LOCAL SPECIALIZATION AND LANDSCAPE-LEVEL INFLUENCE ON HOST USE IN AN HERBIVOROUS INSECT.
MICHAEL SINGER 
ILKKA HANSKI 
Abstract. We studied host plant abundance, host use, and oviposition preference in metapopulations of the butterfly Melitaea cinxia within an area of 3500 [km.sup.2] in the Aland islands, southwestern Finland. In the study area, M. cinxia has [sim]400 small local populations on dry meadows with the larval host plants, Plantago lanceolata and Veronica spicata. Plantago lanceolata occurs in practically all meadows otherwise suitable for the butterfly, whereas the distribution of V. spicata is largely restricted to the northwestern part of the study area. Based on observations of 6500 prediapause larval groups during 1993-1996, we document spatial variation in host plant use in relation to their abundance (electivity). The fraction of larval groups found on V. spicata increased disproportionally with the relative cover of V. spicata in the habitat patches. Additionally, the probability of Veronica use in a population increased with increasing number of larval groups found on Veronica in the surrounding populati ons but decreased with increasing use of Plantago in the neighborhood. This regional effect on host use at the scale of migrating butterflies could be caused either by spatial variation in the insect (in either preference or performance) or by spatial variation in plants (in resistance to attack by the butterflies). To study the first possibility, we conducted oviposition preference experiments using butterflies from five metapopulations located 2-55 km from each other and characterized by differences in host plant availability and host use. We found clear genetic differences in oviposition preference between the five metapopulations consistent with the observed host use patterns in the field. We conclude that the spatial host use patterns of M. cinxia in the study area are driven both by direct effects of local host abundance and by indirect effects mediated through metapopulation-level adaptation to the regionally more abundant host plant.
Key words: host plant availability; host plant use; local adaptation; Melitaea cinxia; metapopulation; oviposition preference; Plantago lanceolata; plant-insect association; Veronica spicata
The distribution of an insect species across a set of host plants at the landscape level can be described entomocentrically as "pattern of host use." Such patterns are influenced not only by spatial variation in host plant abundance (Wiklund 1974, Courtney and Forsberg 1988) but also by spatial variation in host plant quality and local adaptation in insect populations (reviews in Denno and McClure 1983, Jaenike 1990, Via 1994, Mopper and Strauss 1998; examples in Singer and Parmesan 1993, Bossart and Scriber 1995, Mayhew 1997). The host plant most often used locally by an oligophagous insect is not necessarily the most preferred one when the favored host is rare (Singer et al. 1989). Conversely, the most abundant host is not necessarily used most frequently, for instance when the insect prefers another host species over the most abundant one (e.g., Singer 1983, Singer et al. 1989).
The proportion of resources in the diet of a consumer as a function of their availability is known as electivity (Ivlev 1961). If insects encounter hosts in proportion to their abundances, and the probability of accepting each host type does not change with host abundance, then electivity will be constant across a landscape with varying relative abundances of host types. On the other hand, if electivity is spatially variable, insect diet varies in a manner that cannot be simply predicted from host availability. This could be caused by spatial variation in either plant traits or in insect traits, or both (Singer and Parmesan 1993, Mopper 1998, Strauss and Karban 1998). Plants could vary spatially in their suitability (ability to support insect growth and survival) or in their acceptability (attractiveness to ovipositing females or to feeding larvae when larvae are mobile). Insects could vary spatially in host preference and in performance (survival and growth) of the developmental stages. Spatial variation is known to occur in both insect traits (preference and performance) and in plant traits (acceptability and suitability; reviews in Denno and McClure 1983, Jaenike 1990, Thompson and Pellmyr 1991, Via 1994, Mopper and Strauss 1998).
Here, we document spatial variation in electivity in the Glanville fritillary butterfly, Melitaea cinxia (L.) (Nymphalidae), across a landscape measuring 50 X 70 [km.sup.2]. This insect inhabits a naturally highly fragmented environment in the Aland Islands in southwestern Finland, where its metapopulation structure and dynamics have been intensively studied (Hanski et al. 1994, 1995a, b, 1996, Kuussaari et al. 1998, Saccheri et al. 1998). In Aland, M. cinxia uses two host species, Plantago lanceolata (Plantaginaceae) and Veronica spicata (Scrophulariaceae). We examine here how these two host species are distributed in space and how the caterpillars are distributed among the host plants. We then ask whether spatial variation in diet can be simply explained by relative host abundance or whether electivity is spatially variable. Having documented spatial variation in electivity, we describe experiments investigating spatial variation in insect preference that could account for the observed patterns of host use and electivity in the field.
