## Abstract

“Bottom-up” factors strongly influenced the spatial and temporal pattern of survival of Asphondylia atriplicis Townsend (Diptera: Cecidomyiidae) on Atriplex canescens (Pursh) Nutall (Chenopodiaceae) at three locations in Phoenix, AZ. In contrast, “top-down” effects of natural enemies did not influence the pattern of A. atriplicis mortality. A. atriplicis induces a fleshy, multilocular, rounded stem-gall near the apical meristems of A. canescens. A. atriplicis survival increased as gall size increased, and as the depth of larva in the gall increased. Larval mortality from unknown factors on A. atriplicis decreased with gall size, but the overall interval parasitism rate did not change significantly with gall size. The interval parasitism rate for the eurytomid parasitoid group with the shortest ovipositor was negatively correlated with gall size, but interval parasitism by all other parasitoids was not influenced by gall size. Gall size was strongly influenced by the bottom-up forces of environmental and plant heterogeneity. Gall size varied among seasons, sites, and plants. Larval survival and gall-size covaried in each season and site and among plants.

## Introduction

CURRENTLY,THERE IS A DEBATE as to whether “bottom-up”forces (McNaughton 1976, Schindler1978, Schultz and Baldwin 1982), or “top-down” forces (Hairston et al. 1960, Carpenter and Kitchell 1985, Sinclair 1985) determine the population dynamics of species and the structure of ecological communities. The population dynamics of phytophagous insects can be influenced by top-down forces, such as natural enemies (Bernays and Graham 1988), or bottom-up forces, such as environmental variation (Davidson and Andrewartha 1948, White 1978) and plant resource limitation (Ohgushi and Sawada 1985, Price 1990). Hunter and Price (1992) developed a conceptual model where variation in both bottom-up and top-down forces at each trophic level determine how populations are regulated and communities are structured. They argue, however, that in terrestrial systems bottom-up forces predominate. The model uses primary producers as a template and variation at lower trophic levels “cascades up” to influence variation among herbivores and natural enemies.

Host plant variation is crucial in determining the effects that cascade upward to determine the variation in herbivores and natural enemies. The importance of genotypic host plant variation in influencing herbivore performance has been demonstrated in both agricultural (Hare 1992) and natural systems (Moran 1981, McCrea and Abrahamson 1987, Karban 1989, Simms and Rausher 1989, Fritz and Simms 1992, Hare 1992, Strong 1993, Stiling and Rossi 1996, Craig et al. 1999). The importance of phenotypic host plant variation on herbivores has been demonstrated by studies that show that some herbivores do better on stressed plants (White 1969, 1993; Mattson and Haack 1987), some do better on plants of intermediate vigor (Mc-Kinnon et al. 1999), and others perform better on vigorous plants (reviewed by Price 1991).

The influence of variation in host plant traits that cascade upwards to influence natural enemies has been demonstrated in many studies of tritrophic interactions (Price et al. 1980, Heinrich and Collins 1985, Price and Clancy 1986, Gross and Price 1988, Keating et al. 1990, Barbosa et al. 1991, Hare 1992, Hunter and Schultz 1993). Gall-inducing insects are ideal systems in which to study the factors influencing survival because galls preserve a record of the entire life history from egg to adult emergence (Van Driesche 1983, Price and Clancy 1986). Galls are abnormal plant tissue that is induced by insects that can provide refuges for phytophagous insects from parasitoids. Gall characteristics are determined by interactions among the plant, the herbivore, and the environment (Weis and Abrahamson 1986), and variation in these bottom-up characters can have effects that cascade upwards and structure parasitoid communities (Craig 1994). Parasitism rates can be directly influenced by gall characters such as size (Weis and Abrahamson 1985a, Price and Clancy 1986), gall toughness (Craig et al. 1990), gall type (Hawkins and Goeden 1984), and the number of hosts per gall (Weis 1982, Romstock-Volkl 1990, Ehler and Kinsey 1990). Galls also can mediate parasitoid responses to host density and indirectly affect herbivore survivorship (Price 1988).

The goal of this study was to simultaneously measure the top-down and bottom-up forces influencing the survival of the gall-forming midge Asphondylia atriplicis Townsend (Diptera: Cecidomyiidae) that induces tumor stem-galls on four-winged saltbush, Atriplex canescens (Pursh) Nutall (Chenopodiaceae). We hypothesized that gall characters could influence A. atriplicis larval survival both from the bottom-up by determining food quality and protection from the abiotic environment, and from the top-down by determining larval susceptibility to parasitoids. We examined the possibility that larval survival in these multilocular galls could be influenced both by morphological traits shared by all occupants and by their position within the gall.

We measured how the environment and individual plant characteristics affected gall morphological traits and how, in turn, these traits influenced larval survival. We asked five specific questions: 1) Is there variation in A. atriplicis larval survival over seasons, among sites, and among plants? 2) Is there variation in gall morphological traits over seasons, among sites, and among plants? 3) Do gall morphological traits affect A. atri-plicis larval survival? 4) Is attack on A. atriplicis galls by its natural enemies mediated by gall morphological traits? 5) Do gall morphological traits and larval performance covary within seasons, among sites, and among plants? To test these hypotheses, we chose seasons and sites where we predicted there would be strong variation in gall characters. We predicted that there would be differences among seasons because of the large variation in weather between seasons, and among the same season in different years (Mullen et al. 1998). We predicted that there would be differences among sites because of differences in the phenotypic appearance of the plants at each site. We hypothesize that this is primarily due to the differences in drainage patterns that determined how much water and the length of time that the soil retained water, although we did not measure these characteristics in this study.

## Natural History

A. canescens is widely distributed. It ranges from eastern Texas to California and from Zacatecas, Mexico, northward to Alberta, Canada. It is a woody perennial that is highly drought resistant. It is usually associated with alkaline soils, but it grows in a variety of soils and climates (Benson and Darrow 1981).

The midge A. atriplicis oviposits on four-winged saltbush and induces a fleshy, rounded stem-gall near the apical meristems of the plant. Each A. atriplicis larva is contained within its own chamber within the gall. From 1 to 70 larvae per gall have been reported in southern California (Hawkins et al. 1986). Gall shape varies considerably, ranging from small spherical galls to large, slightly irregular, ellipsoid galls. The inner surface of the larval chamber is lined with an obligate fungus upon which A. atriplicis larvae feed (Highland 1964, Shorthouse and Rohfritsch 1992).

