†Author to whom correspondence should be addressed. E-mail: email@example.com
1. The ovigeny index, previously identified as both a significant fitness variable in parasitoid wasps and an important factor in parasitoid–host population dynamics, is the proportion of the maximum potential lifetime complement of eggs that is mature when the female emerges into the environment following pupal development. We tested the hypothesis that ovigeny index varies with female body size in parasitoid wasps. Body size measurements were obtained for 40 species in 13 families, representing a broad taxonomic and morphological diversity of parasitoid wasps. There was an almost 18-fold difference in size between the smallest and the largest species.
2. Ovigeny index is shown to be negatively correlated with body size across species – smaller wasps have a higher proportion of eggs mature at emergence than do larger wasps – a result supporting the hypothesis. This relationship has previously been observed within species.
3. The previously reported cross-species negative correlation between life-span and ovigeny index is robust, as it still holds when variation in body size is controlled for.
4. We discuss the likely selective factors in the evolution of a link between ovigeny index and body size across species.
Within an insect order there may be marked variation in the degree to which the female's lifetime potential egg complement is mature when she emerges into the environment following pupal development (Flanders 1950). For example, the orders Lepidoptera and Hymenoptera each include not only species that emerge with a fully developed lifetime egg complement, but also species whose complement entirely comprises immature oocytes. An ‘ovigeny index’– the ratio (expressed as a proportion) of the mature egg load upon emergence to the lifetime potential fecundity – was recently devised to quantify variation in the degree of egg development shown by insects both interspecifically and intraspecifically (Jervis et al. 2001). An index of 1 indicates that all the oocytes are mature upon emergence while an index of 0 denotes emergence with no mature eggs. In common with certain other life-history traits (Blackburn 1991a,b), a continuum of ovigeny index exists among Hymenoptera (Jervis et al. 2001).
The ‘ovigeny’ concept could contribute to our broader understanding of life-history evolution and diversity in insects. The ovigeny index can be employed as a simple, convenient measure of the allocation of resources to reproduction at the start of adult life (see below), and therefore be used to seek trade-offs, e.g. between current and future reproduction and between reproductive effort and survival, predicted by life-history theory (see Bell & Koufopanou 1986, Van Noordwijk & De Jong 1986; Smith 1991). Indeed, in the insects studied to date (parasitoid wasps) ovigeny index and life-span are, in agreement with life-history theory, negatively correlated both within species (Jervis et al. 2001; using data in Ellers & van Alphen 1997) and across species (Jervis et al. 2001). At least within species, this relationship reflects the differential allocation of carried-over resources to initial egg load and initial reserves (fat body storage) (Ellers & van Alphen 1997).
Furthermore, the ‘ovigeny’ concept could aid understanding of the integration of life-history traits (Boggs 1992), since it links resource capital (larval resources carried-over to the adult), income (through adult feeding) and expenditure (egg production and somatic maintenance) across an insect's life cycle (Jervis et al. 2001). Support for this view comes from comparative life-history data on Lepidoptera (e.g. in Miller 1996; Boggs 1997). The ovigeny concept parallels that of ‘income’ vs. ‘capital’ breeding, a theoretical framework that does not employ an index or other quantitative measure for general application in inter- or intraspecific comparisons (e.g. see Drent & Daan 1980; Boggs 1992, 1997; Stearns 1992; Tammaru & Haukioja 1996).
