INTRODUCTION
Any scientific study, including those on parasite ecology, reveals some patterns or processes. However, a question always remains: how general are these patterns and processes and to what extent do they apply to taxa, settings, or times other than those that were the subject of study? The findings of a particular study should invariably be validated by studies in other geographical locations or on other taxa, if we are to uncover any general law (Poulin, Reference Poulin2007). The identification of such patterns would suggest that apparently diverse and idiosyncratic assemblages may have common and self-organizing principles. Ultimately, the goal of such a comparative approach should be to identify these processes underpinning any observed universal pattern.
For example, studies of variation in parasite abundance and/or species richness across different populations of the same host species have demonstrated that these parameters, on the one hand, represent genuine host species traits. This was found, for example, for nematodes (abundance) (Arneberg et al. Reference Arneberg, Skorping and Read1997) and fleas (abundance and species richness) (Krasnov et al. Reference Krasnov, Shenbrot, Mouillot, Khokhlova and Poulin2005, Reference Krasnov, Shenbrot, Khokhlova and Poulin2006) in mammalian hosts and endoparasites of teleost fish (abundance) (Poulin, Reference Poulin2006), but not for intestinal helminths of mammals (species richness) (Poulin and Mouillot, Reference Poulin and Mouillot2004). On the other hand, these parameters might be also substantially affected by environmental parameters and be considered as characteristic of a geographical locality (Krasnov et al. Reference Krasnov, Shenbrot, Khokhlova and Poulin2006). Furthermore, when taxonomic composition of parasite assemblages rather than mere number of species was taken into account, it appeared that this parameter was repeatable in helminth endoparasites of mammals (Poulin and Mouillot, Reference Poulin and Mouillot2004) but was as variable across, as within, host species in fleas (Krasnov et al. Reference Krasnov, Shenbrot, Mouillot, Khokhlova and Poulin2005). Thus, at first glance, one may conclude that the repeatability of parasite abundance within host species emerges as a general rule while species richness and taxonomic diversity do not, although whether it applies universally within localities across host species remains to be further validated.
When the percentage of variation among samples accounted for by differences among host species is considered, it appears that the relative strength of the effect of host identity on parasite species richness varies greatly among parasite and host taxa. For example, the difference between mammalian hosts as opposed to that among populations within a host explained 32·8% of the variation for flea assemblages on mammals (Krasnov et al. Reference Krasnov, Shenbrot, Mouillot, Khokhlova and Poulin2005), 32% for larval trematodes in snails (Poulin and Mouritsen, Reference Poulin and Mouritsen2003), but only 14·5% for intestinal helminths in mammals (Poulin and Mouillot, Reference Poulin and Mouillot2004). On the other hand, the proportion of the variance in parasite abundance or prevalence that occurred among host species, as opposed to within, was 24·1% for mammalian hosts and fleas (for abundance), whereas this value attained 23% for larval trematodes in snail hosts (for prevalence) (Poulin and Mouritsen, Reference Poulin and Mouritsen2003) and only 13% for various metazoan parasites in fish hosts (for abundance) (Poulin, Reference Poulin2006). This suggests that the effect of host identity on infestation parameters of different parasite taxa depends on some peculiarities of the relationships in a particular host-parasite association. For example, assemblages of fleas were found to be affected little by morphological and physiological features of a host species, but much more strongly by the parameters of the host environment (Krasnov et al. Reference Krasnov, Shenbrot, Khokhlova and Degen2004). As mentioned above, in this parasite-host association, the proportion of the variation in parasite abundance and species richness among host samples associated with differences between host species was not particularly high (Krasnov et al. Reference Krasnov, Shenbrot, Mouillot, Khokhlova and Poulin2005, Reference Krasnov, Shenbrot, Khokhlova and Poulin2006).
Another taxon of haematophagous arthropods, gamasid mites, display quite different patterns in their relationships with host features (Korallo et al. Reference Korallo, Vinarski, Krasnov, Shenbrot, Mouillot and Poulin2007). When the diversity of these parasites was considered among roughly the same set of host species in roughly the same geographical area as with fleas, it appeared to be strongly affected by species-specific host features and much less by parameters of the host environment. Consequently, mite assemblages are expected to depend more strongly on host identity than was the case for fleas.