Study system and host plants of Melitaea cinxia
In the Aland islands Melitaea cinxia is structured into several hundred small local populations, which occur on dry meadows with at least one of the larval host plants (Hanski et al. 1995a). Small local populations go extinct frequently, but unoccupied suitable meadows are also often colonized from nearby populations (Hanski et al. 1995a, b). Movements between nearby populations are common, yet the majority of butterflies spend their entire life in the natal patch (Hanski et al. 1994, Kuussaari et al. 1996). Groups of adjacent local populations function as metapopulations, with much longer persistence times than the persistence times of individual local populations (Hanski et al. 1995a, b, 1996).
Every autumn since 1993 we have surveyed the suitable habitat patches for M. cinxia and counted the conspicuous larval groups within an area of 50 X 70 [km.sup.2], covering the entire range of the butterfly in Finland (see Hanski et al. 1995a and 1996 for details of the survey method). A total of 1677 suitable habitat patches were located during 1993-1996, of which 835(50%) were occupied at least once during the four-year period. The habitat patches are generally small, the mean, median, and maximum patch sizes being 0.16, 0.04, and 7.39 ha, respectively. Because of the small size of the habitat patches the local populations tend to be small, typically consisting of only a few larval groups per local population (Hanski et al. 1995a).
A total of 6557 prediapause larval groups were recorded during the autumn surveys in 1993-1996. Most of the larval groups (99.8%) were found on two host plants, Plantago lanceolata (83.9%) and Veronica spicata (15.9%). The few existing records on other host plants in the Aland islands have been restricted to the genera Plantago (P. media, P. major, and P. maritima) and Veronica (V. chamaedrys. V. serpyllifolia, and V. longifolia), but in no local population has any of these other plant species been used regularly as a host.
Host availability and host use
The percent coverages (0-100%) of P. lanceolata and V. spicata were estimated in each habitat patch. The coverages were estimated during two autumnal surveys, in 1995 and 1996. In both years, two surveyors first estimated the coverages of each host plant species in the patch independently. The average value of these two estimates was used as the coverage estimate of each host species in a particular patch in a particular year. In order to have one abundance estimate for each host species in each patch, we calculated the mean value of the estimates for the two years. In the analysis of the factors affecting host plant use, we used relative abundance of Veronica spicata (varying from 0 to 1) as one independent variable. This measure was determined by dividing the coverage of Veronica by the sum of the coverages of the two host species.
During the larval surveys, we always recorded the species of host plant on which each larval group was found. In order to have reasonable sample sizes per local population in the analysis of the factors affecting host use, we pooled the data on the occurrence of the larval groups in each local population for the four years 1993-1996.
To study large-scale spatial variation in host availability and in host use, we divided the study area into 127 semi-independent patch networks (SINs; Hanski et al. 1996), which are expected to have relatively independent metapopulation dynamics (little migration between the SINs). Habitat patches were assigned to different SINs if there was [greater than or equal to]1 km of forest or open water, or [greater than or equal to]1.5 km of some other type of unsuitable habitat between the patches. We calculated for each SIN relative host plant coverage and total numbers of larval groups found on the two host plants during 1993-1996.
We used a general linear model to ask to what extent the probability of a larval group being found on Veronica in a particular local population (binomial response variable, logit link function) was explained by the following variables: (1) relative abundance of Veronica spicata in the focal patch, (2) geographic location of the patch (along an axis directed from northwest towards southeast), and (3) connectivity of the patch to the use of the two hosts in neighboring areas (see Methods: Connectivity indices). In this analysis, we included only the largest local population with both host species available for each SIN in order to exclude many populations located close to each other.
To illustrate the significant relationships demonstrated by the analysis described above, we constructed two new variables: a measure of electivity in the focal population and a measure of relative host use in the surroundings of the focal population. Electivity was measured as arcsin (square root (relative Veronica use)) -- arcsin (square root (relative Veronica cover)). In this calculation, relative Veronica use number of larval groups on Veronica/total number of larval groups in the habitat patch. Positive values of electivity indicate disproportionally many larval groups on Veronica and negative values an excess of groups on Plantago in relation to relative host cover. Relative host plant use in the neighborhood of the focal population was measured as [S.sub.v] - [S.sub.p] (where subscripts V and P refer to Veronica and Plantago, respectively; see Methods: Connectivity indices), positive values indicating predominant use of Veronica and negative values predominant use of Plantago in the surroundings of t he focal population. The relationship between electivity and relative host plant use in the neighborhood of the focal population was analyzed using linear regression, the material consisting of the local populations with both hosts available and [greater than or equal to]10 larval groups found in 1993-1996 (the latter to have a reasonably accurate estimate of host use in the local population).