Why might ovigeny index vary with body size across species? The available evidence suggests that within species, ovigeny index declines with increasing body size (Jervis et al. 2001; Ellers & Jervis 2003). This contrasts with what happens, intraspecifically, to most other fitness-related life-history variables: in general, they show a positive correlation with body size in parasitoid wasps (fat and glycogen reserves, egg size, ovariole number, initial mature egg load, lifetime potential fecundity, lifetime realized fecundity, longevity (see References in Ellers & Jervis 2003; for exceptions, see Hardy, Griffiths & Godfray 1992; Corrigan & Lashomb 1990; Visser 1994; Kazmer & Luck 1995; West, Flanagan & Godfray 1996; King 1998; Rivero & West 2002). However, given that in foraging models a low ovigeny index, in most circumstances, confers higher lifetime fitness than a high ovigeny index (Ellers et al. 2000), a negative correlation between ovigeny index and body size makes intuitive sense. A dynamic programming model designed to study optimal egg load in relation to habitat quality in synovigenic parasitoid wasps (i.e. ovigeny index < 1) (Ellers et al. 2000), and modified to incorporate variation in both ovigeny index and body size (Ellers & Jervis 2003), explains the inverse correlation between ovigeny index and body size within species in terms of the differential allocation of carried-over resources to reproduction and survival. As body size increases, the total amount of resources increases, and the increase in allocation of these to initial egg load (the numerator in the ovigeny index) is proportionately smaller than the increase in allocation to initial reserves (which contribute to lifetime potential fecundity, the denominator in the index), leading to a decrease in the ovigeny index. The adaptive explanation for this effect is that small, short-lived individuals experience the environment as more stochastic, because they typically sample only a few patches compared with large, longer-lived individuals. In more stochastic environments the optimal egg load exceeds the expected number of hosts found (Ellers et al. 2000), therefore small parasitoids need to allocate a larger proportion of their resources to initial egg load than large individuals, even though small and large parasitoids encounter patches from the same host distribution. For large individuals the optimal initial egg load also exceeds the expected number of hosts found, but, especially in low-quality habitats, it only slightly exceeds the load of smaller individuals. The allocation strategy in this case ensures the maximum probability of finding a patch (initial reserves fuel somatic maintenance) and confers reproductive plasticity.
A negative correlation between ovigeny index and body size could potentially exist also across species, depending on the degree of influence that other factors have on: (i) the relative investments in initial egg load and initial reserves, and (ii) lifetime potential fecundity, to which initial reserves can contribute. For example, it is highly likely that lifetime potential fecundity does not vary with body size across species. This is the case with lifetime realized fecundity (Blackburn 1991a), with which lifetime potential fecundity is closely correlated (Godfray 1994). If initial egg load is positively correlated with body size (see above), then ovigeny index will show a positive, as opposed to a negative, correlation.
Extrinsic mortality acting upon the adult female parasitoid is another potential factor in the evolution of a negative correlation between ovigeny index and body size, since it could select for a shift towards earlier reproduction (i.e. a higher ovigeny index) (Jervis et al. 2001). The available field data (Visser 1994) tentatively indicate that such mortality is most intense among smaller individuals of a species. If this is the case, then the potential exists for ovigeny index to decrease with increasing body size, and for carried-over resource allocation to vary accordingly.
We test the hypothesis that ovigeny index varies with body size across parasitoid wasp species. We also seek to establish whether a link between the two life-history variables confounds the negative correlation between life-span and ovigeny index previously observed by Jervis et al. (2001). The latter interpecific relationship is highly interesting, because it is indicative of a life-span cost of concentrating reproductive effort into the early part of adult life. Empirical evidence for a trade-off between survival and reproductive effort is largely confined to the intraspecific level (e.g. Partridge & Farquhar 1981; Nur 1984, 1988; Tatar et al. 1993; Landwer 1994; see also Bell & Koufopanou 1986; Stearns 1992). Body size is known to influence life span across species (at least intragenerically) when females are starved (Eijs & van Alphen 1999). Whereas the life-span data contained in Jervis et al.'s (2001) database are for adults that were supplied with food, the diet for some of the species may nevertheless have been suboptimal with respect to fueling somatic maintenance, as laboratory-held parasitoid adults are typically provided with substitutes for their natural diet (e.g. sucrose solutions, diluted honey). Thus the possibility exists, with that database, for body size to influence life span, and for the relationship between life span and ovigeny index also to be affected.
Materials and methods
We used a subset of data on ovigeny index and its life-history correlates, taken from a database on parasitoid wasp species (Jervis et al. 2001). Data were abstracted only if we were also able to obtain body size measurements for the females.
Data on potential lifetime egg complement were used both as the denominator in calculating the ovigeny index, and in testing for a relationship between lifetime fecundity and body size. Where possible, we used either the maximum average potential fecundity or a close approximation to this – the maximum average realized fecundity, depending on the type of information available (‘maximum’ in the latter case refers to the highest mean value obtained in an experimental study involving several treatments).
We did not investigate variation in reproductive effort sensu stricto (the proportion of the available resource input that is allocated to reproduction over a defined period of time, Begon, Harper & Townsend 1990), owing to a lack of data. To establish precisely the relationship between it and ovigeny index would require at the very least measurements of female biomass and egg mass, in addition to initial egg load, for each species.
