Phenotypic plasticity in foraging behavior of sawfly larvae

Antti Kause


Phenotypic plasticity, an ability of a genotype to express different phenotypes in response to different environments, may be adaptive or purely environmentally constrained. Despite some debate in the literature (Via 1993), plasticity studies have proven to be an efficient approach for understanding the causes of variation in foraging, growth, and life history (reviewed in Schlichting 1986, 1989, Stearns 1989, Stearns et al. 1991, Newman 1992, Scheiner 1993a, Gotthard and Nylin 1995, Nylin and Gotthard 1998). Plasticity has been studied within populations, among populations, and among species, but most research has concentrated on the benefits of plasticity. There remains a lack of knowledge about its limits and costs (DeWitt et al. 1998).

Traits linked to fitness may be canalized (i.e., buffered against genetic or environmental perturbations [Waddington 1942, Schmalhausen 1986, Scharloo 1991, Stearns et al. 1995]) by phenotypic plasticity. This may occur if traits form correlated networks that respond to environmental variation as an entity, rather than individual traits responding with independent reactions (e.g., Oka 1976, Schlichting 1986, 1989, Rollo 1994). For instance, to minimize detrimental effects on a high-level trait closely connected to fitness, environmentally forced reduction in one low-level trait may be compensated for by adjustments of other low-level traits. Thus, such correlated plasticity can serve as a buffer against environmental variation. In addition to possible benefits, correlated plastic responses may lead to costs; e.g., trade-offs between functionally connected traits may constrain the possible set of plastic responses. Adaptive plasticity should be of particular importance to organisms with restricted movement or a sessile life-style, because they must necessarily cope with variation in local environmental conditions.

Insect herbivores show plastic responses in foraging behavior along leaf quality gradients, and often maintain surprisingly steady growth rates on diets that vary markedly in chemical and physical parameters. In fact, larval growth rate is generally considered as a key life history trait in insects, which is strongly correlated with fitness (reviewed in Nylin and Gotthard 1998). Changes in diet quality can induce at least two different phenotypic responses in foraging behavior. Certain herbivore species are able to increase consumption on nutrient-diluted or otherwise low-quality diets, thus compensating for reduced postingestive physiological efficiency (reviewed in Slansky 1993). Other species may reject a low-quality diet and search for high-quality leaves within the foliage (Schultz 1982, 1983, Edwards and Wratten 1983, Edwards et al. 1991, Barker et al. 1995). As a consequence, foraging behavior and postingestive physiology both influence larval growth, and seem likely to be an intercorrelated set of traits that can respond integrally to diet quality. Yet it remains unclear whether the two foraging strategies show mutual trade-offs setting limits or constraints to possible plastic responses.

Hanhimaki et al. (1995) showed that the larval growth rates of 15 species of herbivorous insects feeding on mountain birch were affected in similar ways by intertree variation in leaf quality. Thus, good trees for one herbivore species tended to be good trees for other species. Hanhimaki et al. (1995) also hypothesized that the similarities among species in growth rates (a high-level trait) were the result of their differing capacities to exploit leaves of various qualities, possibly through differing responses in foraging behavior and digestive physiology (the low-level traits). In this paper, we tested for differences among species, in responding to variation in leaf quality. We studied diet-induced changes in distribution of feeding sites, relative consumption rate, postingestive physiology, and relative growth rate in six species of leaf-chewing sawfly larvae (Hymenoptera: Symphyta) feeding on mountain birch, Betula pubescens ssp. czerepanovii (Orlova). The study included species whose larvae feed during the early, mid, and late season, and therefore encounter different environmental conditions. In early summer, the initially high chemical and physical quality of mountain birch leaves decreases rapidly during leaf maturation (Haukioja et al. 1978, Ayres and MacLean 1987, Hanhimaki et al. 1995, Nurmi et al. 1996), and owing to shoot- and leaf-specific leaf growth schedules and leaf ages, extensive within-canopy variation exists in leaf quality (Suomela and Ayres 1994, Suomela et al. 1995a, b). In contrast, during late summer and autumn, leaf quality is generally low and foliage is relatively homogeneous within mountain birch canopies, since all leaves are mature and tough, and no young expanding leaves are available (Haukioja et al. 1990). Consequently, we were interested in whether sawfly species with different phenologies have different plastic behavioral responses to leaf quality variation, in the form of either compensatory consumption or dispersion of feeding sites. Our multispecies approach enabled us to evaluate the possible benefits and costs of the two behavioral strategies, and evaluate whether or not trade-offs constrain the degree of plastic responses (DeWitt et al. 1998).


Study site and objects

The study was conducted at the station of the Kevo Subarctic Research Institute in northern Finland (69 [degrees] N, 27 [degrees] E) in 1996. The forests at the study site are almost pure mountain birch stands, Betula pubescens ssp. czerepanovii (Orlova) Hamet-Ahti (= tortuosa [Ledeb.] Nyman). Mountain birch forms the tree line in the area. The leaf quality of mountain birch as a resource for insect herbivores displays large spatial variation (Senn et al. 1992, Hanhimaki et al. 1994, 1995, Suomela et al. 1995a, b, Nurmi et al. 1996) and temporal variation (Haukioja et al. 1978, Ayres and MacLean 1987, Hanhimaki et al. 1995, Nurmi et al. 1996). In 1996, birch leaves attained their full size after 12 July and senescence took place in mid-September.

Unlike most other taxa, sawflies (Hymenoptera: Symphyta) attain their greatest species diversity in northern latitudes (Kouki et al. 1994). The birch feeding guild in Finnish Lapland includes some 40 species (Koponen 1978). In our study area most sawflies have a univoltine life cycle, but some species show an extra long diapause (S. Hanhimaki, unpublished data). Sawflies overwinter as prepupae and the adults emerge in the spring, when leaves are large enough for oviposition. Females oviposit in or on young mountain birch leaves; due to species-specific differences in adult emergence and length of the egg stage, the larval feeding periods of different species span the whole growing season, from spring to late autumn.

