Scaling of Metabolic Rate in Teleost Fish PDF

Journal of Animal Ecology 1999, Scaling of metabolic rate with body mass and 68, 893±905 temperature in teleost ®sh ANDREW CLARKE and NADINE M. JOHNSTON British Antarctic Survey, High Cross, Madingley Road, Cambridge CB3 0ET, UK...

Journal of Animal Ecology 1999, Scaling of metabolic rate with body mass and 68, 893±905 temperature in teleost ®sh ANDREW CLARKE and NADINE M. JOHNSTON British Antarctic Survey, High Cross, Madingley Road, Cambridge CB3 0ET, UK Summary 1. We examined published studies relating resting oxygen consumption to body mass and temperature in post-larval teleost ®sh. The resulting database comprised 138 studies of 69 species (representing 28 families and 12 orders) living over a tem- perature range of c. 40  C. 2. Resting metabolic rate (Rb; mmol oxygen gas h±1) was related to body mass (M; wet mass, g) by Rb = aMb, where a is a constant and b the scaling exponent. The model was ®tted by least squares linear regression after logarithmic transformation of both variables. The mean value of scaling exponent, b, for the 69 individual spe- cies was 0´79 (SE 0´11). The general equation for all teleost ®sh was 1nRb = 0´80(1nM) ± 5´43. 3. The relationship between resting oxygen consumption and environmental tem- perature for a 50-g ®sh was curvilinear. A typical tropical ®sh at 30 C requires approximately six times as much oxygen for resting metabolism as does a polar ®sh at 0 C. This relationship could be ®tted by several statistical models, of which the Arrhenius model is probably the most appropriate. The Arrhenius model for the resting metabolism of 69 species of teleost ®sh, corrected to a standard body mass of 50 g, was 1nRb = 15´7 ± 5´02.T±1, where T is absolute temperature (103  K). 4. The Arrhenius model ®tted to all 69 species exhibited a lower thermal sensitivity of resting metabolism (mean Q10 = 1´83 over the range 0±30  C) than typical within-species acclimation studies (median Q10 = 2´40, n = 14). This suggests that evolutionary adaptation has reduced the overall thermal sensitivity of resting meta- bolism across species. Analysis of covariance indicated that the relationships between resting metabolic rate and temperature for various taxa (orders) showed similar slopes but signi®cantly di€erent mean rates. 5. Analysis of the data for perciform ®sh provided no support for metabolic cold adaptation (the hypothesis that polar ®sh show a resting metabolic rate higher than predicted from the overall rate/temperature relationship established for temperate and tropical species). 6. Taxonomic variation in mean resting metabolic rate showed no relationship to phylogeny, although the robustness of this conclusion is constrained by our limited knowledge of ®sh evolutionary history. Key-words: body mass, ®sh, metabolic cold adaptation, respiration, scaling, tem- perature. Journal of Animal Ecology (1999) 68, 893±905 Introduction order from variety. One of the most powerful descriptions has proved to be the striking manner in Faced with the complexity of nature, ecologists have which many aspects of physiology and ecology scale long sought broad patterns as a means of drawing with body size, and a huge body of information has been accumulated on this topic. Most of these data have, however, been concerned with the terrestrial Correspondence: A. Clarke, Marine Life Sciences environment. Despite their being by far the most Division, British Antarctic Survey, High Cross, Madingley # 1999 British Road, Cambridge CB3 0ET, UK. Tel: 01223 221591; Fax species-rich chordate group (Nelson 1994), relatively Ecological Society 01223 221259; e-mail: [email protected] little is known of scaling in ®sh and in his review of 894 scaling relationships in nature, Schmidt-Nielsen determine any taxonomic variability in this relation- Scaling of (1984, p. 72) bemoaned the lack of any general rela- ship. metabolic rate in tionship between metabolic rate and body size in 2. Derive an overall relationship between resting teleost ®sh ®sh. Neither of the detailed compilations by Peters metabolic rate and temperature for teleost ®sh, and (1983) and Calder (1984) were able to provide any to determine any taxonomic variability in this rela- general scaling relationships for metabolic rate in tionship. ®sh. 3. Determine whether this relationship revealed The metabolic rate of all ectothermic organisms is any evidence for metabolic cold adaptation (sensu strongly dependent on temperature, as well as body Krogh). size. This relationship tends, however, to be clearer in aquatic organisms, where the large thermal mass Materials and methods of water bu€ers the rate of change of environmental temperature experienced by organisms. The evolu- DATA COMPILATION tionary signal is thus easier to detect against the There are a great many reports of teleost metabolic noise induced by rapid environmental thermal varia- rate in the scienti®c literature, but not all were suita- bility. Although relationships between metabolic ble for inclusion in this study. To obtain a master rate and environmental temperature have long been list of studies we used a combination of published established for aquatic (primarily marine) inverte- reviews and searches of electronic databases. The brates (for example, Ivleva 1977, 1980; Ikeda 1985), original publication for each study was then exam- there have been few summary relationships provided ined and accepted for our analysis only if it matched for ®sh. Even recent text-books and reviews of tele- all of a number of criteria. ost physiology have based their discussion of meta- 1. All ®sh were post-larval. bolic rate and temperature largely on the general 2. The experimental protocol was unlikely to have picture established over three decades ago by in¯uenced the measured respiration rate (for exam- Scholander and colleagues (Scholander et al. 1953). ple, a sucient period was allowed for the ®sh to The relationship between metabolic rate and tem- acclimatize to the experimental chamber). perature in teleosts is of particular interest in rela- 3. All experimental ®sh were unfed or post-absorb- tion to the polar environment. A now classic tive, to avoid the e€ects of feeding on metabolic rate experiment by Ege & Krogh (1914) led to the sug- (Jobling 1981). gestion that the resting metabolic rates of polar ®sh 4. Measurements were of resting metabolism. Often would be elevated relative to the rates predicted by referred to as standard metabolism, this is the the extrapolation to polar temperatures of the rest- respiration rate of an unfed ®sh resting quietly in ing metabolic rate of temperate water species the experimental chamber and may be taken as the (Krogh 1914, 1916). This concept of metabolic cold best experimental approximation to true basal or adaptation was enormously in¯uential and, although maintenance metabolic rate (Clarke 1991, 1993). criticized on both experimental (Holeton 1973, 5. Experimental temperature re¯ected that experi- enced by the ®sh in its natural habitat, was well con- 1974) and theoretical grounds (Clarke 1980, 1983, trolled and reported. 