To analyze whether host use in a local population was affected by host use in the surrounding populations, we calculated indices measuring the degree of connectivity (inverse of isolation) of each local population to the use of the two host plants in the surrounding local populations. Connectivity of population with respect to host x (Plantago [P] or Veronica [V]), [S.sub.xi], was calculated as (modified from Hanski 1994)
[S.sub.xi] = [[sigma].sub.j[not equal to]i] [e.sup.-[[alpha]d.sub.ij]][N.sub.xj] (1)
where [N.sub.xj] is the number of larval groups observed on host x in population j, and [d.sub.ij] is the distance between populations i and j (in km). This measure takes into account the distances from population i to all the other local populations in which host x was used. This measure of connectivity is meant to reflect the expected numbers of immigrants to patch i which have used host x as a larva. Parameter [alpha] sets the distribution of migration distances in Eq. 1. We used the value 1, as suggested by our previous work on M. cinxia (Hanski et al. 1996). Consequently, the number of potential immigrants decreases rapidly with increasing distance from the focal population. With [alpha] = 1, populations located [greater than]2 km from the focal population have a small effect, and the effect of populations located [greater than]5 km from the focal population is negligible. We calculated the value of [S.sub.xi] separately for each year in 1993-1996 using the yearly numbers of larval groups observed on the two host plant species. The mean value of the [S.sub.xi] for the four years was used in the analysis.
To study the effect of regional host use patterns on the oviposition preference in a metapopulation, we used an index of metapopulation-level connectivity ([M.sub.x]), calculated as the mean [S.sub.x] value (Eq. 1) for host x within a metapopulation. This index measures the mean connectivity to the use of host x in the surroundings of an average local population within a particular metapopulation.
To study oviposition preference, we sampled a total of 17 local populations from five semi-independent metapopulations located 2-55 km from each other (Fig. 2). These metapopulations represent the entire range of the existing variation both in the availability of host plants (from localities with only Plantago available to a locality with both hosts abundant) and in host use (from metapopulations using only Plantago to a metapopulation using 84% Veronica; Table 1).
The material for the oviposition preference experiment originated from 72 independent families collected in the field. By an independent family we mean the offspring of one female. During the winter diapause we moved all the larvae to Austin, Texas, USA, where we reared them to adults on P. lanceolata in a greenhouse in January-March 1996 (plants grown from seeds collected from suitable meadows for M. cinxia in the Aland islands).
We allowed the adult females to mate soon after emergence, after which we tested their postalighting oviposition preference in a greenhouse using P. lanceolata and V. spicata plants grown from seeds collected from North Aland. The seeds were collected from an area where both host plants are used by the butterflies, but not close to any of the butterfly metapopulations tested in the oviposition preference experiment. To test oviposition preference we used the manipulative technique developed by Singer (1982, 1986). Each female was repeatedly offered the two host plants in alternation, but she was not allowed to oviposit. The females usually passed through three phases: a rejection phase, when neither host plant was accepted; a discrimination phase, when one host plant was consistently accepted and the other one was consistently rejected; and a final phase when both host plants were accepted until oviposition was allowed. The preferred host is the one that was accepted during the discrimination phase. When the two hosts were both rejected at first, but next both were accepted in consecutive trials, no discrimination phase was detected and the butterfly was considered to have no preference.
We tested a total of 104 females originating from 47 families, 17 local populations and five metapopulations (Table 1). Each butterfly was tested on a single pair of the two host plant species, and a total of 29 pairs of host plants was used in testing the 104 females. The plants represented the same growth form as they have in the field in Aland with leaves close to the ground. Different plant pairs were used equally often and the plant pairs for butterflies from different metapopulations were randomly chosen. For each butterfly we recorded the direction of preference, Plantago or Veronica, or no detectable preference. We also estimated the strength of the preference as the length of the discrimination phase. We used a negative sign for the discrimination phase to indicate Veronica preference and a positive sign to indicate Plantago preference. When more than one female was tested per family, we used the median direction of preference and the mean discrimination time for the family as an independent data po int in the analyses.
We tested the difference in preference among local populations and among metapopulations by comparing the numbers of Plantago and Veronica preferring families (omitting families with no preference) using contingency tables. We excluded Kumlinge from the contingency tests, because only two families were available. To test the difference in the mean discrimination times (the strength of preference) between local populations and between metapopulations we used Kruskal-Wallis one-way ANOVA. Because the butterflies tested in the oviposition preference experiments were reared from larvae under common conditions in the laboratory, any differences in adult oviposition preference among populations are likely to be genetic.
