Discontinuous gas exchange and the significance of respiratory water loss in scarabaeine beetles

SUMMARY Respiratory water loss in insects is a controversial topic. Whilst earlier studies considered respiratory transpiration a significant component of overall water loss, to the extent that it was thought to be responsible not only for the evolution of discontinuous gas exchange cycles (DGCs) but also for variation in DGC patterns, later work repeatedly questioned its importance. In particular, investigations of the proportional contribution of respiratory transpiration to total water loss in species showing DGCs suggested that respiratory transpiration was unlikely to be important in these species. In turn, these studies have been criticized on analytical grounds. In this study we investigated variation in cuticular and respiratory water loss rates in five Scarabaeus dung beetle species, all of which show discontinuous gas exchange cycles, to ascertain the significance of respiratory water loss using modern analytical techniques. In particular, we determined whether there is variation in water loss rates amongst these beetles, whether both respiratory and cuticular water loss rates contribute significantly to variation in the former, and whether metabolic rate variation and variation in the duration of the DGC periods contribute significantly to variation in respiratory water loss rate. Total water loss rate varied such that species from arid areas had the lowest rates of water loss, and both cuticular and spiracular transpiration contributed significantly to variation in overall water loss rate. Moreover, variation in metabolic rate and in the duration of the DGC periods contributed significantly to variation in respiratory water loss rate. By contrast, examination of proportional water loss revealed little other than that it varies between 6.5% and 21%, depending on the species and the temperature at which it was examined. Cuticular water loss scaled as mass0.721, but did not differ from that expected from geometric considerations alone. By contrast, respiratory water loss scaled as mass0.531, suggesting that gas exchange takes place by diffusion and convection. Our results provide direct evidence that respiratory water loss forms a significant component of water balance, and that changes in both metabolic rate and DGC characteristics contribute to modulation of respiratory water loss.

, see also Lighton et al., 1993). Based on comparative data and on laboratory selection experiments, numerous studies have also shown or argued that metabolic rates in species or populations from dry environments are lower than those in species or populations from more mesic environments, thus reducing water loss under the xeric conditions (Juliano, 1986;Hoffmann and Parsons, 1989a;Gibbs et al., 1997;Chown and Gaston, 1999;Davis et al., 2000;Gibbs, 2002a;Gibbs et al., 2003). In addition, largescale comparative studies by Zachariassen and his colleagues (Zachariassen et al., 1987;Zachariassen, 1996), strongly suggest that respiratory transpiration must account for a significant proportion of water lost by arid-environment insects during dehydration, and have led these authors to conclude that modification of this loss could represent a significant fitness benefit.
By contrast, in those studies that have examined the proportional contribution of respiratory transpiration to total water loss, the general conclusion has been that respiratory water loss is low, contributing 3-15% of the total, and with little relationship to the environment (mesic or xeric) occupied by the species in question Lighton et al., 1993;Quinlan and Hadley, 1993;Quinlan and Lighton, 1999). Indeed, Hadley (1994a,b) pointed out that in the majority of species examined, respiratory water loss constitutes such a small proportion of total water loss that it '…is difficult to see how even major changes in its relative contribution would affect the water status of these animals'. The importance of respiratory water loss has been questioned by findings that insects subjected to desiccation stress often abandon discontinuous gas exchange Quinlan and Hadley, 1993;Chappell and Rogowitz, 2000;Rourke, 2000), and that some species from xeric areas apparently do not exhibit this gas exchange pattern at all (Lighton and Berrigan, 1995;Lighton, 1996). Moreover, several studies have shown that neither a change in gas exchange pattern nor a reduction in metabolic rate are accompanied by a reduction in water loss (Djawdan et al., 1997;Bradley et al., 1999;Rourke, 2000;Shelton and Appel, 2001a,b), and that even substantial changes in metabolic rate might not markedly improve survival time via water conservation (Bosch et al., 2000).
