Does "enemy-free space" exist? Experimental host shifts of an herbivorous fly.
Most plant-feeding insects are specialists, feeding on a few closely related plants within one family (Ehrlich and Raven 1964, Strong et al. 1984, Mitter and Farrell 1991). This empirical observation has generated considerable interest, in both applied and basic fields, in understanding factors that promote or prevent the acquisition of novel plants into an herbivore's diet. Historically, studies have placed much emphasis on host plant chemistry as a determinant of plant use (Dethier 1954, Ehrlich and Raven 1964, Berenbaum 1990). Despite the obvious importance of plant chemistry on in.sect behavior and host choice (Roitberg and Isman 1992), it is often the case that herbivores use plants or plant parts on which larval feeding performance is suboptimal (Damman 1987, Thompson 1988). The weak correspondence between patterns of host plant use and larval performance in herbivores could, in part, be due to nonhost plant quality effects, such as competition, predation, or parasitism, on net fitness (Gilbert 1979, Bernays and Graham 1988, Courtney and Kibota 1990). A potentially important feature of the insect-plant association is how the interactions with the third trophic level change at the onset of a host expansion. It has been suggested that diet breadth in herbivorous insects may be more a function of the biotic interactions between the herbivore and its natural enemies, via "enemy-free space," than of host plant chemistry (Gilbert and Singer 1975, Bernays and Graham 1988). However, evidence supporting this hypothesis is scant (Berdegue et al. 1996).
Natural enemies have long been recognized as important mortality agents and potential selective forces on herbivorous insect populations (Hairston et al. 1960). Natural enemies could be important in molding niche characteristics of herbivores as they escape to "enemy-free space," and may thereby promote the diversification of insect lineages (Van Valen 1973, Lawton 1986). Jeffries and Lawton (1984) suggested that any change in an herbivore's behavior and/or life history that would mitigate the effects of natural enemies could be highly advantageous. One way to escape natural enemies would be to utilize novel plants that provide an ecological refuge from natural enemies (Jeffries and Lawton 1984). Despite the efforts of Bernays and Graham (1988) to rekindle interest in understanding the role that escape from natural enemies plays in determining diet breadth in herbivorous insects, there has been a paucity of studies that have rigorously examined this concept and, in particular, how it pertains to insects on plants. Berdegue et al. (1996) examined 41 studies that were purported to have studied enemy-free space in arthropods: although only 19 explicitly tested the appropriate hypotheses, 84% showed evidence for enemy-free space. This pattern suggests that escape from natural enemies is a potentially important, yet understudied, feature of plant-insect interactions in the context of diet breadth.
In addition, there have been important limitations to many studies of enemy-free space. Often, extant associations of closely related herbivores and their plants are examined and inferences are made regarding the role of natural enemies in determining diet breadth (Gross and Price 1988, Denno et al. 1990, Ohsaki and Sato 1994). In these interspecific studies, it is difficult to distinguish between the causes and consequences of host shifting and escape from natural enemies. If differences in herbivore mortality between normal and novel hosts are not observed, it does not exclude the possibility that differences in enemy pressure once existed and contributed to current patterns of host use (Rausher 1992). Conversely, if differences do currently exist, we do not know if these are the same natural enemies and mortality effects that were present when the host shift first occurred. Intraspecific comparisons of enemy-free space, on the other hand, have often been observational rather than experimental (Brown et al. 1995, Feder 1995).
In this study, we posed the question: does an herbivore experience decreased mortality from natural enemies when it first experiences a novel plant? To test this hypothesis, we took the unique approach of experimentally simulating a host shift, using the dipteran leafminer, Liriomyza helianthi Spencer (Diptera: Agromyzidae). To the best of our knowledge, the study reported herein constitutes the first example of larvae experimentally introduced into novel hosts and subsequently brought to the field for exposure to natural enemy attack. This technique avoids the problems associated with examining extant plant-insect interactions. By experimentally creating a host shift, we were able to mimic enemy-related mortality effects that might exist following a natural host shift, to empirically test for the predicted escape from natural enemies.
MATERIALS AND METHODS
Liriomyza helianthi is a specialist leafminer found exclusively in native wild sunflower (Helianthus annuus) and cocklebur (Xanthium strumarium) (Spencer 1981; C. Gratton, personal observation) in roadside ditches, disturbed, and low-lying areas in Pacific coast states of the United States (Whitson 1992, Hickman 1993). L. helianthi has a life cycle typical of many leafmining agromyzids: females lay eggs in the leaves of their host plant, eggs hatch, and larvae feed in the upper mesophyll layer of the leaf. Larvae progress through three instars in [approximately]5-6 d, under field conditions, before cutting a characteristic crescent-shaped exit hole in the bottom surface of the leaf. Larvae exit the mine and fall to the soil where they pupate. Adults emerge after 10-14 d and return to their host plants to mate and oviposit. L. helianthi is multivoltine during the summer months, with an egg-to-adult cycle of [approximately]21-28 d.
Typical of many herbivores with concealed feeding habits (Hawkins 1994), L. helianthi is vulnerable to a diverse assemblage of hymenopteran parasitoids that accounts for most of the larval mortality from natural enemies. Over 11 species of parasitic hymenoptera have been reared from larvae of this leafminer (mostly Eulophidae); these parasitoids cause 70-90% mortality (C. Gratton, unpublished data). Dominant parasitoid species include endo- and ectoparasitoids, both idiobionts and koinobionts (Askew and Shaw 1986). Koinobionts, usually endoparasitoids that allow continued development of their hosts after parasitism (before consuming them), normally have a more narrow host range than idiobionts. In contrast, idiobionts are most often ectoparasitoids that kill or permanently parasitize their hosts upon attack, and are usually thought to have broader host ranges (Askew and Shaw 1986). Generalist predators such as chrysopids, anthochorids, and predatory mites are observed, but are rarely seen feeding on leafminers (C. Gratton, personal observation).
Experiments were conducted in two fields sites in the Sacramento River Delta of California, United States. In 1994 and 1995, experiments were conducted in a 0.25-ha abandoned field 1 km east of Rio Vista, Sacramento County. This site was flooded until early June, when H. annuus and X. strumarium seeds germinated. Plants flowered by late July, set seed, and died by late September-October. L. helianthi and its parasites were present throughout this period. In 1996, the Rio Vista site was converted to agriculture and a new field site at the Stone Lake Wildlife Refuge (Sacramento County Parks and Recreation/U.S. Fish and Wildlife, 3 km east of Hood, California, on Hood-Franklin Road) was used for the study. A 7.5-ha patch of wild H. annuus just west of the refuge entrance along a dry drainage channel was the primary study area in 1996. Plant phenology at Stone Lake followed a pattern similar to that in Rio Vista.
