- Top of page
- Materials and methods
A long-standing goal in ecology and evolutionary biology has been to understand factors that determine diet breadth, and herbivorous insects have figured prominently in this effort (Futuyma & Moreno, 1988; Jaenike, 1990). Variation in diet breadth among herbivore species has commonly been discussed in terms of female choice (preference) and offspring fitness (performance) (Jaenike, 1978; Thompson, 1988; Gripenberg et al., 2010). These traits are often studied in the laboratory, in conditions that exclude many of the interspecific interactions experienced by herbivores in the wild (Agosta & Klemens, 2009; Forister & Wilson, 2013). In particular, traditional assays of preference and performance do not account for tradeoffs in host use that potentially involve natural enemies such as predators and parasitoids. Recently there has been a call to include more trophic levels when studying host use in order to examine how interaction complexity influences plant–herbivore dynamics (Singer & Stireman, 2005; Forister et al., 2012). Here we use the Melissa blue butterfly, Lycaeides melissa (W. H. Edwards), to examine interactions with mutualists and parasitoids in butterfly populations associated with multiple host plants.
Lycaeides melissa is a well-studied and widespread lycaenid butterfly native to western North America (Nice & Shapiro, 1999; Fordyce et al., 2002; Lucas et al., 2008; Gompert et al., 2010; Forister & Scholl, 2012; Scholl et al., 2012). Lycaeides melissa uses a variety of plants in the pea family (Fabaceae) as hosts, including the exotic host alfalfa (Medicago sativa L.), which was introduced to North America within the last 200 years (Michaud et al., 1988). In general, alfalfa is a poor host for L. melissa caterpillars: individuals reared experimentally on alfalfa can be up to 70% smaller than individuals reared on a common native host, Astragalus canadensis L. (Forister et al., 2009). As with many herbivorous insects, smaller size has a fitness cost because smaller females lay fewer eggs and are less attractive to males (Forister et al., 2009; Forister & Scholl, 2012). Yet, larval performance experiments have been carried out in the laboratory, while larval fitness may be very different in the field where caterpillars are exposed to interactions with other species that may be enemies, competitors or mutualistic partners.
About three-fourths of lycaenid butterflies with well-studied life histories associate with ants; these relationships range from parasitic to mutualistic, and from facultative to obligate (Pierce et al., 2002). Lycaeides melissa caterpillars are facultatively myrmecophilous, providing secretions rich in sugars and amino acids that attract the attention of ants that tend and presumably protect caterpillars. Forister et al. (2011a) found that the presence of ants reduced rates of attack by generalist predators of L. melissa on the novel host alfalfa, but not on the native host A. canadensis. Other studies have shown that ants can also protect lycaenids from parasitoids (Pierce & Mead, 1981; Pierce & Easteal, 1986; Weeks, 2003).
We investigated the importance of ant association for protection of L. melissa caterpillars against parasitoids by asking if ant-tended caterpillars are less likely to be attacked by parasitoids. We asked this question using populations associated with two different native hosts and one exotic host. In addition, we laid the foundations for future studies of interaction ecology using L. melissa by identifying species of parasitoids and species of ants tending caterpillars. With respect to the primary objective, we hypothesised that caterpillars found in association with tending ants would have lower parasitism rates than those found without attending ants. With respect to the natural history of ant and parasitoid interactions, we expected to find relatively generalised interactions. Lycaeides melissa caterpillars are small and never particularly abundant (C. F. Scholl, pers. obs.), and thus it is reasonable to expect that ants and parasitoids might not be involved in specialised relationships, but would opportunistically interact with L. melissa.
Materials and methods
- Top of page
- Materials and methods
Different host plants of L. melissa vary considerably in the communities of insects they support (Forister et al., 2011a). Thus, for the present study, we sampled caterpillars from different sites and host plants, both native and exotic, with the intent of maximising our chances of capturing a diversity of interactions among caterpillars, ants and parasitoids. Field sites were located either in typical Great Basin habitat, where the dominant plant species are sagebrush (Artemisia spp.) and rabbitbrush (Ericameria spp.), or in invaded habitats (i.e. alfalfa along roadsides). Our goal was to collect caterpillars and ants from 20 plants at each of three localities where L. melissa uses native hosts and at three localities where L. melissa uses alfalfa, across 2 years.
