between a leaf-mining and

Quantitative behavioural traits associated with egg-laying, such as the level of selectivity for host-supports and the size of egg clutches, are generally thought to be of great importance for the subsequent survival and development of offspring. These quantitative traits, however, are often difficult to assess reliably by direct observation in the field. This is particularly the case when the insects are very tiny, which is the case for most galling and leaf mining insects. However, a new approach, the "Melba" procedure, allows the indirect inference of these quantitative traits, using easily recorded field-data only. Application of this diagnostic procedure to a large series of samples of beech leaves (Fagus silvatica), harbouring either a leaf miner, Phyllonorycter maestingella (Lepidoptera: Gracillariidae) or one or the other of two galling insects, Mikiola fagi or Hartigiola annulipes (Diptera: Cecidomyiidae) indicates that the leaf miner differs significantly from the two species of galling insect in term of combined values of host-acceptance ratio and average clutch-size, while the two gall-inducing species remain substantially undistinguishable from each other according to these traits. Thus, the galling insects (i) show stronger selectivity for a host than does the miner at any given average clutch-size and (ii) show larger average clutch-size at any given level of selectivity. That is, for at least these three species, the galling insects show a greater level of selectivity when choosing leaves to oviposit on but, then, tend to lay larger egg-clutches. These differences may be due (i) to the gall-inducing process requiring far more of leaf tissues than being simply palatable, which makes it likely that galling species will be more selective in their choice of leaves than leaf miners and (ii) to the capacity of galls to become nutrient sinks, which may help explain why the galling insects laid larger egg clutches. However, whether these trends can be regarded as general rather specific to this particular case, depends on the outcome of future studies on other groups of insects with similar life histories.