Hawkins etal. (1986) described the general biology of the A. atriplicis complex. The female midge oviposits into the plant stem. The egg is the diapause stage of the fly. Gall development is initiated and chambers begin to form around the egg before the egg hatches, but most eggs hatch within a week after the beginning of gall development. A. atriplicis has three instars, the first two each lasting a week, and the third lasting from 2 to 3 wk. Most larval growth occurs in the third instar. To emerge from the gall, the prepupa bores a channel from its chamber and breaks through to the gall surface. With the pupa extruding from the gall, ecdysis occurs and the adult fly emerges from the pupal case. The populations we sampled were bivoltine with generations coinciding with the two rainy seasons in the Sonoran desert (R.J.M. and T.P.C., unpublished data). They diapause in the egg stage after oviposition until rains stimulate plant growth, which apparently causes them to break diapause.

Hawkins and Goeden (1984) reported 37 common species of parasitoids, predators, and inquilines associated with the A. atriplicis complex. Of these, they reared 14 chalcidoid species from A. atriplicis galls on four-winged saltbush. The phenology of attack is not well known for these natural enemies. Hawkins and Goeden (1984) report that Torymus capillaceus Huber (Hymenoptera: Torymidae) feed externally on stages from early second instars to pupae and that Rileya tegularisGahan (Hymenoptera: Eurytomidae) attacks first to early to second instars. Tetrastichus cecidobroter Gordh and Hawkins (Hymenoptera: Eulophidae) induces an endogall within a midge larval chamber (Hawkins and Goeden 1984), suggesting that it attacks early instars.

The presence of T. cecidobroter is negatively correlated with parasitism by other species; thus, it is an important organizer of the parasitoid community structure in tumor stem-galls (Hawkins and Goeden 1984). This wasp oviposits between one and nine eggs in the wall of an A. atriplicis larval chamber. The gall tissue surrounding each egg grows mitotically, forming a cluster of endogalls within the larval chamber. The wasp larva feeds on the tissue of its endogall, and it does not consume the midge larva, but endogall growth eventually crushes it.

## Materials and Methods

### Study Sites

We studied A. atriplicis on four-winged saltbush at three sites in Maricopa County, Arizona, in the Lower Colorado subdivision of the Sonoran Desert. Two of the sites, Adobe Dam and Jomax Road, were located in northern Phoenix and consisted of mixed stands of four-winged saltbush and Atriplex polycarpa (Torrey) Watson (Chenopodiacedae). Both of these populations of Atriplex occur near the base of old earthen dams. The Jomax Road site was adjacent to Cave Creek Dam (112° 05' W 33° 44' N). The Adobe Dam site was located on the Adobe Dam Recreational Area (112° 10' W 33° 41' N). The third site (V&P Nurseries), was located in Phoenix (111°95' W 33° 25 N'). It was comprised of four-winged salt-bush mixed with some A. polycarpa adjacent to an outer fence enclosing V&P Nurseries. The shrubs were outside the nursery property and seemed to be a remnant population of the original desert vegetation.

### Data Collection

We collected galls from the Adobe Dam and Jomax Road sites in spring (March) 1995, fall (October–November) 1995, and spring (March– April) 1996. We attempted to collect galls as close to the time of emergence as possible. The midges were in the late third instar or pupal phase when we collected them. In the spring and fall collection of 1995, 60 four-winged saltbush plants at each site were chosen randomly. All galls were harvested and reared in the laboratory in the manner described below. In spring 1996, we added the V&P Nursery site. We harvested galls from randomly selected plants at each of the three sites. Galls were randomly collected by throwing a plastic tape over a plant with the observerʼs back to the plant and eyes closed, and sampling every gall within 10 cm of the tape. This technique was continued until a sufficient sample was obtained. However, we limited the harvest to three galls from each plant and no >40 plants per site to equalize sample sizes. In total, 1,219 galls were collected from seven samples over three seasons.

### Laboratory Rearing

We measured gall diameter for each of the three axes of each gall with digital calipers. Gall volume (V) was calculated as V = 4/πace, where a is the radius of the largest axis, c is the radius of the next largest axis perpendicular to a, and e is the radius of the next largest axis perpendicular to a and c. Mean gall diameter was measured as the average diameter of the two smallest axes of the gall perpendicular to each other and the long axis. Emerging insects were reared in shell vials and collected daily, identified, and preserved. Chalcidoidea were identified to subfamily using the key in Grissel and Schauff (1990). The most common natural enemies were identified to genus or species. After rearing was complete, each gall was dissected by cutting along the longest axis to identify the number of chambers within the gall.

## Dissections

A subset of 61 galls containing a total of 205 A. atriplicis larvae was used to determine the insect species emerging from each chamber within each gall. We recorded emergence of each insect from each gall. We identified the chamber of origin of each insect by inserting a felt pen tip back through the emergence hole and marking the inside of the chamber. We used unique colors for each chamber when chamber depth was measured from multiple chambers located in one gall. Chambers were coded for the presence or absence of midge emergence. Gall-size was determined in the same manner as in the laboratory rearing. We then dissected the galls and recorded the number of larval chambers per gall, chamber depth, and chamber size for each chamber. Chamber depth was measured as the shortest distance from the center of the chamber to the surface of the gall. Insertion of the ovipositor at the shortest distance from the gall surface to a larva has been reported for parasitoids attacking gall-inducers (Craig et al. 1990). Chamber volume (V) was calculated for each chamber within each gall by measuring the diameter of each chamber (V =4/3πr3). Chamber density per gall was determined by dividing the number of chambers per gall by gall size (cubic millimeters).

## Statistical Analysis

To determine which gall morphological traits were useful predictors of A. atriplicis survival, we fitted a logistic regression model for survival data by using maximum likelihood (SAS Institute 1990). This type of analysis is appropriate for use on presence–absence data (Neter et al. 1989). The gall traits examined as predictors of larval survival were: gall size (cubic millimeters), mean gall diameter (millimeters), larval chambers per gall, chamber density per 100 (cubic millimeters), chamber depth (millimeters), and chamber size (cubic millimeters). These traits were entered into a stepwise logistic regression, and the improvement χ2 test was used to determine whether the addition of a variable improved the predictive ability of the regression.

