• Open Access

Are mice calorically restricted in nature?


Steven N. Austad, Department of Biological Sciences, PO Box 3051, University of Idaho, Moscow, ID 83844–3051, USA. Tel.: +1 208 885 6598; fax: +1 208 8857905; e-mail: austad@uidaho.edu
E-mail for D. M. Kristan: kristand@uidaho.edu


An important question about traditional caloric restriction (CR) experiments on laboratory mice is how food intake in the laboratory compares with that of wild mice in nature. Such knowledge would allow us to distinguish between two opposing views of the anti-aging effect of CR – whether CR represents, in laboratory animals, a return to a more normal level of food intake, compared with excess food consumption typical of laboratory conditions or whether CR represents restriction below that of animals living in nature, i.e. the conditions under which house mice evolved. To address this issue, we compared energy use of three mouse genotypes: (1) laboratory-selected mouse strains (= laboratory mice), (2) house mice that were four generations or fewer removed from the wild (= wild-derived mice) and (3) mice living in nature (= wild mice). We found, after correcting for body mass, that ad libitum fed laboratory mice eat no more than wild mice. In fact, under demanding natural conditions, wild mice eat even more than ad libitum fed laboratory mice. Laboratory mice do, however, eat more than wild-derived mice housed in similar captive conditions. Therefore, laboratory mice have been selected during the course of domestication for increased food intake compared with captive wild mice, but they are not particularly gluttonous compared with wild mice in nature. We conclude that CR experiments do in fact restrict energy consumption beyond that typically experienced by mice in nature. Therefore, the retarded aging observed with CR is not due to eliminating the detrimental effects of overeating.


Some researchers have speculated that the senescence-retarding effect of caloric restriction (CR) on laboratory rodents is in reality an artefact of overfeeding under captive conditions (Cherkin, 1979; Hayflick, 1994). For example, Cutler (1982) asserted that CR returns ‘the animal back to the aging rate it would normally have in its natural ecological niche’ and does not extend lifespan beyond the ‘normal genetic potential for the animal’. More generally, this argument posits that wild mice are chronically calorically restricted due to the difficulty of finding food in nature. Therefore, the typical laboratory protocol of restricting animals to 60% of their ad libitum (ad lib) food intake may more realistically replicate life in the field – the conditions under which mouse physiology has been selected. The hypothesis concludes that typical laboratory experiments, instead of comparing control with restricted animals, are in actuality comparing overfed animals with adequately fed ones, and, not surprisingly, the overfed ones develop a host of pathologies and die younger.

Overfeeding is assumed to be a result of unnatural continuous food availability, possibly exacerbated by the inadvertent selection of animals that are particularly gluttonous. Increased food consumption by mice leads to faster growth, earlier sexual maturation, larger body size and enhanced fertility in adulthood (Eisen et al., 1980; Singleton et al., 2001), and it is well documented that laboratory mice have been selected for enhanced reproductive rate compared with wild mice (Berry, 1969; Clark & Price, 1981; Miller et al., 2000, 2002). Thus the conditions in a standard commercial mouse breeding facility will favour genetic variants that reproduce well and also eat more than other genotypes. This overfeeding response could conceivably be associated with decreased longevity and the development of numerous late-life pathologies (e.g. increased occurrence of spontaneous tumour formation), because of the diminished power of natural selection to winnow out alleles with detrimental effects that only become manifest late in reproductive life (Medawar, 1952; Rose, 1991). Overeating by laboratory mice compared with wild mice has been called the ‘laboratory glutton’ hypothesis of the CR effect (Austad, 2001).

There exists abundant evidence that wild mammals, including house mice, indeed eat less than they would prefer. Although it is not possible to restrict the caloric intake of animals in wild populations in a controlled manner, it is possible to supplement their food and thus potentially increase their caloric intake. Boutin (1990) summarized the results of 70 such studies in wild mammals, including five on house mice. Taken together, the studies find that supplemental feeding usually (but not always) increases the body mass of animals in the study area but always increases breeding intensity, either by lengthening the breeding season, increasing the fraction of reproducing females, or both. In all the house mouse studies, breeding intensity increased with supplemental feeding at least during some part of the calendar year. So wild mice apparently could, and would, eat more food if it were available, and this increased food availability would be beneficial at least in terms of reproduction.