INTRODUCTION
Optimising the survival and future success of offspring in insects arguably implies that mothers lay eggs where there are the appropriate resources for the development of their larvae (Vuorisalo et al., 1989;Thompson & Pellmyr, 1991;Réale & Roff, 1992;Awmack & Leather, 2002;Fox & Czesak, 2002;Obermaier et al., 2008;Fry et al., 2009;Morrison & Quiring, 2009;Trager et al., 2009; although sometimes it is difficult to determine a clear relationship: Gripenberg et al., 2007b).The likely reason for this is that, usually, the larval stage is both the least mobile and the one mainly devoted to feeding in insects.Accordingly, the location of larval development, if it depends upon the place where eggs were laid, is likely to affect larval growth and the resulting insect fitness.Even the more "classical" herbivores (free-feeders such as chewers or more "subtle" borers and skeletonizers) have more or less particular food requirements, with many being to some extent oligotrophous or even (sub-) monophagous (Jolivet, 1983;Carrière, 1998).This, evidently, is especially the case when the larval stage must remain within a host or part of a host, such as (endo-) parasitoids and galling or mining insects.In galling and mining insects the larvae complete their development often within one discrete, clearly-bounded host-part, hereafter designated a "host-unit".This may be a leaf, bud, flower, fruit or particular section of shoot of a given host-species (equivalent to individual host of similar age, belonging to the same species, in parasitoid insects).
In quantitative terms, the degree of selectivity of egglaying mothers among available host-units may be quantified by the acceptance ratio " " , i.e. the proportion of visited host-units that are accepted for egg-laying by a gravid mother.The notion that insects select from among the host-units available, quantified in terms of the acceptance ratio, makes sense because the host-units, for example, the leaves of a given plant or tree, are generally of unequal "quality" regarding future larval development (resources, protection, etc.).Thus, the geometrical, physical, chemical and other characteristics of the same type of host-unit may often vary in space and time (Morrison & Quiring, 2009;Sarfraz et al., 2009;Zehnder et al., 2009), for example, leaves may vary in quite an unsuspected manner in many traits within the canopy of an individual tree or even within the same branch (Faeth, 1991;Kozlov & Koricheva, 1991;Gripenberg, 2007).At least, some of these characteristics may affect the development of the larvae growing upon or within a host-unit.Apparently, gravid females are able to discriminate according to these characters and select opportunely between adjacent hostunits (Vuorisalo et al., 1989;Heimpel & Rosenheim, 1995;Cope & Fox, 2003;Gibbs et al., 2005;Gripenberg & Roslin, 2005;Roslin et al., 2006;Gripenberg et al., 2007a, b;Sarfraz et al., 2009;Trager et al., 2009).While Abstract.Quantitative behavioural traits associated with egg-laying, such as the level of selectivity for host-supports and the size of egg clutches, are generally thought to be of great importance for the subsequent survival and development of offspring.These quantitative traits, however, are often difficult to assess reliably by direct observation in the field.This is particularly the case when the insects are very tiny, which is the case for most galling and leaf mining insects.However, a new approach, the "Melba" procedure, allows the indirect inference of these quantitative traits, using easily recorded field-data only.Application of this diagnostic procedure to a large series of samples of beech leaves (Fagus silvatica), harbouring either a leaf miner, Phyllonorycter maestingella (Lepidoptera: Gracillariidae) or one or the other of two galling insects, Mikiola fagi or Hartigiola annulipes (Diptera: Cecidomyiidae) indicates that the leaf miner differs significantly from the two species of galling insect in term of combined values of hostacceptance ratio and average clutch-size, while the two gall-inducing species remain substantially undistinguishable from each other according to these traits.Thus, the galling insects (i) show stronger selectivity for a host than does the miner at any given average clutch-size and (ii) show larger average clutch-size at any given level of selectivity.That is, for at least these three species, the galling insects show a greater level of selectivity when choosing leaves to oviposit on but, then, tend to lay larger egg-clutches.These differences may be due (i) to the gall-inducing process requiring far more of leaf tissues than being simply palatable, which makes it likely that galling species will be more selective in their choice of leaves than leaf miners and (ii) to the capacity of galls to become nutrient sinks, which may help explain why the galling insects laid larger egg clutches.However, whether these trends can be regarded as general rather specific to this particular case, depends on the outcome of future studies on other groups of insects with similar life histories.
such variations in quality among units on a given hostplant are usually gradual (Dajoz, 1993), females, however, are compelled to a two-fold "go -no go" response, either accepting or not accepting a given unit for subsequent egg-laying.
Then, after having selected an appropriate unit, another decision, of no less importance for future larval development, is the number of eggs a gravid female will deposit during its visit to this unit.The average number of eggs, deposited as a single bout, during a visit to an acceptable host-unit, is commonly referred to as the "clutch size" (hereafter designed as "nc").
For leaf miners and galling insects, selectivity for hostpart quality is likely to be more important for the latter, since for them the selection of a suitable host-unit for oviposition needs to take into account not only its future quality in terms of a source of food but also, and decisively, whether it can be induced by the insect to produce a suitable gall.Accordingly, the relationships between the insect and its host are likely to be even more interconnected for galling insects than leaf miners.Note that whether the mother induces the development of the gall or the hatching larva itself (as in Cecidomyiidae) is immaterial to the fact that the choice of suitable host unit is made by the mother prior to egg-laying.
Comparative studies of leaf miners and galling insects are best done on species exploiting the same type of hostunit on the same host-species.This, indeed, is not difficult since most gall-supporting plant-species may also harbour mining insects in the same host-units.On the other hand, the results and trends derived from this study may not be general but specific to the species studied.
In the present study, the selectivity of two galling insects and one leaf miner exploiting the same host-units on the same species of host, the leaves of beech-trees Fagus sylvatica, were compared.The gall-inducing species were Mikiola fagi (Hartig) and Hartigiola annulipes (Hartig) (Diptera: Cecidomyiidae) and the leaf mining species Phyllonorycter maestingella (Müller) (Lepidoptera: Gracillariidae).All three species are common and consequently, large censuses may be conducted in the field at many locations, which is desirable from a statistical point of view.
The following questions regarding these three common herbivore-insect species feeding on beech leaves are addressed: (1) Do the two gall-inducers substantially differ from the leaf miner in terms of quantitative aspects of their behaviour prior to and during oviposition, that is, do the gall-inducers and leaf miner show significantly different ranges of values for either (or both) , the "leaf acceptance ratio" and nc, the average "clutch size"?
(2) For each of the three species studied, how large is the range in variability of the leaf acceptance ratio , a question that may be connected to differences in the selective behaviour of the mothers and also to the degree of the between tree variability in terms of average leaf quality.