To determine the relationship between larval survival and gall size, we examined the proportion of midge larvae surviving until emergence per gall across gall sizes. Galls were divided into 14 gall size categories at 0.5-mm-diameter intervals. We calculated the mean proportion of A. atriplicis larvae that survived until emergence per gall for each gall size class. The relationship of percentage of A. atriplicis survivorship per gall to gall size was analyzed using a Spearmanʼs rank correlation coefficient.

We determined the proportion of emergence per gall by each natural enemy. For statistical analysis, the data on natural enemies was pooled into four groups based on the feeding biology, temporal sequence of attack, and female ovipositor length. For each group, the relationship of percentage of parasitism per gall to gall size was analyzed using a Spearmanʼs rank correlation coefficient.

To determine the parasitism rate by the larval parasites mentioned above, we calculated the interval parasitism rate (Price and Clancy 1986). Interval parasitism is the number of hosts parasitized divided by the number of larval hosts available (×100). Because galls contain a complete record of mortality throughout the life history of the gall-inducer, we could calculate the number of larvae available to be parasitized by eliminating those individuals that died during the egg stage. This approach avoids the problems of interpreting percentage of parasitism that can occur in studies of free-living insects where the availability of the susceptible stage to parasitoids can be difficult to obtain (Van Driesche 1983).

To test for season, site, and plant effects on A. atriplicis survival, we performed a mixed model analysis of variance (ANOVA) (SAS Institute 1990). Seasons and sites were treated as fixed effects, and the effect of plants was nested within each site and season and treated as a random effect. The proportion of midge larvae to survive until emergence per gall was transformed as $$p' = \sqrt {\left[ {\left( {x + 3/8} \right)/\left( {n + 3/4} \right)} \right]}$$ where X/n is the proportion of emerging flies and larvae per gall (Anscombe 1948).

To test for season, site, and plant effects on gall size, gall diameter, larvae (chambers) per gall, and larval density per cubic millimeter of gall tissue, we performed a multivariate ANOVA (SAS Institute 1990). To control for individual plant effects, a nested two-way multiple ANOVA (MANOVA) was used to test for differences in gall morphology among populations (sites) and between seasons. When there was multivariate statistical significance, we analyzed the individual variables by using ANOVA to determine which contributed to the significance of the MANOVA.

Goodness-of-fit analysis was used to test whether emergence was random with respect to gall size for each of the species groups. The expected distribution of emergence by each group was determined by multiplying the number of chambers in each gall size category by the proportion of total chambers parasitized by each group, respectively. Galls were divided into gall size categories as described above.

To examine the covariance of larval survival and gall size among plants, we used data from the Jomax Road and Adobe Dams sites for the first two seasons of this study. We first ran the analysis of covariance (ANCOVA), including the plant identity by gall size interaction term. Where the interaction term was not significant, we reanalyzed the data without this term.

## Ovipositor Measurement

We dissected at least 15 females of the three most common parasitoid species that attack A. atriplicis to measure their ovipositors. All specimens were taken from the Jomax site in spring, 1996, from galls that were not included in the previous samples. The base of the ovipositor was exposed after dissecting the abdomen of each specimen. The ovipositors were measured by ocular micrometer from the base of the ovipositor to the tip.

## Results

### Environmental Effects onA. atriplicis Larval Performance

A. atriplicis emerged from 60% of the galls and 42% of the chambers (2,087 flies emerging from 1,219 galls and 4,968 chambers). Larval performance, as measured by the proportion of A. atriplicis larvae per gall to survive until adult emergence, ranged from 0 to 100%.

Larval survival significantly differed among seasons, sites, and plants, but the interaction between site and season was not significant (Table 1). Survival rates in each of the seasons significantly differed from the others: it was highest in fall 1995 and lowest in spring 1996 (Table 2). Survival rates varied significantly among sites, with Adobe Dam having more than twice the survival rate of V&P Nurseries (Table 2). The Jomax Road site also had a higher survival rate than V&P Nurseries (Table 2).

Table 1

Analysis of the proportion of A. atriplicis per gall that survive until adult emergence

### Environmental Effects on Gall Morphological Traits

Individual galls showed a large range of variation in several morphological traits. Gall volume varied over 3 orders of magnitude, with galls ranging in volume from 5.7 to 3343.5 mm3. Gall diameters ranged from 1.9 to 12.8 mm. Galls contained from 1 to 24 larvae. Larval density per 100 mm3 of gall tissue ranged from 0.158 to 17.684/100 mm3.

Multivariate analysis of gall morphological traits showed that there was significant variation among seasons, sites, and plants but that the interaction between season and site was not significant (Table 3). Gall volume and diameter were larger in spring 1996 than in other seasons, and gall diameter was larger in the fall 1995 than in spring 1995 (Table 2). Larval density per 100 mm3 of gall tissue also differed among seasons, but the number of larval chambers (larvae) per gall did not vary among seasons (Table 2). In seasons when galls were larger, the density of larvae in the gall decreased, because as gall size increased, there was no significant change in the number of larval chambers per gall.

Table 2

Mean α SE for percentage of A. atriplicis per gall to survive until adult emergence, and for gall traits by season and site

Means within the same column followed by different letters are significantly different (P <0.05, HSD method).

Gall volume, gall diameter, and larval density per 100 mm3 all varied among sites (Table 2) but not the number of larval chambers per gall (Table 2). Galls had a smaller volume and diameter at Jomax Road (Table 2) than at other sites. Smaller galls at this site had higher larval densities, because the number of larval chambers per gall did not differ among sites.