Life-span data were used both to test for a relationship with body size and to determine whether the negative correlation between life-span and ovigeny index recorded by Jervis et al. (2001) is confounded by variation in body size; this effect was not considered in that paper. In all cases, laboratory data were used and, where possible, they were for females given hosts as well as food. This allows for the possibility that oviposition has a life-span cost and also for the known habit of the females of some species to consume host blood and/or tissues and use the nutrients so obtained for egg development. Where more than one mean value was available for life-span (longevity varied, for example, with dietary regime), the highest mean value was used.
Where possible, the body sizes of at least 10 specimens of each species were measured. Our chosen proxy of body size was forewing length, measured from the tip of the wing to the tegulum. Wing length is theoretically no better or worse a measure of body size than is body length, but it is considerably easier to measure accurately. Body length measurements (tip of head to tip of gaster) in wasps are more frequently confounded by physical distortions, in particular the telescoping of the gastral segments, a common postmortem effect of drying (Gauld & Fitton 1987).
Of the 67 species used in the analysis by Jervis et al. (2001), body size measurements were obtained for 40, mostly from specimens among the collections of the Natural History Museum, London (Table 1, Fig. 1). The 40 species were from 13 families (all the families listed in Table 2 of Jervis et al. 2001; with the exception of the Eucharitidae and the Tiphiidae), representing a broad taxonomic and morphological diversity of parasitoid wasps. Measurements could not be made of the other species whose ovigeny index was known, because the females are either apterous (Hemithynnus hyalinatus Westwood), the wings of all the females in a series of specimens were badly distorted, or specimens were not available. Moreover, when taxonomic descriptions of these species were consulted, wing measurements were either not present or they were taken using a method incompatible with our own.
Table 1. Body size (wing length) and ovigeny index in 40 species of parasitoid wasp
Wing length (mm) (geometric mean)
Cephalonomia stephanoderis Betrem
Goniozus legneri Gordh
Goniozus nephantidis (Muesebeck)
Goniozus nigrifemur Ashmead
Bathyplectes anurus graecator Aubert
Bathyplectes curculionis (Thomson)
Pimpla turionellae (L.)
Aphidius picipes (Nees)
Aphidius rhopalosiphi De Stefani-Perez
Asobara tabida (Nees)
Biosteres arisanus (Sonan)
Biosteres vandenboschi (Fullaway)
Cotesia plutellae (Kurdjumov)
Cotesia flavipes (Cameron)
Microctonus hyperodae Loan
Microplitis croceipes (Cresson)
Orgilus obscurator Nees
Trioxys complanatus Quilis Perez
Leptopilina boulardi Barbotin
Trybliographa rapae (Westwood)
Anagrus erythroneurae Trjapitzin & Chiappini
Aphytis melinus DeBach
Coccophagus lycimnia Walker
Encarsia formosa Gahan
Encarsia perniciosi (Tower)
Trichogramma chilonis Ishii
Trichogramma evanescens Westwood
Cheiloneurus noxius Compere
Copidosoma floridanum Ashmead
Microterys flavus Compere
Ooencyrtus johnsoni (Howard)
Brachymeria lasus (Walker)
Chrysocharis laricinellae (Ratzeburg)
Tetrastichus atriclavus Waterston
Tamarixia radiata (Waterston)
Catolaccus grandis (Burks)
Pachycrepoideus vindemiae Rondani
Amitus hesperidum Silvestri
Gryon pennsylvanicum (Ashmead)
There was a 17·7-fold size difference between the smallest species (the egg parasitoid Trichogramma chilonis Ishii, arithmetic mean wing length ± SD: 0·46 ± 0·06 mm, geometric mean = 0·46 mm, n = 6) and the largest species (the pupal parasitoid Pimpla turionellae[L.], arithmetic mean wing length = 8·15 ± 1·88 mm, geometric mean = 7·95 mm, n = 13).