Spatial variation in host availability and host use
Fig. 1 shows the proportions of different kinds of habitat patches in terms of host plant species composition in each patch network (SIN) across the Aland islands. Plantago lanceolata is available in practically all patches (98.8% of the 1532 patches recorded in 1995-1996), whereas the occurrence of V. spicata (available in 25% of the patches in 1995-1996) is restricted to the northwestern part of the study area, where it is present in most patches (Fig. 1). Relative host coverage within the 127 SINs varies from 100% coverage by P. lanceolata on the eastern islands to 94% coverage of V. spicata in northwestern parts of the main Aland island. Among the 58 SINs with both hosts available, 34% of the total host coverage was accounted for by V. spicata.
During 1993-1996, P. lanceolata was used by M. cinxia in 89 and V. spicata in 32 of the 91 patch networks occupied by the butterfly (Fig. 2). The majority of the larval groups was found on P. lanceolata, as expected from its greater availability, but there was much variation in host use among the SINs that had both hosts available (from 0% to 100% of larval groups on Veronica among the SINs with both hosts available in 1993-1996; Fig. 2). In relation to the relative host coverage, disproportionally many larval groups were found on V. spicata in the northwestern parts of the study area (Fig. 3), suggesting that V. spicata may be preferred over P. lanceolata in the areas where V. spicata is most abundant. The reverse was true about the rest of the study area, with P. lanceolata used more than expected from its relative abundance.
Factors affecting host use
All the four factors examined had a significant effect on the probability of a larval group being found on V. spicata (Table 2). Firstly, increasing relative cover of V. spicata in a habitat patch increased the fraction of larval groups on V. spicata. Secondly, the fraction of larval groups on Veronica increased from southeast towards northwest, indicating a broad geographic pattern in host use. Thirdly, an effect of host use in the surroundings of the focal population was indicated by significant coefficients for the connectivity indices [S.sub.p] and [S.sub.V]. Thus the probability of Veronica use in the focal patch increased with increasing Veronica use in the surrounding patches ([S.sub.V]), whereas it decreased with increasing use of Plantago in the surrounding patches ([S.sub.P]). Additionally, two interaction terms were significant, involving the connectivities to Veronica and Plantago use and Veronica cover and the geographic location (Table 2). Because the two connectivity measures ([S.sub.P] and [S.sub.V]) are influenced by the host use pattern only within a few km around the focal population, they explain variation in host use at a relatively small spatial scale, typically within a metapopulation. Fig. 4 illustrates the effects of the relative cover of V. spicata within the focal patch and the relative use of the two host plants in the surrounding populations on the proportion of larval groups on Veronica in the focal population.
Spatial variation in oviposition preference
Fig. 5 gives the laboratory results on oviposition preference for the five metapopulations. The direction of the oviposition preference could not be tested among several metapopulations simultaneously because of small expected cell frequencies, but the strength of the preference differed significantly among the five metapopulations (Kruskal-Wallis H = 19.9, df = 4, P [less than] 0.001; using family means as data points). Excluding Kumlinge, for which only two families were available, four of the six pairwise comparisons of the direction of preference were significant at P [less than or equal to] 0.05 (using Fisher's exact test). Salis, which had the largest fraction of Veronica preferring butterflies, differed from all other metapopulations (P values for the comparisons between Salis and Postad, Jomala and Foglo were 0.01, 0.02, and [less than]0.001, respectively), and Postad differed from Foglo (P = 0.05), which had the largest fraction of Plantago preferring butterflies. Pairwise comparisons of the mean ra nks for the strength of preference indicated that Salis was different from Foglo (P [less than] 0.05).
There were no significant differences in oviposition preference among local populations within metapopulations, though the observed differences were in the expected direction among the three local populations within the Postad metapopulation for which the most extensive material is available. The mean discrimination times were + 13 min (n = 6) in the patch with only Plantago, +3 min (n = 6) in the patch with both hosts, and 0 min (n = 1) in the patch with only Veronica available.
Oviposition preference, host availability, and host use
Spatial variation in oviposition preference agrees closely with spatial variation in host availability and host use in the five metapopulations studied (Pearson's r = 0.97, n = 5, P = 0.005 for the relationship between the fraction of Veronica-preferring butterflies and the fraction of larval groups found on Veronica; Figs. 1, 2, and 5). None of the families from the islands of Kumlinge and Foglo, where P. lanceolata is the only host available, preferred Veronica. A minority of the families from Jomala and Postad, where V. spicata occurs in some habitat patches, preferred Veronica. In Salis, where Veronica occurs in most of the patches and most larval groups have been found on V. spicata, none of the families preferred Plantago.