Table 3

Multivariate analysis of variance of environmental and plant effects on morphological traits of A. atriplicis galls

### Plant Effects on Gall Traits and Larval Performance

The two-factor-nested analyses of variance showed that season, site, and plant identity all had strong effects on several gall morphological characters (Table 4). Gall volume, gall diameter, larval density per 100 mm3 of gall tissue, and the number of larval chambers per gall all varied among plants (Table 4). Plants were the only level of variation that showed variation in the number of larval chambers per gall. ANOVA was performed using the residuals of a regression of gall size on the number of larval chambers per gall for data from fall 1995 and for pooled data from spring 1995 and spring 1996. Plant identity explained 34.99% of the residual variance in gall size in fall 1995 (F56220 =2.12; P < 0.001) and 47.29% of the residual variance in gall size for the pooled data for spring 1995 and spring 1996 (F155,768 =4.45; P <0.0001). Larval survival also varied significantly among plants (Table 1).

### Gall Size-DependentA. atriplicis Survival

A. atri-plicis had higher survival rates when they were deep inside galls and in large galls. We performed a stepwise logistic regression on factors predicting A. atriplicis larval survival until emergence from the gall. The improvement chi-square tests whether the addition of a variable significantly improves the predictive ability of the regression. The probability that a larva would survive until emergence significantly increased with chamber depth (Fig. 1; χ2 =27.502, df =1, P <0.0001). The predictive ability of the regression also was significantly increased with the addition of gall size to the model (x2 =9.0986, df =1, P <0.0026). Mean chamber depth per gall increases with gall diameter (y = 0.171 +0.262x; F146 =24.78, r2 =36%; P <0.0001).

Fig. 1.

Probability that an A. atriplicis will survive until emergence from its gall as a function of larval chamber depth. Arrows indicate the maximum gall depth that the ovipositor could reach for three of the major parasitoids: (a) R. tegularis, (b) T. cecidobroter, and (c) T. umbilicates.

Fig. 1.

Probability that an A. atriplicis will survive until emergence from its gall as a function of larval chamber depth. Arrows indicate the maximum gall depth that the ovipositor could reach for three of the major parasitoids: (a) R. tegularis, (b) T. cecidobroter, and (c) T. umbilicates.

In each season, larval survival was positively correlated with gall size (Fig. 2; spring 1995: rs =0.32, df =627, P <0.0001; fall 1995: rs =0.32, df =283, P <0.0001; and spring 1996 rs =0.23, df =309, P <0.001), but the shape of the relationship varied among seasons. In spring 1995 and spring 1996, there was a positive linear trend between larval survival and gall size. In fall 1995, there was a rapid increase in larval survival followed by a gradual increase once galls exceeded 4.5 mm (Fig. 2).

Fig. 2.

Mean proportion per gall survivorship (±1 SE) of A. atriplicis until adult emergence for the indicated gall diameter classes for three seasons. The means include data from all three all sites.

Fig. 2.

Mean proportion per gall survivorship (±1 SE) of A. atriplicis until adult emergence for the indicated gall diameter classes for three seasons. The means include data from all three all sites.

Table 4.

Two-factor nested analysis of variance of environmental and plant effects on gall morphological traits of A. atriplicis galls

Survival of A. atriplicis larvae at all three sites was correlated with gall size (Fig. 3; Adobe Dam: rs =0.24, df =186, P <0.001; Jomax Road: rs =0.37, df =919, P <0.0001; and VP Nurseries: rs =0.34, df =114, P <0.001). The pattern of survival varied among sites. At the Adobe Dam site, there was a threshold for survival: there was almost no A. atriplicis survival until gall diameter exceeded 4.5 mm. Larval survival then increased to nearly 50% survival per gall and showed a continuing gradual increase with increasing gall diameter. At the Jomax Road site, there was a positive linear relationship between larval survival and gall diameter. At the V&P Nurseries site, there was no larval survival in galls <6.5 mm, but once gall diameter exceeded 6.5 mm larval survival increased with gall size.

Fig. 3.

Mean proportion per gall survivorship (±1 SE) of A. atriplicis until adult emergence for the indicated gall diameter classes for three seasons. The means includes data from all three seasons.

Fig. 3.

Mean proportion per gall survivorship (±1 SE) of A. atriplicis until adult emergence for the indicated gall diameter classes for three seasons. The means includes data from all three seasons.

### Covariation of A. atriplicis Larval Survival and Gall Size among Plants

At Adobe Dam in spring 1995, the larval survival per gall ranged from 0 to 92% and significantly differed among plants (Table 5). The covariate gall size had a strong effect on larval survival (Table 5). Among plants in this site, gall size and larval survival displayed a clear pattern of covariance, more larvae survived in larger galls. There was no significant plant by gall size interaction at this site. The sample size for the Adobe Dam site for fall 1995 was small with a total of only 17 galls harvested from 10 plants. The ANCOVA model (not shown) for this site did not yield significant results.

Table 5.

Analysis of covariance of the proportion of A. atriplicis larvae per gall surviving until adult emergence on the Adobe Dam and Jomax Road sites for spring and fall 1995

Where the interaction term was not significant (NS), the data were reanalyzed without this term.

At Jomax Road, in both spring and fall of 1995, larval survival significantly differed among plants (Table 5) ranging from 0 to 100% survival per gall. Plant identity and gall size explained 28% of the variance in larval survival per gall in spring 1995, and 51% of the variance in survival in fall 1995. Gall-size had a significant effect on A. atriplicis survival in both seasons (Table 5). There was a significant plant by gall size interaction effect on larval survival in both seasons on this site (Table 5). Although overall more larvae survived in larger galls, the strength of the relationship between larval survival and gall size varied among plants (Fig. 4). The amount of variance in survival explained by gall diameter ranged from nonsignificant on plant 1 to r2 =60% on plant 16. Among the plants where there was a significant relationship between gall diameter and survival, the slopes of the relationship varied widely.

Fig. 4.

Relationship of larval survival to gall size for four plants growing on the Jomax Road site in spring 1995. Individual plants differed in the relationship between gall diameter and larval survival, and a range of relationships are shown from a nonsignificant relationship (A) to highly significant relationships with a range of slopes (B–D).

Fig. 4.

Relationship of larval survival to gall size for four plants growing on the Jomax Road site in spring 1995. Individual plants differed in the relationship between gall diameter and larval survival, and a range of relationships are shown from a nonsignificant relationship (A) to highly significant relationships with a range of slopes (B–D).