The six sawfly species studied were chosen haphazardly from among the common birch-feeding species, but in such a way that their combined larval periods covered the whole summer. In the following, we use the terms early, mid, and late-season species according to the timing of the larval period of the sawfly species. We used two early-summer species, Amauronematus amplus Konow and A. kevoensis Vikberg; one midsummer species, Nematus brevivalvis Thomson; and three late-summer species, Priophorus pallipes Lepeletier, Arge sp. Schrank, and Dineura pullior Schmidt and Walter. Arge sp. belongs to the family Argidae, while the other species are tenthredinids. D. pullior was previously known as D. virididorsata (Retzius); the species were recently separated by Schmidt and Walter (1995). Arge sp. is a member of the A. ustulata-clavicornis group, and in some earlier studies the name A. fuscinervis Lindqvist had been used for this species (Hanhimaki et al. 1994, 1995). The larvae of all the species are solitary external feeders.

Larval material

The experimental larvae were the offspring of individuals collected from the local natural population during the year preceding the experiment. Pairs of adults of both sexes were enclosed in mesh bags on the branches of nonexperimental mountain birches. The enclosed larvae were allowed to spend most of their larval period in net enclosures in trees in the field, and were brought into the laboratory only shortly before the experiments. In the laboratory, the larvae were reared individually in 100-mL transparent plastic vials and given freshly detached short-shoot leaves from nonexperimental mountain birch trees every 3 d. The positions of the vials in the vial frames were randomized. To synchronize pre-experimental larval development within a species, larger individuals were held at a temperature of 1 [degrees] C. By this means we slowed down their development, apparently without any detrimental effects on the larvae of these subarctic species. Similar short-term temperature treatments did not change the performance of the autumnal moth, Epirrita autumnata (Bkh), caterpillars (M. Ayres, unpublished data). All species were tested at the time of their natural occurrence; otherwise they were treated in a similar manner, and there were no visible differences in the viability of the species. Depending on the species, the experiments were conducted with fourth- or fifth-instar larvae.

The experiment

We studied plastic responses in larval foraging behavior, postingestive physiology, and relative growth rates by rearing larvae on leaves from trees having known different qualities for herbivores. The experiments were run on 19 July (A. amplus), 23 July (A. kevoensis), 29 July (N. brevivalvis), 12 August (P. pallipes), and 16 August (Arge sp. and D. pullior). We used 4-6 larval broods from each species. In the laboratory we individually reared 4-6 larvae (each from a different brood) on short shoot leaves of each of the 20 experimental trees. Short shoot leaves were used because within a tree they are of equal age, unlike long shoot leaves. The larvae were offered intact short shoots, picked haphazardly from the canopy, carrying on average three leaves. At the beginning of the experiment, the larvae were weighed to the nearest 0.1 mg, allowed to feed for exactly 24 h at 12 [degrees] C, and reweighed. After the experiment, the leaves given to the larvae were collected and pressed. The numbers and sizes of meals and the leaf areas fed were measured using an image analysis computer (MCID, M4, Imaging Research Incorporated, Ontario, Canada). To transform the areas consumed into fresh and dry masses, an additional sample of five short shoot leaves per tree was collected (on 19 and 29 July, 12 and 16 August), weighed fresh, pressed, and dried at 60 [degrees] C for a week. The sample was then reweighed, and the leaf areas were recorded. Mean specific leaf dry masses for each tree were also calculated. The leaf sample for 19 July was used for both A. amplus and A. kevoensis.

TABLE 1. Means and their variances ([[micro]gram]/[mm.sup.2]) in

specific leaf dry masses of short shoot leaves on each sampling


Sampling date Mean Variance

19 July 75.33 0.0728

29 July 78.55 0.0479

12 August 79.48 0.0630

16 August 76.12 0.0632

To attain large variation in leaf quality among our experimental trees, we chose 20 trees, including the best and the worst, from a pool of 40 trees previously studied (Nurmi et al. 1996, Ossipov et al. 1997). This is possible because quality differences among mountain birch individuals remain consistent between successive years (Senn et al. 1992, Hanhimaki et al. 1995, Suomela et al. 1995b). In the present paper, we take the tree-specific specific leaf dry mass as an index of leaf quality. Specific leaf dry mass measures the amount of leaf matter per unit of leaf area; it correlates positively with leaf toughness (Ayres and MacLean 1987) and with concentrations of potentially important resistance compounds. Our measure of leaf quality is characterized by chemical data for a leaf sample taken from the experimental trees 3 d before the first bioassay. Specific leaf dry mass correlated significantly with soluble proanthocyanidins (r = 0.70, P = 0.0006), marginally significantly with cell wall-bound proanthocyanidins (r = 0.41, P = 0.08), and negatively with total nitrogen (r = -0.58, P = 0.007), sucrose (r = -0.48, P = 0.03), and the proportion of water (r = -0.68, P = 0.001), but not with total carbohydrates (r = -0.29, P = 0.21, n = 20 trees in all correlations). In mountain birch leaves, cell wall unpalatability to insect herbivores is due in part to proanthocyanidins (Nurmi et al. 1996, Ossipov et al. 1997), which efficiently impede larval growth (V. Ossipov et al., unpublished manuscript). In most woody plants, carbohydrates are largely translocated as sucrose (Kramer and Kozlowski 1979), and high foliar water and nitrogen content supports high larval growth rates in insects (Scriber and Slansky 1981, Mattson and Scriber 1987). Thus, a high specific leaf dry mass indicates both poor palatability and low levels of nutritional quality.