METHODS
Although the "acceptability" or "non acceptability" of hostunits, in this case leaves, by ovipositing mothers makes sense, estimating the proportion of acceptable units (the acceptance ratio ) is far from straightforward in practice.While the number of actually accepted leaves is easy to determine, they form just a part of the set of potentially acceptable leaves, a part which is dependent upon the density of ovipositing insects.Therefore, the proportion of leaves accepted is a poor predicator, an undefined underestimate of the proportion of potentially acceptable leaves.It is the proportion of the potentially acceptable leaves (not those actually accepted) that is meaningful in terms of mothers' requirements.Consequently, the proportion of the leaves accepted may definitely not serve as a surrogate for determining the leaf-acceptance ratio .The same holds true for the total number of eggs laid on a given leaf, an easily recorded but evidently poor predicator of clutch size nc, since several batches of eggs may be successively laid on the same leaf.
Of course, good estimates of and nc can be obtained from well defined, tedious experimental protocols (see for example Gripenberg, 2007).Alternatively, far easier and rapid determinations of and nc may be obtained indirectly by using an appropriate model implemented with easily recorded field census data.Accordingly, this second approach is especially interesting when it is planned to analyze numerous sets of data.Indeed, such extensive surveys, involving many different samples from different locations are now considered of fundamental importance to ensure a comprehensive grasp of plant-animal relationships.This is because contextual variability may substantially influence and appreciably modulate more deterministic ecological relationships (Trager et al., 2009;Zehnder et al., 2009).As quoted by Gripenberg & Roslin (2007), "the mere spatial location of a site will have major impact on local insect populations", and thus, "descriptive studies of individual habitats or sites will fall short of generality".
Thus, an indirect but more rapid procedure for determining and nc, namely the "Melba" procedure was used (for a detailed description of the procedure see Béguinot, 2005Béguinot, , 2009a, b and the Appendix hereafter for the particular treatment of the average clutch-size).This procedure relates the unknown parameters and nc to the easily recorded distribution ( ) of the total number of eggs laid per leaf (or the resulting artefacts such as mines or galls) all over the entire set of leaves within the sample.Briefly, the procedure is based on the fact that, thanks to acceptable simplifying assumptions, the distribution ( ) should theoretically be of a specific type with and nc as the adaptable parameters.Conversely, the inferred values for and nc are those that result in the best fit between the recorded distribution ( ) and the corresponding theoretical distributions ( ), parameterised in terms of and nc.Here, the field data is the distribution of singly-occupied galls or mines (which reflects that of those eggs that have successfully hatched), rather than the distribution of eggs, since here, the former is more easily recorded.Accordingly, it should be remembered that, hereafter, the term "clutch-size" nc will apply exclusively to the number of artefacts (galls or mines) that resulted from one clutch of eggs; the "artefact-clutch" size thus corresponds to the "eggs-clutch" size reduced in the proportion of eggs that have successfully hatch [the "Melba" procedure remains equally appropriate in this case as it is independent of the proportion of eggs that successfully hatch: see Béguinot (2005Béguinot ( , 2009a, b), b)].
Field data used in this study (  1. Leaf-acceptance ratio, , and average clutch size, nc, (= number of artefacts resulting from one egg-clutch), for leaf gallers (Hartigiola annulipes, Mikiola fagi) and a leaf-miner (Phyllonorycter maestingella) based on a series of samples of leaves collected from Fagus silvatica.Contextual parameters: nu = number of leaves in the sample, na = total number of galls or mines recorded in the sample, nfa = number of leaves with mines or galls in the sample.Oviposition pressure: na /( nu), mines/galls density: na/nu.
Alpes, during the years 2005 to 2009, for each of the three species considered (Béguinot, unpubl. data).Two supplementary samples from Dajoz (1981) comply well with our more recent samples and were also included in this study.In total, more than 10,000 mines or galls were recorded on the ca.36,000 leaves of Fagus silvatica surveyed.Individual samples usually consisted of all the leaves collected from one or several adjacent branches (as a rule however, leaves of the second aestival flush were duly not included in the census).

RESULTS
Sixty-six samples (respectively 17, 27, 22 samples for Hartigiola annulipes, Mikiola fagi and Phyllonorycter maestingella) were successively analyzed using the Melba procedure in order to quantitatively evaluate the following behavioural traits: the leaf-acceptance ratio and the average clutch size nc in term of the mines/galls that resulted from the eggs that successfully hatched.

Combined distributions of the behavioural traits and nc
Considering the results for the three species taken altogether, the leaf-acceptance ratio extends homogeneously across almost its entire possible range, from 0.05 to 1.00 and the average clutch size, nc, ranges from 1.00 to 1.38.Considering the species separately however, indicates that Phyllonorycter maestingella differs from the two other species, Hartigiola annulipes and Mikiola fagi, in terms of the ranges of values of its behavioural traits.The trend, however, is weak, with large overlaps when the traits and nc are considered separately, but is clear when combined values of and nc are considered, which reveals that the respective domains of the galling insects Hartigiola and Mikiola on the one hand and the leaf miner Phyllonorycter on the other hand are strikingly distinct from each other (Fig. 1).The domain of the combined values { , nc} for the leaf miner Phyllonorycter is confined, with only one exception out of 22, within a small triangle of comparatively high host acceptance ratios and low average clutch sizes and thus clearly separate from the common domain of Hartigiola and Mikiola (Chi 2 test for separation between the domains of combined values { , nc} for galling insects and leaf miner respectively: 2 = 57 (with Yates correction), p < 0.001).This trend in the mutually exclusive domains of the combined values { , nc}, depending on species identity (or guild membership), is all the more interesting because these behavioural traits, especially the leaf acceptance ratio, is very variable intra-specifically.