### Gall Size-Dependent Mortality

The relative importance of mortality factors changed with gall size. As gall size increased, the arcsine-square root-transformed proportion of mortality due to sources other than predators or parasitoids decreased significantly (y =1.02 -0.089x; r2 =81.7%, df =11, P <0.0001). We refer to this category as mortality from unknown factors, and this could include mortality from desiccation, disease, inadequate nutritional resources, or the interaction between fungus and plant. Total interval parasitism, which is the rate of parasitism among hosts available during the interval when parasitoid attack occurs, did not change significantly with gall size (Fig. 5).

Fig. 5.

Mean proportion per gall survivorship of A. atri-plicis larvae, and mortality due to unknown factors and parasitism for the indicated gall diameter classes for three sites over all seasons.

Fig. 5.

Mean proportion per gall survivorship of A. atri-plicis larvae, and mortality due to unknown factors and parasitism for the indicated gall diameter classes for three sites over all seasons.

The relationship between gall size and parasitism rate varied among parasitoid groups. There was no relationship between interval parasitism rate and gall size for the Torymidae or Eulophidae species groups. The arcsine square root-transformed interval parasitism rate of Eurytomidae was strongly negatively related to gall diameter (y =0.499 -0.0289x;r2 =49.4%, df =11, P <0.001).

### Ovipositor Length among Parasitoids

Ovipositor lengths of T. umbilicatus (3.68 ±0.11 mm), T. cecido-broter (1.96 ±0.08), andR. tegularis (1.14 ±0.04 mm) were all significantly different (F253 =283.34; P <0.0001; P <0.05; Tukeyʼs honestly significant difference [HSD] method). Based on these measurements, we estimated the percentage of chambers that would be available to each parasitoid (Fig. 1) to be 100% for T. umbilicatus, 34.14% for T. cecidobroter, and 2.4% for R. tegularis. We assumed that the ovipositor could be fully extended and that attack occurred after the gall had reached its maximum size.

### Discussion

Bottom-Up Forces Regulate the Population Dynamics of A. atriplicis. A. atriplicis survival was strongly influenced by gall size, indicating that bottom-up forces predominate in determining the intra-and interpopulation variation in survival that we observed in this system. The smaller the gall, the higher the probability was of early mortality from unknown factors. Subsequent overall mortality due to natural enemies did not differ among gall size categories, although larvae in small galls were more vulnerable to parasitism by the eurytomid group. Gall size was regulated by the bottom-up forces of environmental and plant variation, indicating that bottom-up forces cascade upward (sensu Hunter and Price 1992) to influence herbivore density and natural enemy community composition. The success of the eurytomid parasitoid group was mediated by gall size, which was controlled from the bottom-up.

The impact of bottom-up environmental factors in controlling A. atriplicis was demonstrated by the influence of plant identity on overall larval survival (Table 4). In each season and site, gall size covaried with A. atriplicis larval survival (Figs. 2 and 3), and gall size covaried with A. atriplicis larval survival among plants. At the Adobe Dam site in spring 1995, plant identity and the covariate gall size explained 41% of the variance in larval survival per gall. At Jomax Road, in both spring and fall of 1995, plant identity interacted with gall size to influence larval survival. Gall size did not affect larval survival in some plants, yet it accounted for a large percentage of the variance in larval survival in other plants. Plant and gall size explained 28% of the variance in larval survival per gall in spring 1995, and 51% of the variance in survival in fall 1995.

Increased larval depth was associated with increased larval survival. Larval depth was determined by a combination of gall size and position in a gall. Gall size increase was due to an increase in the amount of plant tissue surrounding the chambers and not an increase in the number of chambers per gall. Females could potentially increase the number of chambers by increasing the number of eggs oviposited to increase the number of larvae initiating the gall, but there is no evidence that this occurred. Instead, the amount of tissue surrounding the chambers varied as the result of the plant where the gall was formed and environmental variation among sites and seasons. A strong environmental influence on gall growth is indicated, but whether there is also genotypic variation for gall size among plants has not been determined. We also have not explored whether there was individual variation among A. atriplicis females in the number of eggs that they oviposited or in the strength of the stimulus for gall formation. Individual variation among herbivores can have a strong impact on survival and has important implications for the evolution of the interaction (Singer 1983, Singer et al. 1988).

### Weak Top-Down Forces Act onA. atriplicis

We did not find density-dependent natural enemy mortality that would indicate top-down population regulation. Because A. atriplicis develops confined in galls, natural enemies find larvae in discrete and variable patches. Parasitoids should respond to high-density patches by searching these patches more frequently and by staying longer to exploit the increased availability of hosts (Cook and Hubbard 1977), increasing parasitism rates. Large galls contained higher numbers of available hosts, but the interval parasitism rate was not related to gall size (Fig. 5).

The “ovipositor limitation” hypothesis is that when parasitoid ovipositors are too short to reach all larvae, gall size mediates parasitism rates (Weis and Abraha-mson 1985b, Price and Clancy 1986, Stiling and Rossi 1996, Dixon et al. 1998). This hypothesis was supported by the strong negative correlation between gall size and interval parasitism by the eurytomid group. Large galls had greater mean chamber depths, so that as gall size increased an increasing proportion of the larvae may be too deep to be reached by the short eurytomid ovipositors. The torymid and eulophid groups with longer ovipositors showed no relationship between gall size and interval parasitism rate. Our estimate of the proportion of larvae available to each species may be imprecise for at least two reasons. First, ovipositor length may overestimate the distance to which a parasitoid can reach into a gall with her ovipositor, because the entire ovipositor may not be available during oviposition (Stiling and Rossi 1994, Dixon et al. 1998). Second, some of the parasitoids may attack before gall growth is complete. Because R. tegularis primarily attacks first instars when galls may be smaller than the final diameter, a larger proportion of larvae may be available than we estimated. A detailed study of parasitoid attack phenology and gall growth is needed to rigorously test the ovipositor limitation hypothesis. Overall, interval parasitism rate did not increase with gall size, so the increase in larval survival with larval chamber depth must involve factors other than vulnerability to parasitoids.