Our results might have been affected by differences among experiments in the means and magnitudes of variation in specific leaf dry masses of short shoot leaves. To test for this, repeated measures ANOVA was used to check whether the means varied among the leaf sampling dates (Procedure GLM with type III sum of squares, SAS Institute 1990). The same experimental trees were sampled four times (Table 1), and the sampling date was thus regarded as a within-subject variable. The tree was regarded as a random independent effect, and hence the mean squares of the tree-by-sampling-date interaction were used as an error term for the date-main effect (Zar 1984). The sampling date was treated as a fixed effect (Mead 1988). The data satisfied the no-sphericity assumption (Mauchly’s criterion = 0.97, chi-square = 2.07, df = 5, P = 0.84; PRINTE option in GLM procedure), and the residuals of the model did not deviate from the normal distribution. Levene’s test was used to check whether variances differed among the leaf sampling dates. In addition, differences among the coefficients of variation in the samples of specific leaf dry masses were checked after logarithmic transformation of the original observations by Levene’s test for homogeneity of variances (Sokal and Brauman 1980). The samples did not differ among the different sampling dates (Table 1), as indicated by nonsignificant differences in means (sampling date-main effect: F = 1.79, df = 3, 57, P = 0.16), variances (F = 0.15, df = 3, 76, P = 0.93), and CVS (F = 0.21, df = 3, 76, P = 0.89) of specific leaf dry masses of short shoot leaves.

Traits measured

For each larva, we recorded four preingestive behavioral traits: number and size of meals (based on consumed fresh leaf mass), and relative consumption rates based on fresh ([]) and dry ([]) leaf masses consumed. In addition, a postingestive trait (efficiency of conversion of ingested food to larval biomass, ECI) and the relative larval growth rate (RGR) were calculated. The following formulae were used to calculate food utilization indices: RCR = consumption/initial larval mass, ECI = growth/consumption, and RGR = loge(final mass) – [log.sub.e](initial mass); units of these indices are in milligrams per milligram. All gravimetric indices were based on dry larval masses, and time was not included in formulae because all larvae were allowed to feed for 24 h. Fresh larval masses were transformed to dry masses using species-specific regressions calculated from samples of larvae (n = 2050 per species), which were weighed, dried, and reweighed. The regression of A. amplus was used for its close relative A. kevoensis due to the lack of spare larvae for the latter species. The statistical analyses based on fresh larval masses gave similar results to ones using dry masses. A hole or a notch in a leaf was regarded as a meal. The number of meals describes the dispersion of feeding sites and movement of the larva. The consumption rate based on fresh mass describes the actual mass consumed and is important in characterizing behavioral responses of larvae to leaf quality when the treatment foods differ in dry mass concentration or water content, as in this study (Slansky 1993).

Statistical analyses

Analyses of variance. – To test for species-specific differences in the use of the same individual trees in meals, relative consumption rates, physiological processing of leaf material, and relative growth rates, we ran two-way parametric ANOVAs in which a significant species-by-tree interaction implies among-species differences in host use. To control the significance levels for multiple tests, we first performed MANOVA using all larval traits as dependent variables, and only proceeded with univariate ANOVAs if MANOVA indicated a significant species-by-tree interaction (procedure GLM with type III sums of squares, SAS Institute 1990). This approach is referred to as a protected ANOVA by Scheiner (1993b). The tree, the sawfly species, and interaction were included in both the MANOVA and ANOVA models as independent variables. Both the tree and the species were regarded as random effects, since they represent a sample of the natural variation within the local tree population and the sawfly guild. Thus, in the MANOVA, the matrix, and in the ANOVA, the mean squares of tree-by-species interactions were used as an error term for the main effects (Zar 1984). The residuals of the models did not deviate from the normal distribution, but dry- and fresh-mass-based relative consumption rates, size, and number of meals were [log.sub.e] and RGR square-root transformed to homogenize the variances.

Path analyses. – We used both correlation and path analysis to describe the detailed mechanisms whereby larval traits responded to the gradient in specific leaf dry mass, our composite measure of leaf quality. A correlation coefficient describes the total relationship between two traits measured. Therefore, a correlation between specific leaf dry mass and a given larval trait summarizes the total plastic response in a given larval trait along the leaf quality gradient. By means of a path analysis, correlations can be subdivided into direct effects and effects via other traits (Mitchell 1993). Direct effects (i.e., path coefficients) are standardized partial regression coefficients that describe the effect of a given variable on another, while holding other traits statistically constant.

The larval traits form a hierarchical network. When a larva starts to disperse its feeding bouts, the movement of the larva interrupts the current meal, leading to decreased meal size. Moreover, both the number and size of the meals affect the relative consumption rate. The relative growth rate is mainly a product of RCR and ECI (Scriber and Slansky 1981). RGR can also be influenced by larval movements, and ECI may function as a feedback mechanism to leaf quality, partially controlling the consumption of leaf material. Thus, dispersion of meals (indicating larval movements) and meal size form the base level, RCR and ECI the second level, and growth rate the highest level in the hierarchical network linking leaf quality and larval growth. This trait network was used as a basic model in the analyses.

The CALIS procedure of SAS and its RAM statement were used to carry out the path analyses (SAS Institute 1990). In all the analyses, tree-specific trait means were used to avoid pseudoreplication. To avoid multicollinearity, the fresh-mass-based RCR was used to describe realized consumption, and the corresponding dry-mass-based measure was therefore excluded from the analyses. No other larval traits were excluded, since they all contribute strongly to the structure of the phenomena studied. The path analyses were conducted for each species separately according to the following rules. First, we started from a basic model in which all paths were considered equally important. We removed the weakest paths one by one until the removal caused a significant difference between the models. To compare two models, the difference between their goodness-of-fit chi-square values was computed, which is distributed as a chi-square with degrees of freedom equal to the difference in degrees of freedom between the two models (Mitchell 1993). Significant or marginally significant paths were not removed from the analyses. An inherent constraint emerged from the structure of the data: the relationship between specific leaf dry mass and number of meals was obligatory in all models. Otherwise, the program would have automatically connected the traits to each other with a correlation. The residuals of the models were checked as described in Mitchell (1993) and found not to deviate from the normal distribution. Multicollinearity was tested by constructing the path diagrams using multiple regressions (REG procedure and its STB, VIF, and COLLIN options). The largest variance inflation factors were [less than] 3, indicating lack of multicollinearity. In addition, the largest condition index value was 28, which is just below the critical value (Petraitis et al. 1996). The maximum likelihood option of the CALIS procedure was used to calculate the chi-square goodness-of-fit of the models, and high P values indicate that the path model did not deviate from the observed correlation structure. Unanalyzed causes, i.e., influences on each endogeneous variable that are unexplained by the model, were calculated as [-square root of 1 – [R.sup.2], where [R.sup.2] denotes the total explained variance in the variable.