Gamasids are characterized by extremely high interspecific variation in their ecology and feeding modes. They include soil-dwelling and nidicolous predators, and both facultative and obligate vertebrate ecto- and endoparasites (see Radovsky, Reference Radovsky and Kim1985 for review). However, here we focused on haematophagous species collected from host bodies. These mites use their hosts both as food sources and as dispersal vehicles, and, thus, the association between these mite species and their hosts is assumed to be very intimate (Radovsky, Reference Radovsky and Kim1985).
The aim of this study was to test whether the abundance and diversity of gamasid mites parasitic on small mammals from 26 different geographical regions of the Palaearctic are determined mainly by host identity rather than by parameters of the abiotic environment. The focus of our analyses is on the contemporary patterns that can be observed, rather than on their underlying co-phylogenetic historical origins. We evaluated the repeatability of estimates of mite abundance and diversity across populations of the same host species, to determine if the abundance and diversity are repeatable within host species; i.e. if the values of abundance and diversity are more similar among populations of the same host species or regions than among different host species or regions, respectively. In addition, we searched for correlations between the abiotic characteristics of a region and mite abundance and diversity, separately for several host species.
MATERIALS AND METHODS
Data set
Data on gamasid mites collected from the bodies of small mammals (Soricomorpha, Lagomorpha and Rodentia) in 26 different regions of the Palaearctic were obtained from published surveys and unpublished data that reported the number of mites of a particular species found on each given small mammal species in a particular location (Table 1). We used data on host species that (a) occurred and were found infested with gamasid mites in at least 2 regions and (b) were represented in a regional survey by at least 3 individuals. This amounted to 42 host species (31 rodents, 10 soricomorphs and 1 lagomorph) occurring in 26 regions and comprising 237 host-region associations.
Table 1. Data on small mammals from 26 regions used in the analyses
(Numbers in parentheses represent the total numbers of sampled individuals.)
Abundance and diversity estimates
For each host species in each region we calculated abundance of all mite species as well as 2 measures of the diversity of mite assemblages, namely species richness and taxonomic distinctness (Δ+). Abundance of mites was calculated as mean number of mites per individual host. Other measures of infection level, such as prevalence and intensity of infestation, were not available for most of the regions considered. Mite abundance correlated weakly, albeit significantly, with host sampling effort (number of host individuals examined) (r 2=0·02, F 1,234=5·6, P<0·05; after log transformation). Consequently, to control for the confounding effort of unequal sampling, the original values of mite abundance were substituted with their residual deviations from the regression on sampling effort in log-log space.
The two measures of mite species diversity we used were (a) the number of mite species found on a host species, or species richness and (b) average taxonomic distinctness (Δ+) of the mite species present. Estimates of parasite species richness may be biased if some hosts are examined more intensively than others (Morand and Poulin, Reference Morand and Poulin1998). Indeed, mite species richness appeared to be strongly affected by sampling effort (r 2=0·44, F 1,234=186·3, P<0·001). Consequently, each value of mite species richness was then substituted by its residual deviation from a regression on the number of hosts examined in log-log space. This provided a measure of mite species richness that is independent of sampling effort.
When these mite species are placed within a taxonomic hierarchy, the average taxonomic distinctness is the mean number of steps up the hierarchy that must be taken to reach a taxon common to 2 mite species, computed across all possible pairs of mite species (Clarke and Warwick, Reference Clarke and Warwick1998, Reference Clarke and Warwick1999; Warwick and Clarke, Reference Warwick and Clarke2001; Poulin and Mouillot, Reference Poulin and Mouillot2004). The greater the taxonomic distinctness between mite species, the higher the number of steps needed, and the higher the value of the index Δ+. Using the taxonomic classification of Bregetova (Reference Bregetova1956), Radovsky (Reference Radovsky and Kim1985), and Halliday (Reference Halliday1998), all mite species were fitted into a taxonomic structure with 4 hierarchical levels above species, i. e. genus, subfamily, family and superfamily (Dermanyssoidea). The maximum value that the index Δ+ can take is thus 4 (when all mite species belong to different families), and its lowest value is 1 (when all mite species belong to the same subgenus or species group). However, since the index cannot be computed for hosts exploited by a single mite species, we assigned a Δ+ value of 0 to these host species, to reflect their extremely species-poor mite assemblages. The number of mite species exploiting a host species was significantly positively correlated (albeit weakly) with Δ+ (r 2=0·37, F 1,234=140·8, P<0·001), indicating that this measure was influenced by the number of species in an assemblage. Therefore, in subsequent analyses Δ+ was corrected for mite species richness in an assemblage by substituting the original values with their residual deviations from the regression on mite species richness in log-(log+1) space.