L. helianthi larvae do not survive well in plants outside the family Asteraceae (C. Gratton, unpublished data). Hence, only plants within the family were used as novel hosts in transfer experiments. In 1994 and 1996, Centaurea solstitialis L. (yellow star thistle) was used as the sole novel plant. C. solstitialis was naturally abundant at both field sites in all years of the experiments, although it was not used by L. helianthi. In 1995, three additional novel plant species were used to represent increasing taxonomic difference from the normal host [ILLUSTRATION FOR FIGURE 1 OMITTED]: within the same genus as the normal host, Helianthus maximilianii Schrader (Maximilian's sunflower); within the same tribe (Heliantheae), Ambrosia artemisiifolia L. (common ragweed); and in a different subfamily (Cichorioideae), Taraxacum officinale Wigg. (dandelion, tribe Lactuceae). Using plants of introduced origin, such as C. solstitialis and T. officinale, increased the "novelty" of those plants because there is a lower possibility that either the leafminer or the parasitoids have had any prior association with the plants in the past. A cultivated "small black" variety of H. annuus (Turner Seeds, Breckenridge, Texas, USA) was used as the control host plant for all experiments. Field tests showed no difference in natural enemy attack rates between H. annuus grown from wild-collected seed and the cultivated variety, and feeding trials demonstrated no developmental differences between wild and cultivated sunflower (Gratton 1997, Gratton and Welter 1998).
All plants were grown under greenhouse conditions in 15-cm pots and were used before they reached the flowering stage. Sunflower plants (H. annuus and H. maximilianii) were at the 8-10 leaf stage and [approximately]50 cm in height. Taraxacum officinale and C. solstitialis were at the rosette stage, with average basal diameters of 35-45 cm. Ambrosia artemisiifolia averaged 40 cm in height and 10-12 leaves. Plants were transplanted into larger 4-L pots with a 2:1 soil to vermiculite mixture. In the field, pots were wrapped with either moistened white muslin cloth or aluminum insulating wrap to prevent overheating. In 1996, 15-cm pots were placed into white 1-L pots lined with plastic that served as a water reservoir and heat shield. Plants were watered as needed throughout the day.
Larval transfers involved removing a larva of L. helianthi from a donor plant and inserting it into a recipient plant (Sehgal 1971, Gratton and Welter 1998). Larvae used for transfers came from a laboratory colony maintained on cultivated H. annuus (Gratton and Welter 1998). Larvae were transferred 12-24 h after egg hatch, when they averaged 0.5 [+ or -] 0.1 mm in length (mean [+ or -] 1 SD, n = 794). Larvae were collected from donor leaves (H. annuus) by peeling back the epidermis of leaves with forceps and removing tunneling larvae with the aid of a fine brush. The larvae were then carefully introduced head first into a small pocket created with a minuten pin (a short, ultrafine pin, frequently used for pinning small insect specimens) in a recipient leaf. Transfers were checked 1-6 h later for mining activity. Movement of mouth hooks and midgut indicated a successful transfer. Unsuccessful transfers were repeated once. Transfers were performed over a one or two day period as necessary. For each plant, five leafminers per plant were transferred, one per leaf, in the distal third of the leaf near the midvein. Larvae transferred into both normal and novels plants were able to develop to pupation and emerge as adults (Gratton and Welter 1998).
Field exposure and determination of larval mortality
Within 12-24 h of the larval transfers, plants were brought to the field site, where naturally occurring H. annuus and X. strumarium, as well as leafminers and parasitoids, were present in abundance. Experimental plants were arranged haphazardly in replicate groups so that each group consisted of one normal plant with transferred larvae accompanied by one plant of each novel species containing transferred larvae. In 1994 and 1996, only two species were paired in each group (the normal H. annuus and the novel C. solstitialis), in 1995, four novel species (H. maximilianii, A. artemisiifolia, T. officinale, and C. solstitialis) and one normal species (H. annuus) were placed in groups in the field. Ten replicate species groups were used for all experiments in 1994, 7-8 replicate species groups in 1995, and 16 replicate species groups in 1996.
Plants were fully exposed in the field by 0900 Pacific Standard Time, and were returned to the laboratory after 33 or 57 h (2 or 3 d) of exposure. Larval development of transferred larvae in the field takes [approximately] 4 d. Differences in larval mortality between plants did not appear to be affected by exposure length in the field (Gratton 1997). Experiments were repeated on consecutive weeks (referred to as "trials"), four times in 1994 (Exp. A-D; 15 August-5 September) and 1995 (Exp. E-H; 27 August-17 September), and five times in 1996 (Exp. I-M; 5 August-9 September).
Due to a high density of grasshoppers at the Stone Lake site, plants were enclosed in cages to prevent total defoliation. Cages (50 x 50 x 110 cm) constructed of nylon netting with [approximately]3.1 mm diameter openings (Ace weave, Nylon Net Company, Memphis, Tennessee, USA) were used to exclude grasshoppers, yet allow movement of parasitoids in and out of the cage. The eulophid parasitoids that attack L. helianthi have cross-sectional diameters of [less than]1 mm. Separate experiments (Gratton 1997) showed no difference between caged and uncaged plants in mortality from parasitoids.
After field exposure, plants were brought back to the laboratory. Using a dissecting microscope (10-60x), leafminers were examined and dissected to determine if they were alive (A), exited (E), parasitized (P), or host fed/stung by parasitoids (HF). Destructive host feeding occurs when a parasitoid female probes a larva with her ovipositor and proceeds to imbibe fluids that are released from the larva's body through the puncture wound, but the parasitoid does not lay an egg (Heinz and Parrella 1989, Kato 1989). Paralyzing venom is usually injected into the host during the initial sting and probing. This stinging behavior causes the larva to enter a suspended state of animation that is eventually fatal. Larvae that have been attacked for host feeding often appear flattened and dry, with melanized puncture wounds clearly visible. Although most larvae were recovered before exiting, scoring exited mines (E) as unparasitized would underestimate mortality in cases where larval-pupal parasitoids are present. However, larval-pupal parasitoids were usually rare, with the exception of 1994. Larvae that died from unknown causes (i.e., not caused by natural enemies: natural death, desiccation) or did not survive the larval transfer were not used in the analysis ([approximately]34% of larvae across all plants).
For each plant, percentage mortality of larvae was calculated as the number killed (P + HF) divided by total available hosts (A + E + P + HF) times 100. This estimate includes mortality from both host feeding/stinging behavior and from parasitism by natural enemies. Analyses using percentage parasitism (number parasitized (P)/(parasitized (P) + alive (A)), a different estimate of mortality that uses only cases in which larvae or parasitoids were present, gave similar results (Gratton 1997).