### Possible Bottom-Up Mechanisms

A number of alternative hypotheses could explain why unknown mortality decreases with gall size. The “nutrition” hypothesis states that gall induction increases the availability of nutrients to larvae (Price et al. 1986). The immobile larvae do not have to search for nutritious plant tissue, and this can result in high growth and assimilation efficiencies in gallmakers (Stinner and Abrahamson 1979). Nutritional quality may increase with gall size, and this may increase survivorship in large A. atriplicis galls. Dipteran-induced galls concentrate nitrogen (Paclt and Hassler 1967) and act as sinks for nutrients (Abrahamson and McCrea 1986). Raman and Abrahamson (1995) showed that there was higher calorific concentration in gall tissue induced by cecidomyiids than in normal stem tissue. Weis and Kapelinski (1984) demonstrated that galls induced on Salix ordata Michaux (Salicaceae) by Rhabdophaga strobiloides Osten Sacken (Diptera: Cecidomyiidae) draw photosynthate from distant locations in the plant. A stronger stimulus to the host plant was reported for galls containing larger numbers of larvae than galls with fewer larvae (Lalonde and Shorthouse 1985). It is possible that A. atriplicis galls act as nutrient sinks, and it also is plausible that large galls provide more resources to larvae than small galls.

The “microenvironment” hypothesis (Price et al. 1986) states that gall tissue buffers larvae against environmental variation and that this increases survival. In harsh arid environments, galls likely reduce hygro-thermal stress for larvae residing within gall tissue (Price et al. 1986, Tscharntke 1994), and hygrothermal stress may decrease with increasing gall size. In xeric environments both tropical and temperate gallers had higher population densities, higher larval density per gall, and lower mortality than in mesic environments (Fernandes and Price 1992). In the Sonoran Desert, protection against dehydration is an important selective factor. A. atriplicis gall tissue had higher percentage water content than ungalled stem tissue (Hawkins and Unruh 1988). Higher water content in gall tissue protects against hygrothermal stress, reducing the need for thick larval skins. This could reduce energy use associated with weight loss due to molting, and increase larval survival. Hygrothermal stress in larger galls may decrease with gall size.

### Evolution of Gall Size

The positive correlation between gall size and A. atriplicis survival in all of the environments that we studied indicates directional selection for increased gall size. We have not identified a factor that selects for small gall size induction by the gall midge. In contrast, Weis et al. (1992) found that there was stabilizing selection on gall size in the tephritid gall-inducer Eurosta solidaginis (Fitch), due to opposing selection by different natural enemies. Gall size cannot increase infinitely: so what limits gall size? A. atriplicis could potentially increase gall size either by increasing the number of eggs oviposited or by increasing the larval stimulus for gall growth that is exerted on the plant. There may be a trade-off between increasing the probability of offspring survival by increasing the number of eggs in one gall and the increased risk of placing a large proportion of offspring in one site. Selection may act to reduce the risk of “putting all your eggs in one basket” and favor placing a smaller number of eggs in a larger number of galls. The mechanisms determining the amount of tissue around each chamber is not understood for this insect, and an understanding of the mechanisms that induces gall growth is only beginning to be understood for a few insects (Sopow et al. 2003). Selection to increase the amount of tissue in a gall may be opposed by selection on the plant to limit the resources lost to gall formation. Alternatively, physiological limits on plant growth in these tissues that have evolved for other reasons may set an upper limit to gall growth.

In conclusion, this study provides additional support for the importance of bottom-up forces in the ecology and evolution of terrestrial communities. The seemingly complex variation in A. atriplicis survival and parasitism rates among sites, seasons, and individual plants can be largely explained by variation in a single variable: gall size. Our results support the hypothesis that the interaction between gall morphology and herbivore survival can be the key to understanding a gallmakerʼs population dynamics and community structure (Weis and Abrahamson 1985b, Price and Clancy 1986, Craig et al. 1990, Craig 1994, Stiling and Rossi 1996). Variation in gall phenotype depends on both insect and plant genotype and their interactions with environmental factors (Weis and Abrahamson 1986). Accordingly, selection on both the plant and insect can drive the evolution of gall morphology and the response of natural enemies to this variation.

We thank S. Scheiner, S. Faeth, M. Douglas, P. Price, and J. Itami for valuable comments and recommendations on earlier drafts of this manuscript. S. Lindgren and two anonymous reviewers provided helpful editorial comments. K. Dixon provided help in identifying rare predators and parasitoids. We thank R. Lerma for providing field assistance at various times. B. Hawkins provided valuable advice about working with Asphondylia. We thank Arizona State University West for financial support of this research.

## References Cited

Abrahamson
W. G.
McCrea.
K. D.
1986
.
Nutrient and biomass allocation in Solidago altissima: effects of two gallmakers, fertilization, and ramet isolation
.
Oecologia (Berl.)
.
68
: 174 –180.
Anscombe
F. J
.
1948
.
The transformation of Poisson, binomial, and negative data
.
Biometrika

35
:
246
254
.
Barbosa
P
Gross
P.
Kemper
J.
.
1991
.
Influence of allelochemicals on tobacco hornworm and its parasitoid, Cotesia congregata.
Ecology

72
:
1567
1575
.
Benson
L
Darrow
R. A
.
1981
.
Trees and shrubs of the southwestern desert
.
The University of Arizona Press
,
Tucson, AZ
.
Bernays
L.
Graham
M
.
1988
.
On the evolution of host specificity in phytophagous arthropods
.
Ecology

69
:
886
892
.
Carpenter
S. A.
Kitchell
J. F.
1985
.
Cascading trophic interactions and lake productivity
.
Bioscience

35
:
634
639
.
Cook
R. M.
Hubbard
S. F.
1977
.
Adaptive searching strategies in insect parasitoids
.
J. Anim. Ecol
.
45
:
115
125
.
Craig
T. P
.
1994
.
Effects of intraspecific plant variation on parasitoid communities, pp. 205–227
.
In B. D. Hawkins and W. Sheehan [eds.], Parasitoid communities
.
Oxford University Press
,
Oxford, UK
Craig
T. P.
Abrahamson
W. G.
Itami
J. K.
Horner.
J. D.
1999
.
Oviposition preference and offspring performance of Eurosta solidaginis on genotypes of Solidago altissima.
Oikos

86
:
119
128
.
Craig
T. P.
Itami
J. K.
Price.
P. W.
1990
.
The window of vulnerability of a shoot-galling sawfly to attack by a parasitoid
.
Ecology