We sampled five sites during the summers of 2010 and 2011. In the second year, we were able to resample four of the five sites from 2010. The four sites that were studied in both years included two sites where the native A. canadensis is used as a host, Washoe Lake (WLA, 39.2331°N, −119.7795°W) and Silver Lake (SLA, 39.6497°N, −119.9263°W), and two sites in and around Verdi, Nevada (VCP, 39.5141°N, −119.9950°W and VER, 39.5145°N, −119.9927°W) where alfalfa is used as a host. The fifth site, Beckwourth Pass (BWP, 39.7797°N, −120.0734°W), could not be resampled in 2011 due to intensive livestock grazing. This site has both the native host, A. canadensis, and alfalfa, so data for this site (from 2010) were separated by host. In 2011, a site in Bishop, California (BHP, 37.3585°N, −118.3906°W) was added, where the native host used is Glycyrrhiza lepidota (Nutt.) Pursh.
The identity of host plants is inherently confounded with geography in this system (there is only one site where two hosts are present, and we were only able to sample it in the first year). As a result, our analyses (described later) control for variation among sampling locations, but do not address host plant as a unique factor. All sampling was performed in August or the first week of September, when caterpillars can be readily found by locating damage on plants, and when ant and other insect abundances are relatively high (C. F. Scholl and M. L. Forister, pers. obs.). Sampling occurred at the same time of day across sites, between 08.00 hours and 12.30 hours.
Collection of caterpillars and ants
In order to keep sampling as even as possible across sites, our sampling design was based on the collection of larvae from 20 plants at each site. Because larvae on individual plants potentially experience similar enemy pressure and similar potential to interact with ants, the individual plant is our unit of replication throughout the collections and analyses (described later). The plants were sampled along transects extending out from the centre of a site, which was identified as the location with the highest concentration of adult activity and caterpillar damage evident on plants, based on both observations in previous years and observations at the initiation of the present study. Individual plants to be searched were chosen in the following way: a line was extended out from the centre of the area along a random compass direction, and walked until that line hit a host plant. Each plant was then inspected thoroughly for caterpillars. If no caterpillars were found, sampling was continued in the given direction until either a plant with caterpillars was found or the edge of the sampling area was reached at 25 m (a distance that we have found tends to encompass ‘patches’ of host plants utilised by L. melissa in Great Basin habitat). As plants were searched for caterpillars, any tending ants were collected before a caterpillar was removed from the plant. It is important to note that there is little ambiguity in tending behaviour: tending ants stay very close to caterpillars and are in frequent physical contact with the caterpillars.
After the foliage of each plant was thoroughly searched for caterpillars and ants, the base of the plant was checked for larvae and pupae. The size of each plant (from which caterpillars were collected) was found by measuring the height, the width at the widest point, and the width perpendicular to the widest part. Plants were flagged to avoid resampling. Each site was sampled until either 20 plants with larvae were found or all the plants within the 25 m radius were searched.
Our recording of caterpillar and ant data was slightly different in the 2 years of the study. In 2010, data were collected strictly by plant: caterpillars and ants were labelled with reference to the plant from which they were collected, and we noted whether any caterpillars were being tended (thus, ‘tended’ or not is a categorical variable associated with the caterpillar–ant community on a plant). Following the first year of collection, we adjusted our sampling design to allow for greater resolution of ant tending. In the second year, data were collected for each caterpillar (with caterpillars and ants labelled together). Thus the ‘tending’ factor for analyses changed to a continuous variable: the percentage of caterpillars found being tended on a given plant. In all cases, we made the assumption that the observation of tending ants at the time of collection is a useful indicator of the ant–caterpillar interaction for some meaningful period of caterpillar development, as has been assumed by others (e.g. the observational component of Pierce & Mead, 1981). More specifically, we assumed some consistency of interaction at the level of whole plants. In other words, plants that include more tending ants at the time of collection will, we predict, harbour more caterpillars that will have been protected for a longer period of time than plants without tending ants.