When parasitoids were found on larvae, they were further categorized as idiobiont (suspending leafminer development) ectoparasitoid, idiobiont endoparasitoid, or koinobiont endoparasitoid (allowing continued development of the leafminer host before eventual host death). Endo- and ectoparasitoids differed in where female parasitoids placed their eggs and where larvae developed relative to the leafminer host. Because dissection of larval hosts usually precluded the successful rearing of parasitoids, additional hosts were collected weekly from neighboring wild H. annuus plants surrounding experimental plants. These hosts were used to determine the species identity of the different parasitoid guilds attacking L. helianthi.
Host size effects on mortality differences between plants
Differential developmental rates in different host plants can be a potentially confounding factor in determining differences in larval mortality between plants. For example, if parasitoids prefer to oviposit in larger hosts and ignore small hosts (Heinz and Parrella 1989), lower larval mortality in plants where development times are slower may be due to larval size differences between plants. To circumvent this potential problem, in 1994 all larval transfers into the novel host, C. solstitialis, were performed before transfers into H. annuus. Laboratory experiments have shown that a 12-24 h head start for larvae in C. solstitialis, for a 2-d field exposure, almost completely eliminated the developmental advantage of larvae in H. annuus (Gratton 1997). In contrast, in 1995 and 1996, a random block design was used for larval transfers, with each block having all plant species represented. Unlike 1994, this design allowed larval size distributions to vary among plants. Length of recovered larvae, measured using an ocular micrometer, was used as a index of the size of larvae when attacked by parasitoids. Heinz (1996) found that, for Chrysocharis nephrus, a [TABULAR DATA FOR TABLE 1 OMITTED] eulophid parasitoid attacking the lepidopteran leafminer Cameraria jacintoensis, the size of a paralyzed leafminer larva is a good correlate of the actual larval size when it was attacked in the field. Because most parasitoids in 1995 and 1996 were idiobionts, which halted leafminer development upon attack, we were able to determine the fate of different-sized larvae upon discovery by parasitoids in the field.
For each replicate group of plant species, a difference in percentage mortality between normal and novel plant species was calculated: [Delta]% mortality = % mortality (novel) - % mortality (normal = H. annuus). Therefore, negative differences indicate lower mortality in novel hosts. Nonparametric tests were used for all mortality analyses, although data are reported as means [+ or -] 1 SD. In 1994 and 1996, a one-tailed signed-ranks test was used to determine if there was significantly greater mortality in the normal host than in the novel host, as is predicted by the enemy-free space hypothesis. In 1995, mortality data were analyzed using Friedman's test (Zar 1984) to assess the effect of plant species and replicate groups on larval mortality. Missing data (some plants either were destroyed or did not contain viable larvae from which to estimate mortality) were estimated iteratively following Zar (1984). Results were not significantly affected if groups with missing cells were omitted from the analysis altogether. When significant plant species effects were observed, multiple comparisons between normal and novel plants were performed, using Dunnett's test on ranks of percentage mortality with H. annuus as the control group. To compare mortality for the combined novel plants vs. normal plants, a nonparametric version of Scheffe's multiple contrast was used (Zar 1984). Data across experimental trials were pooled to get an overall mean, because there was no effect of experimental trial on mortality (Kruskal-Wallis, 1994: [[Chi].sup.2] = 7.46, df = 3, P = 0.06; 1995: [[Chi].sup.2] = 1.37, df = 3, P = 0.71; 1996: [[Chi].sup.2] = 0.978, df = 4, P = 0.91). Differences in final larval lengths were tested by Tukey hsd comparisons. Analyses were performed using JMP 3.1 (SAS Institute 1995).
Parasitism rates on normal and novel hosts
Percentage mortality results are presented separately for each of the three years (1994-1996) in Tables 13. For two of the three years of the study, 1994 and 1996 ([ILLUSTRATION FOR FIGURE 2A, C OMITTED], Tables 1 and 3), there were consistent, significant reductions in mortality for leafminers in novel plants compared to normal host plants. In 1994, differences in percentage mortality from natural enemies ranged from - 15% to -68% and were statistically significant in two of four individual trials. The overall average difference in larval mortality between normal and novel plant species was -41% in 1994, which represented the greatest overall decline seen for any of the three years [ILLUSTRATION FOR FIGURE 2 OMITTED]. Pooling data across all experimental trials in 1996 also resulted in a highly significant overall difference in larval mortality between normal and novel plant species of -13% [ILLUSTRATION FOR FIGURE 2C OMITTED]. Although average larval mortality was greater for normal than novel hosts in each of the five trials conducted in 1996, no single experimental trial in 1996 displayed a statistically significant difference in larval mortality between normal and novel plants [ILLUSTRATION FOR FIGURE 2C OMITTED]. It should be noted that levels of mortality from natural enemies were high in 1996, despite the use of cages (Table 3), suggesting minimal cage effects.
In contrast to the results for 1994 and 1996, there were no differences in larval mortality between normal and novel plants in 1995 ([ILLUSTRATION FOR FIGURE 2B OMITTED], Table 2). This was true irrespective of the four different novel plant species used by leafminers in 1995 tests (Friedman's test for effects of plant species on percentage mortality, Exp. E: [[Chi].sup.2] = 0.656, df = 4, P [greater than] 0.05; Exp. G: [[Chi].sup.2] = 3.03, df = 4, P [greater than] 0.05; Exp. H: [[Chi].sup.2] = 4.44, df = 4, P [greater than] 0.05). Only in the trial started on 3 September (Exp. F) was mortality higher in H. annuus than in T. officinale (Friedman's [[Chi].sup.2] = 15.59, df = 5, P [less than] 0.05, Dunnett's q = 2.29, P [less than] 0.05; [ILLUSTRATION FOR FIGURE 2B OMITTED]). Combining all novel plants together for analysis still did not reveal any significant mortality differences compared to the normal host in 1995 (Scheffe's S = 0.13, 6.74, 0.04, 0.18, in Exp. E-H, respectively, df = 4, P [greater than] 0.05). The overall mean difference of 3% was not statistically different from zero (signed-ranks = -15, P = 0.67; [ILLUSTRATION FOR FIGURE 2B OMITTED]).
Variation in parasitoid guilds
The difference among years in larval mortality between normal and novel host plants was related to the variability in parasitoid guilds. Dissection of parasitized larvae in different host plants found a general pattern of dominance of endoparasitoids in 1994 and 1996, and the appearance of a significant ectoparasitoid guild in 1995 [ILLUSTRATION FOR FIGURE 3 OMITTED]. When the percentage of ectoparasitoids in the guild was correlated to the difference in percentage larval mortality between normal and novel plants across all trials over the 3-yr study, there was [TABULAR DATA FOR TABLE 2 OMITTED] a positive relationship [ILLUSTRATION FOR FIGURE 4 OMITTED]. The correlation was marginally significant using parametric methods (r = 0.55, P = 0.063) and highly significant using a nonparametric measure of association (Spearman's [Rho] = 0.67, P = 0.016). This suggests that when ectoparasitoids represented a larger proportion of the parasitoid guild, the relative advantage to leafminer larvae developing in novel plants decreased.