7
:
1471
1482
.
Davidson
J
Andrewartha
H. G.
.
1948
.
The influence of rainfall, evaporation and atmospheric temperature on fluctuations in the size of natural populations of Thrips imaginis (Thysanoptera).
J. Anim. Ecol
.
17
:
200
222
.
Dixon
K. A.
Lerma
R. R.
Craig
T. P.
Hughes.
K. A.
1998
. Gall morphology and community composition in Asphondylia flocossa (Cecidomyiidae) galls on Atriplex polycarpa (Chenopodiaceae). Environ. Entomol.
27
:
592
599
.
Ehler
L. E.
Kinsey.
M. G.
1990
.
Influence of gall size on survival and fecundity of Rhopaloymia californica (Diptera: Cecidomyiidae), a biological control agent for Baccharis halimifolia (Asteraceae)
.
Environ. Entomol
.
19
:
1558
1565
.
Fernandes
G. W.
Price.
P. W.
1992
.
The adaptive significance of insect gall distribution: survivorship of species in xeric and mesic habitats
.
Oecologia (Berl.)
.
90
:
14
20
.
Fritz
R. S.
Simms.
E. L.
1992
.
Plant resistance to herbivores and pathogens
.
University of Chicago Press
,
Chicago, IL
.
Grissel
E. E.
Schauff.
M. E.
1990
.
A handbook of the families of Neartic Chalcidoidea (Hymenoptera)
.
Entomol. Soc
.
Wash.
,
Washington, DC
.
Gross
R. S.
Price.
P. W.
1988
.
Plant influences on two leafminers: a test of enemy-free space
.
Ecology

69
:
1506
1516
.
Hairston
N. G.
Smith
F. E.
Slobodkin
L. B.
.
1960
.
Community structure, population control and competition
.
Am. Nat
.
94
:
421
424
.
Hare
J. D
.
1992
.
Effects of plant variation on herbivore-natural enemy interactions, pp. 278–298
.
In R. S. Fritz and E. L. Simms [eds.], Plant resistance to herbivores and pathogens
.
University of Chicago Press
,
Chicago, IL
.
Hawkins
B. A.
Goeden.
R. D.
1984
.
Organization of a parasitoid community associated with a complex of galls on Atriplex spp. in southern California
.
Ecol. Entomol
.
9
:
271
292
.
Hawkins
B. A.
Unruh.
T. R.
1988
.
Protein and water levels in A. atriplicis (Diptera: Cecidomyiidae) galls
.
Southwest. Nat
.
33
:
114
117
.
Hawkins
B. A.
Goeden
R. D.
Gagne´
R. J.
.
1986
.
Ecology and taxonomy of the Asphondylia spp. (Diptera: Cecidomyiidae) forming galls on Atriplex spp. (Chenopo-diaceae) in Southern California
.
Entomography

4
:
55
107
.
Heinrich
B
Collins
S. L.
.
1985
.
Caterpillar leaf damage and the game of hide-and-seek with birds
.
Ecology

64
:
592
602
.
Highland
H. A
.
1964
.
Life history of Asphondylia ilicicola (Diptera: Cecidomyiidae) a pest of American holly. J
.
Econ. Entomol
.
57
:
81
83
.
Hunter
M. D.
Price.
P. W.
1992
.
Playing chutes and ladders: heterogeneity and the relative roles of bottom–up and top-down forces in natural communities
.
Ecology

73
:
724
732
.
Hunter
M. D.
Schultz.
J. C.
1993
.
Induced plant defenses breached: phytochemical induction protects an herbivore from disease
.
Oecologia (Berl.)
.
94
:
1195
1203
.
Karban
R
.
1989
.
Fine-scale adaptation of herbivorous thrips to individual host plants
.
Nature (Lond.)

340
:
60
61
.
Keating
S. T.
Hunter
M. D.
Schultz.
J. C.
1990
. Leaf phenolic inhibition of gypsy moth nuclear polyhedrosis virus: role of polyhedral inclusion body aggregation. J. Chem. Ecol.
16
:
1445
1457
.
Lalonde
R. G.
Shorthouse.
J. D.
1985
.
Growth and development of larvae and galls of Urophora cardui (Diptera: Tephritidae) on Cirsium (Compositae)
.
Oecologia (Berl.)
.
65
:
161
165
.
Mattson
W. J
Haack
R. A.
1987
.
The role of drought and stress in provoking insect outbreaks, pp. 365–407. In P. Barbosa and
J. C. Schultz [eds.], Insect outbreaks
.
,
New York
.
McCrea
K. D.
Abrahamson.
W. G.
1987
.
Variation in herbivore infestation: historical and genetic factors
.
Ecology

68
:
822
827
.
McNaughton
S. J
.
1976
.
Serengeti migratory wildebeest: facilitation of energy flow by grazing
.
Science (Wash DC)

191
:
92
94
.
McKinnon
M. L.
Quiring
D. T.
Bauce
F.
.
1999
.
Influence of tree growth rate, shoot size and foliar chemistry on the abundance and performance of a galling adelgid. Funct
.
Ecol
.
13
:
859
867
.
Moran
N
.
1981
.
Intraspecific variability in herbivore performance and host quality: a field study of Uroleucon caligatum (Homoptera: Aphididae) and its Solidago hosts (Asteraceae)
.
Ecol. Entomol
.
6
:
301
306
.
Mullen
S. L.
Schmitz
J. T.
Renno.
N. O.
1998
. In-traseasonal variability of the summer monsoon over Southeast Arizona. Mon. Weather Rev.
126
:
3016
3035
.
Neter
J
Wasserman
W.
Kutner.
M. H.
1989
. Applied linear regression models. Irwin, Homewood, IL.
Ohgushi
T
H.
.
1985
.
Population equilibrium with respect to available food resource and its behavioural basis in an herbivorous lady beetle Henosepilachna niponica.
J. Anim.Ecol
.
54
:
781
796
.
Paclt
J
Hassler
J.
.
1967
.
The concentration of nitrogen in plant galls
.
Phyton