1991), it has yet to disappear from the literature. 6. Measurements were made over a sucient range There have, however, been almost no comparative of body mass to allow a precise estimation of the studies of ®sh metabolism across the ecological tem- scaling parameters within that species. perature spectrum (polar to tropical) since the pio- These criteria meant that only a small subset of neering work of Scholander et al. (1953). A recent published studies were included in our analysis. exception was a detailed comparative study of the These studies were compiled into three separate metabolic rate of six species of sedentary marine ®sh databases. The largest data set (FULL) comprised from tropical, temperate and polar habitats, which 138 studies covering 12 orders, 28 families and 69 demonstrated a curvilinear relationship between species living over a temperature range of 1 40 C. resting metabolic rate and temperature (Johnston, Many species were, however, represented by studies Clarke & Ward 1991a). This relationship was similar at more than one temperature and sometimes by to that previously found for marine invertebrates more than one study at that temperature. This leads and it provided no support for the concept of meta- to statistical problems in that species represented by bolic cold adaptation. more than one study are unduly weighted in the As an adjunct to this detailed comparative experi- analysis and can thereby bias the overall relation- mental study we have undertaken an analysis of lit- ship. A smaller data set (ONESTUDY) was there- # 1999 British erature data on respiration rate in teleost ®sh. The fore constructed with each species represented by a Ecological Society aims of this study were to: single study. The criteria for selection were, ®rst, an Journal of Animal Ecology, 1. Derive an overall scaling relationship between experimental temperature most representative of 68, 893±905 metabolic rate and body mass in teleost ®sh, and to that experienced in the wild, and secondly, the 895 widest range of body size or the largest number of this non-linearity of Q10 is small over the normal A. Clarke & individuals. physiological temperature range (see discussion in N.M. Johnston The third database (POLAR) included all studies Clarke 1983). of polar ®sh, including those which did not satisfy All statistical analyses were carried out using the the criteria for a range of body sizes. This database standard PC-based packages MINITAB (version 10´1; thus included some studies excluded from the FULL Minitab Inc, Pennsylvania) and GENSTAT 5 (Payne and ONESTUDY databases, but was used only for et al. 1993). Where logarithmic transformations were an examination of speci®c questions concerned with applied, natural (Naperian) logarithms were used, metabolic cold adaptation. with relevant corrections where original studies used log10 transformation. It is frequently commented that in studies of scaling, least-squares (model I), STATISTICAL ANALYSIS regression is inappropriate and some form of model II regression is needed (see, for example, Ricker All literature data for metabolic rate were expressed 1973; Laws & Archie 1981); full discussions of this in terms of oxygen uptake, but this was reported in point are those by LaBarbera (1989) and Harvey & a wide range of units and often expressed in mass- Pagel (1991). In the study reported here the indepen- speci®c terms. For this analysis all oxygen data were dent variables (body mass, temperature) were converted to absolute (per individual) rates and usually measured with relatively small percentage molar units (mmol oxygen gas h±1). Conversions error and plots of the residuals (not shown here) from volumetric units were made on the assumption indicated that the error distributions were indepen- of STP (1 mol oxygen gas occupies 22´4 L). For dent of body mass or temperature. Model I regres- incorporation into energy budgets metabolic rates sion has the bene®t of providing residuals that are should strictly be expressed in units of power. A uncorrelated with the independent variable (in this general conversion may be made on the basis of a study, body mass or temperature), whereas model II mean oxycalori®c coecient (1 mmol oxygen gas regression does not (Harvey & Pagel 1991). Model I equates to the utilization of 434 J), but this involves regression was therefore used in this study. an assumption of the substrates being oxidized. Furthermore, most metabolic scaling data are reported in units of oxygen, for this is how they are measured. For these reasons we have not converted literature oxygen utilization data to estimated power Results consumption. The scaling equation relating resting oxygen con- SCALING OF RESTING METABOLIC RATE sumption (Rb, mmol h±1) to wet body mass (M, g) is WITH BODY MASS Rb = aMb A summary plot of the scaling relationships where a is a constant and b the scaling exponent obtained for individual species (Fig. 1a) suggested (Peters 1983; Calder 1984; Schmidt-Nielsen 1984). that, although species di€ered markedly in their The scaling exponent was estimated by the slope of metabolic rates, this variation was underlain by a a least squares linear regression following logarith- general overall scaling relationship between resting mic transformation of both variables. metabolic rate and body mass. The frequency histo- The Arrhenius relationship used to model the grams of the scaling parameter b for individual stu- relationship between absolute temperature, T, and dies showed a Gaussian distribution (Fig. 1b), with is: mean values of 0´791 (SE = 0´011, n = 138) for the FULL data set and 0´793 (SE = 0´011, n = 69) for Rb = A exp ( ± m/RT) the ONESTUDY data set. where A is a constant, m the Arrhenius constant and Although 110 studies (80% of the FULL data set) R the universal gas constant. Thus, a plot of 1nRb reported scaling exponents in the range 0´65±0´95, against T±1 yields a straight line of slope ±m/R. An individual values ranged from 0´40 to 1´29. Analysis alternative measure of temperature sensitivity is Q10. of variance indicated a statistically signi®cant varia- In this study, Q10 was estimated over the range 0± tion between di€erent families (F = 3´50, P < 0´001) 30 C. The ®tted Arrhenius model was used to esti- and orders (F = 3´48, P < 0´001). At the level of mate the resting metabolic rate of a 50-g ®sh at 0 C family these di€erences were caused mainly by low [Rb(0)] and 30 C [Rb(30)]. Q10 was then calculated values of the scaling exponent for Bathydraconidae from (n = 1) and Gobiidae (n = 1), and high values for # 1999 British Myctophidae (n = 3) and Ictaluridae (n = 1). At the Ecological Society [Rb(30)/Rb(0)]10/30 level of order, the di€erences were caused mainly by Journal of Animal Ecology, A system exhibiting Arrhenius behaviour has a high mean values for Myctophiformes (n = 3) and 68, 893±905 Q10 which varies with temperature, but the e€ect of Salmoniformes (n = 13). 