Environmental factors
For each region, we computed climatic variables (annual, winter and summer precipitation, mean surface air temperature of January, mean surface air temperature of July, and mean annual surface air temperature) and elevation parameters using the Global Ecosystems database (Kineman et al. Reference Kineman, Hastings, Ohrenschall, Colby, Schoolcraft, Klaus, Knight, Krager, Hayes, Oloughlin, Dunbar, Ikleman, Anderson, Burland, Dietz, Fisher, Hannaughan, Kelly, Boyle, Callaghan, Delamana, Di, Gomolski, Green, Hochberg, Holquist, Johnson, Lewis, Locher, Mealey, Middleton, Mellon, Nigro, Panskowitz, Racey, Roake, Ross, Row, Schacter and Weschler2000). These variables where calculated for a buffer of 100×100 km around the centre of each region (because it was not possible to pinpoint the precise sampling area for some of the regions). Because some of these variables strongly correlated with each other, we substituted them with the scores of principal components calculated from these 7 variables. The resulting 3 factors explained 91·5% of the variance, and their eigenvalues were 3·58, 1·59 and 1·23. The first factor (F1) represented an increase in air temperature, whereas the second (F2) and third (F3) factors represented an increase in (a) mean elevation and winter precipitation and (b) annual and summer precipitation, respectively (Table 2).
Table 2. Linear correlation (r) between each of the principal components (factors F1, F2 and F3) and each of 7 environmental variables calculated for each of 26 geographical regions
Data analysis
To determine whether abundance and species diversity of mite assemblages, expressed either as mite species richness or average taxonomic distinctness among mites (Δ+), are geographically invariant, i.e. a parameter that varies less among populations of the same host species than among host species, we performed a repeatability analysis (see Arneberg et al. Reference Arneberg, Skorping and Read1997; Poulin and Mouritsen, Reference Poulin and Mouritsen2003; Poulin and Mouillot, Reference Poulin and Mouillot2004; Krasnov et al. Reference Krasnov, Shenbrot, Mouillot, Khokhlova and Poulin2005, Reference Krasnov, Shenbrot, Khokhlova and Poulin2006). Using host species which occurred in at least 2 regions, we analysed the variation in the mite abundance, number of mite species and taxonomic distinctness among mites (Δ+) in 3 separate one-way ANOVAs in which host species was the independent factor. A significant effect of host species would indicate that the measures are repeatable within host species, i.e. that they are more similar to each other than to values from other host species. We estimated the proportion of the total variance originating from differences among host species, as opposed to within species, following Sokal and Rohlf (Reference Sokal and Rohlf1995). To assess whether mite abundance and diversity are determined also by a complex of environmental conditions, we performed the repeatability analyses using regions where at least 2 of 42 hosts occurred (as a proxy for geographical differences in a set of environmental conditions) instead of host species as the single factor (25 regions). A significant effect of region would indicate that the mite abundances and/or diversities are repeatable within region, i.e. that they are more similar among populations of different host species within the same region than among regions.
We analysed the effect of environmental parameters (expressed as 3 composite variables extracted from original environmental measures using principal component analysis, see above) on variation in abundance and diversity of mite assemblages across regions, within each of 18 host species which occurred in at least 6 regions using Generalized Linear Models (GLM) with a normal distribution and power-link function, and searched for the best model using the Akaike's Information Criterion. Then, we tested the significance of the parameter estimates in each best model using the Wald statistic.
We did not apply the Bonferroni adjustment of alpha-level as this approach has been increasingly criticized by statisticians and ecologists in recent years, because it often leads to the incorrect acceptance of the false null hypothesis when multiple comparisons are in fact independent of one another (Rothman, Reference Rothman1990; Perneger, Reference Perneger1998, Reference Perneger1999; Moran, Reference Moran2003; Garcia, Reference Garcia2004) as is the case in our study.
RESULTS
Each of the 42 host species in the data set was recorded from between 2 and 17 regions. The repeatability analysis for these host species demonstrated that mean mite abundance per host individual was repeatable within host species. Abundances of mites on the same host species were more similar to each other than expected by chance, and varied significantly among hosts (F 41,194=6·6, P<0·0001), with 50·3% of the variation among samples explained by differences between host species (Fig. 1A). At the same time, abundances of mites were also repeatable among host species within a region (F 24,210=2·7, P<0·0001), although only 16·3% of the variation among samples was accounted for by differences between regions (Fig. 1B).