Nevertheless, the view that respiratory water loss forms an important component of total transpiration continues to permeate the modern literature. For example, in revisiting the comparative analysis of water loss undertaken by Zachariassen and his colleagues, Addo-Bediako et al. (2001) concluded that respiratory water loss is of considerable importance in xeric insect species, although they did not find support for reduced metabolic rate as a means by which a reduction in transpiration might be effected. Similarly, in a comparative analysis of water balance in Drosophila species from mesic and xeric habitats, Gibbs et al. (2003) concluded that lower rates of water loss in the xeric species are achieved primarily by reduction in respiratory losses associated with a reduction in metabolic rate and activity levels, and improved spiracular control. Based on investigations of discontinuous gas exchange in beetles, Duncan and Dickman (2001) and Duncan et al. (2002a) supported the water conservation hypothesis for the DGC because the other hypotheses seemed not to apply to the species they investigated (see also Vogt and Appel, 2000). Indeed, Duncan (2003) went on to conclude that every facet of the DGC can be altered (presumably in an adaptive fashion) to effect respiratory water savings. Thus, it is clear that at present there is little consensus regarding the significance of respiratory water loss in insect water balance. Whilst many arguments have been raised against its likely significance, it continues to engender support. This state of affairs was perhaps best summed up by Quinlan and Lighton (1999) who argued that the '…interrelationships between gas exchange and water balance are still largely a matter of conjecture'.
Whilst reviewing respiratory water loss in insects, Chown (2002) proposed several ways in which the current polarization of findings might be overcome. Amongst these, the most significant included a move away from the use of proportions for expressing respiratory water loss (see also Boardman, 1988, 1999), the more frequent use of comparative analyses that involve work on species in which respiratory and cuticular transpiration can be clearly distinguished, and clear a priori statements of the null expectation. In the latter case, arguments regarding the proportional contribution of respiratory loss to total transpiration, the relationship between metabolic rate and respiratory water loss rate, and the contribution of variation in components of the DGC to water conservation can be restated as: (1) variation in respiratory water loss is unrelated to variation in total water loss, (2) variation in metabolic rate is unrelated to variation in respiratory water loss rate and (3) there is no covariation in DGC period characteristics and respiratory water loss. In general, the alternatives to these null expectations are straightforward, although in the case of DGC period characteristics there are two, non-exclusive alternative hypotheses. These are that covariation between C-and/or Fperiod duration and water loss rate is negative (Lighton, 1990;Davis et al., 1999), and/or that covariation between O-period duration and water loss rate is positive (Lighton, 1990;Duncan et al., 2002a).
In this study, we examine these three major hypotheses in an analysis of water loss in five species of Scarabaeus dung beetles that exhibit discontinuous gas exchange (Davis et al., 1999). We determine whether there is variation in total water loss rate across the individuals in our sample, whether respiratory water loss contributes significantly to this variation, whether metabolic rate and water loss rate covary, and whether DGC period durations covary with water loss rate. In doing so our aim is not only to investigate patterns of water loss in these species, but also to provide an example of the type of analysis Respiratory water loss in dung beetles that could be useful for overcoming the stalemate that currently reigns in the field of respiratory water loss in insects.

Materials and methods
Study material The five, diurnally active species of Scarabaeus (Coleoptera, Scarabaeidae, Scarabaeinae) comprised two flightless species, with fused elytra, that were collected on the arid west coast of South Africa, and three volant species that were collected from more mesic eastern habitats in the country (Table·1). Both of the west coastal species were collected in shrubland on sandveld near Kleinzee. Scarabaeus gariepinus was collected on very coarse-grained dune sands and shows a distribution northwards into the Namib Desert (Davis et al., 2000), whereas Scarabaeus striatus was collected closer to the coast on less coarse-grained sands and shows a distribution centred in Namaqualand, north-west South Africa. Both species feed on dry dung pellets that are rehydrated after burial (Scholtz, 1989). Of the three north-easterly species, S. galenus was collected in open woodland on sandy soils in the hot dry lowlands near Crocodile Bridge in the Kruger National Park, S. rusticus was collected at moderate altitude in grassland on sandy loam from Rustenburg Nature Reserve, which is situated on the cooler rocky hillsides of the Magaliesberg, and S. westwoodi was collected from grassland on finer-grained soils at high altitude along the Sani Pass on the cool, wet lower slopes of the Drakensberg. For further details of geographical distribution, climatic influences and diet of the species, see Davis et al. (1999Davis et al. ( , 2000.

Experimental design
Individuals were held at 25°C under a 12·h:12·h L:D photoperiod for at least 5 days before respirometry commenced. Individual beetles were starved for 24-48·h before a trial to avoid excretion during the experiment. Each individual was weighed prior to and after a trial using a Sartorius R 200 D digital balance (Epsom, UK), and mean mass was used in data analyses. A single beetle was placed in the cuvette during the morning (±10:00·h) to allow it to settle before data measurement over the following night, when diurnal Scarabaeini are relatively inactive (Davis, 1996(Davis, , 2002. The cuvette was placed inside a water jacket connected to a Grant LTD20 water bath (Cambridge, UK) that controlled the selected temperature to within ±0.2°C. The laboratory was maintained in darkness throughout each trial.