896 Scaling of metabolic rate in teleost ®sh Fig. 1. (a) Relationship between resting oxygen consumption (mmol h±1) and body mass (g) for teleost ®sh. Each line repre- sents the scaling relationship determined for a particular species, plotted between the maximum and minimum body masses used in that study. Data from ONESTUDY data set, so each species is represented only once. A small number of studies have been omitted for clarity. (b) Frequency histogram of literature values of the within-species scaling exponent, b, for tele- ost ®sh. FULL data set (n = 138 studies, 69 species). Despite this natural variability, the overall rela- SCALING OF RESTING METABOLIC RATE tionships shown in Fig. 1(a) and Fig. 1(b) suggest WITH TEMPERATURE that a log/log model with a scaling exponent of 0´79 is a suitable statistical model to provide a ®rst-order Before examining the relationship between resting description of the overall relationship between rest- metabolic rate and temperature it is necessary ®rst to remove the e€ect of body mass. Two widely used ing metabolic rate and body mass in teleost ®sh. techniques are either to analyse the residuals around Such statistical models, whilst undoubtedly obscur- ing some important biological variability, are of great heuristic value in drawing broad ecological conclusions. The overall relationship was 1nRb = 0´80(1n M) ± 5´43. Scaling exponents determined within species may, however, di€er from those between species, for both evolutionary and statistical reasons (see discussion by Harvey & Pagel 1991). An estimate of the between-species scaling exponent for teleost ®sh was derived by calculating the mean resting metabolic rate of a ®sh of median size for each experimental species and ®tting an overall scaling relationship to logarithmically transformed data (Fig. 2). The slope of the model I regression line was 0´801 (SE = 0´062, n = 69; data from the ONESTUDY data set). The between species and the mean within- species mass exponents in teleost ®sh are thus e€ec- tively identical. Although there is clearly a strong scaling relation- ship between resting metabolic rate and body mass # 1999 British in teleost ®sh, the data show considerable scatter Fig. 2. Between-species scaling relationship between resting Ecological Society oxygen consumption (Rb, mmol h±1) and wet body mass about the regression line (Fig. 2). An important fac- Journal of Animal (M, g) in teleost ®sh. Each species is represented by a single Ecology, tor in explaining this remaining variance is environ- data point (ONESTUDY data set). The overall model 1 68, 893±905 mental temperature. regression is 1nRb = 0´80(1nM) ± 5´43. 897 the metabolic rate/body mass relationship (Bennett ably more oxygen for resting metabolism than does A. Clarke & 1987) or to correct data for each species to a given a polar ®sh at 0 C. N.M. Johnston size. These techniques produce similar qualitative Ecologists typically prefer to work with linear results; the data presented here are for metabolic rather than curvilinear relationships, and there are rate corrected to a standard size of 50 g (close to the several transformations that are e€ective at lineariz- median size of all ®sh in the study, which was 47 g). ing the relationship between resting metabolic rate This correction utilized the species-speci®c mass and temperature. These include log/linear (Fig. 3b), exponent determined in the original study. The qua- log/log (Fig. 3c) and Arrhenius models (Fig. 3d). All litative patterns were the same if the mean between- of these statistical models predict a similar increase species exponent was used for all studies. in resting oxygen requirement for tropical ®sh rela- A linear plot of the data (Fig. 3a) reveals a mono- tive to that of polar ®sh (range 5´9 to 6´2) and tonic curvilinear relationship between metabolic rate all explain similar amounts of the overall variation and temperature in teleost ®sh. This is strikingly in resting metabolic rate (55±59%). The remaining similar to relationships established previously for variance will include both experimental error, and aquatic invertebrates (Ivleva 1977, 1980; Ikeda 1985; any e€ects of phylogeny, ecology or life-style. The summarized in Clarke 1991) and it indicates that a latter three in¯uences are, of course, likely to be typical tropical ®sh living at 30 C requires consider- linked. # 1999 British Fig. 3. Relationship between resting oxygen consumption (Rb, mmol h±1) and temperature for teleost ®sh, with di€erent Ecological Society ®tted statistical models. All data corrected to a size of 50 g wet mass; data shown are from ONESTUDY data set, although Journal of Animal the FULL data set gave similar relationships. (a) Linear plot; note elevation of abscissa for clarity. (b) Log/linear plot with Ecology, ®tted least-squares regression line. (c) Log/log plot with ®tted least-squares regression line. (d) Arrhenius plot, with ®tted 68, 893±905 least-squares regression line. 898 The in¯uence of phylogeny on resting metabolic ing metabolic rates were shown by gadoids and the Scaling of rate was examined for two of the linear statistical lowest by eels (Table 1). metabolic rate in models, the double logarithmic (Fig. 3c) and Although ANCOVA revealed signi®cant di€erences teleost ®sh Arrhenius (Fig. 3d). The results were closely similar in mean resting metabolic rate between di€erent and only the analysis of the Arrhenius data is dis- taxa of teleost ®sh, the large variance indicates a cussed here. high degree of overlap in the data (Fig. 4). The pooled Arrhenius model (Fig. 3d), thus provides a valid ®rst-order statistical model for the overall rela- tionship between resting metabolic rate and tem- TAXONOMIC VARIATION IN RESTING perature in teleost ®sh. This relationship is METABOLIC RATE 1nRb = 15´7 ± 5´02.T±1. Analysis of covariance (ANCOVA) was used to exam- ine the relationship between resting metabolic rate DO POLAR FISH SHOW METABOLIC COLD and temperature for di€ering levels of taxonomic ADAPTATION? aggregation, using data from the ONESTUDY data set. Pooling of data by family provided insucient The concept of metabolic cold adaptation is that data for ANCOVA; too many families were repre- polar ®sh should show a resting metabolic rate sented by only a single study or the temperature greater than the rate predicted by extrapolation to range represented in the data was too low (< 10 C). polar temperatures of the resting metabolic rate of Analysis of data pooled by order did, however, temperate or tropical ®sh. Although originally