstor.org/terms

21

tal galls parasitized by E. gigantea was estimated at 20.3% of the population. E. gigantea is estimated to have con- sumed E. obtusiventris in 21.9% of all galls originally con- taining E. obtusiventris. Mordellistena should have con- sumed E. obtusiventris in 28.6% of all galls originally con- taining E. obtusiventris. Total galls originally containing E. obtusiventris were estimated at 10.5% of the population.

Discussion

The most notable finding of this study was the net upward selection pressure on Eurosta gall diameter created by Eur- osta's natural enemies. Larvae in small galls have low rela- tive fitnesses whereas larvae in intermediate sized galls have high relative fitnesses. Thus, there is a stabilizing compo- nent to the selection pressures created by the natural ene- mies in this system. At the same time, the selection is direc- tional since the gall size with peak survival is higher than the average of the gall size distribution. Thus, even though some of the larvae of the larger galls are eaten by birds, pressure from E. gigantea is often strong enough that selec- tion is pushing in an upward direction favoring some slight increase in gall diameter.

The selection intensities reported here indicate the mag- nitude and direction of the selection pressures created by the gallmaker's natural enemies. They do not, however, in- dicate the evolutionary responses to those pressures. The potential evolutionary responses are determined by both the heritable variation in gallmakers for gall size and the selection intensities. Although gall size has a heritable com- ponent to the insect (Weis and Abrahamson 1986), plant genotype has an even stronger genetic effect on gall size. The limited genetic control of gall size by the gallmaker will markedly restrain the amount of evolutionary response possible. We will report on the responses of these gallmaker populations to natural selection in another paper (Weis et al., unpublished work).

The selection intensities we measured were highly vari- able from field to field. The parasitoid E. gigantea created an upward selection pressure in every field we examined, but not every field had a downward selection pressure im- posed by birds. Since the birds involved in gallmaker attack are primarily woodland species, bird attack is dependent on proximity of appropriate bird habitat.

The occupation of the smallest-gall subset by E. gigantea was expected given that it attacks after galls have reached maximum size. Thus, E. gigantea is excluded from galls whose wall thicknesses are greater than their ovipositor length (Weis and Abrahamson 1985; Weis et al. 1985). Al- though Mordellistena can attack all gall sizes because of the beetle's early invasion of gall tissue, it attacked small galls more frequently. On average, E. obtusiventris occupies gall sizes larger than the population mean, but this is likely the result of differential attack on E. obtusiventris by E. gigantea and Mordellistena in small galls. Thus, E. gigantea and Mordellistena reduce they number of E. obtusiventris in small galls in the same way the do Eurosta. E. obtusiven- tris can be found in galls of all diameters because its oviposi- tion involves egg deposition on Eurosta larvae prior to gall formation (Abrahamson unpublished data). This natural enemy is probably the only natural enemy that does not exert selection on Eurosta for gall size.

The mean of galls successfully attacked by birds was larger than surviving Eurosta galls and the mean of all galls

not attacked by birds. However, the frequency of bird at- tack on galls was independent of field gall density but it was positively correlated with field mean gall diameter. This suggests that birds select galls based on gall size rather than gall abundance. This finding is similar to that reported by Confer and Paicos (1985).

In our study, successful bird attack accounted for 14.5% of Eurosta mortality whereas E. gigantea parasitism ac- counted for 20.3% of Eurosta mortality (this figure includes galls originally occupied by E. gigantea but ultimately con- sumed by Mordellistena). Uhler (1961) found that E. gigan- tea accounted for 5.6% of Eurosta mortality while birds accounted for only 2.3%. Cane and Kurczewski (1976), however, reported that bird predation can be as high as 20%. These findings and our own data (Weis et al. unpubl. data) lead us to conclude that site to site and yearly fluctua- tions in bird predation rates are common.

Annual variation in attack by Eurosta's natural enemies have the potential to cause shifts in the mean gall diameters of the surviving Eurosta population. One possible scenario is that gall size increases do occur as a consequence of attack by E. gigantea. The resulting population of larger galls could become more attractive to birds and if so, bird predation could increase, causing a shift toward a smaller mean gall size. These two counter-balancing selection forces could create a dynamic equilibrium in gall diameter that would fluctuate in different fields in various years. An alter- native scenario would see shifts in selection intensity as primarily the result of environmental effects. In years when climate promotes vigorous gall growth, the average gall size would be nearer the peak of the fitness function, and thus selection would be weak. Under poorer climatic condi- tions (e.g., drought) the mean gall size would be substan- tially less than the size conferring peak fitness, hence the intensity of selection would be strong. Environmental ef- fects on gall development may mask genetic differences in some generations but not others, thus causing the evolu- tionary response to selection to occur intermittently, rather than as a smooth and continuous function (Weis et al., un- published work). Our study found that interactions among the insect members of the Eurosta natural enemy guild was complex and frequent.

Acknowledgements. We thank Chris Abrahamson, Robert Bertin, Amy Ershler, Irene Kralick, Wayne McDiffett, and Gary Metzger for their assistance. Financial support was provided by Bucknell University and NSF grants DEB-8205856 and BSR-8614768 to W.G.A. A portion of this research was part of a M.S. thesis submit- ted to Bucknell University by J.F.S.