Effect of population connectivity on oviposition preference
In spite of the limited amount of data on spatial variation in oviposition preference across metapopulations, we detected an indication of the effect of population connectivity on oviposition preference. When the five metapopulations were ranked according to the index of metapopulation-level connectivity ([M.sub.v]) from areas where V. spicata was used, the rank order of the metapopulations was exactly the same as when the five metapopulations were ranked according to their mean discrimination times. The probability of getting the same rank order by chance is 0.03.
In attempting to understand the mechanisms producing spatial variation in host use in our study system, we first considered the relationship between insect diet and host plant abundance. The relative abundances of the two plant species in the study area show a clear spatial pattern, with Plantago omnipresent and Veronica present only in the western half of the study area. The abundance of Veronica declines from northwest to southeast, with some local exceptions (Fig. 1).
A null model of regional host use is that there is no spatial variation among the insects nor among conspecific plants. Each butterfly has the same constant probability of accepting Plantago whenever encountered, and a possibly different but constant probability of accepting Veronica when encountered. In this case the observed geographical cline in diet would result simply from differences in the rates of host encounter, stemming from the cline in relative host abundance. If this were true, electivity (the proportion of larvae on each host as a function of relative host abundance) would be constant across the system. Our results clearly falsify this model: butterflies showed greater use of Plantago, relative to its local abundance, in habitat patches surrounded by areas where more larval webs had previously been found on Plantago, and likewise, electivity was biased towards Veronica in patches surrounded by areas with a history of high Veronica use. Electivity itself shows a spatial trend in the direction of the trend in relative plant abundance, and thereby the spatial pattern in diet amplifies the pattern in plant abundance.
Thus, there is an apparent effect of regional diet on electivity in a focal habitat patch. This effect must be caused by spatial variation in the quality of the butterflies, the plants, or both. The present results show conclusively that there is variation among butterfly metapopulations, and furthermore that this variation could at least qualitatively account for the observed spatial variation in electivity. Butterflies tended to prefer the regionally more abundant of the two host species. Since the oviposition preference was tested with insects raised on Plantago, the differences are almost certainly genetic, as is the case for postalighting oviposition preference in butterflies in general (Thompson and Pellmyr 1991, Bossart and Scriber 1995), including Euphydryas editha, a melitaeine species closely related to M. cinxia. Euphydryas editha shows heritable variation of the discrimination phase both within (Singer et al. 1988) and among populations (Singer et al. 1991, Singer and Parmesan 1993). Manipulation of encounter rates of E. editha with two host species had no detectable effect on acceptance of either host after alighting (Thomas and Singer 1987).
Some butterfly species improve their rate of host-finding by learning to search for the most abundant host species available, using leaf-shape as a cue (Rausher 1978, Papaj and Prokopy 1989). Such learning would create a pre-alighting preference for the most abundant host, with resulting spatial bias of electivity towards the locally more abundant host species. However, no such effects were found in E. editha: naive insects were efficient at locating their principal host species on their very first alighting of the very first search, and experienced insects were no more efficient than naive ones (Parmesan et al. 1995). The result of this fixed alighting bias was that E. editha searched efficiently for traditional hosts but inefficiently for hosts that had recently been incorporated into the diet of a population (Mackay 1985, Parmesan 1991).
Our preliminary results suggest that the spatial pattern in electivity is not explained by two other possible mechanisms apart from spatial variation in oviposition preference. In an experiment conducted in 1998, no indication was detected for spatial variation in plant quality either for ovipositing females or for larval development (M. Singer, unpublished data). In other words, P. lanceolata growing in the region where V. spicata is common was not less accepted, nor supported worse larval growth, when compared in an experiment with P. lanceolata from the region dominated by this plant species. The analogous result was obtained for V. spicata.
With the present information on M. cinxia, it is thus possible that genetic variation in oviposition preference, spatially autocorrelated among neighboring metapopulations, is the sole or the principal mechanism causing spatial variation in electivity. In other words, variable preference interacts with encounter rates (local host plant abundance) to generate the diet. Spatial variation in preference could, in turn, be a consequence of host abundance: natural selection may have led to evolution of preference for the locally most abundant host plant. Such local adaptation at the level of relatively small areas is consistent with our results on migration distances and hence gene flow in this species. Empirically we have estimated that [sim]90% of migrating butterflies move [less than]1.5 km (Hanski et al. 1994). Within a set of 240 colonization events observed in 1993-1996 in the entire Aland system, the longest observed colonization distance was 4.7 km (Hanski 1999). Preliminary genetic results on M. cinxia sho w aggregated spatial distribution of rare microsatellite alleles, demonstrating limited gene flow (I. Saccheri, unpublished data).