pro- reveal a signi®cant e€ect. posed by Krogh (1914, 1916), this idea was most ANCOVA provided no evidence to reject the null convincingly espoused by Scholander et al. (1953), hypothesis that the slope of the relationship between Wohlschlag (1960) and Dunbar (1968). So much so metabolic rate and temperature was the same for that for a long time metabolic cold adaptation was each order of teleost ®sh (F = 1´74, P > 0´10). There the de®ning paradigm for investigations of evolu- were, however, highly signi®cant di€erences in eleva- tionary adaptation in polar marine organisms (see tion (F = 12´7, P < 0´001) indicating that the resting discussions by Clarke 1980, 1983, 1991). metabolism of ®sh from di€erent taxonomic groups A number of studies of metabolic cold adaptation varied signi®cantly (Fig. 4). The highest mean rest- have compared the metabolic rate of polar ®sh with Fig. 4. Analysis of covariance for the relationship between resting metabolic rate and temperature for teleost ®sh, with data pooled by taxonomic order. Data are from the FULL data set, with ®tted Arrhenius statistical models. Symbols represent # 1999 British di€erent orders, with regression lines from ANCOVA shown; in order of decreasing metabolic rate the lines are for Ecological Society Gadiformes (.), Pleuronectiformes (*), Perciformes ( & ), Salmoniformes (~), Cypriniformes (!), and Anguilliformes (r). Journal of Animal Analysis of the ONESTUDY data set revealed a similar pattern, except that two data sets (Anguilliformes, Gadiformes) Ecology, had to be excluded because of insucient data. ANCOVA using a double logarithmic statistical model ®tted to both the 68, 893±905 ONESTUDY and FULL data sets gave closely similar results. 899 Table 1. Taxonomic variation in the relationship between resting oxygen consumption (Rb, mmol h±1) and temperature A. Clarke & in teleost ®sh. Taxonomy follows Nelson (1994). The individual scaling relationships between resting metabolic rate (mmol h±1) and body mass (g) were used to estimate the oxygen consumption of a 50-g ®sh. These data were then ®tted by N.M. Johnston an Arrhenius model, with regression lines ®tted individually to each order by ANCOVA (Fig. 4). These relationships were then used to predict the resting metabolic rate of a 50-g ®sh at 15 C. The six orders included in the ANCOVA include 55 of the 69 species examined in this study. Note that the standard error (SE) has been back-transformed from natural loga- rithms Number of Predicted metabolic rate Order Families Genera Species Mean SE Gadiformes 1 3 4 0´417 0´078 Pleuronectiformes 5 8 8 0´173 0´298 Salmoniformes 2 4 6 0´231 0´033 Perciformes 10 17 30 0´193 0´013 Cypriniformes 2 5 5 0´104 0´013 Anguilliformes 1 1 2 0´086 0´016 that of temperate species (reviewed by Macdonald, Discussion Montgomery & Wells 1987). Whilst Johnston et al. (1991a) attempted to control for ecology and life- BODY MASS SCALING EXPONENT style, no comparative studies to date have controlled The mean within-species (intraspeci®c) mass scaling for phylogeny. This is potentially a problem because exponent for teleost ®sh was 0´79 (SE 0´11, the Southern Ocean ®sh fauna is most unusual in n = 138). Although a scaling exponent of about 0´75 being dominated ecologically by the radiation of a is commonly reported for the between-species rela- single perciforme suborder, the notothenioids tionship between resting or basal metabolic rate, (Eastman 1993; Clarke & Johnston 1996; Eastman and body mass in ectotherms, there is no generally & Clarke 1998). Given the signi®cant variation of accepted theoretical explanation for this value. mean resting metabolic rate with phylogeny (Fig. 4; Discussion of this topic started over a century ago Table 1) the only valid test of metabolic cold adap- tation is to con®ne the analysis to perciformes and compare the resting metabolic rates of polar notothenioids with the metabolic rates of non-polar perciforme ®sh. When this is done there is no evi- dence that the resting metabolic rates of notothe- nioids is detectably di€erent from that predicted from the relationship between metabolic rate and temperature of non-polar perciforme ®sh (Fig. 5). Perciformes are the most speciose teleost order (Nelson 1994), and comprise ®sh with a wide range of morphology and ecology. The variance about the regression line (Fig. 5) also indicates that there is a considerable diversity of resting metabolic rate. Zoarcids (eel-pouts) in particular have a somewhat sedentary life-style and a low metabolic rate. Wohlschlag (1963) described the Antarctic Rhigophila dearborni as having an unusually low metabolism, but Holeton (1974) was able to show that this was a feature of zoarcids as a taxon and that a comparative study of zoarcid metabolism Fig. 5. A test of metabolic cold adaptation. The resting a€orded no support for metabolic cold adaptation. metabolic rate of notothenioid ®shes from the Southern The only other teleost order for which there are Ocean (.) compared with that of other (non-notothenioid) sucient polar and non-polar data to test for meta- perciforme ®sh from warmer waters (*). The data are pre- bolic cold adaptation is Gadiformes (true cods). sented as an Arrhenius plot, and the line ®tted to all non- # 1999 British These show no evidence for metabolic cold adapta- polar data. The polar data show no evidence of signi®cant Ecological Society metabolic cold adaptation (sensu Krogh: a tendency for tion (data not shown) although the robustness of polar species to lie above the ®tted line). The polar data Journal of Animal Ecology, this conclusion is limited by the few data available include several studies not included in either the FULL or 68, 893±905 (Holeton 1974). ONESTUDY data sets (see methods). 900 (Rubner 1883) and many of the key arguments were must ®nd six times as much food per unit time sim- Scaling of explored by Zeuthen (1947, 1953). Nevertheless, ply to fuel resting metabolism. The di€erence will metabolic rate in despite attempts based on surface area/volume con- likely be even greater for overall power budgets teleost ®sh siderations, the incorporation of physiological time, incorporating the annual costs of locomotor activity biological similarity and strict dimensional analysis and growth, both of which are generally greater in (see discussions in Peters 1983; Schmidt-Nielsen tropical ®sh (Bennett 1985; Clarke & North 1991; 1984; and Calder 1984) nobody has been able to Johnston et al. 1991a; Kock & Everson 1998). derive a convincing or universally accepted predic- Although any one of the four models shown in tion of the value of the scaling exponent from ®rst Fig. 3 would allow the resting metabolism of a ®sh principles. A recent suggestion has been that of to be estimated on the basis of its mass and environ- Kooijman (1993), who used dynamical energy bud- mental temperature, we disagree with Peters (1991) get theory to conclude that the exponent is a that these may in any way be construed as theory. weighted sum of surface area and volume. The various plots are simply statistical models Kooijman argued that the actual value thus lies which summarize observations in a convenient man- within the bounds 0´66 and 1, depending on the role ner. It remains a major diculty in thermal ecology of energy metabolism in thermoregulation, with pre- that we have as yet no theory which can explain or dictions of lower values (1 0´66) for endotherms, predict from ®rst principles the relationship to be and higher values (closer to 1) for pure ectotherms. expected between whole organism resting metabo- Most recently, West, Brown & Enquist (1997) have lism and environmental temperature in ectotherms deduced a value of 1 0´75 for inter-speci®c allome- (Clarke 1991, 1993). try based on physiological constraints associated The Arrhenius relationship provides an excellent with the architecture of branching vascular systems. description of the relationship between reaction rate The diculty with many of these suggestions is that and temperature for a simple chemical system. of deciding which feature of a tightly integrated Furthermore, the form of this relationship can be physiology represent the true evolutionary con- deduced from ®rst principles for it is based ®rmly straint, and which features are co-adaptations opti- on statistical thermodynamics. It might therefore be mized to a constraint elsewhere. regarded as a valid model of whole-animal thermal Whereas an intra-speci®c scaling exponent is use- physiology, if it could be assumed that the key phy- ful for studies of within-species energetics, general siological processes involved in energy metabolism models applicable to all ®sh require a between-spe- form an integrated whole that behave in the same cies (inter-speci®c) scaling exponent. These two way with respect to temperature as do the isolated exponents may be di€erent, for both statistical and component systems (a suggestion which originated evolutionary reasons (see discussion by Harvey & at least as far back as Krogh 1916). It is not clear to Pagel 1991), and Kozl/ owski & Weiner (1997) have what extent this assumption can be justi®ed, other recently suggested that inter-speci®c scaling re¯ects than by the existence of extensive and subtle con- the overall pattern of body size optimization at the trols that regulate whole-animal physiology. On the individual species level. Calder (1984) also cautions other hand, the log/log and log/linear (exponential) against mixing ontogenetic and phylogenetic data in models are no more than traditional statistical trans- studies of scaling. In the study reported here, the formations which have proved e€ective in linearizing inter-speci®c scaling exponent was 0´80, which is sta- many biological relationships. In the absence of any tistically indistinguishable from the mean within- alternative theory or explanation, the Arrhenius species value. This mean value is also very similar to relationship is therefore probably the most appro- that calculated by Schmidt-Nielsen (1984) from a priate statistical description of whole-animal thermal much smaller data set and to the overall mean value physiology. calculated for endotherms by Peters (1983), which The Arrhenius statistical model ®tted to the was 0´74 (SE = 0´11, n = 146). ONESTUDY data set (Fig. 3d) is equivalent to a between-species Q10 of 1´83 over the temperature range 0±30 C. (Systems exhibiting Arrhenius beha- RESPIRATION AND TEMPERATURE viour have a Q10, which varies inversely with tem- The resting oxygen consumption of teleost ®sh was perature, although the e€ect is small over the found to vary signi®cantly with temperature (Fig. 3). ecological temperature range: Clarke 1983.) This is The estimate of the factorial increase in resting smaller than typical within-species acute Q10 values metabolism comparing a representative tropical reported in the literature. Of the 138 studies exam- (30 C) and polar (0 C) teleost depended on the sta- ined here, 14 reported Q10 values; these ranged from # 1999 British tistical model ®tted, and ranged from 5´9 (log/log 0´45 to 3´41 with a median of 2´40 (the frequency Ecological Society model) to 6´2 (Arrhenius model). This indicates that distribution being negatively skewed). This suggests Journal of Animal Ecology, a typical tropical ®sh consumes roughly six times as that evolutionary temperature adaptation has pro- 68, 893±905 much oxygen as does a typical polar ®sh and, hence, duced a between-species relationship that has a 901 lower thermal sensitivity than is typical of within- or polar ®sh, an e€ect now thoroughly con®rmed by A. Clarke & species relationships. Although there remains a sig- both time-course (Morris & North 1984) and hor- N.M. Johnston ni®cant di€erence between the rate of resting meta- monal (Egginton 1994) studies. Slow gut-passage bolism in typical polar and tropical teleosts, this is times in polar ®sh also extend the period over which less than would be expected on the basis of the ther- post-prandial increases in metabolic rate are mal sensitivity of individual species determined in observed (Johnston & Battram 1993; Boyce & experimental studies. Clarke 1997). Neither the careful experimental work of Holeton (1973, 1974) nor theoretical arguments (Clarke METABOLIC COLD ADAPTATION 1980, 1991, 1993) resulted in the demise of the con- cept of metabolic cold adaptation. Recent studies In considering the evolutionary adaptation of teleost have concluded that there may be a small elevation ®sh to environmental temperature, probably the sin- of resting metabolic rate in polar ®sh (perhaps by a gle most contentious aspect has been that of meta- factor of 2: Forster et al. 1987; Torres & Somero bolic cold adaptation. The pioneering Danish 1988), but these comparative studies have usually respiratory physiologist August Krogh undertook a involved only a limited range of species. The now classic experiment in which a single gold®sh, detailed examination of the resting metabolic rate of Carassius auratus, was exposed to a series of tem- tropical, temperate and polar ®sh of similar size and peratures and its oxygen consumption measured life-style (Johnston et al. 1991a), and the comparison (Ege & Krogh 1914). The experiment produced a here for perciform ®sh (Fig. 4) have, however, con- roughly exponential decrease in respiration rate as ®rmed the conclusion ®rst promulgated by Holeton the temperature was lowered, and this relationship (1974): metabolic cold adaptation (sensu Krogh) became known as Krogh's normal curve. does not exist. Krogh (1914, 1916) made the intuitively reason- able prediction that ®sh which had evolved to live at polar temperatures would be expected to have THE INFLUENCE OF PHYLOGENY evolved some form of compensation for