Caterpillars from the field were kept cool and brought back to a laboratory at the University of Nevada, Reno. Before being put individually into Petri dishes (90 × 12 mm) for rearing, caterpillar length was measured, because it can be used as a correlate of age. Instar or developmental stage at time of collection is relevant, as certain parasitoids preferentially attack specific life-history stages of their caterpillar hosts (Neveu et al., 2003). However, subsequent analyses found no indication of an association between age and parasitoid attack in our samples, and this issue is not discussed further.
Our goal in rearing was to maximise caterpillar survival in order to maximise the chance of observing a successfully emergent parasitoid. To this end, the following host plants were used in rearings: Lupinus polyphyllus Lindley in 2010 and Lotus nevadensis S. Watson in 2011, both native hosts of the closely related butterfly species, Lycaeides anna (Edwards) (formerly L. idas anna). In laboratory trials, these hosts support more reliable caterpillar survival and development relative to both the native host of L. melissa, A. canadensis, and the novel host, alfalfa (Scholl et al., 2012). While it is known that larval diet can impact parasitoid development and caterpillar immune response (Smilanich et al., 2009), switching all caterpillars to the same host should not affect comparisons across or within populations. Lupinus polyphyllus was collected approximately 7 miles north of Truckee, California, off State Route 89 (39.3233°N, −120.5999°W) and L. nevadensis was collected at the Hwy 20 exit off I-80 (39.4332°N, −120.2037°W). The use of different rearing hosts in the different years simply reflects evolving laboratory protocols, but does not affect results since analyses (described below) are conducted within but not across years.
Plant material was kept in a refrigerator until needed, with new plant material collected at least once a week. Caterpillars were checked daily and new plant material was provided as needed, generally every second or third day. Caterpillars were reared until emergence as an adult or until a parasitoid emerged (or until death from other causes). All dishes were kept at room temperature (20–23 °C) on laboratory benches with overhead lights set on a LD 12:12 h cycle. Parasitoids were identified by specialists: John O. Stireman III (Wright State University) identified the Tachinidae, David Wahl (American Entomological Institute) the Ichneumonidae, Scott R. Shaw (University of Wyoming) the Aleiodes and James B. Whitfield (University of Illinois) the Cotesia. Ants in the Fusca group of Formica were identified by us using Francoeur (1973), while all other species were identified using Fisher and Cover (2007) and Wheeler and Wheeler (1986).
Our primary objective was to address the influence of ant tending on parasitoid attack. Because data collection methods differed in the 2 years (see earlier discussion) and because not all sites visited in 2010 could be resampled in 2011, all analyses were run separately for the 2 years. We employed a model comparison approach using generalised linear models with binomial error (appropriate for data bounded between 0 and 1) to compare parasitism across sites (Burnham & Anderson, 2002) using the mumin package in r 2.14.0 (R Development Core Team, 2011). For 2010 we modelled percentage parasitism of caterpillars on a plant as a function of the presence or absence of tending ants, site, size of plant and caterpillar density on a plant. In 2011, the ant covariate changed (as discussed earlier) to the percentage of caterpillars found on a plant being tended by ants. The other covariates, site, size of plant and caterpillar density, remained the same, as did the response, percentage parasitism of caterpillars on a plant. For all analyses, plant size was calculated by multiplying the height by the two width measurements. Caterpillar density was found by dividing the number of caterpillars found on a plant by plant size. Caterpillars were removed from analyses if they died in the laboratory without a parasitoid emergence or if they pupated but failed to emerge as adults with no parasitoid evident. Occasional deaths that were not associated with parasitism tended to occur because of infection, causing the larvae to turn black and liquify, making dissection for parasitoid detection difficult.
For all of our analyses involving model comparison, a model was considered competitive if it was within 2ΔAICc of the top model (AICc, or corrected AIC, is the Akaike information criterion adjusted for small sample sizes). Akaike weights or normalised relative likelihoods (wi) were also calculated to corroborate model selection and to judge the relative importance of each factor (by summing the Akaike weights for all the models with a given covariate; Burnham & Anderson, 2002). A null model with no variables was used for comparison; support for the null model that is comparable to other models suggests that there are no significant predictors of the response variable.