The reliability of the significance tests of individual path coefficients provided by the CALLS procedure was checked in several ways. For two reasons, a path analysis performed using multiple regressions (procedure REG, SAS Institute 1990) is more reliable. First, the significance tests of individual path coefficients provided by the CALIS procedure are approximate. Second, a low ratio of sample size to number of estimated parameters in a path analysis may influence significance levels of individual path coefficients (Bollen 1989), and this ratio is increased when multiple regression is applied. In no case was a significant path coefficient calculated using the CALIS procedure nonsignificant, when multiple regression (procedure REG) was applied, suggesting that the significance levels were not spurious in our path diagrams. A disadvantage of regression analysis is that an overall fit of a path model to the observed correlation structure cannot be tested. For each individual path, we further calculated the ratio of the path coefficient estimate to its standard error (Bollen 1989). In four species, this ratio was low for the path that was obligatory and nonsignificant in the models, i.e., the path between specific leaf dry mass and number of meals. All other paths showed ratio values near or over 2, indicating no problems with the reliability of the individual components (Bollen 1989).

Interspecific correlations. – ECI, together with the RCR, forms the plastic response of RGR. This occurs because the relative growth rate is mainly a product of relative consumption rate and ECI (Scriber and Slansky 1981). Accordingly, for each species we calculated the ability to counterbalance reduced ECI in the leaf quality gradient as the difference between the plastic responses in ECI and RGR, i.e., the correlations of ECI and RGR with specific dry leaf mass. The ability to counterbalance reduced ECI describes the species’ capability to buffer larval growth rate against leaf quality variation. To demonstrate the benefit of compensatory consumption, we calculated a Pearson correlation between the species-specific plastic response in fresh leaf mass based relative consumption rate, i.e., a correlation of RCR with specific dry leaf mass, and ability to counterbalance reduced ECI. To describe trade-offs among plastic responses in foraging behavior, we plotted pairs of the species-specific correlations between larval traits and leaf quality, i.e., total plastic responses (Schlichting 1986, 1989). For instance, a trade-off between two plastic responses is reflected by a negative relationship between the total plastic responses of the traits.


A significant species-by-tree interaction in the MANOVA justified the use of univariate ANOVAs to identify out which of the larval traits created this result (Pillai’s Trace, F = 2.18, df = 570, 2658, P = 0.0001). Univariate ANOVAs revealed that foraging behavior traits, but not postingestive physiology and growth rate, showed significant species-by-tree interactions, establishing the existence of species-specific foraging behavior on the leaves of different quality (Tables 2 and 3). The larvae consumed and distributed their feeding sites in species-specific ways on the leaves of the experimental trees, as indicated by significant species-by-tree interactions in number and size of meals and relative consumption rates. Interestingly, however, both the physiological utilization of leaves and the relative growth rates of larvae did not indicate any species-specific patterns, as shown by the lack of significant species-by-tree interactions in ECI and RGR (Table 2). Thus, the species had similar larval growth rates across trees of variable leaf quality, but they exhibited species-specific plasticity in foraging behavior.

Among-species differences in plasticity

The correlations and path diagrams described the differences among species in foraging plasticity and [TABULAR DATA FOR TABLE 2 OMITTED] the general similarity among species in the postingestive plasticity [ILLUSTRATION FOR FIGURE 1 OMITTED]. Species were similar in that high specific leaf dry mass tended to reduce the conversion efficiency (i.e., ECI). In the path diagrams, the impact of leaf quality on larval physiology is reflected by negative path coefficients between specific leaf dry mass and ECI. In contrast, there were fundamental differences in foraging plasticity among species with different larval phenologies; on low-quality leaves, larvae of the early-season species significantly dispersed their feeding sites, the midseason species exhibited only weak changes in their behavior, and the three late-season species exhibited significant compensatory consumption [TABULAR DATA FOR TABLE 3 OMITTED] [ILLUSTRATION FOR FIGURE 1 OMITTED]. The larvae of A. amplus, the earliest species, had significantly more feeding bouts and screened (sampled and ingested small amounts of) more leaves on poor-quality trees with high specific leaf dry masses, indicating increased dispersion of feeding sites. This was shown by a significant positive path coefficient from specific leaf dry mass to number of meals [ILLUSTRATION FOR FIGURE 1 OMITTED], and by a significant correlation between specific leaf dry mass and number of leaves tasted by the larvae (r = 0.45, P = 0.05, n = 20). The plastic response in number of meals of A. kevoensis, the other early species, showed a pattern similar to that of A. amplus, although the path from leaf quality to number of meals was nonsignificant. In contrast, larvae of the midsummer species, N. brevivalvis, as well as those of the three phenologically latest species, P. pallipes, Arge sp., and D. pullior did not display strong systematic plastic responses in the dispersion of feeding sites, as indicated by small path coefficients between specific leaf dry mass and number of meals.

Significant positive correlations between specific leaf dry mass and fresh leaf mass based relative consumption rates revealed that larvae of the three phenologically latest species, P. pallipes, Arge sp., and D. pullior showed significantly higher relative consumption rates on poor-quality trees, i.e., those with a high specific leaf dry mass (r = 0.57, P = 0.0088; r = 0.51, P = 0.02; r = 0.44, P = 0.05; n = 20, respectively). These correlations were partitioned into direct and indirect effects in the path analyses, which indicated that compensatory consumption was achieved in different ways by the late-season species. In P. pallipes, compensatory consumption resulted because meal size was increased by both reduced conversion efficiency and by low leaf quality, as shown by the negative paths from both ECI and specific leaf dry mass to meal size [ILLUSTRATION FOR FIGURE 1 OMITTED]. In D. pullior, a feedback mechanism was implied by a path from ECI to relative consumption rate: RCR was directly, although marginally significantly, increased by reduced ECI. In Arge sp., low leaf quality slightly and directly increased both meal size and relative consumption rate [ILLUSTRATION FOR FIGURE 1 OMITTED], as reflected by the paths from specific leaf dry mass to meal size and RCR. In contrast to the late-season species, larvae of the two early-season species, A. amplus and A. kevoensis, and the midsummer species, N. brevivalvis, did not exhibit compensatory consumption; in all species, the correlation between specific leaf dry mass and fresh leaf consumption rate was weak (r = 0.08, P = 0.73; r = 0.11, P = 0.66; r = -0.00, P = 0.99; n = 20, respectively). Consequently, the path analyses showed no arrows from leaf quality to either meal size or to relative consumption rate [ILLUSTRATION FOR FIGURE 1 OMITTED].