The taxonomic distributions of ovigeny index and body size suggest that these traits are, to a significant degree, phylogenetically constrained. For example, members of the family Mymaridae typically have an ovigeny index of 1, whereas all members of the Bethylidae and Ichneumonidae for which we have data have an index of 0 (Table 1). All mymarids are tiny insects, whereas ichneumonids are typically medium- to large-bodied (Gauld & Bolton 1988) (e.g. see Table 1). Thus, to avoid taxonomic pseudoreplication in our analyses, we used statistical techniques that take account of the potential effects of phylogeny upon variation in trait values (Grafen 1989; Harvey & Pagel 1991), as in our earlier study (Jervis et al. 2001). When investigating a relationship between two continuous variables (ovigeny index and body size; life-span and ovigeny index), we used Harvey & Pagel's (1991) method of independent contrasts. When investigating the effects of more than two variables simultaneously (the relationship between life-span and ovigeny index, controlling for body size), we used Grafen's (1989) phylogenetic regression, as this provides the extra flexibility required. Although these two methods assign branch lengths differently (there is insufficient information to specify them individually), in no case did we find any disagreement between the results obtained by the two methods. The information on phylogenetic relationships necessary to make such comparisons was compiled from Dowton & Austin (2001) (for Hymenoptera: Apocrita), Ronquist (1999) (for Cynipoidea), Campbell et al. (2000) (for Chalcidoidea), Gauthier et al. (2000) (for Eulophidae), Belshaw et al. (2000) (for Braconidae) and Quicke et al. (2000) (for Ichneumonidae). Some authors present several alternative trees for a particular higher taxon, in which case we based our choice of tree, where possible, on the opinions expressed by the authors, either in the publications themselves or communicated directly to us.
Using the aforementioned sources, a composite tree (Fig. 1) was produced. Where relationships could not be resolved, means for the unresolved taxa were used. All variables were transformed to logarithms to achieve normality. Unless stated to the contrary, all tests applied were two-tailed.
Because of the controversy currently surrounding phylogeny-based techniques in comparative studies (Ricklefs & Starck 1996; Price 1997; Björklund 1997; see also Charnov 1993) we supplemented these with traditional cross-species analyses that do not take phylogeny into account. Below, we distinguish between results obtained by phylogeny-based statistical techniques (‘PT’) and results obtained by traditional techniques (‘TT’). We attach greater biological significance to the results of the PT analyses, given the evident strong phylogenetic bias both in ovigeny index and in body size, among parasitoid wasps.
The relatively small size of our database, compared with that used by Blackburn (1991a,b) in his analyses of parasitoid wasp life-history traits, warrants explanation. Data on fecundity and longevity have for a long time been routinely obtained in studies of parasitoids, as measures of the reproductive capacity (and the assumed pest control potential) of a species and, more recently, as measures of fitness (Jervis & Copland 1996). Hence, data on this combination of life-history variables exist for a relatively large number of species (474, Blackburn 1991a,b). By contrast, the behavioural, ecological and evolutionary significance of both egg load and ‘ovigeny’ in parasitoid wasps has been appreciated only relatively recently (Iwasa, Suzuki & Matsuda 1984; Jervis & Kidd 1986; Kidd & Jervis 1989). Therefore, what information exists for calculating ovigeny index is presently confined to relatively few species (< 70). Of that number, there are some species for which body size measurements cannot be readily made (see below). Copies of the main database are available from M. A. Jervis.
body size and ovigeny index
Figure 2 shows the distribution of body sizes in relation to ovigeny index. Ovigeny index is significantly and negatively correlated with body size according to a phylogeny-based analysis (PT, independent comparisons, line forced through the origin: F1,37 = 5·65, P = 0·023, r2 = 13·33, Fig. 3), while the result of a traditional regression analysis is not quite significant (TT: F1,38 = 3·42, P = 0·072). On these bases, we conclude that there is a link between ovigeny index and body size, i.e. our hypothesis is supported.
life-span, initial egg load and lifetime potential fecundity
Life-span does not vary significantly with body size (PT: F1,24 = 0·93, P > 0·2; TT: F1,24 = 0·88, P > 0·2), in agreement with Blackburn's (1991a) finding. Thus, despite there being the potential for life span to be responsive to body size in laboratory-held parasitoids (see Introduction), such effect was not detected either among the parasitoid species examined here or among the taxonomically more diverse set of species investigated by Blackburn (1991a). This suggests that the laboratory diet of most of the species in each database was adequate in terms of fuelling somatic maintenance, despite the high probability of it having been artificial in most cases.
Even with a much reduced data set compared with that used in Jervis et al. (2001), we have again found that life-span is negatively correlated with ovigeny index, both by a phylogeny-based analysis and by a traditional analysis (PT: F1,24 = 5·18, P = 0·025–0·050; TT: F1,25 = 10·99, P = 0·003). Controlling for body size variation, this relationship is confirmed by both phylogenetic regression (Fig. 4) and traditional regression (PT: one-tailed: F1,23 = 4·39, P = 0·025–0·050; TT: F1,25 = 8·39, P = 0·008). So, despite the evidence that ovigeny index is negatively correlated with body size, the previously recorded (Jervis et al. 2001) negative correlation between life-span and ovigeny index still holds.