Variability in the leaf acceptance ratio within insect species
Even within a species there is a large variability in the leaf acceptance ratio (Fig. 1).If this variability would mainly reflect large differences in leaf quality requirements of individual conspecific insects, it would be surprising because it is an important behavioural trait.Therefore, variation in other factors that could affect the leaf acceptance ratio needs to be checked.Thus, the variability in the local population density of insects (and the resulting oviposition pressure) and the variability in average leaf quality between samples (i.e. between individual trees) were also considered.
Figs 2 and 3 provide information relevant to these suggestions.
There is no correlation between the leaf acceptance ratio, , and the number of mines or galls per potentially acceptable leaf, which is used here as a measure of oviposition pressure: coefficient of determination r 2 = 0.06, 0.03, 0.07 for Hartigiola, Mikiola and Phyllonorycter respectively (Fig. 2).On the other hand, the leaf acceptance ratio, , has a close association with the overall density of mines/galls, (number of mines or galls per leaf, considering all the leaves in the sample whether they were potentially acceptable or not): r 2 = 0.72, 0.65, 0.38 for Hartigiola, Mikiola and Phyllonorycter, respectively (Fig. 3).As will be discussed further later on, this linkage between and could actually reflect their common dependence upon between-tree heterogeneity in term of average leaf quality and thus support the likely influence of between-tree heterogeneity upon the large variability of the acceptance-ratio within a same insect species.

Combined distributions of the oviposition behavioural traits and nc
The distributions of the leaf acceptance ratio, , and the average clutch size, nc, (Fig. 1) indicate significantly distinct patterns for Hartigiola annulipes and Mikiola fagi, the two gall insects, and Phyllonorycter maestingella, the leaf miner, although all three species feed on beech leaves: the two galling insects show stronger levels of selectivity for host quality than the leaf miner, at any given value of average clutch size; and, reciprocally, the two galling insects have a larger average clutch size than the leaf miner, at any given level of leaf selectivity.In short, at least for these three species, the two galling insects are clearly more selective when choosing leaves on which to oviposit and lay more eggs per visit than the leaf miner.Strong selection for high host-plant quality in galling insects is previously reported by Craig et al. (1986), Burnstein & Wool (1993, see also references therein), Stein & Price (1995), Fernandes et al. (2000), but see also Larsson & Ekbom (1995).
The ability of mothers to select among leaves prior to egg laying is all the more likely given that the variability in leaf quality seems generally to be even greater within than between trees (Suomela & Ayres, 1994;Wallin & Raffa, 1998;Gripenberg & Roslin, 2005;Roslin et al., 2006;Gripenberg et al., 2007a, b).Therefore, a substantial part of the whole range in leaf quality occurs within a limited area, thus making it easy for insects to "shop" for those leaves that are the best for their needs.Selection for quality is commonly reported in insects (Vuorisalo et al., 1989;Gripenberg & Roslin, 2007;Gripenberg et al., 2007a).It is a logical, although not always experimentally supported assumption (Mayhew, 2001), that this selection of a high quality host prior to egg-laying is associated with improved fitness of their offspring, especially when the latter are more or less dependent on one host-unit only for their larval development (see Mayhew, 1997).
Note that as the resulting galls/mines were recorded here instead of eggs, it might be argued that the clumped distribution of galls or mines could proceed a posteriori from a difference in the ratio of eggs to mines/galls depending on leaf characteristics, rather than the consequence of the selection of oviposition sites by the insects a priori.In both cases however, it is easily verified that the expression and estimate of the acceptance ratio, , would remain strictly unchanged (Béguinot, 2009a, b).
Actually, while the selection by insects based on host quality has occasionally been challenged (Burnstein & Wool, 1993), it is generally accepted, particularly for gall inducing insects, as mentioned above.
As it is generally accepted that insects select where they lay their eggs it should prove revealing to proceed in further comparisons regarding the levels of host-quality selection either within insect species in which the populations are differently structured and located in different places, or between species and especially between groups of species, such as free-feeders, leaf miners or galling insects, which are likely to have substantially different requirements.
This would entail very large investigations, involving both extensive field surveys of large sets of insect species and a wide range in contextual variation of the hosts (Berenbaum, 1981;Brewer & Gaston, 2003).Studying such a wide and comprehensive framework requires a convenient and rapid method for evaluating host acceptance, such as the "Melba" procedure.