Fig. 1. Rank plot of mite abundance across 42 hosts (A) and 25 regions (B). The 42 host species recorded in at least 2 regions (A) and 25 regions where at least 2 host species occurred (B ) are ranked according to mean log-transformed mite abundance values corrected for host sampling effort, with rank 1 given to the host or region with the lowest mean mite abundance; all sample estimates are plotted for each host species or region. If variation is small within compared to between host species or region, we expect the points of the plot to stretch from the lower left to the upper right corner, with few or no points in either the upper left or lower right corner.
Host- and region-related patterns of variation in mite species diversity differed from those for abundance. Although mite species richness was repeatable both within host species among regions and among host species within a region (F 41,194=2·6 and F 24,210=4·7, respectively; P<0·01 for both; Fig. 2), the percentage of the variation among samples explained by differences between host species was slightly lower than that accounted for by differences between regions (22·4% versus 29·5%, respectively). In contrast, taxonomic distinctness of mite assemblages was only weakly, but significantly, repeatable within host species (F 41,194=1·7, P<0·05) with only 10·3% of the variation among samples accounted by differences between hosts, but it was not repeatable among host species within a region (F 24,210=1·3, P>0·1) (Fig. 3).
Fig. 2. Rank plot of mite species richness across 42 hosts (A) and 25 regions (B). See Fig. 1 for explanations.
Fig. 3. Rank plot of mite taxonomic distinctness (Δ+) across 42 hosts (A) and 25 regions (B). See Fig. 1 for explanations.
In 15 of 18 host species (except for Apodemus agrarius, Micromys minutus and Mus musculus), at least 1 of the parameters characterizing mite assemblages was correlated (positively or negatively) with at least 1 environmental factor. Of them, mite abundance was affected by the environment in 13 hosts, whereas mite species richness and taxonomic diversity were affected by the environment in 7 hosts each (although these two sets of host species were different) (Tables 3 and 4). In addition, as can be seen from Table 4, mite abundance was affected by air temperature (F1; alone or in interaction with other factors) in 11 hosts (positively in 6 cases and negatively in 5 cases), by elevation (F2; alone or in interaction with other factors) in 7 hosts (positively in 4 cases and negatively in 3 cases) and by precipitation (F3; alone or in interaction with other factors) in 7 hosts (positively in 2 cases and negatively in 5 cases) (see an illustrative example with Arvicola amphibius in Fig. 4). Similarly, mite species richness was affected by factor F1 in 4 hosts (positively in all cases), by factor F2 in 3 hosts (positively in 2 cases and negatively in 1 case) and by factor F3 in 6 hosts (positively in 2 cases and negatively in 4 cases) (see an illustrative example with Microtus arvalis in Fig. 5). Taxonomic distinctness of mite assemblages was affected by factor F1 in 4 hosts (positively in 1 case and negatively in 3 cases), by factor F2 in 2 hosts (positively in 1 case and negatively in 1 case) and by factor F3 in 4 hosts (positively in 1 case and negatively in 3 cases). In other words, abundance of mites was affected mainly by air temperature, whereas mite species richness was affected mainly by precipitation (see an illustrative example with Ondatra zibethica in Fig. 6).
Fig. 4. Relationship between total mite abundance and scores of factor F1 across populations of Arvicola amphibus.
Fig. 5. Relationship between species richness of mite assemblages and scores of factor F3 across populations of Microtus arvalis.
Fig. 6. Relationship between taxonomic distinctness of mite assemblages and scores of factor F3 across populations of Ondatra zibethica.
Table 3. The significant (P<0·05) best models explaining variance in abundance (A), species richness (SR) and taxonomic distinctness (Δ+) of gamasid mite assemblages on 15 small mammalian species
(The modelling was carried out using a Generalized Linear Model with the application of Akaike's Information Criterion (AIC) for the best model selection. F1, F2 and F3 are composite variables extracted from 7 environmental variables calculated for each region (see text and Table 2 for explanations).)
Table 4. Parameter estimates for the significant best models (see Table 3) explaining variance in abundance (A), species richness (SR) and taxonomic distinctness (Δ+) of gamasid mite assemblages on 15 small mammalian species
(F1, F2 and F3 are composite variables extracted from 7 environmental variables calculated for each region (see text and Table 2 for explanations). All parameters are significant (P<0·05).)