References

Abrahamson WG, Armbruster PO, Maddox GD (1983) Numerical relationship of the Solidago altissima stem gall insect-parasitoid guild food chain. Oecologia 58:351-357

Abrahamson WG, Weis AE (1987) Nutritional ecology of arthro- pod gallmakers. In: Slansky F Jr., Rodriquez JG (eds) Nutri- tional ecology of insects, mites, spiders, and related inverte- brates. John Wiley & Sons, Inc., Publishers, New York, pp 235- 258

Askew RR (1961) On the biology of the inhabitants of oak galls of Cynipidae (Hymenoptera) in Britain. Trans Soc for British Ent 14:237-268

Beck EG (1947) Some studies of the Solidago gall caused by Eur- osta solidaginis Fitch. Ph.D. dissertation, University of Michi- gan, Ann Arbor

This content downloaded from 64.107.81.200 on Thu, 02 May 2019 15:56:48 UTCAll use subject to https://about.jstor.org/terms

22

Bergman JM, Tingey WM (1979) Aspects of interaction between plant genotypes and biological control. Bull Entomol Soc Am 25:275-279

Cane JT, Kurczewski FE (1976) Mortality factors affecting Eurosta solidaginis (Diptera: Tephritidae). J New York Entomol Soc 84:275-282

Carango P, McCrea KD, Abrahamson WG, Chernin MI (1988) Induction of a 58000 dalton protein during goldenrod gall for- mation. Biochem Biophys Res Comm 152:1348-1352

Confer JL, Paicos P (1985) Downy woodpecker predation on gold- enrod galls. Field Ornithol 56: 56-64

Efron B (1981) Nonparametric estimates of standard error: the jackknife, the bootstrap, and other methods. Biometrika 68:19-29

Endler JA (1986) Natural selection in the wild. Princeton Universi- ty Press, Princeton, NJ, USA, p 337

Falconer DS (1981) Introduction to quantitative genetics. Second edition. Longman, Harlow, UK

Fitch ASA (1855) First report on the noxious, beneficial and other insects of the state of New York. Transact NY State Agric Soc 14:66-67

Harrington WH (1895) Occupants of the galls of Eurosta solidagin- is. Canad Entomol 27:197-198

Judd WW (1953) Insects reared from goldenrod galls caused by Eurosta solidaginis Fitch (Diptera: Tephritidae) in southern On- tario. Canad Entomol 85:294-296

Miller WE (1959) Natural history notes on the goldenrod ball gall fly, Eurosta solidaginis (Fitch) and on its parasites, Eury- toma obtusiventris Gahan and Eurytoma gigantea Walsh. J Tenn Acad Sci 34:246-251

Milne LJ (1940) Autecology of the goldenrod gall fly. Ecology 21: 101-105

Price GR (1970) Covariance and selection. Nature (London) 227: 520-521

Price PW, Bouton CE, Gross P, McPheron BA, Thompson JN,

Weis AE (1980) Interactions among three trophic levels: influ- ence of plants on interactions between insect herbivores and natural enemies. Annu Rev Ecol Syst 11:41-65

Price PW, Clancy KM (1986) Interactions among three trophic levels: gall size and parasitoid attack. Ecology 67:1593-1600

Schlichter L (1978) Winter predation by black-capped chickadees and downy woodpeckers on inhabitants of the goldenrod ball gall. Canad Field Nat 92:71-74

Sokal RR, Rohlf Fl (1981) Biometry. W.H. Freeman and Com- pany, San Francisco, CA

Spofford SH (1977) Poplar leaf stem gall insects as food for war- blers and woodpeckers. Wilson Bull 89:485

Stinner BR, Abrahamson WG (1979) Energetics of the Solidago canadensis-stem gall insect-parasitoid guild interaction. Ecology 60:918-926

Uhler LD (1951) Biology and ecology of the goldenrod gall fly, Eurosta solidaginis (Fitch). Cornell Univ Agric Exp Stat Mem 300:1-51

Uhler LD (1961) Mortality of the goldenrod gall fly, Eurosta solida- ginis in the vicinity of Ithaca, New York. Ecology 42:215-216

Varley GC (1947) The natural control of population balance in the knapweed gall fly (Urophora jaceana). J Anim Ecol 16:139-187

Weis AE, Abrahamson WG (1985) Potential selective pressures by parasitoids on the evolution of a plant-herbivore interaction. Ecology 66:1261-1269

Weis AE, Abrahamson WG (1986) Evolution of host plant manip- ulation by gallmakers: ecological and genetic factors in the Solidago-Eurosta system. Am Nat 127:681-695

Weis AE, Abrahamson WG, McCrea KD (1985) Host gall size and oviposition success by the parasitoid Eurytoma gigantea. Ecol Entomol 10:341-348

Received September 7, 1988

This content downloaded from 64.107.81.200 on Thu, 02 May 2019 15:56:48 UTCAll use subject to https://about.jstor.org/terms