Costs and benefits of behavioral plasticity

The relationships between foraging behavior and relative larval growth rate were also species specific. In A. amplus and D. pullior, increased movement, as evidenced by their large number of meals, reduced relative larval growth rates, as reflected by the significant paths from number of meals to relative growth rate [ILLUSTRATION FOR FIGURE 1 OMITTED]. The effect was presumably a direct one, as the statistical procedure we used eliminated effects via all other traits we measured. However, the reasons for this possible cost of increased movement were different: unlike A. amplus, the increased number of meals in D. pullior was not a consequence of poor leaf quality [ILLUSTRATION FOR FIGURE 1 OMITTED]. The species exhibiting strong compensatory consumption on poor-quality trees (the three late-season species) counterbalanced the generally low physiological efficiency on poor-quality trees better than the other species, as indicated by a significant positive relationship between the correlations of specific leaf dry mass with relative consumption rate (i.e., the plastic responses in []) and the ability to counterbalance reduced ECI [ILLUSTRATION FOR FIGURE 2 OMITTED]. Assuming that a stable growth rate is advantageous, this result appears to reveal a benefit of compensatory consumption.

The relationships among species-specific correlations of specific leaf dry mass with behavioral traits demonstrated trade-offs among the plastic responses. There was a significant trade-off between dispersion of feeding sites and meal size; the species whose larvae dispersed their feeding sites on low-quality leaves consumed smaller meals [ILLUSTRATION FOR FIGURE 3A OMITTED]. Consequently, the largest species-specific increases in relative consumption rates along the leaf quality gradient were related to a lack or reduced dispersion of feeding sites [ILLUSTRATION FOR FIGURE 3B OMITTED]. Without N. brevivalvis the same correlation is much stronger (r = -0.99, P = 0.002, n = 5), indicating that the foraging behavior of N. brevivaivis larvae was different from that of the other species, and was relatively insensitive to leaf quality. The important biological conclusion is that none of the larvae for any of these species were able to simultaneously increase their consumption rate and disperse their feeding sites on poor-quality trees with a high specific leaf dry mass.


The results of the experimental manipulations provide evidence that, when reared on mountain birch trees whose leaf quality for herbivores varied from good to poor, the sawfly guild shows species-specific plastic responses in larval foraging behavior. This is in spite of the species maintaining relatively similar larval growth rates across trees of variable leaf quality. This was proposed by Hanhimaki et al. (1995) and is now confirmed by our findings. To our knowledge, these are the first observations showing that dispersion of feeding sites, exhibited by the early-season species, and compensatory consumption, displayed by the late-season species, tend to be mutually exclusive plastic foraging strategies for maintaining stable growth rates on diets of variable quality.

Dispersion of feeding sites

In the earliest species, A. amplus, the plastic behavioral response to low-quality leaves was to disperse feeding sites [ILLUSTRATION FOR FIGURE 1 OMITTED]. Although we studied the effects of among-tree variation in leaf quality on behavior, we suggest that the mechanism of increased movement is also applicable to within-tree variation. As proposed by Schultz (1982, 1983) and Edwards and Wratten (1983), dispersion of feeding sites may be beneficial to larvae if it leads them to higher quality leaves within a canopy. Indeed, mountain birch canopies are of heterogeneous quality to insect herbivores (Suomela and Ayres 1994, Suomela et al. 1995a, b), particularly during early summer, due to the large within-canopy variation in biochemical and physical leaf characteristics. For instance, within-canopy variation in foliar water content accounts for 44% (among-tree variation accounts for 21%) of the total observed variation within the local mountain birch population. Corresponding values are 42 and 12% for leaf toughness, 12 and 64% for total nitrogen, and 21 and 64% for total phenolics (Suomela and Ayres 1994, Suomela et al. 1995b). As a result, larvae of early-season herbivores on mountain birch may benefit from increased movements because they may be able to move from low- to high-quality leaves.

However, dispersion of feeding sites reduced relative growth rate in A. amplus [ILLUSTRATION FOR FIGURE 1 OMITTED]. This possible cost appeared not to be unique to diet-induced plasticity in feeding site dispersion, but also included larvae of D. pullior, which had high overall mobility (Table 3) without any systematic plasticity in number of meals [ILLUSTRATION FOR FIGURE 1 OMITTED]. Hence, cost of increased number of meals is not necessarily a cost of plasticity, since a species with fixed behavior pays the same cost (DeWitt et al. 1998). In a field situation, increased movement may also increase the mortality risk by actions of natural enemies (den Boer 1971, Schultz 1982, 1983, Edwards and Wratten 1983, Bergelson and Lawton 1988), although feeding per se may also be risky (Bernays 1997).

There appeared to be a gradual seasonal transition from dispersion of feeding sites to compensatory consumption. Larvae of A. kevoensis, the other early-season species, showed plastic responses in foraging behavior that were similar, but weaker, to those of A. amplus, and the midsummer species, N. brevivalvis, exhibited no strong indications of plastic behavioral responses.