DISCUSSION
The results of this study support our expectation that abundance and taxonomic diversity of gamasid mite assemblages depend more on host identity than on environmental parameters, although the species richness of mites appeared to be almost equally dependent on host identity and environmental factors. In other words, the abundance and taxonomic diversity of mites can be considered as genuine host species characters with some host species harbouring consistently higher numbers of mites representing more higher taxa than other host species. Furthermore, the species richness of the component communities of gamasid mites may instead represent a local characteristic, with some localities being characterized by higher mite species richness in all host species than other localities.
These results support the idea that part of the parasite community observed on a host is due to its identity, as a direct result of the co-phylogenetic history of hosts and their parasites, whereas another part is due to its specific geographical location (Kennedy and Bush, Reference Kennedy and Bush1994). In the case of gamasid mites, the source of variation associated with host identity must derive from interspecific differences in host biology. For example, mite diversity has been shown to correlate with host body mass (Korallo et al. Reference Korallo, Vinarski, Krasnov, Shenbrot, Mouillot and Poulin2007). However, the direction of this correlation depends on which higher host taxon is considered. Larger rodents harboured less diverse mite assemblages, whereas the opposite was true for soricomorphs. Differences in basal metabolic rate can also play a role. In general, rodent hosts with higher basal metabolic rates harbour more diverse mite assemblages than hosts with lower BMR (Korallo et al. Reference Korallo, Vinarski, Krasnov, Shenbrot, Mouillot and Poulin2007). Both these features, body mass and basal metabolic rate, are rather conservative within-species (Peters, Reference Peters1983; Degen, Reference Degen1997).
The likely source of geographical variation in mite assemblages is the local diversity of the host's biotic and abiotic environment. For example, a richer community of co-habitating hosts increases the probability of lateral transfer of parasites and, thus, affects richness and composition of a parasite assemblage (Caro et al. Reference Caro, Combes and Euzet1997; but see Korallo et al. Reference Korallo, Vinarski, Krasnov, Shenbrot, Mouillot and Poulin2007 for gamasid mites). The abiotic environment external to a host, such as air temperature and precipitation or substrate texture, can also affect parasite species composition (Galaktionov, Reference Galaktionov1996; Krasnov et al. Reference Krasnov, Shenbrot, Medvedev, Khokhlova and Vatschenok1998). Indeed, gamasid mites, both parasitic (e.g. Carrol et al. Reference Carrol, Young and Bruce1992) and free-living (e.g. Sjursen et al. Reference Sjursen, Michelsen and Jonasson2005), are strongly affected by temperature with different species having different temperature preferences (e.g. Avdonin and Striganova, Reference Avdonin and Striganova2005). Another factor that may strongly affect mite abundance is relative humidity (e.g. Mašan and Stanko, Reference Mašan and Stanko2005). Furthermore, humidity tolerance varies among mite species. For example, Ophionyssus galloticolus is more tolerant of low humidity than Ophionyssus natricis (Bannert et al. Reference Bannert, Karaca and Wohltmann2000).
Differential environmental preferences by different species may be a reason behind the inconsistent relationships between various aspects of mite assemblages and environmental variables across host species. Indeed, environmental factors were often correlated with one or more parameters of mite assemblages either positively or negatively, but no distinct prevailing trend could be distinguished. The repeatability of mite species richness and their taxonomic distinctness suggests that, in general, every host species harbours a mite assemblage of a certain composition independently of its geographical locality. If, for example, most mite species in a host-specific assemblage prefer relatively low temperature, then the abundance of mites would decrease with increasing air temperature, whereas the opposite would be true if most mite species in a host-specific assemblage preferred relatively high temperatures. However, no data supporting this explanation are available. This is because environmental preferences for the vast majority of mite species in our data set are unknown.
A comparison of the amount of variance among samples accounted for by differences among host species or regions as opposed to those within hosts and regions suggests that the abundance and taxonomic distinctness of mites is mainly determined by host identity, whereas their species richness depends almost equally on both host identity and geographical locality. A reason for repeatable mite species richness among populations within a host species may be associated with some host constraints on how many mite species it can harbour. For example, there can be a limit to a host's ability to cope with multiple mite species, such that the host manages to maintain mite pressure (expressed as a number of parasite species) at a ‘tolerable’ level (Combes, Reference Combes2001). The repeatability of mite species richness among host species within a region may be due to variability in the external environment that can lead to, for example, extinction of certain mite species in some regions due to unsuitable microclimatic conditions in host burrows.