We sought to establish which of the two components of ovigeny index – the numerator (initial egg load) or the denominator (lifetime potential fecundity) – was mainly responsible for the negative correlation between ovigeny index and body size, to gain insight into differential resource allocation by females (see Introduction). Initial egg load declines significantly with increasing body size when phylogeny is taken into account, but not when it is ignored (PT: F1,24 = 5·31, P = 0·025–0·050; TT: F1,32 = 0·42, P > 0·2). When egg-yolk richness (anhydropy, hydropy, presumed measures of investment per egg, see Le Ralec 1995; Jervis et al. 2001) is considered, the decline in initial egg load falls just short of being significant (PT; F1,31 = 4·00, P = 0·05–0·1; TT: F1,30 = 4·09, P = 0·052). Lifetime potential fecundity does not vary significantly with body size according to either a phylogeny-based or a traditional analysis (PT: F1,20 = 0·01, P > 0·2; TT: F1,21 = 1·22, P > 0·2), in agreement with Blackburn's (1991a) finding for lifetime fecundity (although it should be noted that Blackburn does not specify whether his data were for realized fecundity or potential fecundity). Even when egg-yolk richness is taken into account, fecundity is still not linked to body size (PT; F1,19 = 0·08, P > 0·2; TT: F1,19 = 1·22, P > 0·2). The interaction between body size and egg-yolk richness was also included in these analyses. Neither in the case of initial egg load nor in the case of lifetime potential fecundity were the interaction terms significant (P > 0·2 in all cases). The questions of precisely how much of the observed cross-species variation in ovigeny index is explained by variation in initial egg load, and whether, like ovigeny index, initial egg load is negatively correlated with life span, are the subject of another study (M. A. Jervis, P. N. Ferns & G. E. Heimpel, manuscript in preparation).
Note that we carried out an earlier analysis based on a markedly different composite tree (more closely resembling trees based on morphological characters; notably, it did not include Campbell et al.'s 2000 findings). The results of all the analyses were qualitatively identical to those given above. A copy of that tree can be obtained from Peter Ferns.
Whereas intraspecifically lifetime fecundity and life-span generally vary significantly with body size in parasitoids (see Introduction), neither of them do so interspecifically (Blackburn 1991a; this study; but see Eijs & van Alphen's 1999 finding for starvation life-span). The lack of an interspecific body size effect on these two variables is probably attributable to the overriding importance of host availability, extrinsic mortality and intrinsic life-history factors that are idiosyncratic to particular parasitoid–host systems (Price 1975; Gauld 1987; Blackburn 1991a,b; Godfray 1994; Weisser, Völkl & Hassell 1997; Mayhew & Blackburn 1999; Pexton & Mayhew 2002). The ovigeny index includes lifetime fecundity, so it too will be moulded by these factors. Therefore, at the interspecific level, a link to body size will be weak. Nevertheless, there is a detectable relationship between ovigeny index and body size even amidst the considerable differences that exist in the suites of selection pressures acting upon parasitoid species.
How can we explain, in functional terms, the observed negative correlation between ovigeny index and body size that we have found? Explanations for this relationship must invoke ecologically and/or physiologically based parameters that lead to consistent size-related shifts in optimal egg maturation schedules. Because of the high variability, among species, in selection along multiple gradients, we expect these effects to be strong for their influence to be detected in such a restricted sample as ours. We have pointed to habitat quality (host availability) as one candidate factor exerting such shifts (see Introduction), but there is insufficient information, in the primary literature, on host species-characteristic patterns of abundance and distribution, to enable us to assess its role in determining interspecific body size-related variation in ovigeny index. Extrinsic mortality is the other candidate factor: a higher level of extrinsic mortality for smaller species could, in principle, select for shifts in the egg maturation schedule towards the start of adult life (i.e. a higher initial egg load and therefore a higher ovigeny index). Mortality rates of minute parasitoids can indeed be extremely high in the field, with life-expectancy measured in hours rather than days or weeks (Heimpel, Rosenheim & Mangel 1997; Rosenheim 1998). No clear differences were found in field predation rates suffered by adult aphelinid and aphidiine braconid parasitoids (Rosenheim 1998) despite the fact that aphelinids are much smaller than aphidiines (∼ three times, e.g. see Table 1). In an intraspecific analysis, however, Visser (1994) showed that smaller individuals of Aphaereta minuta (Nees) (an alysiine braconid) suffer a higher mortality risk under field conditions. The causes of size-related death in A. minuta were not identified, although one likely cause is desiccation.