Importantly, although long-term caloric restriction provides many benefits to non-reproductive laboratory mice, these mice typically have ceased oestrous cycling (Nelson et al., 1985). Because all existing wild mice populations must reproduce or become extinct, presumably wild mice are eating relatively more than restricted laboratory mice during at least part of the year. Alternatively, wild mice may respond to low calorie intake differently to laboratory mice, and therefore be capable of continued reproduction under energetic conditions in which laboratory mice are infertile.

Caloric intake of mice

For clarity and concision, we term house mice that have lived under laboratory conditions for more than 20 generations ‘laboratory mice’. Mice that have lived in the laboratory for four generations or fewer (too short a time for inadvertent selection to have substantially altered their genetic make-up in comparison with their wild ancestors) we call ‘wild-derived mice’, and mice living in nature we call ‘wild mice’.

The central issue for the laboratory glutton hypothesis is not whether wild mice eat less than they would prefer, but whether laboratory mice eat substantially more than wild mice, such that a typical laboratory restriction regime really approximates the food consumption of mice in nature. The answer is not obvious a priori. Wild mice might not eat as much as they would like (Boutin, 1990, and references therein), but they may still eat as much or more than laboratory mice, because wild mice have many energetic demands, such as those associated with foraging, thermoregulation and avoiding predators, that laboratory mice do not. Wild mice must eat at least enough to meet these demands or they will suffer fitness consequences (e.g. decreased reproduction, death). Whether the amount that captive ad lib fed mice consume is less than, equal to or more than what is consumed by wild mice needs to be assessed empirically. Therefore, we compared, using data from the literature and from our own empirical studies, ad lib food intake of laboratory vs. wild-derived mice, in order to determine whether food intake of laboratory mice has been increased as a correlated response to selection for enhanced reproduction. To determine whether mice in nature eat less than mice fed ad lib in captivity, we also assessed energy use or intake of wild mice vs. captive mice (both laboratory mice and wild-derived mice).

Measuring energy intake in nature

With the advent of the doubly labelled water technique to measure metabolic rate of wild animals (Speakman, 1997), we can now compare energy use by wild animals with that of captive ones. The doubly labelled water technique tracks the rate of disappearance of hydrogen and oxygen isotopes (2H or 3H and 18O) in blood samples after a known amount of doubly labelled water is injected into an animal. Oxygen in the body water rapidly equilibrates with oxygen in respiratory carbon dioxide. Thus, as oxygen leaves the body in both CO2 and water, the radioactive oxygen signal declines. By contrast (making some reasonable assumptions), labelled hydrogen leaves the body only in water. CO2 production can therefore be easily calculated (and metabolic rate thus estimated) from the difference between the rates of disappearance of the oxygen and hydrogen isotopes (Speakman, 1997).

Assuming that animals over a short period will often be at, or near, a state of energy balance (energy expenditure equals energy intake), we can directly compare the energetics of mice in the laboratory and field. The assumption about animals in the field being at, or near, energy balance derives from the fact that small mammals have metabolic rates too high to be supported by stored fat for prolonged periods (Bronson et al., 1991). Therefore, small mammals in the field need to receive about as much energy as they use most of the time. While 24 h may not be long enough for wild animals to show energy balance (Speakman et al., 1994), 2–2.5 days (the duration of doubly labelled water experiments used in our analysis) may approach the time over which small rodents must show approximately balanced energy budgets.