Compensatory consumption

Consumption and postingestive physiology formed an intercorrelated set of traits that responded to leaf quality variation as a whole, forming a buffer against food quality variation. Larvae of the three late-season species, P. pallipes Arge sp., and D. pullior, displayed active plastic increases in fresh leaf biomass consumption on poor-quality trees [ILLUSTRATION FOR FIGURE 1 OMITTED], thus compensating for reduced physiological efficiency of processing leaf material on poor-quality trees [ILLUSTRATION FOR FIGURE 2 OMITTED]. This compensatory consumption is common among lepidopteran larvae. Nutritional studies conducted using artificial diets have shown responses in consumption rates to individual nutrients that are important to insect growth (reviewed in Slansky 1993). However, the mechanisms of compensatory consumption are not the same in all species; in our data, compensatory consumption was achieved in species-specific ways [ILLUSTRATION FOR FIGURE 1 OMITTED]. Growth rate has been considered to be strongly correlated with fitness in herbivorous insects (reviewed in Scriber and Slansky 1981, Nylin and Gotthard 1998). Stable growth rates in sawfly larvae presumably have similar consequences as in Epirrita autumnata (Bkh), a birch-feeding geometrid, whose decreased larval growth rate leads to long larval periods and small final size (Tammaru 1998). In E. autumnata and in many other insects, a long larval period and small size have negative fitness consequences via decreased fecundity (Haukioja and Neuvonen 1985, Tammaru et al. 1996a, b; for a general review see Leather 1988, Honek 1993, Nylin and Gotthard 1998) and low survival (Haukioja et al. 1988).

Our analysis revealed a significant trade-off between compensatory consumption and dispersion of feeding sites; redistribution of feeding sites on low-quality leaves constrained the ability to consume larger meals and to increase relative consumption rate [ILLUSTRATION FOR FIGURES 3A, B OMITTED]. Compared to dispersion of feeding sites, which is the strategy used by early-season species, compensatory consumption may be a better foraging strategy during the late summer with low within-canopy heterogeneity in leaf quality. This may occur because in the late summer, after cessation of long shoot growth and the production of new leaves, all the leaves are relatively old (Haukioja et al. 1990). In addition, mature leaves are generally less nutritious and tougher than young foliage (Haukioja et al. 1978, Ayres and MacLean 1987), even though the contents of soluble low-molecular phenolics decrease with the season, presumably because they become permanently bound to cell walls, thereby increasing leaf toughness (Nurmi et al. 1996, Ossipov et al. 1997).

To test whether the among-species differences in plasticity of foraging behavior are adaptive, a reciprocal experiment should be conducted. Evidence for adaptive plasticity would be provided if the plastic response of an early-season species would lead to higher fitness compared to that of a late-season species, when both species are tested at the same time during early summer, and vice versa (Gotthard and Nylin 1995). It should be noted that the two early-season species belong to the same genus, so their similar behavior may have resulted from common phylogeny. In contrast, the late-season species belong to several distinct genera, but they still all showed similar plastic responses in foraging behavior. Moreover, the observed among-species differences in behavior cannot be explained by differences in the average size of larvae (Table 3); for instance, despite a more than six-fold difference in the average larval size, all late-season species behaved similarly.

Implications for plant-herbivore interactions

In resistance against herbivores, low leaf quality has been suggested to disperse rather than to decrease the amount of leaf biomass removed by herbivores, leading to less detrimental effects on plant growth and reproduction (Schultz 1982, 1983, Edwards and Wratten 1983). On the other hand, low-quality leaves may not be advantageous for plants if the low quality leads to increased leaf damage levels due to compensatory consumption (Moran and Hamilton 1980). Recent studies have demonstrated that simulated dispersed leaf damage produces fewer harmful effects on plants, compared to local, concentrated defoliations (Marquis 1988, 1992, Price and Hutchings 1992, Mauricio et al. 1993, Honkanen and Haukioja 1994, Coleman and Leonard 1995), especially when the leaves are young (Tuomi et al. 1988, 1989, Ruohomaki et al. 1997). Consequently, the different foraging strategies observed in this study may have differing effects on mountain birch. However, to determine the true relevance of different feeding patterns on plants, future studies should directly test whether variation in insect feeding behavior affects plant performance.

Our multispecies approach showed the diversity among species in larval foraging behavior along the leaf quality gradient. This diversity calls for caution in extrapolating the results of studies based only on a single species. A guild level study is needed to unravel the generality of patterns and mechanisms observed in herbivore responses to rapidly changing hosts.


Irma Saloniemi gave encouraging advice whenever needed. Virpi Lummaa, Kai Ruohomaki, Toomas Tammaru, Mark A. McPeek, Christian Peter Klingenberg, Matt Ayres, and an anonymous reviewer gave valuable comments on earlier versions of the paper. K. Ruohomaki also kindly let us use his field-collected Dineura individuals. The staff at the Kevo Institute supported us in many ways. The center of Biotechnology in Turku provided an image-analyzer for leaf measurements. Janne Henriksson, Outi Seppala, and Miia Tanhuanpaa helped in collecting the larvae. The English was checked by Ellen Valle. The study was financed by the Academy of Finland. We are grateful to all of them.


Ayres, M. E, and S. E MacLean, Jr. 1987. Development of birch leaves and the growth energetics of Epirrita autumnata (Geometridae). Ecology 68:558-568.

Barker, A.M., S. D. Wratten, and P. J. Edwards. 1995. Wound-induced changes in tomato leaves and their effects on the feeding patterns of larval lepidoptera. Oecologia 101:251-257.

Bergelson, J. M., and J. H. Lawton. 1988. Does foliage damage influence predation on the insect herbivores of birch? Ecology 69:434-445.

Bernays, E. A. 1997. Feeding by lepidopteran larvae is dangerous. Ecological Entomology 22:121-123. Bollen, K. A. 1989. Structural equations with latent variables. John Wiley and Sons, New York, New York, USA.

Coleman, J. S., and A. S. Leonard. 1995. Why it matters where on a leaf a folivore feeds. Oecologia 101:324-328.

den Boer, M. H. 1971. A colour-polymorphism in caterpillars of Bubalus piniarius (L.) (Lepidoptera: Geometridae). Netherlands Journal of Zoology 21:61-116.