We found spatial variation in the direction of preference as well as in the strength of preference. Butterflies from Salis strongly preferred Veronica over Plantago, while those from Foglo and Kumlinge strongly preferred Plantago over Veronica. This is unusual, with some exceptions (e.g., Bossart and Scriber 1995). Most studies of herbivorous insects have reported variation in the strength of preference within species and populations, but variation in the preference rank only among insect species. Examples where intraspecific variation is apparently restricted to degree rather than direction include fruit flies (Courtney et al. 1989) and swallowtail butterflies (Wehling and Thompson 1997). Although unusual, our finding of genetic variation in preference rank within a relatively small area (50 X 70 [km.sup.2]) is not surprising, since such variation is rampant among, and occasionally within, populations of the closely-related Euphydryas editha (Singer and Parmesan 1993, Singer et al. 1994, Thomas and Singer 19 98). Preference tests of the type used here were applied to E. editha captured in the act of ovipositing (or attempting to oviposit) on different plant species in the same area. Females captured ovipositing exhibited the same choice when they were subjected to preference tests, regardless of whether the discriminations were made among conspecific plants (Rausher et al. 1981) or heterospecific plants (Singer 1983). These experiments with E. editha demonstrate that postalighting preference trials such as those used here do indeed measure variation of oviposition preference that is relevant to the decisions that females make in the field in multihost communities.
The spatial variation in electivity and in oviposition preference that we have reported here may have consequences for the dynamics of the mostly very small local populations into which M. cinxia metapopulations are structured in the Aland (Hanski et al. 1995a). Many previous studies have demonstrated how gene flow influences local or regional adaptation (Holt 1995, Dias 1996, Pulliam 1996, Mopper and. Strauss 1998). Our results raise the intriguing possibility that the reverse might also occur and that gene flow and the establishment of new local populations may be influenced by regional adaptations.
We thank Juha Poyry and the students who have participated in the autumn surveys in Aland during 1993-1996, and Jerry Drummond, Camille Parmesan, Eileen Vandenburgh, and Sharon Zhang for their help in preference testing in Austin in 1996. Christian Kotkavuori and Tapio Gustafsson helped to produce the maps in Figs. 1-3. Bob O'Hara gave valuable help with Generalized Linear Models. Edward Connor, Saskya van Nouhuys, and Chris Thomas are thanked for helpful comments on the manuscript.
(1.) Department of Ecology and Systematics, Division of Population Biology, P.O. Box 17 (Arkadiankatu 7), FIN-00014 University of Helsinki, Finland
(2.) Department of Zoology, University of Texas, Austin, Texas 78712 USA
(3.) Present address: Finnish Environment Institute, Nature and Land Use Division, P.O. Box 140, FIN-00251 Helsinki, Finland. E-mail: firstname.lastname@example.org
Bossart, J. L., and J. M. Scriber. 1995. Genetic variation in oviposition preference in tiger swallowtail butterflies: interspecific, interpopulation and interindividual comparisons. Pages 183-193 in J. M. Scriber, Y. Tsubaki, and R. C. Lederhouse, editors. The swallowtail butterflies: their ecology and evolutionary biology. Scientific Publishers, Gainesville, Florida, USA.
Courtney, S. P., G. K. Chen, and A. Gardner. 1989. A general model for individual host selection. Oikos 55:55-65.
Courtney, S. P., and J. Forsberg. 1988. Host use by two pierid butterflies varies with host density. Functional Ecology 2:65-75.
Denno, R. F., and M. S. McClure, editors. 1983. Variable plants and herbivores in natural and managed systems. Academic Press, New York, New York, USA.
Dias, P. 1996. Sources and sinks in population biology. Trends in Ecology and Evolution 11:326-330.
Hanski, I. 1994. A practical model of metapopulation dynamics. Journal of Animal Ecology 63:151-162.
Hanski, I. 1999. Metapopulation ecology. Oxford University Press, Oxford, UK.
Hanski, I., M. Kuussaari, and M. Nieminen. 1994. Metapopulation structure and migration in the butterfly Melitaea cioxia. Ecology 75:747-762.
Hanski, I., A. Moilanen, T. Pakkala, and M. Kuussaari. 1996. The quantitative incidence function model and persistence of an endangered butterfly metapopulation. Conservation Biology 10:578-590.