A final set of analyses involved an investigation of parasitism within sites as a complement to the analyses described earlier that were among sites. For these analyses, we focused only on sites where parasitism had been observed, using generalised linear models with binomial error to model parasitism rates (for two sites in 2010 and four sites in 2011). Candidate factors in these models were the same as described earlier, with the obvious exception of site that was not used for these within-site analyses.
- Top of page
- Materials and methods
We collected a total of 526 caterpillars, 424 of which survived and were included in analyses (213 caterpillars survived in 2010, and 211 in 2011). Caterpillars were collected from 185 plants, with an average of 2.3 caterpillars per plant, ranging from one individual to 17 found on a single A. canadensis plant at BWP in 2010. We collected 288 ants associated with caterpillars in the wild, and a total of 27 parasitic wasps and 21 parasitic flies were reared from caterpillars in the laboratory (Fig. 1). In each case, only a single parasitoid emerged from each of the 48 parasitised caterpillars. Tending ants were found at all the sites. Results from the model comparisons for 2010 show the top model as ‘site + ant’ (Table 1). There are two other models within 2ΔAICc of the top model, both of which contain site as a factor. The relative importance of site was found to be 1 by summing all the Akaike weights of all models with site as a factor. The presence of ants was weakly but positively associated with percentage parasitism in two of the top three models. The model-averaged coefficient associated with the importance of ant tending was β = 1.76 ± 1.10, Z-value = 1.57, P = 0.12 (the positive coefficient associated with ants, which is contrary to expectation, is discussed further below). For 2011, the top model contained only site (Table 1). The next top model was 1.79ΔAICc from the top model and included the factors site and percentage tended. The relative importance of site was again found to be 1. For 2011 the model-averaged coefficient associated with the importance of ant tending was β = 0.633 ± 0.915, Z-value = 0.682, P = 0.50.
Figure 1. (a) The number of larvae collected and parasitism rates for 2010 by site; and (b) the number of plants studied and rates of ant-tending in the same year. (c) The number of larvae collected and parasitism rates for 2011 by site; and (d) ant tending rates for individual larvae for the same year. SLA, Silver Lake (39.6497°N, −119.9263°W); WLA, Washoe Lake (39.2331°N, −119.7795°W); BWP, Beckwourth Pass (39.7797°N, −120.0734°W); BHP, Bishop, California (37.3585°N, −118.3906°W); VCP, Verdi, Nevada (39.5141°N, −119.9950°W); VER, Verdi, Nevada (39.5145°N, −119.9927°W).
Download figure to PowerPoint
Table 1. Results from model comparisons across sites for 2010 and 2011, with percentage parasitism per plant as the response
|Model||ΔAICc||wi||No. of parameters|
|Site + ant||0||0.34||7|
|Site + plant size + ant||1.59||0.15||8|
|Site + ant + caterpillar density||2.17||0.12||8|
|Site + plant size||2.62||0.09||7|
|Site + caterpillar density||3.84||0.05||7|
|Site + plant size + ant + caterpillar density||4.00||0.05||9|
|Site + plant size + caterpillar density||4.26||0.04||8|
|Ant + caterpillar density||24.73||0||2|
|Site + percentage tended||1.79||0.16||6|
|Site + caterpillar density||2.04||0.14||6|
|Site + plant Size||2.25||0.13||6|
|Site + percentage tended + caterpillar density||3.95||0.06||7|
|Site + plant size + percentage tended||4.09||0.05||7|
|Site + plant size + caterpillar density||4.37||0.05||7|
|Site + plant size + percentage tended + caterpillar density||6.35||0.02||8|
|Plant size + caterpillar density||48.94||0||2|
The results from the within-site model comparisons for 2010 (Table 2) showed that the presence of tending ants, plant size and caterpillar density were not significant predictors of percentage parasitism. For the 2011 within-site comparisons, the percentage of caterpillars being tended, plant size and caterpillar density were, likewise, not significant predictors of parasitism (Table 2). The results from the within-site analyses are presented in a different way from those of the across-site analyses. Because each site is treated as a separate factor in across-site analyses, results are given using an AICc table in order to judge the importance of site itself (rather than the different levels of site) as a predictor along with the other covariates. For the within-site analyses (Table 2), estimates of coefficients from the generalised linear model are given for each covariate.