Rearing of parasitoids from experimental plants, as well as naturally occurring H. annuus, showed that idiobiont ectoparasitoids were represented by two species of Diglyphus, D. begini (Ashmead), and D. pulchripes (Crawford) (Eulophidae), which were especially common in 1995. In 1994, a significant percentage of parasitism was caused by a koinobiont endoparasitoid, Opius dimidiatus (Ashmead) (Braconidae). This braconid species was nearly absent in larvae collected in 1995 and 1996, to be replaced by Chrysocharis ainsliei Crawford. In 1994, endoparasitoids were represented mostly by Closterocerus utahensis Crawford and C. cinctipennis Ashmead (of which males cannot be distinguished from each other) and Neochrysocharis arizonensis Crawford, with Neochrysocharis becoming more prevalent in 1995. In 1996, Neochrysocharis was the dominant parasitoid overall, and Closterocerus was completely absent.
Overall mortality levels on normal and novel hosts
Another factor that appeared to be related to differences in larval mortality between normal and novel plants was the overall level of parasitoid-inflicted mortality. Mean percentage larval mortality in H. annuus (% mortality [normal]) was correlated with the mean difference in percentage larval mortality between the [TABULAR DATA FOR TABLE 3 OMITTED] novel hosts and normal hosts ([Delta]% mortality) using data from all years. We used the mean percentage mortality in H. annuus control plants for each experiment as an index of "background" mortality levels and general parasitoid abundance in the field. A mean of the estimates was calculated for each experiment because the inherent autocorrelation between the difference in percentage mortality (novel - normal) and percentage mortality (normal) within each plant pair would force a negative correlation between the variables. For 1995 data, which had multiple novel plants, an average of all novel plants was calculated (Table 2). Data from Exp. D (5 September 1994) were excluded as outliers because they fell outside the 95% confidence interval of the joint bivariate distribution of [Delta]% mortality and % mortality (normal). The correlation between estimates of percentage mortality was positive and statistically significant ([ILLUSTRATION FOR FIGURE 5 OMITTED]; r = 0.61, P = 0.034, [Rho] = 0.61, P = 0.038). The observed positive correlation between the variables suggests that the pattern was not an artifact of the potential autocorrelation. This result indicates that, as overall levels of mortality increase in the field, the advantage to larvae developing in novel plants decreases relative to normal plants.
A multiple linear regression analysis was performed to analyze the joint effects of overall mean mortality levels (% mortality [normal]) and mean ectoparasitoid representation (by percentage) in the natural enemy guild on the difference in percentage mortality between normal and novel plants [ILLUSTRATION FOR FIGURES 4 AND 5 OMITTED]. There was a highly significant effect of both percentage mortality (normal) ([F.sub.1,9] = 9.19, P = 0.014) and percentage of ectoparasitoids in the guild ([F.sub.1,9] = 7.34, P = 0.024). These two factors explained almost two-thirds of the variation in mean larval escape between novel and normal plants ([R.sup.2] = 0.656). In addition, there was no significant relationship between overall mortality levels and ectoparasitoid guild representation (r = 0.0353, P = 0.913), suggesting that these factors operate independently to influence larval escape from parasitism in normal vs. novel plants.
Size-related parasitism differences between normal and novel plants
Overall, larval sizes were smaller in the novel host, C. solstitialis, than in the normal host ([ILLUSTRATION FOR FIGURE 6 OMITTED], length of recovered larvae was 1.192 [+ or -] 0.525 mm (mean [+ or -] 1 SD) in C. solstitialis, and 1.507 [+ or -] 0.507 mm in H. annuus; t = -5.5, df = 365, P [less than] 0.0001). This suggests that development was retarded, to a certain extent, for L. helianthi in C. solstitialis (see also Gratton and Welter 1998). If parasitoids tend to ignore smaller sized hosts for attack, then this could account for the significant reduction in larval mortality on novel hosts. However, we found that smaller larvae in either host plant were not ignored by parasitoids; rather, they were used for host feeding instead of oviposition [ILLUSTRATION FOR FIGURE 6 OMITTED]. In addition, both host feeding and oviposition resulted in larval death. Length measurements of larvae demonstrated that parasitoids responded to smaller larvae present in C. solstitialis with a tendency to oviposit into smaller larvae in C. solstitialis than in H. annuus (Tukey's hsd P [less than] 0.05). On both C. solstitialis and H. annuus, the smallest larvae encountered were used for host feeding/stinging, and larger larvae were used for oviposition [ILLUSTRATION FOR FIGURE 6 OMITTED]. In addition, larvae used for oviposition in C. solstitialis and those used for host feeding in H. annuus were approximately the same length ([ILLUSTRATION FOR FIGURE 6 OMITTED]; Tukey's hsd P [greater than] 0.05). Moreover, Heinz and Parrella (1989) suggest that only very small leafminer larvae (first instar) are overlooked by one leafminer parasitoid, Diglyphus. By the time plants were brought to the field in our experiment, larvae were entering the second instar, making them large enough for detection by foraging parasitoids. Hence, although size did influence whether a leafminer larva was used for oviposition or host feeding, small larvae were not ignored by parasitoids. The net result of lower attack rates on novel hosts was, therefore, a function of lower visitation rates or rejection of hosts by parasitoids before stinging, and not due simply to larval size differences between plants.
Does enemy-free space exist?
Larvae of Liriomyza helianthi experimentally transferred to a novel host incurred [approximately]17% less mortality inflicted by parasitoids than when larvae were placed in normal hosts. These results strongly suggest that enemy-free space is available on novel hosts and may provide the impetus for host range expansion. However, we did detect significant temporal variation in larval mortality levels across years: in 1994 and 1996, escape from natural enemies was significant and large (-41% and -13% less on novel plants, respectively), whereas in 1995, there was little enemy-free space available on novel hosts. Variability in enemy escape was influenced by host plant species [ILLUSTRATION FOR FIGURE 2 OMITTED]. That is, feeding on novel hosts promoted escape from natural enemies relative to feeding on normal hosts. Enemy escape was further modulated by the relative composition of the parasitoid guilds present at different times [ILLUSTRATION FOR FIGURES 3, 4 OMITTED]. Finally, levels of overall parasitoid-inflicted mortality, an index of relative natural enemy abundance in the field, also explained some of the variation in enemy-free space [ILLUSTRATION FOR FIGURE 5 OMITTED]. This variation makes enemy-free space a dynamic feature of plant-herbivore interactions and may help to promote shifts to novel host plants.