Flow-through respirometry based on a Sable Systems (Henderson, Nevada, USA) turnkey system was used to measure both CO2 and H2O production (fully described in Davis et al., 1999Davis et al., , 2000. Air was passed through columns of Drierite and soda lime to remove both water vapour and CO2. The scrubbed air was then passed through an automated baselining system, the 340·ml cuvette containing the beetle, and a Li-Cor CO2/H2O Analyzer Model Li 6262 at a flow rate of 150·ml·min -1 . Sable Systems DATACAN V software was used for data capture and analysis, and all measurements were corrected to standard temperature and pressure. In each of the five Scarabaeus species, temperaturemodulated changes in the amounts of both CO2 and H2O production were recorded at 4°C intervals across a temperature range from 16°C to 32°C (with the exception of S. striatussee Table·2). Whereas 10 or more individuals were measured for S. westwoodi and S. rusticus, fewer individuals were available for the other three species (Table·2). For each individual, measurements of CO2 and H2O release were made at 5·s intervals during an overnight period of 9-10·h commencing at dusk (18:00-19:00·h). Periodic observations during initial respirometry work revealed that DGCs were indicative of immobile individuals (Davis et al., 1999), and therefore provided standardized data for intra-and interspecific comparisons. Where possible, DGC measurements were only entered into the analysis for data recorded after the first 4·h, and for each DGC, measurements of both CO2 and H2O were made over the same range of samples. Because the DGC periods were less readily discerned in the water vapour trace than in the CO2 recording, the water vapour data were divided into O-period and combined Closed/Flutter (CF) period. Cperiod emission rate of CO2 was very close to the baseline (see also Davis et al., 1999), suggesting that there is minimal escape of gasses during the C-period, indicating that CO2 is lost . H∑O for the O-period (cuticular plus respiratory water loss rate), respiratory water loss rate (O-period only), total water loss rate (V . H∑O), volume of cuticular water loss, volume of respiratory water loss, total water loss volume, and proportional contribution of respiratory water loss were calculated. The data for each individual were generally derived from the mean data for at least four DGCs (though in some instances fewer DGCs were used), and in all subsequent analyses these mean values for each individual were used as the primary data. Although some authors use values from each cycle in a discontinuous gas exchange of a single individual as primary, independent data points (e.g. Duncan et al., 2002a,b), we consider this pseudoreplication and did not do so.

Data analysis
Generalized linear modelling (see McCullagh and Nelder, 1989;Quinn and Keough, 2002) was used to obtain best-fit models for total V . H∑O, respiratory V . H∑O and cuticular V . H∑O. The independent variables included in these models were: (1) treatment temperature and species identity for total V . H∑O, to determine if there is variation in total water loss rate amongst species; (2) treatment temperature, log10(mass), log10(cuticular water loss rate), and log10(spiracular water loss rate) for total V . H∑O, (3) treatment temperature, log10(mass), log10(CF-period duration), log10(O-period duration), and log10(mean V . CO∑) for respiratory V . H∑O; (4) treatment temperature, and log10(mass) for log10(cuticular V . H∑O).
In all cases a normal distribution and identity link function were specified, and the Akaike Information Criterion (AIC) was used to identify the best subset of explanatory variables. In cases where the AIC was similar for different subsets of variables, the model with the fewest variables was chosen (Quinn and Keough, 2002). These models were run again and Type III likelihood tests were used to confirm the significance of each variable. Because Davis et al. (2000) found that wing status (flying vs. flightless) explained much of the variation in V . CO∑ in these Scarabaeus species (see also Reinhold, 1999;Addo-Bediako et al., 2002), and because flightlessness is reputedly a means by which beetles are able to effect a water savings (Draney, 1993;Chown et al., 1998), the latter (three) analyses were repeated including a dummy variable for wing status.