Table 2. Results from generalised linear model comparisons of parasitism within sites for 2010 and 2011
| ||Estimate||Standard error||P|
| Caterpillar density||5.083e–6||3.844e–5||0.895|
| Plant size||−1.160e–5||3.816e–5||0.761|
| Caterpillar density||7.330e–7||3.804e–6||0.847|
| Plant size||7.489e–7||1.555e–6||0.630|
| Percentage tended||2.785||2.000||0.164|
| Caterpillar density||1.902e–6||4.309e–5||0.965|
| Plant size||7.230e–6||1.444e–5||0.617|
| Percentage tended||0.9473||1.996||0.635|
| Caterpillar density||1.730e–5||1.038e–4||0.868|
| Plant size||2.262e–5||6.538e–5||0.729|
| Percentage tended||−17.57||3956||0.996|
| Caterpillar density||1.985e–5||3.159e–5||0.530|
| Plant size||1.386e–5||2.431e–5||0.568|
| Percentage tended||1.807||2.736||0.509|
| Caterpillar density||4.417e–6||1.480e–5||0.765|
| Plant size||−1.901e–5||1.649e–5||0.249|
The 48 parasitoids that emerged encompassed two species of parasitic fly in the family Tachinidae and three species of parasitic wasps in two families, Ichneumonidae and Braconidae (Fig. 2). The two fly species were Aplomya theclarum Scudder and Patelloa cf. fuscimacula Aldrich and Webber; the three species of wasps were Anisobas bicolor Cushman (Ichneumonidae), Aleiodes cultrarius Shaw & Marsh and Cotesia sp. (both Braconidae). Two sites had parasitoids in 2010 (WLA and VCP) while parasitoids were found at four sites in 2011 (SLA, WLA, BHP and VCP). The diversity of parasitoids increased in 2011, with all five species of parasitoids represented, instead of only A. bicolor and A. theclarum. All five species of parasitoid were found on the native hosts, while only two were found on the exotic host (Fig. 2a).
Figure 2. (a) Abundance of parasitoid species found across hosts; (b) abundance of ant species found tending across hosts. The ant species Camponotus vicinus was found on the exotic host and Tapinoma sessile was found on a native host (the two smallest bars at the right-hand end of the graph).
Download figure to PowerPoint
We collected and identified 288 ants from 10 species in five genera actively tending caterpillars (149 from 2010 and 139 from 2011, Fig. 2b). Across sites in 2010, 60% of plants with caterpillars had tending ants; in 2011, 42.7% of the caterpillars found were tended. Ants in the genus Formica made up the vast majority of ants found tending (96.4%). Of these, Formica oreas Wheeler was most common, comprising 60% of the ants collected. This species was found at all but one (SLA) of the sites sampled. Formica lepida Wheeler was the next most common species, making up 19.4% of the ants found tending, and was found at every site except for WLA in 2011. Ants from other genera, Camponotus vicinus Mayr, Myrmecina americana Emery, Monomorium minimum Buckley and Tapinoma sessile Say, were each found only at one site during 1 year of sampling. Formica oreas may have enslaved colonies of F. lepida at several sites. The genus Formica is split into two groups, Rufa and Fusca, and members of the Rufa group are known to enslave members of the Fusca group (Fisher & Cover, 2007). Both species, F. oreas and F. lepida, were often encountered tending on the same plant, and in one case members of both species were found tending the same larva. We found variation across years and sites in the species of ants tending, with five species only on native hosts (Formica aerata Francoeur, Formica gnava Buckley, Formica neoclara Emery, M. minimum Buckley and T. sessile Say) and two species (Camponotus vicinus Mayr and Myrmecina americana Emery), only on the exotic host (Fig. 2b).