Parasitoid guilds and parasitism levels
Our results suggest several mechanisms that may be responsible for enemy-free space on novel plants. One possibility is that differences in larval mortality between normal and novel plants reflect an inability of parasitoids to recognize novel plants as potential sites for host occurrence (sensu Vinson 1976). Leafminer parasitoids have been shown to respond to host plant volatiles (Finidori-Logli et al. 1996, Olivera and Bordat 1996), and differences in secondary plant chemistry exist among the Asteraceae hosts used in our study (Heywood et al. 1977). Thus, variation in plant chemistry could influence host searching and acceptance behavior of parasitoids in such a way as to reduce the probability of detection and utilization of L. helianthi larvae on novel hosts (Campbell and Duffey 1979). Some Centaurea species, for example, as well as other plants in the Asteraceae, contain toxic sesquiterpene lactones with known antiherbivore effects (Mabry et al. 1977, Mabry and Gill 1979, Landau et al. 1994). Such compounds could have adverse effects on the third trophic level as well, by either repelling and/or inhibiting parasitoid foraging and probing behavior on these novel plants. Denno et al. (1990) found that Phratora vitellinae (Chrysomelidae) that fed on the salicylate-rich Salix fragilis and S. dasyclados repelled coccinellid predators more effectively than when they consumed the salicylate-poor S. viminalis, suggesting that sequestration of plant secondary compounds may play a role in escape from natural enemies. On the other hand, if parasitoids use host cues rather than plant cues to locate prey (Kato 1989, Nelson and Roitberg 1993), then feeding on novel plants may not provide much enemy-free space. Gross and Price (1988) found that a presumed host shift to a novel host by a lepidopteran leafminer (Tildenia inconspicuella, Lepidoptera: Gracillaridae) resulted in increased parasitism from parasitoids. In this system, natural enemies were more influenced by the feeding mode of the herbivore (e.g., internal vs. external feeding) than by differences between the two host plant species. Nevertheless, it is likely that the decrease in mortality of L. helianthi larvae on novel hosts is due, in part, to differences in secondary plant chemistry between these two plants.
Variation in enemy-free space could also be brought about by changes in the parasitoid guild structure between years. For example, in 1995, when ectoparasitoids were most common [ILLUSTRATION FOR FIGURE 3 OMITTED], there was no evidence for enemy-free space on novel plants [ILLUSTRATION FOR FIGURE 2B OMITTED]. In addition, our correlative data suggest that in trials when ectoparasitoids were relatively more common, escape to enemy-free space via host shifting was less likely [ILLUSTRATION FOR FIGURE 4 OMITTED]. It has been proposed that ectoparasitoids have broader host ranges than endoparasitoids and are, in general, less influenced by host plant factors than are ectoparasitoids (Askew and Shaw 1986, Godfray 1994). The association between an increase in prevalence of the ectoparasitoid guild, represented by Diglyphus begini and D. pulchripes, and a decrease in differences in larval mortality between novel and normal plants [ILLUSTRATION FOR FIGURE 4 OMITTED] could be due to the generalist nature of Diglyphus, a solitary idiobiont ectoparasitoid. Diglyphus begini, for example, a common parasitoid of Liriomyza spp., has been found to attack a variety of agromyzid leafminers (also one tephritid) in at least 12 plant families (Heinz and Parrella 1990). Our findings [ILLUSTRATION FOR FIGURE 4 OMITTED] are consistent with the idea that ectoparasitoids are relatively less sensitive to host plant effects (Askew 1994).
Unlike the ectoparasitoid Diglyphus, which has been extensively studied and used for the biological control of greenhouse leafminers in the United States and Europe (Heinz et al. 1988, van Lenteren and Woets 1988, Heinz and Parrella 1989, Heinz and Parrella 1990), the host ranges and biologies of the endoparasitoids reared in this study have not been examined in detail. Neochrysocharis and Closterocerus and their congeners, for example, are found to parasitize mostly dipteran leafminers, as well as some lepidopteran and coleopteran leafminers, on various host plants (LaSalle and Parrella 1991, Hansson 1994, 1995). The braconid, Opius dimidiatus, and the eulophid, Chrysocharis ainsliei, endoparasitoids have been reared mostly from agromyzid hosts. In the absence of more information on the host range of these endoparasitoids, we are left with the general pattern that endoparasitoids typically have more restricted diet breadth and are also more influenced by host plants (although Shaw  urges caution when making sweeping generalizations). Ultimately, experimental data will be necessary to conclusively show that the endoparasitoids found attacking L. helianthi, (e.g., N. arizonensis) are relatively more sensitive to differences between host plants than are the ectoparasitoids (Diglyphus spp.), and that this difference is responsible for some of the observed variation in enemy-free space.
In addition to changes in guild structure between years, another factor that could influence variation in enemy-free space is changes in abundance of parasitoids. We used levels of mortality in the normal host as a proxy for natural enemy abundance in the field. Our data showed that there was an increase of larval mortality in novel plants relative to normal plants (i.e., less enemy-free space) as the overall "background" activity of parasitoids increased [ILLUSTRATION FOR FIGURE 5 OMITTED]. It is possible that, at times of high natural enemy activity, available leafminer hosts in normal plants were depleted, driving parasitoids to forage on less favored novel plants. In addition, the fact that overall mortality rates and the representation of ectoparasitoids in the community were not correlated suggests that these factors operate independently to influence larval mortality. Hence, the ratio of the different parasitoid guilds present (ecto vs. endoparasitoids), in conjunction with their relative abundance at different times, influence escape from natural enemies.
Size-related parasitism differences between normal and novel plants
Developmental differences between larvae in normal and novel plants could be a possible confounding effect influencing mortality differences. For example, lower mortality in the novel host could be simply due to parasitoids ignoring smaller larvae feeding in these inferior host plants. However, the two different approaches taken to account for developmental differences between plants (synchronizing development and size measurements) showed that differences in larval mortality between plants were due to factors other than size alone. In 1994, larval transfers were staggered so as to ensure similar size distributions of larvae in both normal and novel plants. Thus, the significant differences in larval mortality observed between normal and novel plants [ILLUSTRATION FOR FIGURE 2A OMITTED] cannot be attributed to differences in larval size distributions between plants. Data from 1995 and 1996, when larval sizes were allowed to vary between hosts, showed differences between C. solstitialis and H. annuus in larval size used for host feeding and oviposition. It is important to note that small larvae in C. solstitialis were not ignored, but were used for host feeding and, therefore, were scored as killed in the mortality calculations. Thus, any differences in mortality between plants are due to lower visitation/discovery rates of the novel plants rather than larval size differences between plants.