Hanski, I., T. Pakkala, M. Kuussaari, and G. Lei. 1995a. Metapopulation persistence of an endangered butterfly in a fragmented landscape. Oikos 72:21-28.
Hanski, I., J. Poyry, T. Pakkala, and M. Kuussaari. 1995b. Multiple equilibria in metapopulation dynamics. Nature 377:618-621.
Holt, R. D. 1995. Food webs in space: an island geographic perspective. Pages 313-323 in G. A. Polis and K. O. Winemiller, editors. Food webs: integration of patterns and dynamics. Chapman and Hall, London, UK.
Ivlev, V. S. 1961. Experimental ecology of the feeding of fishes. Yale University Press, New Haven, Connecticut, USA.
Jaenike, J. 1990. Host specialization in phytophaguous insects. Annual Review of Ecology and Systematics 21:243-273.
Kuussaari, M., M. Nieminen, and I. Hanski. 1996. An experimental study of migration in the Glanville fritillary butterfly Melitaea cinxia. Journal of Animal Ecology 65:791-801.
Kuussaari, M., I. Saccheri, M. Camara, and I. Hanski. 1998. Allee effect and population dynamics of the Glanville fritillary butterfly. Oikos 82:384-392.
Mackay, D. A. 1985. Prealighting search behavior and host-plant selection by ovipositing Euphydryas editha butterflies. Ecology 66:142-151.
Mayhew, P. J. 1997. Adaptive patterns of host-plant selection by phytophagous insects. Oikos 79:417-428.
Mopper, S. 1998. Local adaptation and stochastic events in an oak leafminer population. Pages 139-155 in S. Mopper and S. Y. Strauss, editors. Genetic structure and local adaptation in natural insect populations: effects of ecology, life history and behavior. Chapman and Hall, New York, New York, USA.
Mopper, S., and S. Y. Strauss, editors. 1998. Genetic structure and local adaptation in natural insect populations: effects of ecology, life history and behavior. Chapman and Hall, New York, New York, USA.
Papaj, D. R., and R. J. Prokopy. 1989. Ecological and evolutionary aspects of learning in phytophagous insects. Annual Review of Entomology 34:315-350.
Parmesan, C. 1991. Evidence against plant "apparency" as a constraint on evolution of insect search efficiency (Lepidoptera: Nymphalidae). Journal of Insect Behavior 4:417-430.
Parmesan, C., M. C. Singer, and I. Harris. 1995. Absence of adaptive learning from the oviposition foraging behavior of a checkerspot butterfly. Animal Behaviour 50:161-175.
Pulliam, H. R. 1996. Sources and sinks: empirical evidence and population consequences. Pages 45-69 in O. E. Rhodes, Jr, R. K. Chesser, and M. H. Smith, editors. Population dynamics in ecological space and time. University of Chicago Press, Chicago, USA.
Rausher, M. D. 1978. Search image for leaf shape in a butterfly. Science 200:1071-1073.
Rausher, M. D., D. A. Mackay, and M. C. Singer. 1981. Pre-and post-alighting host discrimination by Euphydryas editha butterflies: the behavioural mechanisms causing clumped distributions of egg clusters. Animal Behaviour 29:1220-1228.
Saceheri, I., M. Kuussaari, M. Kankare, P. Vikman, W. Fortelius, and I. Hanski. 1998. Inbreeding and extinction in a butterfly metapopulation. Nature 392:491-494.
Singer, M. C. 1982. Quantification of oviposition preference by manipulation of oviposition behavior in the butterfly Euphydryas editha. Occologia 52:224-229.
Singer, M. C. 1983. Determinants of multiple host use by a phytophagous insect population. Evolution 37:389-403.
Singer, M. C. 1986. The definition and measurement of oviposition preference. Pages 65-94 in J. Miller and T. A. Miller, editors. Plant-insect interactions. Springer-Verlag, Berlin, Germany.
Singer, M. C., D. Ng, and R. A. Moore. 1991. Genetic variation in oviposition preference between butterfly populations. Journal of Insect Behavior 4:531-535.
Singer, M. C., D. Ng, and C. D. Thomas. 1988. Heritability of oviposition preference and its relationship to offspring performance within a single insect population. Evolution 42:977-985.
Singer, M. C., and C. Parmesan. 1993. Sources of variation in patterns of plant-insect association. Nature 361:251-253.
Singer, M. C., C. D. Thomas, H. L. Billington, and C. Parmesan. 1989. Variation among conspecific insect populations in the mechanistic basis of diet width. Animal Behaviour 37:751-759.