the rate- depressing e€ect of temperature he had observed in It is now well recognized that to achieve a true evo- his gold®sh. Speci®cally, the prediction was that the lutionary perspective in comparative studies it is metabolic rate of ®sh that had evolved to live at necessary to control for phylogeny (Harvey & Pagel polar temperatures would be higher than the rates 1991). Closely-related species are likely to have exhibited by temperate ®sh cooled to polar tempera- diverged from a common ancestor only recently tures (or the rates predicted by extrapolation to and, hence, to have had less time to accumulate dif- polar temperatures of the metabolic rate/tempera- ferences than more distantly related species. This ture relationships established for warmer water ®sh), can lead to a phylogenetically-based pseudo-replica- the end result of this evolutionary process being tion in comparative studies and unless phylogeny is termed metabolic cold adaptation. Although the con- controlled for statistically, results in an increased cept was established explicitly in terms of respira- incidence of Type 1 errors. tory physiology, it was subsequently extended to In the absence of a detailed phylogeny for the spe- measures such as the activity of individual metabolic cies examined in this study, we have instead under- pathways or enzymes. taken an analysis of covariance. Data were Early measurements of the metabolic rate of polar aggregated at the level of order for ANCOVA; when ®sh appeared to con®rm Krogh's prediction, thereby data were aggregated at lower taxonomic levels establishing the concept of metabolic cold adapta- (family, genus) there were too many missing obser- tion ®rmly in the literature of comparative and evo- vations to permit statistical analysis. This analysis lutionary physiology (Scholander et al. 1953; revealed statistically signi®cant di€erences in resting Wohlschlag 1960, 1964; and many subsequent text- metabolic rate between the six orders for which we books and reviews). had sucient data (Fig. 4; Table 1). This result is The next development was an exceptionally care- similar to that obtained by Pauly (1980) for the ful series of measurements by Holeton (1973, 1974), mortality rate of di€erent taxa of ®sh living at dif- which showed clearly that there was no convincing ferent temperatures. experimental evidence for an elevated metabolic rate Although there are signi®cant di€erences between in polar ®sh. Holeton explained the early results di€erent teleost orders in mean resting metabolic which appeared to support the concept of metabolic rate, there is also substantial variation within major cold adaptation as being caused partly by genuinely taxa. Thus, di€erent species within a given order # 1999 British elevated metabolic rates induced by the experimental may vary markedly in metabolic rate, even after Ecological Society protocol. Holeton (1974) was the ®rst to recognize controlling for body mass and temperature. This Journal of Animal Ecology, that polar ®sh need much longer periods of acclima- variation is correlated with ecology and, particu- 68, 893±905 tion to experimental apparatus than do temperate larly, with broad patterns of activity. Thus, in polar 902 notothenioids, resting metabolic rate increases from ciated with di€erences in lifestyle (and particularly Scaling of less active to more active species (Morris & North in activity) between species, possibly mediated metabolic rate in 1984) and this is true of polar ®sh in general through factors such as di€erences in mitochondrial teleost ®sh (Zimmerman 1997). It would thus appear that the density and associated physiological costs. These evolution of particular lifestyles brings with it a patterns would suggest that broad di€erences concomitant associated metabolic cost, as has between major lineages in physiology have no phy- been shown in muroid rodents (Koteja & Weiner logeny-speci®c component. Phylogenetically-linked 1993). di€erences appear merely as epiphenomena because By analogy with data emerging for birds (Daan, the proximal causes of di€erences in physiology, Masman & Groenewold 1990), it would seem likely such as di€erences in anatomical and physiological that broad di€erences in resting metabolic rate of organization, tend to be more similar within lineages ®sh are caused by changes in the relative propor- than between lineages. tions of body organs with di€ering inherent mass- The ANCOVA also suggests that when comparison speci®c metabolic demands. Because the anatomical is made between di€erent taxonomic orders (for con®guration of species is likely to be more con- example, gadoids and eels: Fig. 4) there is a tendency served within lineages than between lineages, this for resting metabolic rate at representative environ- would explain the observed di€erences in mean rest- mental temperatures to be more similar than the ing metabolic rate between major taxa. overall relationship across the species within an It would therefore be instructive to map the order. The degree of temperature sensitivity thus observed resting metabolic rates onto a secure phy- decreases in the sequence species > within orders > logeny to undertake a thorough analysis of evolu- between orders. tionary trends. Unfortunately, there is no phylogeny available for the species used in this study and even THE EVOLUTION OF METABOLIC RATE the topology at the level of order is not well resolved. Nevertheless, it is clear that there is no The correlation of resting metabolic rate with envir- strong relationship between higher level phylogeny onmental temperature (Fig. 3) and the variation and resting metabolic rate, since the euteleost clade between major taxa (Fig. 4) pose the question of incorporates taxa with the highest and one of the what are the key driving forces behind the evolution lowest resting metabolic rates (Fig. 6). of a particular metabolic rate. What dictates that Taxonomic (and by inference, phylogenetic) dif- one type of ®sh has a higher or lower resting meta- ferences in the resting metabolic rate of ®shes are bolic rate than another, and why is resting metabolic probably related to lineage-speci®c variations in rate so much higher for a tropical than a polar ®sh? body architecture and, speci®cally, the relative pro- The lower thermal sensitivity of resting metabolic portions of the major body organs. Variations in rate between species (Q10 = 1´83) compared with resting metabolic rate within major taxa are asso- that observed typically in acclimation studies (med- # 1999 British Ecological Society Fig. 6. Estimated resting oxygen consumption of a 50-g ®sh at 15 C from the di€erent teleost orders examined in this study, Journal of Animal superimposed on a working phylogeny. Taxonomy and cladogram topology from Nelson (1994). The metabolic rate data Ecology, plotted are mean (line), one standard error (stippled bar) and 95% con®dence intervals for the mean (open bar), estimated 68, 893±905 from ANCOVA. 