Results
In the case of total water loss rate, treatment temperature (χ 2 (1)=153.5, P<0.0001) and species identity (χ 2 (4)=285.4, P<0.0001) contributed significantly to the model, verifying that there is indeed considerable variation amongst species in the total rate of water loss (Fig.·1). Mean rates of water loss for each species covaried significantly with the mean annual rainfall of the area in which each species was collected (r s=0.9, P=0.037). Summary statistics for the DGC and water loss characteristics also revealed substantial variation amongst species in these parameters (Table·2). The best-fit model for total water loss rate included only cuticular water loss rate and spiracular water loss rate (full model AIC=441.4, with just these terms AIC=438.1) (Table·3). Therefore, both cuticular water loss and spiracular water loss contribute to variation in total water loss rate (each of the latter are affected by temperature, which has an effect on cuticular transpiration and DGC frequency, but temperature tends not to enter the models proportional contribution of spiracular to total water loss (%), adjusted for body mass, at 20°C in the five Scarabaeus species investigated in this study. Variation in rate of water loss is in the direction expected from the mean annual rainfall of the area from which the species was collected. Values are means ± S.E.M. Gari, S. gariepinus; Stri, S. striatus; Gale, S. galenus; Rust, S. rusticus; West, S. westwoodi. significantly when forced in initially using a Type I approach -data not shown). When wing status was incorporated, the best-fit model included wing status, cuticular water loss rate and spiracular water loss rate, with a lower AIC (424.12) than the model excluding wing status. However, Quinn and Keough (2002) have pointed out that the AIC should be interpreted with caution when both categorical and continuous predictors are included in the model. In this context the reduction in the deviance/d.f. value (Table·3) compared with the model not including wing status suggests that there is little difference between the two models.
The best-fit model for spiracular water loss rate included CF-period duration, O-period duration, and V . CO∑ (AIC=-187.3, compared to -184.8 for the full model) (Table·4). When wing status was considered, the best-fit model included CF-period duration, O-period duration, V . CO∑ and wing status (AIC=-189.2). However, CF-period duration was not significant (χ 2 =2.995, P=0.08). Therefore, the best fit model included wing status, O-period duration and V . CO∑ (Table·4). One explanation for the exclusion of CF-period duration from the model including wing status is the particularly prolonged duration of the CF period in S. gariepinus, and to a lesser extent in S. striatus (see Table·2).
In the case of cuticular water loss, mass and treatment temperature contributed significantly to the model, as did wing status (Table·5). To obtain an indication of the scaling of cuticular and respiratory water loss rate, the relationship between log10(mass) and log10(cuticular water loss), and log10(mass) and log10(spiracular water loss) was investigated for measurements made at 20°C (for which most data were available) using generalized linear models. Cuticular water loss scaled significantly (P=0.002) as mass 0.721±0.234 , which is not significantly different from a value expected from geometric considerations alone (mass 0.667 , t(37)=0.231, P>0.5). Spiracular water loss scaled significantly (P=0.0009) as mass 0.531±0.160 , which is also not significantly different from a value expected from geometric considerations alone (mass 0.667 , t(37)=-0.85, P>0.4). By contrast, V . CO∑ scaled significantly (P=0.0001) as mass 1.284±0.160 .

Discussion
Although there was considerable variation among the Scarabaeus species in total water loss rate and the components thereof, in a direction that might be expected based on the habitats they occupy (see also Davis et al., 2000Davis et al., , 2002Davis, 2002), the variation in the proportional contribution of respiratory water loss to the total was less clear (Fig.·1). Indeed, the proportional contributions were all within the range of (or close to in the case of S. gariepinus), the low values reported previously in the literature, e.g. 15% for the grasshopper Melanoplus sanguinipes (Rourke, 2000), 2% for the ant Camponotus vicinus (Lighton, 1992), and 5% for the dung beetle Aphodius fossor (Chown and Holter, 2000). Moreover, unlike total water loss rate, there was no relationship between rainfall of the site at which the species S. L. Chown and A. L. V. Davis were collected and proportional contribution of respiratory water loss to the total (rs=-0.7, P=0.18). These findings might easily have resulted in a conclusion similar to that reached by Hadley (1994a,b), and many other recent workers (see Introduction), i.e. that respiratory water loss is largely unimportant in insects. However, based on a simple theoretical argument, Chown (2002) argued that analyses of the proportional contribution of respiratory transpiration to water loss are more likely to obscure than to clarify investigations of respiratory water loss. In particular, he argued that if selection were to reduce cuticular and respiratory transpiration in concert, then changes in the proportional respiratory transpiration would not be detected at all. Likewise, depending on other life history requirements, such as those associated with flightless as opposed to volant species (Zera and Denno, 1997;Reinhold, 1999), modulation of respiratory and cuticular water loss might occur independently. Therefore, instead of concluding that respiratory water loss is probably insignificant in these species, we chose to further explore the contributions of cuticular and respiratory transpiration to total water loss. Perhaps unsurprisingly, both cuticular and respiratory water loss contributed significantly to total water loss rate, in keeping with much of the comparative literature on water balance (for reviews, see Edney, 1977;Wharton, 1985;Hadley, 1994b;Addo-Bediako et al., 2001). In other words, the first of our null hypotheses was falsified. In these beetles, variation in respiratory water loss was related to variation in total water loss.