- Top of page
- Materials and methods
Throughout much of the history of ecology as a science, positive interactions between species have been studied noticeably less than negative interactions (Boucher et al., 1982). In recent years, the potential importance of mutualisms in population dynamics and community structure has become more widely appreciated (Bruno et al., 2003; Guimarães et al., 2011). Mutualistic interactions between ants and caterpillars have been of interest for their ubiquity and the ease with which they can be observed. We examined interactions between ants and caterpillars of L. melissa, and did not find support for the hypothesis that the presence of ants decreases parasitism rates of L. melissa caterpillars. We found that parasitism was highly variable by site (Fig. 1); the importance of location in relation to parasitism was reflected in the presence of site as a factor in all top models across years (Table 1). The top models for both years also tended to include either the presence of ants on a plant (in 2010) or the percentage of caterpillars associated with ants (in 2011). Rather than the predicted negative relationship that would have been consistent with ant protection, the inclusion of ant tending in top models appears to have been driven by a weak, positive association between the presence of ants and parasitoids (the same relationship was not evident in within-site analyses that have lower power). For example, in 2010, nine out of 10 plants with parasitised caterpillars had tending ants. It is possible that some individual plants have a more abundant and diverse insect community that includes more ants, more ant mutualists such as aphids and other hemipterans, and also more parasitoids that might be drawn to that rich prey community. Indeed, some wasp parasitoids have adaptations for attacking their host caterpillars specifically in the presence of ants (Thomas et al., 2002). These possibilities suggest complex indirect interactions, which can only be posed at this time as a hypothesis for further study in the L. melissa system.
Other studies have found mixed results for the role of mutualistic ants in protecting caterpillars from parasitoids. Protection from parasitoids by ants has been found by experimental exclusion of tending ants (Pierce & Mead, 1981; Pierce & Easteal, 1986; Weeks, 2003). Pierce and Mead (1981) also collected final instars in the field from unmanipulated plants (similar to our study) and found that those caterpillars that were untended or tended by only one ant were more likely to have been parasitised than those found being tended by three or more ants. DeVries (1991), on the other hand, found that ants protect caterpillars from predatory wasps but not from tachinid flies. Savignano (1994) found that the presence of ants did not reduce parasitism but did reduce predation in Lycaeides samuelis (formerly L. melissa samuelis; Forister et al., 2011b). Combined with the results from Savignano (1994) and a previous study from one of our focal sites (Forister et al., 2011a), it seems likely that the presence of ants reduces predation, but not parasitism in L. melissa.
In the future, with a greater understanding of host–parasitoid dynamics in this system, it may be possible to do a manipulative experiment to test for an effect of ant removal on parasitism. As shown by Forister et al. (2011a), doing a manipulative experiment where ants are removed is possible in this system; however, with the high disappearance and removal rate of experimental larvae [over a third of larvae were lost after being left out for only 6 h in the Forister et al. (2011a) study], it would require a very large number of larvae to be tracked. Also, because of variable parasitism rates across sites, it would be difficult to choose a site where parasitism rates would be high enough to test for an effect of ant attendance.
We found variation in the community of ants tending across sites and years. It has been shown in other ant–caterpillar (Fraser et al., 2001) and ant–plant (Palmer & Brody, 2007; Palmer et al., 2010) mutualisms that different ant species exhibit varying levels of protection. Caterpillars can also elicit changes in the intensity of tending behaviour by varying the amount of reward they produce (Agrawal & Fordyce, 2000; Axén, 2000; Fraser et al., 2001). In the L. melissa system, it would be interesting to better understand and quantify both the larval reward provided and the intensity of tending for different species of ants. It might be the case that some ant species are better at protecting L. melissa larvae from parasitoids than others, and that a failure to account for this source of variation could explain the lack of evidence for protection against parasitoids. In a different system involving acacia trees and symbiotic ants, Palmer et al. (2010) found that the interaction complexity associated with the presence of different ant species was key to understanding the evolution and persistence of the interaction. Another type of complexity that we did not address in our study, but which could be addressed in the future, involves the spatial scale of ant protection. We used individual plants as our unit of replication, although it is possible that quantifying ant–caterpillar interactions and parasitism rates at either a finer or coarser scale (e.g. at the level of host plant patch) would be more informative.