Host shifts: physiological costs vs. enemy-free space
It has been proposed that during the initial stages of a host shift, the benefit that a herbivore receives from escaping natural enemies may act to offset physiologically based fitness costs associated with developing in a novel host (Price et al. 1980, Jaenike 1985, 1990). In laboratory feeding experiments on normal and novel plants, L. helianthi larvae experienced [approximately]66.7% larval survivorship in the novel C. solstitialis compared to 91.3% in the normal host H. annuus (Gratton and Welter 1998). In our field experiments, mortality from natural enemies averaged 46.3% in C. solstitialis compared to 64.2% in H. annuus. based on these data, larval survivorship in novel plants is therefore similar to that in normal hosts, after adjusting for natural enemies (novel: 0.358 = 0.66711 - 0.463]; normal: 0.327 = 0.91311 - 0.642]). Thus, in this system, escape from natural enemies has the potential to offset performance costs incurred when developing on a novel host, thereby promoting a host range expansion. Feder (1995) found similar results in the apple maggot, Rhagoletis pomonella (Tephritidae): despite having higher egg-pupal mortality on the novel apple host (Prokopy et al. 1988), larvae experienced significantly lower rates of parasitism on apple than on ancestral hawthorn plants. When accounting for natural enemies, egg-to-pupal survivorship was about the same between the two hosts, suggesting that escape from natural enemies facilitated the use of novel hosts (Feder 1995). Brown et al. (1995) showed a similar pattern in studying Eurosta solidaginis (Tephritidae), a gallmaker that has expanded its diet breadth beyond the ancestral host, Solidago altissima, to include the novel Solidago gigantea. Although flies that utilized the derived host experienced higher mortality associated with their host plant, survivors experienced lower mortality from natural enemies than did flies in the normal host (Brown et al. 1995). Changes in associations of herbivores with their natural enemies may therefore be a general feature of host shifting. If the result is enemy-free space, there may be a sufficient increase in fitness on a novel host plant to facilitate the inclusion of the novel host into the herbivore's diet.
In addition to escape from natural enemies and larval performance, there may be other factors that could influence herbivore fitness on novel hosts. Host plant phenology (Feder et al. 1993), competitive interactions (Denno et al. 1995), and mate location (Colwell 1986), as well as physiological considerations, such as changes in fecundity (Parrella 1987), and genetic tradeoffs (Jaenike 1990), may further decrease or constrain fitness on novel hosts. For example, developing on some hosts may result in longer development times and potentially increased exposure to natural enemies (Feeny 1976, Price et al. 1980, Clancy and Price 1987, Benrey and Denno 1997). Increases in larval mortality in novel hosts may eliminate survivorship differences between normal and novel plants, even though natural enemies may be biased toward normal hosts. Nevertheless, between-host comparisons of natural enemy-caused mortality due to increased exposure times have given mixed results (Benrey and Denno 1997). For L. helianthi, increased exposure duration did not appear to significantly increase larval mortality in novel hosts (Gratton 1997). In general, however, additional unmeasured fitness costs may tilt the balance toward lower fitness on novel hosts relative to normal hosts, thus inhibiting incorporation of the novel host into the diet of the herbivore.
For some herbivorous insects, many of the essential characteristics for host shifting are present (sensu Berenbaum 1990): the use of novel plants does occasionally occur via oviposition "mistakes" (Wiklund 1975, Fox and Lalonde 1993, Larsson and Ekbom 1995); herbivore larvae often can develop successfully in plants that are not used for oviposition by females (Dethier 1954, Bernays and Chapman 1987, Thompson 1988, Courtney and Kibota 1990, Gratton and Welter 1998); and enemy-free space can exist and may offset physiological costs (Brown et al. 1995, Feder 1995). However, host shifts in nature are usually rare (Bush 1975). If enemy-free space is important in facilitating host shifts, then how much enemy-free space is sufficient, in magnitude, frequency, and duration, to allow a novel host plant to be successfully incorporated into an insect's diet?
Our results have shown that enemy-free space is not an all-or-none phenomenon, but is variable from year to year. It may be useful, therefore, to view enemy-free space as a window of opportunity that can offset physiological hurdles that may be present when utilizing novel hosts (e.g., Hanks et al. 1995). This window, however, is ephemeral (due to variation in natural enemy abundance and guild composition) and may open only briefly and intermittently, making shifts likely only in certain circumstances. To understand how important enemy-free space is in facilitating host shifts, it may be necessary to quantify the distribution of the enemy-free space advantage over a longer period of time. The debate, ultimately, is not whether plant chemistry (i.e., physiological performance, behavioral responses) or natural enemies is more important in determining diet breadth in herbivores (see Bernays and Graham 1988 and responses, Courtney and Kibota 1990). Rather, the issue is the relative balance of these different factors at various stages of the insect's life and at different stages of a host expansion. Alternatively, the rate-limiting step in host range expansion may largely be a function of the behavioral responses of ovipositing females to host plant stimuli (Dethier 1954, Futuyma 1983). The observation that host shifts are, in general, rare in herbivorous insects (Bush 1975) may partly reflect the difficulty in lining up all of the requisite elements for a successful host shift: oviposition behavior, larval performance (physiology), and natural enemies. Identifying and quantifying the various behavioral, physiological, and ecological parameters involved in host shifting will allow for better estimation of the likelihood of host shifts.
In conclusion, experimentally created hosts shifts in the leafmining fly L. helianthi resulted in decreased larval mortality from parasitoids. The benefit of this experimental host shift was diminished in the presence of the more generalist ectoparasitoid Diglyphus spp., as well as by overall high mortality levels. In years when endoparasitoids were dominant and mortality levels were [less than] 75%, there was, on average, a 22% reduction in mortality of larvae feeding in novel hosts. These results show that the extent to which enemy-free space exists for a leafmining herbivore utilizing a novel host depends on variability in the natural enemy complex between years. Enemy-free space is, therefore, an important component of the environment that should be addressed when examining questions of host shifting and diet breadth in herbivorous insects.
We thank Nick Mills, Wayne Sousa, Sujaya Udayagiri, Art Zangerl, and two anonymous reviewers, whose suggestions greatly improved earlier drafts of the manuscript. Special thanks go to Bob Denno, who critiqued many versions of this manuscript, and Brenda Gratton, for illustrations of plants. We also thank the County Parks of Sacramento for allowing access to the Stone Lake Wildlife Refuge. Leafminer identifications were verified by K. Spencer (Exwell Farm, Bray Shop, Callington, PL17 8QJ, Cornwall, England). Parasitoids were identified by J. LaSalle (International Institute of Entomology, UK) and R. Wharton (Texas A&M University). Voucher specimens of L. helianthi and parasitoids were deposited at the University of California, Berkeley, Essig Museum of Entomology. Funding for this study was provided by a UC Berkeley Vice Chancellor's Graduate Research Grant to C. Gratton and a Committee on Research Grant to S. C. Welter.