Table·3. Generalized linear model of the explanatory variables on total water loss rate
Variation in spiracular water loss rate was best explained by a combination of CF-period duration, O-period duration and V . CO∑. The estimates for these parameters indicated that decreasing rates of water loss are associated with an increase in the duration of the CF-period, a decline in the duration of the O-period, and a decrease in metabolic rate (Table·4). These results provide strong support for the hypothesis that alteration of metabolic rate at rest can be used to effect a change in water loss, as has previously been suggested by several authors (Barnhart and McMahon, 1987;Zachariassen et al., 1987Zachariassen et al., , 1988Lighton and Bartholomew, 1988;Hoffmann and Parsons, 1989b;Chown and Gaston, 1999). Therefore, the second of our null hypotheses can be rejected.
In the context of the relationship between metabolic rate and total water loss rate, it is important to note that total rates of water loss measured here differed from gravimetric estimates made for two of the species for which data are available (see Klok, 1994). This suggests that comparisons between studies using short-term water loss rate measurements and those undertaken over longer-term periods (e.g. Zachariassen et al., 1987Zachariassen et al., , 1988Zachariassen, 1996;Addo-Bediako et al., 2001) must be made with caution. Nonetheless, in the context of these kinds of comparative analysis of water loss and metabolic rate, it is important to note that in these Scarabaeus species, the relationship between the residuals of each of the log(rate)-log(mass) relationships (see Addo-Bediako et al., 2001) was positive (r 2 =0.97, P=0.014). This finding provides evidence that the relationship between log(metabolic rate) and log(water loss rate) was not simply a function of the covariation of both variables with body mass. The same result was obtained in a multiple regression with water loss rate and both mass and metabolic rate as independent variables, as recommended by Freckleton (2002). That is, metabolic rate was retained as a significant term (partial correlation t=4.3, P<0.05). Indeed, although larger body size has frequently been identified as an important means by which insects in general (Schoener and Janzen, 1968;Remmert, 1981;Lighton et al., 1994;Le Lagadec et al., 1998; but see also Gibbs and Matzkin, 2001;Chown and Klok, 2003;Gibbs et al., 2003), and dung beetles in particular (Chown et al., 1995), might alter their responses to environmental water availability, it did not enter most of our models as a significant term. Clearly, both cuticular and spiracular water loss rates scale with body mass, but, especially in the latter case, body mass is much less important in explaining variation than are other factors. Moreover, it should be kept in mind that body size can affect desiccation resistance not only via variation in water loss rates, but also via variation in water content Chown et al., 1995).