As with the ant interactions, we observed considerable variation in parasitoid abundance and diversity across space and time. VER and VCP are geographically very close, 200 m apart; however, parasitoids were only found at VCP. The highest rate of parasitism was at WLA in 2011, where 65.7% of caterpillars were parasitised by three different parasitoids. It is interesting to note that there was only one caterpillar found being tended at WLA in 2011; this site has a sandy substrate that is seasonally inundated with water, which might limit ant presence. Other studies of parasitism have also found large differences in parasitism rates across space and time (Pacala & Hassell, 1991; Stireman & Singer, 2002; Heard et al., 2006). Using theoretical population models, this variation has been shown to stabilise host–parasitoid population dynamics (Hassell & May, 1974; Hassell et al., 1991). For example, Hassell et al. (1991) found that parasitoid and caterpillar populations can persist if parasitoid density is sufficiently variable across space. In addition to temporal variation, there is, of course, variation among parasitoids in their ecology and natural history. Many parasitoids are specific in the stage of larvae they attack (Neveu et al., 2003). However, we did not find any evidence that caterpillar age affected parasitism rates (results not shown, as described earlier). Some tachinids lay microtype eggs on the host plant (Stireman et al., 2006) where they are then ingested by feeding caterpillars. This does not appear to have been the case for one of the species found, A. theclarum, as an egg was seen on a 3-mm caterpillar that was then found to be parasitised (C. F. Scholl, pers. obs.). However, the other tachinid species, P. cf. fuscimacula, is indeed assumed to lay microtype eggs on the leaves of host plants (Peigler, 1994).
One of the interesting aspects of the L. melissa system is the recent expansion of host range to include alfalfa, an inferior host for larval development and survival (Scholl et al., 2012). Our study was designed to investigate parasitism and ant tending, as they vary in space and time, but was not designed to test directly for differences among host plant species in those interactions. Nevertheless, it is interesting to note that neither rates of parasitism nor ant tending appear to differ consistently among plant species (Fig. 1). For example, parasitism was detected at one native and one exotic site in 2010, while ant tending was found across all native and exotic sites in that year. By contrast, all five species of parasitoids emerged from larvae on native hosts, while only two species of parasitoids emerged from larvae on alfalfa (Fig. 2a).
Of the four species of parasitoids identified to species, three have previously documented host associations with lyceanid butterflies. Aplomya theclarum is often reared from lycaenids (Pierce & Mead, 1981; Weeks, 2003), including L. samuelis (Savignano, 1994). Anisobas bicolor has also been reared from several other lyceanids, and we present a new host record for this species (Krombein et al., 1979; Pierce & Easteal, 1986). The only previously know host of Aleiodes cultrarius was the Xerces blue butterfly (Glaucopsyche xerces), which went extinct in the early 1940s (Shaw et al., 2006). Because additional A. cultrarius specimens were collected from 1955 to 1967, Shaw et al. (2006) pointed out that the species must have an alternative host, which is confirmed by our results. Having a bright yellow cocoon and being reared from a lycaenid are a rare collection of traits for the genus Cotesia, making it likely that the collected individuals represent a new, specialised species (J. B. Whitfield, pers. comm.). The only generalist parasitoid collected was P. cf. fuscimacula, which has been reared from a wide variety of lepidoptera, but has no previous record of being reared from a lycaenid (Peigler, 1994; Yoo, 2006; J. O. Stireman, pers. comm.).
In conclusion, we found a diversity of parasitoid species and parasitism rates across space and time. We did not find evidence that ants provide L. melissa protection from parasitoid attack, which adds to our knowledge of both interactions in this system and lycaenid–ant interactions in general. However, we have documented novel interactions between caterpillars and tending ant species and parasitoid species, thus laying a foundation for further studies investigating how species interactions, especially interactions with parasitoids, affect the ecology of a common member of the western North American butterfly fauna.