Taxonomic distinctness of mite assemblages was found to be weakly, albeit significantly, repeatable within a host species, as reported for helminths (Poulin and Mouillot, Reference Poulin and Mouillot2004), but not for fleas (Krasnov et al. Reference Krasnov, Shenbrot, Mouillot, Khokhlova and Poulin2005). This means that whenever a new species is added to a host's mite community, this species is not a random addition from the regional pool of mite species but rather is closely related to the mite species that a host already harbours. As in closely-related free-living species that have similar life-history traits (Brooks and McLennan, Reference Brooks and McLennan1991; Harvey and Pagel, Reference Harvey and Pagel1991), closely-related mite species may also have similar environmental and host preferences. Therefore, the suitability of a host species for a new mite species may be indicated by the occurrence of this mite's close relatives in mite assemblages already on this host. This supports our previous findings of a high similarity in the composition of mite communities within a host species at different, sometimes geographically distant, locations (Vinarski et al. Reference Vinarski, Korallo, Krasnov, Shenbrot and Poulin2007). Another explanation for the repeatability of mite taxonomic distinctness may be related to co-evolution of mites with their hosts. In other words, this pattern can arise as a consequence of a co-evolutionary history dominated mainly by co-speciation. If a mite lineage co-evolved tightly with its host, then it would only spread around the Palaearctic together with the host. Although no study on the co-evolution of dermanyssoid mites and their mammalian hosts has been carried out, strong evidence for such a pattern of co-evolution was found in other parasitic mites (Bochkov and OConnor, Reference Bochkov and OConnor2005). In addition, the contrasting patterns of repeatability of taxonomic distinctness obtained for mite and flea assemblages suggest that the ‘tightness’ of association between a parasite and its host matters when within-host geographical variation of parasite assemblage structure is considered. Fleas spend a considerable time off host and are strongly affected by the off-host environment and, although they are obligate haematophages, their larvae are almost never parasitic (Marshall, Reference Marshall1981). The gamasids considered in this study are either obligate or facultative parasites. However, haematophagy is a characteristic feeding mode not only for the imago but also for nymphal stages of many dermanyssoid mites (Radovsky, Reference Radovsky1969, Reference Radovsky and Kim1985). Moreover, some mite species spend their entire life-cycle on the host body (Zemskaya, Reference Zemskaya1969). The dependence of both imago and pre-imaginal stages on the host can be, at least in part, responsible for the tighter association between mites and hosts than is the case for fleas and hosts and, thus, for the repeatability of taxonomic diversity of mite assemblages within a host species. Nevertheless, although this parameter has not been found to be repeatable within a geographical locality in general, it was affected by environmental factors among different populations of 6 host species (Arvicola amphibius, Microtus agrestis, Myodes glareolus, Ondatra zibethica, Sorex araneus and Sorex caecutiens). In most cases, the taxonomic diversity of mites in these hosts decreased with an increase in winter precipitation and/or air temperature. Most of these hosts occupy habitats that are frequently flooded by melting snow which starts earlier in spring in areas with high air temperatures and which can cause the disappearance of those mite lineages that spend their life both on the host body and in host burrows in areas with high levels of snowfall.
Environmental factors clearly influenced mite assemblages: the abundance of mites was mainly affected by air temperature, whereas the diversity of mites was mainly affected by precipitation. These results call for some explanation. It is possible that the effect of air temperature on mites is mainly direct and manifested in the temperature-dependence of survival and developmental rates in mite individuals (Zemskaya, Reference Zemskaya1973), thus resulting in variation in abundance of any mite independently of its species identity. In contrast, the effect of precipitation on mite communities may be mediated by its effect on habitat heterogeneity, which would indirectly cause variation in the structure of mite communities.
To conclude, the results of this study allow us to understand better the lack of generality in some ecological rules governing parasite communities (Poulin, Reference Poulin2007). A particular type of relationship within each parasite-host association, such as the tightness of association between a parasite and its host's body versus the external environment, seems to be of primary importance for the manifestation of any ecological pattern.
We thank Carl Dick and an anonymous referee for their helpful comments on an earlier version of the manuscript. This study was partly supported by the Israel Science Foundation (Grant no. 249/04 to B.R.K). This is publication no. 577 of the Mitrani Department of Desert Ecology and no. 229 of the Ramon Science Center.