DeWitt, T. J., A. Sih, and D. S. Wilson. 1998. Costs and limits of phenotypic plasticity. Trends in Evolution and Ecology 13:77-81.

Edwards, P. J., and S. D. Wratten. 1983. Wound induced defenses in plants and their consequences for patterns of insect grazing. Oecologia 59:88-93.

Edwards, P. J., S. D. Wratten, and R. M. Gibberd. 1991. The impact of inducible phytochemicals on food selection by insect herbivores and its consequences for the distribution of grazing damage. Pages 205-221 in D. W. Tallamy and M. S. Raupp, editors. Phytochemical induction by herbivores. John Wiley and Sons, New York, New York, USA.

Gotthard, K., and S. Nylin. 1995. Adaptive plasticity and plasticity as an adaptation: a selective review of plasticity in animal morphology and life history. Oikos 74:3-17.

Hanhimaki, S., J. Senn, and E. Haukioja. 1994. Performance of insect herbivores on hybridizing trees: the case of the subarctic birches. Journal of Animal Ecology 63:163-175.

Hanhimaki, S., J. Senn, and E. Haukioja. 1995. The convergence in growth of foliage-chewing insect species on individual mountain birch trees. Journal of Animal Ecology 64:543-552.

Haukioja, E., and S. Neuvonen. 1985. The relationship between size and reproductive potential in male and female Epirrita autumnata (Lep., Geometridae). Ecological Entomology 10:267-270.

Haukioja, E., S. Neuvonen, S. Hanhimaki, and P. Niemela. 1988. The autumnal moth in Fennoscandia. Pages 163-178 in A. A. Berryman, editor. Dynamics of forest insect populations. Patterns, causes, implications. Plenum, New York, New York, USA.

Haukioja, E., P. Niemela, L. Iso-Iivari, and E.-M. Aro. 1978. Birch leaves as a resource for herbivores. II. Variation in the suitability of leaves. Reports from the Kevo Subarctic Research Station 14:5-12.

Haukioja, E., K. Ruohomaki, J. Senn, J. Suomela, and M. Walls. 1990. Consequences of herbivory in the mountain birch (Betula pubescens ssp tortuosa): importance of the functional organization of the tree. Oecologia 82:238-247.

Honek, A. 1993. Intraspecific variation in body size and fecundity in insects: a general relationship. Oikos 66:483-492.

Honkanen, T., and E. Haukioja. 1994. Why does a branch suffer more after branch-wide than after tree-wide defoliation? Oikos 71:441-450. Koponen, S. 1978. Tutkimuksia tunturikoivun herbivori-hyonteisista Pohjois-Fennoskandiassa ja Etela-Gronlannissa. Ph. Lic. thesis. (In Finnish.) University of Turku, Turku, Finland.

Kouki, J., P. Niemela, and M. Viitasaari. 1994. Reversed latitudinal gradient in species richness of sawflies (Hymenoptera, Symphyta). Annales Zoologici Fennici 31:83-88.

Kramer, P. J., and T. T. Kozlowski. 1979. Physiology of woody plants. Academic Press, New York, New York, USA.

Leather, S. R. 1988. Size, reproductive potential and fecundity in insects: things aren’t as simple as they seem. Oikos 51:386-388.

Marquis, R. J. 1988. Intra-crown variation in leaf herbivory and seed production in striped maple, Acer pensylvanicum L. Oecologia 77:51-55.

—–. 1992. A bite is a bite is a bite? Constraints on response to folivory in Piper arieianum (Piperaceae). Ecology 73:143-152.

Mattson, W. J., and J. M. Scriber. 1987. Nutritional ecology of insect folivores of woody plants: nitrogen, water, fiber, and mineral considerations. Pages 105-146 in F. J. Slansky and J. G. Rodriquez, editors. Nutritional ecology of insects, mites, spiders, and related invertebrates. John Wiley and Sons, New York, New York, USA.

Mauricio, R., M. D. Bowers, and F. A. Bazzaz. 1993. Pattern of leaf damage affects fitness of the annual plant Raphanus sativus (Brassicaceae). Ecology 74:2066-2071.

Mead, R. 1988. The design of experiments. Statistical principles for practical application. Cambridge University Press, Cambridge, UK.

Mitchell, R. J. 1993. Path analysis: pollination. Pages 211231 in S. M. Scheiner and J. Gurevitch, editors. Design and analysis of ecological experiments. Chapman and Hall, New York, New York, USA.

Moran, N., and W. D. Hamilton. 1980. Low nutritive quality as defense against herbivores. Journal of Theoretical Biology 86:247-254.

Newman, R.A. 1992. Adaptive plasticity in amphibian metamorphosis. What type of phenotypic variation is adaptive, and what are the costs of such plasticity? BioScience 42: 671-678.

Nurmi, K., V. Ossipov, E. Haukioja, and K. Pihlaja. 1996. Variation of total phenolic content and individual low-molecular-weight phenolics in foliage of mountain birch trees Betula pubescens ssp. tortuosa). Journal of Chemical Ecology 22:2023-2040.

Nylin, S., and K. Gotthard. 1998. Plasticity in life-history traits. Annual Review of Entomology 43:63-83.

Oka, H. I. 1976. Mortality and adaptive mechanisms of Oryza perennis strains. Evolution 30:380-392.

Ossipov, V., J. Loponen, E. Haukioja, and K. Pihlaja. 1997. Gallotannins of birch Betula pubescens leaves: HPLC separation and quantification. Biochemical Systematics and Ecology 25:493-504.

Petraitis, P. S., A. E. Dunham, and P. H. Niewiarowski. 1996. Inferring multiple causality: the limitations of path analysis. Functional Ecology 10:421-431.

Price, E. A. C.and M. J. Hutchings. 1992. Studies of growth in the clonal herb Glechoma hederacea. II. The effects of selective defoliation. Journal of Ecology 80:39-47.

Rollo, D. C. 1994. Phenotypes, their epigenetics, ecology and evolution. Chapman and Hall, London, UK.