12
:
173
176
.
Price
P. W
.
1988
.
Inversely density-dependent parasitism: the role of plant refuges for hosts.
J. Anim Ecol
.
57
: 89 –96.
Price
P. W
.
1990
.
Evaluating the role of natural enemies in latent and eruptive species: new approaches in life table construction pp. 221–232
.
In A. D. Watt, S. R. Leather, M. D. Hunter, and N.A.C. Kidd [eds.], Population dynamics of forest insects. Intercept, Andover, UK

Price
P. W
.
1991
.
The plant vigor hypothesis and herbivore attack
.
Oikos

62
:
244
251
.
Price
P. W.
Clancy.
K. M.
1986
.
Interactions among three trophic levels: gall size and parasitoid attack
.
Ecology

69
:
1593
1600
.
Price
P. W.
Bouton
C. E.
Gross
P
McPheron
B. A.
Thompson
J. N.
Weis
A. E.
1980
.
Interactions among three trophic levels: influence of plants on interactions between insect herbivores and natural enemies
.
Annu. Rev. Ecol. Syst
.
11
:
41
65
.
Price
P. W.
Waring
G. L.
Fernandes
G. W.
.
1986
Hypotheses on the adaptive nature of galls
.
Proc. Entomol. Soc. Wash
.
88
:
361
363
.
Raman
A
Abrahamson
W. G.
.
1995
.
Morphometric relationships and energy allocation in the apical rosette galls of Solidago altissima (Asteraceae) induced by Rhopolimyia solidaginus (Diptera: Cecidomyiidae)
.
Environ. Entomol
.
24
:
635
639
.
Romstock-Volkl
M
.
1990
.
Host refuges and spatial patterns of parasitism in an endophytic host-parasitoid systems
.
Ecol. Entomol
.
15
:
321
331
.
SAS Institute
.
1990
.
SAS/STAT userʼs guide, version 6, 4th ed. SAS Institute, Cary, NC
.
Schindler
D. W
.
1978
.
Factors regulating phytoplankton production and standing crop in the worldʼs fresh waters
.
Limnol. Oceanogr
.
23
:
478
486
.
Schultz
J. C.
Baldwin.
I. T.
1982
.
Oak leaf quality declines in response to defoliation by gypsy moth larvae
.
Science (Wash. DC)

217
:
149
151
.
Shorthouse
J. D.
Rohfritsch
O.
.
1992
.
Biology of insect-induced galls
.
Oxford University Press
,
Oxford, UK
.
Simms
E. L.
Rausher.
M. D.
1989
.
The evolution of resistance to herbivory in Ipomea purpurea. II. Natural selection by insects and costs of resistance
.
Evolution

43
:
563
585
.
Sinclair
A.R.E.
1985
.
Does intraspecific competition or predation shape African ungulate community?
J. Anim. Ecol
.
54
: 899 –918.
Singer
M. C
.
1983
.
Determinants of multiple host use by a phytophagous insect population
.
Evolution

37
:
389
403
.
Singer
M. C.
Ng
D.
Thomas.
C. D.
1988
.
Heritability of oviposition preference and its relationship to offspring performance within a single insect population
.
Evolution

42
:
977
985
.
Sopow
S. L.
Shorthouse
J. D.
Strong
W
Quiring
D. T.
2003
.
Evidence for long-distance, chemical gall induction by an insect
.
Ecol. Lett
.
6
:
102
105
.
Stiling
P. D.
Rossi.
A. M.
1994
.
The window of parasitoid vulnerability to hyperparasitism: template for parasitoid complex structure, pp. 228–244
.
In B. A. Hawkins and W. Sheenan [eds.], Parasitoid community ecology
.
Oxford University Press
,
Oxford, UK
.
Stiling
P. D.
Rossi.
A. M.
1996
.
Complex effects of genotype and environment in insect herbivores and their enemies
.
Ecology

77
:
2212
2218
.
Stinner
B. R.
Abrahamson.
W. G.
1979
.
Energetics of the Solidago canadenis-stem gall insect-parasitoid guild interaction
.
Ecology

60
:
918
926
.
Strong
D. R
.
1993
.
Heritability of host plant resistance to herbivory changes with gallmidge density during an outbreak on willow
.
Evolution

47
:
291
300
.
Tscharntke
T
.
1994
.
Tritrophic interactions in gallmaker communities on Phragmites australis: testing ecological hypotheses, pp. 73–92. In P. W. Price, W.
J. Mattson, and Y. N. Baranchikov [eds.], The ecology and evolution of gall-forming insects. Forest Service
,
U.S. Department of Agriculture
,
St. Paul, MN
.
Van Driesche
R. G
.
1983
.
Meaning of “percent parasitism” in studies of insect parasitoids
.
Environ. Entomol
.
12
:
1611
1622
.
Weis
A. E
.
1982
.
Resource utilization patterns in a community of gall-attacking parasitoids
.
Environ. Entomol
.
11
:
809
815
.
Weis
A. E.
Abrahamson
W. G.
.
1985a
.
Host gall size and oviposition success by the parasitoid Eurytoma gi-gantea.
Ecol. Entomol

10
:
341
348
.
Weis
A. E.
Abrahamson
W. G.
.
1985b
.
Potential selective pressures by parasitoids on a plant-herbivore interaction
.
Ecology

66
:
1261
1269
.
Weis
A. E.
Abrahamson.
W. G.
1986
.
Evolution of host-plant manipulation by gall makers: ecological and genetic factors in the Solidago-Eurosta
system. Am. Nat
.
127
:
681
695
.
Weis
A. E.
Kapelinski
A.
.
1984
.
Manipulation of host plant development by the gall-midge Rhobdophaga stro-biloides.
Ecol. Entomol
.
9
:
457
465
.
Weis
A. E.
Abrahamson
W. G.
Andersen
M. C.
.
1992
.
Variable selection on Eurostaʼs gall size, I: The extent and nature of variation in phenotypic selection
.
Evolution

46
:
1674
1697
.
White
T.C.R
.
1969
.
An index to measure weather-induced stress of trees associated with outbreaks of psyllids in Australia
.
Ecology

50
:
905
909
.
White
T.C.R
.
1978
.
The importance of relative shortages of food in animal ecology. Oecologia (Berl.)
.
33
:
71
86
.
White
T.C.R
.
1993
.
The inadequate environment: nitrogen and the abundance of animals
.
Springer, Berlin, Germany
. Plant–Insect Interactions