Askew, R. R. 1994. Parasitoids of leaf-mining Lepidoptera: what determines their host ranges? Pages 177-202 in B. A. Hawkins and W. Sheehan, editors. Parasitoid community ecology. Oxford University Press, Oxford, UK.
Askew, R. R., and M. R. Shaw. 1986. Parasitoid communities: their size, structure and development. Pages 225264 in J. Wagge and D. Greathead, editors. Insect parasitoids. Academic Press, London, UK.
Benrey, B., and R. E Denno. 1997. The slow-growth-high-mortality hypothesis: a test using the cabbage butterfly. Ecology 78:987-999.
Berdegue, M., J. T. Trumble, J. D. Hare, and R. A. Redak. 1996. Is it enemy-free space? The evidence for terrestrial insects and freshwater arthropods. Ecological Entomology 21:203-217.
Berenbaum, M. R. 1990. Evolution of specialization in insect-umbellifer associations. Annual Review of Entomology 35:319-343.
Bernays, E. A., and R. F. Chapman. 1987. Evolution of deterrent responses by phytophagous insects. Pages 159-173 in R. F. Chapman, E. A. Bernays, and J. G. Stoffolano, editors. Perspectives in chemoreception and behavior. Springer-Verlag, New York, New York, USA.
Bernays, E. A., and M. Graham. 1988. On the evolution of host specificity in phytophagous arthropods. Ecology 69: 886-892.
Brown, J. M., W. G. Abrahamson, R. A. Packer, and P. A. Way. 1995. The role of natural-enemy escape in a gallmaker host-plant shift. Oecologia 104:52-60.
Bush, G. L. 1975. Modes of animal speciation. Annual Review of Ecology and Systematics 6:339-357.
Campbell, B.C., and S.S. Duffey. 1979. Tomatine and parasitic wasps: potential incompatibility of plant antibiosis with biological control. Science 205:700-702.
Clancy, K. M., and P. W. Price. 1987. Rapid herbivore growth enhances enemy attack: sublethal plant defenses remain a paradox. Ecology 68:736-738.
Colwell, R. K. 1986. Population structure and sexual selection: lessons from hummingbird flower mites. Pages 406424 in S. Karlin, editor. Evolutionary processes and theory. Academic Press, Orlando, Florida, USA.
Courtney, S. P., and T. T. Kibota. 1990. Mother doesn't know best: selection of hosts by ovipositing insects. Pages 162-188 in E. A. Bernays, editor. Insect-plant interactions. Volume 2. CRC Press, Boca Raton, Florida, USA.
Damman, H. 1987. Leaf quality and enemy avoidance by larvae of a pyralid moth. Ecology 68:87-97.
Denno, R. E, S. Larsson, and K. L. Olmstead. 1990. Role of enemy-free space and plant quality in host-plant selection by willow beetles. Ecology 71:124-137.
Denno, R. E, M. S. McClure, and J. R. Ott. 1995. Interspecific interactions in phytophagous insects: competition reexamined and resurrected. Annual Review of Entomology 40:297-331.
Dethier, V. G. 1954. Evolution of feeding preference in phytophagous insects. Evolution 8:33-54.
Ehrlich, P. R., and P. H. Raven. 1964. Butterflies and plants: a study in coevolution. Evolution 18:586-608.
Feder, J. L. 1995. The effects of parasitoids on sympatric host races of Rhagoletis pomonella (Diptera: Tephritidae). Ecology 76:801-813.
Feder, J. L., T. A. Hunt, and L. Bush. 1993. The effect of climate, host plant phenology and host fidelity on the genetics of apple and hawthorn infesting races of Rhagoletis pomonella. Entomologia Experimentalis et Applicata 69: 117-135.
Feeny, P. 1976. Plant apparency and chemical defense. Pages 1-40 in J. W. Wallace and R. L. Mansell, editors. Biochemical interactions between plants and insects. Plenum Press, New York, New York, USA.
Finidori-Logli, V., A. G. Bagneres, and J. L. Clement. 1996. Role of plant volatiles in the search for a host by parasitoid Diglyphus isaea (Hymenoptera: Eulophidae). Journal of Chemical Ecology 22:541-558.
Fox, C. W., and R. G. Lalonde. 1993. Host confusion and the evolution of insect diet breadths. Oikos 67:577-581.
Futuyma, D. J. 1983. Evolutionary interactions among herbivorous insects and plants. Pages 207-231 in D. J. Futuyma and M. Slatkin, editors. Coevolution. Sinauer, Sunderland, Massachusetts, USA.
Gilbert, L. E. 1979. Development of theory in the analysis of insect-plant interactions. Pages 117-154 in D. J. Horn, G. R. Stairs, and R. D. Mitchell, editors. Analysis of ecological systems. Ohio State University Press, Columbus, Ohio, USA.
Gilbert, L. E., and M. C. Singer. 1975. Butterfly ecology. Annual Review of Ecology and Systematics 6:365-397.
Godfray, H. C. J. 1994. Parasitoids: behavioral and evolutionary ecology. Princeton University Press, Princeton, New Jersey, USA.
Gratton, C. 1997. Ecology of host shifts in the leaf-mining fly, Liriomyza helianthi (Diptera: Agromyzidae). Dissertation. University of California, Berkeley, California, USA.
Gratton, C., and S.C. Welter. 1998. Oviposition preference and larval performance of Liriomyza helianthi (Diptera: Agromyzidae) on normal and novel host plants. Environmental Entomology 27:926-935.
Gross, P., and P. W. Price. 1988. Plant influences on parasitism of two leafminers: a test of enemy-free space. Ecology 69:1506-1516.
Hairston, N. G., F. E. Smith, and L. B. Slobodkin. 1960. Community structure, population control and competition. American Naturalist 94:421-425.
Hanks, L. M., J. C. Millar, and T. D. Paine. 1995. Biological constraints on host-range expansion by the wood-boring beetle Phoracantha semipunctata (Coleoptera: Cerambycidae). Annals of the Entomological Society of America 88:183-188.
Hansson, C. 1994. Re-evaluation of the genus Closterocerus Westwood (Hymenoptera: Eulophidae), with a revision of the Nearctic species. Entomologica Scandinavica 25:1-25.
-----. 1995. Revision of the Nearctic species of Neochrysocharis Kurdjumov (Hymenoptera: Eulophidae). Entomologica Scandinavica 26:27-46.
Hawkins, B. A. 1994. Pattern and process in host-parasitoid interactions. Cambridge University Press, Cambridge, UK.
Heinz, K. M. 1996. Host size selection and sex allocation behavior among parasitoid trophic levels. Ecological Entomology 21: 218-226.
Heinz, K. M., J.P. Newman, and M.P. Parrella. 1988. Biological control of leafminers on greenhouse marigolds. California Agriculture 42:10-12.