Variability in the leaf acceptance ratio within insect species
As already suggested, the large variability in the degree of selectivity shown by a host-specific species (Fig. 1) seems, at first sight, somewhat surprising for a specific behavioural trait that is likely to have large affect on insect fitness.But variations in the leaf acceptance ratio do not necessarily indicate a similar variation in the insect's requirement for leaf quality but may result from variations in environmental conditions as well.Among the environmental factors listed are (i) the influence of oviposition pressure, which may induce females to be less choosy about leaf quality in order to avoid overcrowding and (ii) the influence of the average leaf quality of each individual host, which may mechanically determine the proportion, , of acceptable leaves, i.e. the proportion of  leaves that are, at least, at the supposedly constant threshold level of quality required by the females.
Oviposition pressure may be measured by the density of mines/galls on either all the leaves in a sample or only those that are potentially acceptable.As the host leaves on which the eggs are laid are by definition potentially acceptable leaves, the oviposition pressure should be measured in terms of the potentially acceptable leaves.Fig. 2 shows that, in spite of the appreciably large range of its variation within each species, the leaf acceptance ratio, , remains substantially independent of oviposition pressure so defined.As a consequence, oviposition pressure is not likely to play a significant role in the large intra-specific variability in the acceptance ratio, .Thus, none of the three species studied seems to be sensitive to oviposition pressure, i.e. they do not avoid ovipositing on leaves on which there are already eggs from conspecifics.While this apparent lack of sensitivity or, at least, lack of reaction of females to the presence of previously deposited eggs, is far from general in insects, it is frequently reported for small insects such as galling insects and leaf miners (Auerbach & Simberloff, 1989;Kagata & Ohgushi, 2002;Cornelissen & Stiling, 2006) and was especially clearly demonstrated experimentally for a common and frequently studied leaf miner on Quercus leaves, Tischeria ekebladella (Gripenberg, 2007).
According to Fig. 3, the leaf acceptation ratio, , significantly increases with the overall density of mines/galls, , and there is a similar trend in the results for each of the three species studied.This positive relationship is not likely to result from any direct correlation between and insect density since the leaf acceptance ratio, , is, as indicated above, independent of oviposition pressure.Rather it is determined by the positive dependence of both and (the latter related to the population density of females) on the global level of "quality" of the tree from which the sample was collected.The better the "quality" of the tree, (i) the higher the acceptance ratio, , will be (providing the requirement threshold remains unchanged) and, (ii) the higher the number of females attracted to a tree or the longer they will remain there, the higher will be.Parameters and would then be indirectly connected in this way.It is then suggested that the variability in average leaf quality between trees, or even between leaves on branches of the same tree, may be important in determining the large range in intra-specific variability in the leaf acceptance ratio.
Of course, the results and trends derived above are, a priori, significant only for the host-herbivore system studied.Although complying with the logical arguments put forward previously, the difference in the trends recorded for the two galling insects and the leaf miner feeding on beech leaves have to be compared with that of insects feeding on other host species before any generalisation may be tentatively proposed.Indeed, it is anticipated that results might differ for other herbivores feeding on other species of hosts.For example, the significantly higher content of deterrent substances in some highly defended plants might result in leaf miners becoming as selective for leaf acceptability as gall inducing species feeding on the same host.This, indeed, would remain in keep with the preceding logic.
the recorded and corresponding theoretical distributions, ( ), will be now in terms of adaptable parameters x and , instead of nc and .In addition, as mentioned above, artefacts (mines or galls) resulting from successfully hatched eggs may be recorded and considered, instead of eggs themselves, with the only restriction that the term "clutch size" be applied to the artefacts rather than to the original clutch of eggs.

Fig. 3 .
Fig. 3. Leaf acceptance ratio, , plotted against the overall density of mines/galls in the sample, .

Fig. 2 .
Fig.2.Leaf acceptance ratio, , plotted against oviposition pressure, defined as the number of mines/galls per potentially acceptable leaf.

TABLE
Table 1) include a substantial series of 66 samples, each one from a different tree, collected in the central part of France: Burgundy, Limousin and Hautes-