Ruohomaki, K., E. Haukioja, S. Repka, and K. Lehtila. 1997. Value of leaves: effects of damage to individual leaves on growth and reproduction of mountain birch shoots. Ecology 78:2105-2117.

SAS Institute. 1990. SAS/STAT user’s guide. Fourth edition, version 6. SAS Institute, Cary, North Carolina, USA.

Scharloo, W. 1991. Canalization: genetic and developmental aspects. Annual Review of Ecology and Systematics 22: 65-93.

Scheiner, S. M. 1993a. Genetics and evolution of phenotypic plasticity. Annual Review of Ecology and Systematics 24: 35-68.

—–. 1993b. MANOVA: multiple response variables and multispecies interactions. Pages 94-112 in S. M. Scheiner and J. Gurevitch, editors. Design and analysis of ecological experiments. Chapman and Hall, New York, New York, USA.

Schlichting, C. D. 1986. The evolution of phenotypic plasticity in plants. Annual Review of Ecology and Systematics 17:667-693.

—–. 1989. Phenotypic integration and environmental change. BioScience 39:460-464.

Schmalhausen, I. I. 1986. Factors of evolution. The theory of stabilizing selection. Reprinted from 1949 edition. University of Chicago Press, Chicago, Illinois, USA.

Schmidt, S., and G. H. Walter. 1995. Description of Dineura pullior sp. n. (Hymenoptera: Tenthredinidae), with quantified observations on saw wear. Entomologica Scandinavica 26:385-392.

Schultz, J. C. 1982. Impact of variable plant defensive chemistry on susceptibility of insects to natural enemies. Pages 37-54 in P. A. Hedin, editor. Plant resistance to insects. American Chemical Society Symposium Series 208. American Chemical Society, Washington, D.C., USA.

—–. 1983. Habitat selection and foraging tactics of caterpillars in heterogeneous trees. Pages 61-90 in R. E Denno and M. S. McClure, editors. Variable plants and herbivores in natural and managed systems. Academic Press, New York, New York, USA. Scriber, J. M., and E Slansky, Jr. 1981. The nutritional ecology of immature insects. Annual Review of Entomology 26:183-211.

Senn, J., S. Hanhimaki, and E. Haukioja. 1992. Among-tree variation in leaf phenology and morphology and its correlation with insect performance in the mountain birch. Oikos 63:215-222.

Slansky, F. 1993. Nutritional ecology: the fundamental quest for nutrients. Pages 29-91 in N.E. Stamp and T. M. Casey, editors. Caterpillars: ecological and evolutionary constraints on foraging. Chapman and Hall, New York, New York, USA.

Sokal, R. R., and C. A. Brauman. 1980. Significance tests for coefficients of variation and variability profiles. Systematic Zoology 29:50-66.

Stearns, S. C. 1989. The evolutionary significance of phenotypic plasticity. BioScience 39:436-445.

Stearns, S., G. de Jong, and B. Newman. 1991. The effects of phenotypic plasticity on genetic correlations. Trends in Evolution and Ecology 6:122-126.

Stearns, S. C., M. Kaiser, and T. J. Kawecki. 1995. The differential genetic and environmental canalization of fitness components in Drosophila melanogaster. Journal of Evolutionary Biology 8:539-557.

Suomela, J., P. Kaitaniemi, and A. Nilson. 1995a. Systematic within-tree variation in mountain birch leaf quality for a geometrid, Epirrita autumnata. Ecological Entomology 20: 283-292.

Suomela, J., V. Ossipov, and E. Haukioja. 1995b. Variation among and within mountain birch trees in foliage phenols, carbohydrates, and amino acids, and in growth of Epirrita autumnata larvae. Journal of Chemical Ecology 21:1421-1446.

Suomela, S., and M. P. Ayres. 1994. Within-tree and among-tree variation in leaf characteristics of mountain birch and its implifications for herbivory. Oikos 70:212-222.

Tammaru, T. 1998. Determination of adult size in a folivorous moth: constraints at instar level. Ecological Entomology 23:80-89.

Tammaru, T, P. Kaitaniemi, and K. Ruohomaki. 1996a. Realized fecundity in Epirrita autumnata (Lepidoptera: Geometridae): relation to body size and consequences to population dynamics. Oikos 77:407-416.

Tammaru, T., K. Ruohomaki, and K. Saikkonen. 1996b. Components of male fitness in relation to body size in Epirrita autumnata (Lepidoptera, Geometridae). Ecological Entomology 21:185-192.

Tuomi, J., P. Niemela, I. Jussila, T. Vuorisalo, and V. Jormalainen. 1989. Delayed budbreak: a defensive response of mountain birch to early-season defoliation? Oikos 54: 87-91.

Tuomi, J., T. Vuorisalo, P. Niemela, S. Nisula, and V. Jormalainen. 1988. Localized effects of branch defoliations on weight gain in female inflorescences in Betula pubescens. Oikos 51:327-330. Via, S. 1993. Regulatory genes and reaction norms. American Naturalist 142:374-378.

Waddington, C. H. 1942. Canalization of development and the inheritance of acquired characters. Nature 150:563-565.

Zar, J. H. 1984. Biostatistical analysis. Second edition. Prentice-Hall, Englewood Cliffs, New Jersey, USA.

COPYRIGHT 1999 Ecological Society of America

COPYRIGHT 2000 Gale Group

You May Also Like

On lumping species in food webs

On lumping species in food webs Andrew R. Solow INTRODUCTION A community food web describes the binary feeding relationships b…

Control of gut retention time by secondary metabolites in ripe Solanum fruits

Control of gut retention time by secondary metabolites in ripe Solanum fruits Sofia A. Wahaj INTRODUCTION Despite the importan…

Bird communities in transition: the Lago Guri islands

Bird communities in transition: the Lago Guri islands John Terborgh INTRODUCTION Notwithstanding a substantial literature dedi…

Differential responses of littoral communities to ultraviolet radiation in an alpine lake

Differential responses of littoral communities to ultraviolet radiation in an alpine lake Rolf D. Vinebrooke INTRODUCTION Th…