The inclusion of CF-period and O-period durations in the best-fit model for spiracular water loss also suggests that modifications in the pattern of gas exchange might be important for altering water loss rates. That is, our third hypothesis has also been rejected. Whilst it has long been suspected that there is covariation between gas exchange patterns and water loss rates (Lighton, 1988a(Lighton, , 1990(Lighton, , 1991Lighton et al., 1993;Davis et al., 1999;Bosch et al., 2000;Duncan and Dickman, 2001;Duncan et al., 2002a), few studies have examined the relationships between DGC characteristics and spiracular water loss (for exceptions, see Lighton, 1992;Lighton et al., 1993;Quinlan and Lighton, 1999). For the most part, inferences concerning the importance of modulation of either CF-period or O-period duration for altering spiracular water loss are based on measurements of V . CO∑ only (e.g. Bosch et al., 2000;Duncan, 2003), and then rarely involving investigations of more than one or two species (though for exceptions, see Lighton, 1991;Davis et al., 1999;Duncan and Byrne, 2000). By contrast, our results provide explicit support for the idea that modulation of DGC characteristics and metabolic rate can be used to alter water loss rate. Moreover, these changes are in a direction that is consistent with a response to changes in environmental water availability. That is, species from more arid areas have lower metabolic rates, shorter O-periods and longer CF-periods. Whilst we cannot conclude that these changes are adaptive, mostly because the two wingless species are more closely related to each other than they are to any of the other species (for further discussion, see Davis et al., 2000), our results suggest that this is likely to be the case. In other words, the data provide support for the proposition that by reducing the period for which spiracles remain open, and by prolonging the duration of the closed and flutter periods, species showing DGC can reduce respiratory water loss. Although the contribution of the F-period to water savings hinges on whether gas exchange takes place predominantly by convection or diffusion (Lighton, 1988b;Lighton and Garrigan, 1995;Lighton, 1996), it seems likely that there would be substantial convective airflow in the Fperiod in these species, because this has been found in other dung beetle species (Chown and Holter, 2000;Duncan and Byrne, 2000). Nonetheless, the exclusion of CF-period duration from the model including wing status suggests that alteration of CF-period duration might only be a response to very arid conditions, rather than one found in all species, whereas all species might modulate water loss via changes in O-period duration. Prolonged CF-period durations in other beetle species from extremely arid areas (Bosch et al., 2000;Duncan and Byrne, 2000;Duncan et al., 2002a) provide support for this contention. Although body mass was excluded from the model used to explain respiratory water loss, we undertook two additional analyses to investigate the scaling of both cuticular and spiracular water loss, because scaling of these variables has long been of interest in water balance physiology (Kestler, 1985;Nagy and Peterson, 1988;Lighton and Feener Jr, 1989;Lighton et al., 1994;Lehmann et al., 2000). Although several estimates of the scaling of overall water loss have been made, ranging from 0.484 to 0.943 (Nagy and Peterson, 1988;Zachariassen et al., 1988;Chown, 1993;Lighton et al., 1994), consensus scaling equations for respiratory and cuticular water loss in insects have not been derived (Edney, 1977;Arlian and Veselica, 1979;Peters, 1983;Wharton, 1985;Hadley, 1994b). In part, this must be a consequence of the difficulty of distinguishing respiratory and cuticular water loss, and the problems of catabolism when water loss is measured gravimetrically (for a discussion, see Nicolson, 1980;Edney, 1982;Addo-Bediako et al., 2001). By contrast, the measurements of respiratory and cuticular transpiration made here are not confounded by these problems. The scaling exponent of 0.721 for cuticular water loss did not differ from an expectation of 0.67 based on geometric considerations alone (as is the case for overall water flux in the Drosophila species examined by Lehmann et al., 2000). This finding does not provide support for Kestler's hypothesis (Kestler, 1985) that cuticular water loss should scale as mass 0.33 . Kestler (1985) argued that because cuticular thickness scales as mass 0.33 , cuticular water loss should scale similarly. The fact that our findings and those of Lehmann et al. (2000) do not support this hypothesis is perhaps not unexpected given that it is not only cuticular thickness that determines water loss rates (Gibbs et al., 1991;Gibbs, 1998Gibbs, , 2002bRourke, 2000). Spiracular water loss had a somewhat lower scaling exponent (0.531), although it did not differ from 0.67 either. According to Kestler (1985), in a purely diffusion-based system, respiratory water loss should scale as mass 0.33 , whereas in a convection based system it should scale as mass 1.0 . The intermediate value obtained here suggests that both diffusive and convective water loss take place in these species.
In conclusion, we have rejected all three of our null hypotheses, thereby providing considerable support for previous contentions that modulation of respiratory water loss is important for water balance in insects. Moreover, our results also provide direct, comparative evidence that in species with discontinuous gas exchange cycles, alterations in both metabolic rate and gas exchange pattern contribute to changes in respiratory water loss. Therefore, our work not only provides direct evidence for several theoretically appealing but empirically poorly supported ideas, but also joins a growing body of evidence (e.g. Lehmann, 2001;Gibbs et al., 2003) demonstrating that respiratory water loss cannot be discounted in investigations of insect water balance. In the context of discontinuous gas exchange cycles it also suggests that hypotheses for the origin and maintenance of these cycles, which are predicated on water savings (the hygric and chthonic genesis hypotheses; see Lighton and Berrigan, 1995;Lighton, 1996Lighton, , 1998, should not be discarded just yet.