Singer, M. C., C. D. Thomas, H. L. Billington, and C. Parmesan. 1994. Correlates of speed of evolution of host preference in a set of 12 populations of the butterfly Euphydryas editha. Ecoscience 1:107-114.
Strauss, S. Y., and R. Karban. 1998. The strength of selection: intraspecific variation in host-plant quality and the fitness of herbivores. Pages 156-180 in S. Mopper and S. Y. Strauss, editors. Genetic structure and local adaptation in natural insect populations: effects of ecology, life history and behavior. Chapman and Hall, New York, New York, USA.
Thomas, C. D., and M. C. Singer. 1987. Variation in host preference affects movement patterns within a butterfly population. Ecology 68:1262-1267.
Thomas, C. D., and M. C. Singer. 1998. Scale-dependent evolution of specialization in a checkerspot butterfly: from individuals to metapopulations and ecotypes. Pages 343-374 in S. Mopper and S. Y. Strauss, editors. Genetic structure and local adaptation in natural insect populations: effects of ecology, life history and behavior. Chapman and Hall, New York, New York, USA.
Thompson, J. N., and 0. Pellmyr. 1991. Evolution of oviposition behavior and host preference in Lepidoptera. Annual Review of Entomology 36:65-89.
Via, 5. 1994. Population structure and adaptation in a clonal herbivore. Pages 58-85 in L. Real, editor. Ecological genetics. Princeton University Press, Princeton, New Jersey, USA.
Wehling, W. F., and J. N. Thompson. 1997. Evolutionary conservatism of oviposition preference in a widespread polyphagous insect herbivore, Papilio zelicaon. Oecologia 111:209-215.
Wiklund, C. 1974. The concept of oligophagy and the natural habitats and host plants of Papilia machaon L. in Fennoscandia. Entomologica Scandinavica 5:151-160.
Host use and host availability during 1993- 1996 in the five metapopulations in which oviposition preference was studied. Number of: Fraction of Relative Local groups on Veronica Metapopulation Patches populations Larval group Veronica cover Salis 27 20 190 0.84 0.63 Postad 18 17 247 0.39 0.51 Jomala 38 17 98 0.28 0.33 Foglo 28 26 181 0.00 0.00 Kumlinge 100 47 248 0.00 0.00 Number of tested: Metapopulation Populations Families Females Salis 1 8 13 Postad 3 13 24 Jomala 5 6 13 Foglo 6 18 43 Kumlinge 2 2 11 Notes: Locations of the metapopulations are shown in Fig. 2. General linear model of the probability of a larval group being found on Veronica spicata (binomial response variable, logit link function) in 43 local populations of Melitea cinxia across the Aland Islands. Mean Deviance Independent variable df Deviance deviance ratio Relative cover of Veronica spicata 1 224.8 224.8 109.80 Connectivity to Veronica use during 1993-1996 1 58.9 58.9 28.76 Connectivity to Plantago use during 1993-1996 1 11.5 11.5 5.60 Geographic location [+] 1 15.9 15.9 7.78 Connectivity to Veronica use X Connectivity 1 15.8 15.8 7.73 Plantago use Relative cover of Veronica X Geographic 1 6.9 6.9 3.37 location Residual 36 73.7 2.0 Total 42 407.5 9.7 Independent variable P Relative cover of Veronica spicata [less than]0.001 Connectivity to Veronica use during 1993-1996 [less than]0.001 Connectivity to Plantago use during 1993-1996 0.023 Geographic location [+] 0.008 Connectivity to Veronica use X Connectivity 0.009 Plantago use Relative cover of Veronica X Geographic 0.075 location Residual Total Notes: The response variable is the presence or absence of larval groups on V. spicata in each local population for all larval groups during 1993-1996, which is explained by four significant independent variable and two interaction terms. Connectivity indices [S.sub.V] and [S.sub.P] are the means for the four years. (+.)Geographic location is the location of the focal population along an axis directed from northwest toward southeast.
|Printer friendly Cite/link Email Feedback|
|Author:||KUUSSAARI, MIKKO; SINGER, MICHAEL; HANSKI, ILKKA|
|Article Type:||Statistical Data Included|
|Date:||Aug 1, 2000|
|Previous Article:||MEASURING THE EFFECTS OF POLLINATORS AND HERBIVORES: EVIDENCE FOR NON-ADDITIVITY IN A PERENNIAL HERB.|
|Next Article:||CAN HOST-PARASITOID METAPOPULATIONS EXPLAIN SUCCESSFUL BIOLOGICAL CONTROL?|