Small-bodied parasitoid wasps should run a higher risk of desiccation compared with large-bodied wasps because of their higher surface area to volume ratio. Terrestrial insects lose water by evaporation through the integument even when the cuticle is undamaged (Chapman 1998), and both within and across species (Loveridge 1968 and Edney 1971, respectively) smaller insects have been shown to lose a larger proportion of their original body mass per unit time than larger insects when maintained in dry air. While mortality rates in the parasitoids Metaphycus helvolus (Compere) (Encyrtidae) and Aphytis lingnanensis Compere (Aphelinidae) are similar when wasps are held at 50% r.h., the larger M. helvolus suffers much lower mortality than the minute A. lingnanensis at 0% r.h. (Bartlett 1962). On the other hand, a study of the eulophid Achrysocharoides zwoelferi (Delucchi) recorded no size-related difference in mortality rate under low humidity (a positive size–longevity relationship was detectable under a higher humidity regime) (West et al. 1996). Whatever the nature of the size-related mortality in parasitoid wasps, it would need to be concentrated among older, small females for it to select for a shift towards early egg production (see Stearns & Hoekstra 2000).
Given that a negative correlation between ovigeny index and body size exists interspecifically as well as intraspecifically, the question arises as to whether it is due to the same pattern of carried-over resource allocation (see Introduction). It is reasonable to assume that the absolute amount of resource carry-over is correlated with body size across species, which is known to be the case with other insects (Wickman & Karlsson 1989; Karlsson 1994). However, it is unclear whether the size-related partitioning of these resources accords with the predictions of the Ellers & Jervis (2003) model. Nothing at all is known about the allocation of initial reserves in relation to body size, across parasitoid wasp species. The decline in initial egg load with increasing body size contrasts both with theoretical expectations (see Introduction) and with the clear increase recorded intraspecifically in parasitoid wasps (see Opp & Luck 1986; le Masurier 1991; King & King 1994; Visser 1994; Olson & Andow 1998; Pexton & Mayhew 2002). However, this apparent anomaly can be explained by the cross-species trade-off between egg size (which is positively correlated with body size, see Berrigan 1991; and Blackburn 1991b) and egg load, a relationship predicted by general life-history theory (Smith & Fretwell 1974; Fox & Czesak 2000) and confirmed empirically (Berrigan 1991).
In referring to fitness consequences at the intraspecific level, Godfray (1994) argued that the single factor posing the greatest problem for the development of a quantitative behavioural ecology for parasitoids is the lack of a full understanding of the consequences to an adult of being large or small. A similar argument can be extended to the causes of cross-species diversity in reproductive strategies, including variation in ovigeny index. For the development of a comprehensive theory linking parasitoid size and ovigeny index, both within and among species, all the following need quantifying: size-related variation in the amount of carried-over resources, egg maturation rate, metabolic costs, host encounter rate, variation in habitat quality (hosts and food), stochasticity in host and food availability, and size-related extrinsic mortality. At the comparative level, model parameterization would require data from: (i) physiological studies of cross-species differences in metabolic rate, ovarian dynamics and resource allocation patterns in relation to body size; (ii) field measurements of host species-characteristic patterns of abundance and dispersion; (iii) field measurements of size-specific encounter rates; and (iv) field life-table studies of adult parasitoids, involving the recording, from among a taxonomically diverse array of species, of body size-related data on reproductive success and survival.
We thank: Jacintha Ellers, Hefin Jones and Neil Kidd for commenting on early drafts; the trustees of the Natural History Museum, London, for permitting MAJ to examine the museum's collections of parasitoid wasps; Mike Fitton and NHM colleagues for enabling MAJ to carry out measurements; Mike Copland for supplying pupae of Eretmocerus sp.; Jay Rosenheim for supplying body size data on one species; Carol Boggs, Mark Dowton, George Else, Ian Hardy, Brad Hawkins, Bethia King, David Lees, Peter Mayhew, John Noyes and Donald Quicke for useful advice; Alan Grafen for making available a copy of an implementation (PHYLO.GLM 1·03) of his phylogenetic regression; and two anonymous referees for suggesting very useful improvements to the manuscript.