903 ian Q10 = 2´40) suggests a degree of evolutionary increased costs for tropical species). The question is, A. Clarke & adjustment. It is not clear whether the lower thermal thus, what evolutionary forces are keeping tropical N.M. Johnston sensitivity of the evolutionary relationship is the resting metabolic rates high? At present we do not result of a high resting metabolic rate at lower tem- know, and why the relationship between resting peratures, a reduced resting metabolic rate at higher metabolic rate and temperature is what it is remains temperatures or a combination of both. an intractable problem for physiological ecologists. A subtle but important distinction which needs to One possible factor is the extent to which evolu- be drawn here is that between an overall evolution- tionary adjustment to temperature allowing a ary adjustment to the thermal sensitivity of resting reduced resting metabolic rate has also constrained metabolic rate, and metabolic cold adaptation the capacity to generate metabolic power. If relative (sensu Krogh). The former applies across the com- aerobic scope (the ratio of peak power to resting plete physiological temperature range (roughly ± 2 C power) remains more or less independent of tem- to + 40 C for teleost ®sh), whereas the latter perature, then absolute metabolic power is greatly hypothesis predicted a temperature/resting meta- reduced in polar species. Thus, for a constant rela- bolic rate relationship for polar ®sh di€erent in kind tive aerobic scope of 5 a representative polar spe- from all other ®sh. It is also possible that the cies at 0 C can generate 0´33 W extra power above between-species relationship (Fig. 3) is simply an epi- resting metabolism, whereas a representative tropi- phenomenon resulting from physiological optimiza- cal species at 30 C could generate an extra 2´05 W tion at the species level and thus requires no (data for a representative 50-g ®sh, calculated from evolutionary explanation. the overall Arrhenius statistical model: Fig. 3d). This A major problem for thermal physiologists considerable di€erence may have enormous ecologi- attempting to explain the relationship between tem- cal signi®cance. perature and metabolic rate is that it is not at all It is not yet clear, however, the extent to which clear what physiological processes are operating fas- relative aerobic scope varies with temperature or the ter in a resting tropical ®sh (or slower in a resting extent to which polar species have a reduced capa- polar ®sh). It seems likely that protein turnover, city to generate metabolic power, but this is an membrane homeostasis and cellular ion balance are active area of research (Johnston, Johnson & all important (Hawkins 1991; Clarke 1993), Battram 1991b; Boyce & Clarke 1997; Johnston although neither the relative contribution of each of et al. 1998). these to maintenance metabolism, nor how these vary with temperature or lifestyle, are known with any certainty. Acknowledgements Resting metabolism does, however, represent a We thank Dr Nicky Fenton for undertaking a preli- cost to the organism in that energy utilized in main- minary compilation and analysis of these data, Dr tenance must be met from food or reserves. It can- Sara Boyce for collation of extra data, Alistair not be used in processes that contribute to Murray for statistical advice and for undertaking evolutionary ®tness, such as growth, reproduction the ANCOVA, and both Alistair Murray and or behavioural activity (other than in the trivial Professor Ian Boyd for constructive criticism of sense that a dead organism has zero ®tness). Thus, draft manuscripts, which enhanced signi®cantly the although it is easy to recognize in principle the ®t- clarity of the argument. Dr Monty Priede provided ness advantage to be gained from an enhanced a helpful review. scope for growth or increased scope for activity, it is much more dicult to discern an evolutionary advantage to an elevated resting or maintenance References metabolic rate. This problem was recognized early in the debate over metabolic cold adaptation Bennett, A.F. (1985) Temperature and muscle. Journal of Experimental Biology, 115, 333±344. (Dunbar 1968; Somero, Giese & Wohlschlag 1968), Bennett, A.F. (1987) Interindividual variability: an under- but not resolved. utilized resource. New Directions in Ecological It has long been known for a variety of taxa that Physiology (eds M.E. Feder, A.F. Bennett, within a species there is often an inverse correlation W.W. Burggren & R.B. Huey), pp. 147±169. Cambridge University Press, Cambridge. between resting metabolic rate and growth rate Boyce, S.J. & Clarke, A. (1997) E€ects of body size and (Hawkins 1991). Such variability within a species ration on speci®c dynamic action in the Antarctic provides raw material for evolutionary change, but plunder®sh Harpagifer antarcticus Nybelin, 1947. does not explain why resting metabolic rate varies Physiological Zoology, 70, 679±690. # 1999 British between taxa as it does. The study reported here Calder, W.A. (1984) Size, Function and Life History. Ecological Society Harvard University Press, Cambridge (Massachusetts). shows that there is an energetic advantage to living Journal of Animal Clarke, A. (1980) A reappraisal of the concept of metabolic Ecology, in cold water in terms of decreased maintenance cold adaptation in polar marine invertebrates. 68, 893±905 costs (or alternatively a disadvantage in terms of Biological Journal of the Linnean Society, 14, 77±92. 904 Clarke, A. (1983) Life in cold water: the physiological ecol- Johnston, I.A., Calvo, J., Guderley, H., Fernandez, D. & Scaling of ogy of polar marine ectotherms. Oceanography and Palmer, L. (1998) Latitudinal variation in the abun- Marine Biology: an Annual Review of, 21, 341±453. dance and oxidative capacities of muscle mitochondria metabolic rate in Clarke, A. (1991) What is cold adaptation and how should in perciform ®shes. Journal of Experimental Biology, teleost ®sh we measure it? American Zoologist, 31, 81±92. 201, 1±12. Clarke, A. (1993) Seasonal acclimatization and latitudinal Johnston, I.A., Clarke, A. & Ward, P. (1991a) compensation in metabolism: do they exist? Functional Temperature and metabolic rate in sedentary ®sh from Ecology, 7, 139±149. the Antarctic, North Sea and Indo-West Paci®c Ocean. Clarke, A. & Johnston, I.A. (1996) Evolution and adaptive Marine Biology, 109, 191±195. radiation of Antarctic ®shes. Trends in Ecology and Johnston, I.A., Johnson, T.P. & Battram, J.C. (1991b) Evolution, 11, 212±218. Low temperature limits burst swimming performance Clarke, A. & North, A.W. (1991) Is the growth of polar in Antarctic ®sh. Biology of Antarctic Fish (eds G. di ®sh limited by temperature? Biology of Antarctic Fish Prisco, B. Maresca & B. Tota), pp. 177±190. Springer- (eds G. di Prisco, B. Maresca & B. Tota), pp. 54±69. Verlag, Berlin. Springer-Verlag, Berlin. Kock, K.