Heinz, K. M., and M.P. Parrella. 1989. Attack behavior and host size selection by Diglyphus begini on Liriomyza trifolii in chrysanthemum. Entomologia Experimentalis et Applicata 53:147-156.
Heinz, K. M., and M.P. Parrella. 1990. Holarctic distribution of the leafminer parasitoid Diglyphus begini (Hymenoptera: Eulophidae) and notes on its life history attacking Liriomyza trifolii (Diptera: Agromyzidae) in chrysanthemum. Annals of the Entomological Society of America 83:916924.
Heywood, V. H., J. R. Harborne, and B. L. Turner. 1977. The biology and chemistry of the compositae. Academic Press, London, UK.
Hickman, J. C., editor. 1993. The Jepson manual: higher plants of California. University of California Press, Berkeley, California, USA.
Jaenike, J. 1985. Parasite pressure and the evolution of amanitin tolerance in Drosophila. Evolution 39:1295-1301.
-----. 1990. Host specialization in phytophagous insects. Annual Review of Ecology and Systematics 21:243-273.
Jeffries, M. J., and J. H. Lawton. 1984. Enemy-free space and the structure of biological communities. Biological Journal of the Linnean Society 23:269-286.
Kato, M. 1989. Host-handling behavior of the parasitoid, Diglyphus minoeus (Hym.: Eulophidae), parasitizing the honeysuckle leaf-miner, Phytomyza lonicerae (Dipt.: Agromyzidae). Entomophaga 34:503-509.
Landau, I., H. Muller-Scharer, and P. I. Ward. 1994. Influence of cnicin, a sesquiterpene lactone of Centaurea maculosa (Asteraceae), on specialist and generalist insect herbivores. Journal of Chemical Ecology 20:929-942.
Larsson, S., and B. Ekbom. 1995. Oviposition mistakes in herbivorous insects: confusion or a step towards a new host plant. Oikos 71:155-160.
LaSalle, J., and M.P. Parrella. 1991. The chalcidoid parasites (Hymenoptera, Chalcidoidea) of economically important Liriomyza species (Diptera, Agromyzidae) in North America. Proceedings of the Entomological Society of Washington 93:571-591.
Lawton, J. H. 1986. The effect of parasitoids on phytophagous insect communities. Pages 265-287 in J. Wagge and D. Greathead, editors. Insect parasitoids. Academic Press, London, UK.
Mabry, T. J., and J. E. Gill. 1979. Sesquiterpene lactones and other terpenoids. Pages 502-537 in G. A. Rosenthal and D. H. Janzen, editors. Herbivores: their interactions with secondary plant metabolites. Academic Press, New York, New York, USA.
Mabry, T. J., J. E. Gill, W. C. Burnett, and S. B. Jones. 1977. Antifeedant sesquiterpene lactones in the Compositae. Pages 179-184 in P. A. Hedin, editor. Host plants' resistance to pests. American Chemical Society, Washington, D.C., USA.
Mitter, C., and B. Farrell. 1991. Macroevolutionary aspects of insect/plant interactions. Pages 35-78 in E. A. Bernays, editor. Insect-plant interactions. Volume 3. CRC Press, Boca Raton, Florida, USA.
Nelson, J. M., and B. D. Roitberg. 1993. Factors governing host discrimination by Opius dimidiatus (Ashmead) (Hymenoptera: Braconidae). Journal of Insect Behavior 6:1324.
Ohsaki, N., and Y. Sato. 1994. Food plant choice of Pieris butterflies as a trade-off between parasitoid avoidance and quality of plants. Ecology 75:59-68.
Olivera, C. R., and D. Bordat. 1996. Influence of Liriomyza species (Diptera: Agromyzidae) and their host plants on oviposition by Opius dissitus females (Hymenoptera: Braconidae). Annals of Applied Biology 128:399-404.
Parrella, M.P. 1987. Biology of Liriomyza. Annual Review of Entomology 32:201-224.
Price, P. W., C. E. Bouton, P. Gross, B. A. McPheron, J. N. Thompson, and A. E. Weis. 1980. Interactions among three trophic levels: influence of plants on interactions between insect herbivores and natural enemies. Annual Review of Ecology and Systematics 11:41-65.
Prokopy, R. J., S. R. Diehl, and S.S. Cooley. 1988. Behavioral evidence for host races in Rhagoletis pomonella flies. Oecologia 76:293-295.
Rausher, M.D. 1992. Natural selection and the evolution of plant-insect interactions. Pages 20-88 in B. Roitberg and M. B. Isman, editors. Insect chemical ecology. Chapman and Hall, New York, New York, USA.
Roitberg, B., and M. B. Isman. 1992. A physiological perspective. Chapman and Hall, New York, New York, USA.
SAS Institute. 1995. JMP 3.1. SAS Institute, Cary, North Carolina, USA.
Sehgal, V. 1971. Biology and host-plant relationships of an oligophagous leaf miner Phytomyza matricariae Hendel (Diptera: Agromyzidae). Quaestiones Entomologicae 7: 255-280.
Shaw, M. R. 1994. Parasitoid host ranges. Pages 111-144 in B. A. Hawkins and W. Sheehan, editors. Parasitoid community ecology. Oxford University Press, Oxford, UK.
Spencer, K. A. 1981. A revisionary study of the leaf-mining flies (Agromyzidae) of California. Agricultural Sciences Publications, University of California, Berkeley, California, USA.
Strong, D. R., J. H. Lawton, and T. R. E. Southwood. 1984. Insects on plants. Harvard University Press, Cambridge, Massachusetts, USA.
Thompson, J. N. 1988. Evolutionary ecology of the relationship between oviposition preferences and performance of offspring in phytophagous insects. Entomologia Experimentalis et Applicata 47:3-14.
van Lenteren, J. C., and J. Woets. 1988. Biological and integrated pest control in greenhouses. Annual Review of Entomology 33:239-269.
Van Valen, L. 1973. A new evolutionary law. Evolutionary Theory 1:1-30.
Vinson, S. B. 1976. Host selection by insect parasitoids. Annual Review of Entomology 21:109-133.
Whitson, T. D., editor. 1992. Weeds of the West. Western Society of Weed Science, Laramie, Wyoming, USA.
Wiklund, C. 1975. The evolutionary relationship between adult oviposition preferences and larval host plant range in Papilio machaon. Oecologia 18:185-197.
Zar, J. H. 1984. Biostatistical analysis. Prentice-Hall, Englewood Cliffs, New Jersey, USA.
|Printer friendly Cite/link Email Feedback|
|Author:||Gratton, Claudio; Welter, Stephen C.|
|Date:||Apr 1, 1999|
|Previous Article:||Bottom-up limitation of predaceous arthropods in a detritus-based terrestrial food web.|
|Next Article:||Conditional outcomes of interactions: the pollinator-prey conflict of an insectivorous plant.|