-H. & Everson, I. (1998) Age, growth and maxi- Daan, S., Masman, D. & Groenewold, A. (1990) Avian mum size of Antarctic ®sh revisited. Fishes of the basal metabolic rates; their association with body-com- Antarctic Ocean, a Biological Overview (eds G. di position and energy expenditure in nature. American Prisco, E. Pisano & A. Clarke), pp. 29±40. Springer- Journal of Physiology, 259, R333±R340. Verlag, Berlin. Dunbar, M.J. (1968) Ecological Development in Polar Kooijman, S.A.L.M. (1993) Dynamic Energy Budgets in Regions; a Study in Evolution. Prentice Hall, Biological Systems. Cambridge University Press, Englewood Cli€s, New Jersey. Cambridge. Eastman, J.T. (1993) Antarctic Fish Biology Evolution in a Koteja, P. & Weiner, J. (1993) Mice, voles and hamsters: Unique Environment. Academic Press, San Diego. metabolic rates and adaptive strategies in muroid Eastman, J.T. & Clarke, A. (1998) A comparison of adap- rodents. Oikos, 66, 505±514. tive radiations of Antarctic ®sh with those of non- Kozl/ owski, J. & Weiner, J. (1997) Interspeci®c allometries Antarctic ®sh. Fishes of Antarctica: a Biological are by-products of body size optimization. American Overview (eds G. di Prisco, E. Pisano & A. Clarke), Naturalist, 149, 352±380. pp. 3±26. Springer-Verlag, Berlin. Krogh, A. (1914) The quantitative relation between tem- Ege, R. & Krogh, A. (1914) On the relation between the perature and standard metabolism in animals. temperature and the respiratory exchange in ®shes. Internationale Zeitschrift fuÈr Physikalisch-Chemische Internationale Revue der Gesamten Hydrobiologie und Biologie, 1, 491±508. Hydrographie, 1, 48±55. Krogh, A. (1916) The Respiratory Exchange of Animals and Egginton, S. (1994) Stress response in two Antarctic tele- Man. Longmans, London. osts (Chaenocephalus aceratus Lonnberg and LaBarbera, M. (1989) Analyzing body size as a factor in Notothenia coriiceps Richardson) following capture ecology and evolution. Annual Review of Ecology and and surgery. Journal of Comparative Physiology B, 164, Systematics, 20, 90±117. 482±491. Laws, E.A. & Archie, J.W. (1981) Appropriate use of Forster, M.E., Franklin, C.E., Taylor, H.H. & Davison, regression analysis in marine biology. Marine Biology, W. (1987) The aerobic scope of an Antarctic ®sh, 65, 13±16. Pagothenia borchgrevinki, and its signi®cance for meta- Macdonald, J.A., Montgomery, J.C. & Wells, R.M.G. bolic cold adaptation. Polar Biology, 8, 155±159. (1987) Comparative physiology of Antarctic ®shes. Harvey, P.H. & Pagel, M.D. (1991) The Comparative Advances in Marine Biology, 24, 321±388. Method in Evolutionary Biology. Oxford University Morris, D.J. & North, A.W. (1984) Oxygen consumption Press, Oxford. of ®ve species of ®sh from South Georgia. Journal of Hawkins, A.J.S. (1991) Protein turnover: a functional Experimental Marine Biology and Ecology, 265, 75±86. appraisal. Functional Ecology, 5, 222±233. Nelson, J.S. (1994) Fishes of the World, 3rd edn. John Holeton, G.F. (1973) Respiration of arctic char (Salvelinus Wiley, New York. alpinus) from a high arctic lake. Journal of the Fisheries Pauly, D. (1980) On the interrelationships between natural Research Board of Canada, 30, 717±723. mortality, growth parameters and mean environmental Holeton, G.F. (1974) Metabolic cold adaptation of polar temperature in 175 ®sh stocks. Journal of Du Conseil ®sh: fact or artefact? Physiological Zoology, 47, 137± Permanent International Pour l'Exploration de la Mer, 152. 39, 175±192. Ikeda, T. (1985) Metabolic rates of epipelagic marine zoo- Payne, R.W., Lane, P.W., Digby, P.G.N., Harding, A., plankton as a function of body mass and temperature. Leech, P.K., Morgan, G.W., Todd, A.W., Thompson, Marine Biology, 85, 1±11. R., Tunnicli€e Wilson, G., Welham, S.J. & White, Ivleva, I.V. (1977) Quantitative correlation of temperature R.P. (1993) Genstat 5 Release 3 Reference Manual. and respiratory rate in poikilothermic animals. Polskie Oxford University Press, Oxford. Archiwum Hydrobiologii, 20, 283±300. Peters, R.H. (1983) The Ecological Implications of Body Ivleva, I.V. (1980) The dependence of crustacean respira- Size. Cambridge University Press, Cambridge. tion on body mass and habitat temperature. Peters, R.H. (1991) A Critique for Ecology. Cambridge International Revue der Gesamten Hydrobiologie und University Press, Cambridge. Hydrographie, 65, 1±47. Ricker, W.E. (1973) Linear regression in ®shery research. Jobling, M. (1981) The in¯uence(s) of feeding on the meta- Journal of the Fisheries Research Board of Canada, 30, bolic rate of ®shes: a short review. Journal of Fish 409±434. # 1999 British Biology, 18, 385±400. Rubner, M. (1883) Ueber den Ein¯uss der KoÈrpergroÈsse Ecological Society Johnston, I.A. & Battram, J.C. (1993) Feeding energetics auf Sto€und Kraftwechsel. Zeitschrift fuÈr Biologie, 19, Journal of Animal and metabolism in demersal ®sh species from 535±562. Ecology, Antarctic, temperate and tropical environments. Schmidt-Nielsen, K. (1984) Scaling: Why Is Animal Size So 68, 893±905 Marine Biology, 115, 7±14. Important? Cambridge University Press, Cambridge. 905 Scholander, P.F., Flagg, W., Walters, V. & Irving, L. Wohlschlag, D.E. (1963) An antarctic ®sh with unusually A. Clarke & (1953) Climatic adaptation in arctic and tropical poiki- low metabolism. Ecology, 44, 557±564. lotherms. Physiological Zoology, 26, 67±92. Wohlschlag, D.E. (1964) Respiratory metabolism and eco- N.M. Johnston logical characteristics of some ®shes in McMurdo Somero, G.N., Giese, A.C. & Wohlschlag, D.E. (1968) Cold adaptation of the antarctic ®sh Trematomus ber- Sound, Antarctica. Biology of the Antarctic Seas, Vol 1 nachii. Comparative Biochemistry and Physiology, 26, (ed. M.O. Lee), Antarctic Research Series, pp. 33±62. 223±233. Washington D.C., American Geophysical Union. Zeuthen, E. (1947) Body size and metabolic rate in the ani- Torres, J.J. & Somero, G.N. (1988) Metabolism, enzymic mal kingdom. Compte Rendu Des Travaux Du activities and cold adaptation in Antarctic mesopelagic Laboratoire de Carlsberg, Serie Chimique, 26, 17±65. ®shes. Marine Biology, 98, 169±180. Zeuthen, E. (1953) Oxygen uptake as related to body size West, G.B., Brown, J.H. & Enquist, B.J. (1997) A general in organisms. Quarterly Review of Biology, 28, 1±12. model for the origin of allometric scaling laws in biol- Zimmerman, C. (1997) On the ecology of Arctic and ogy. Science, 276, 122±126. Antarctic ®sh: activity, sensory capabilities and beha- Wohlschlag, D.E. (1960) Metabolism of an Antarctic ®sh viour. Berichte Zur Polarforschung, 231, 1±137. and the phenomenon of cold adaptation. Ecology, 41, 287±292. Received 16 April 1998; revision received 4 December 1998 # 1999 British Ecological Society Journal of Animal Ecology, 68, 893±905