Three studies to date have measured field metabolic rate in house mice. Geographical locations of these studies were southern California, southern Australia and a sub-Antarctic island (Nagy, 1987; Rowe-Rowe et al., 1989; Mutze et al., 1991). Metabolic rate measurements were typically made over a 2- to 2.5-day time interval in both sexes during both breeding and non-breeding seasons (Table 1). To test the laboratory glutton hypothesis we compared energy use of wild mice in two of these studies to food (energy) consumed by captive ad lib fed mice (see below for details). The sub-Antarctic island study is useful in indicating level of food intake in a particularly demanding environment (cold exposure).

Table 1.  Energy consumption/use by laboratory mice, wild-derived mice in captivity (four or fewer generations captive), wild mice in captivity and wild mice in nature
GenotypeMass (g)Daily metabolic rateSexAge (months)Temperature
kJ day−1kJ day−1 (g0.568)−1
  • a

    Mass estimated from graph.

  • b

    Age estimated based on median age of adult wild house mice;

  • c

    age estimated based on mass–age relationship of outbred strains of laboratory mice.

  • d

    Digestible food intake (total food intake multiplied by digestibility);

  • e

    e digestible food intake estimated from published caloric content of diet.

  • f

    Doubly labelled water;

  • g

    g unknown.

  • *

    Classified as separate species of Mus in some phylogenies.

  • Anomalously low ad lib food intake compared with other studies of the same genotype. This is also the only published study in which CR did not extend lifespan in this genotype.

Mice in captivity
 Wild-derived19.5 45.9 8.5M323Food intaked,eJ. Harper & S. N. Austad (unpubl.)
 Wild-derived20.2 44.9 8.2F523Food intakedD. M. Kristan & K. Hammond (unpubl.)
 Wild-derived*14.2 32.7 7.2M4b2224-h respirometryTicu & Stoica (1971)
 Wild-derived*14.4 35.7 7.8F4b2224-h respirometryTicu & Stoica (1971)
 C3B10RF127a 52.8 8.1F4.522Food intaked,eWeindruch et al. (1986)
 MF138.8 69.5 8.7Unkg5c1824-h respirometrySpeakman et al. (1991)
 C57BL/6NNia29a 61.1 9.0M422Food intaked,eTurturro et al. (1999)
 C57BL/6NNia23a 58.6 9.9F422Food intaked,eTurturro et al. (1999)
 C57BL/6J26a 42.1 6.6F4.522Food intaked,eHarrison et al. (1984)
 B6D2F128a 63.6 9.6M422Food intaked,eTurturro et al. (1999)
 B6D2F121a 55.8 9.9F422Food intaked,eTurturro et al. (1999)
 DBA/2JNia26a 63.610.0M422Food intaked,eTurturro et al. (1999)
 DBA/2JNia21a 55.8 9.9F422Food intaked,eTurturro et al. (1999)
 B6C3F128a 75.311.3M422Food intaked,eTurturro et al. (1999)
 B6C3F123a 59.710.0F422Food intaked,eTurturro et al. (1999)
 Swiss Webster33a 60.8 8.3F323Food intakedKristan & Hammond (2000)
Mice in field
 Wild13 39.8 9.3M/F4b DL waterfNagy (1987)
 Wild14 45.310.1M/F4b DL waterfMutze et al. (1991)
In demanding conditions
 Wild (cold)19.3 65.112.2M/F4b 5DL waterfRowe-Rowe et al. (1989)
 Wild derived (lactation)29.3143.421.1F623Food intakedD. M. Kristan & K. Hammond (unpubl.)

Effects of body mass and age on metabolism

One confounding issue in comparing different house mice studies is variation in body mass. Wild mice typically average substantially less than 20 g in body mass, with some mice as light as 9 g, depending on the population studied (Austad, 1996). Even ‘giant’ island mouse populations average less than 30 g in body mass (Berry et al., 1978). By contrast, many laboratory mice grow to more than 30 g and outbred laboratory genotypes may reach peak body masses of as much as 40–50 g (Turturro et al., 1999). In the studies we analysed, body mass varied from 13 to 14 g for wild mice, 14.2 to 20.2 g for wild-derived mice and 23 to 38.8 g for laboratory mice (Table 1). Note there is no overlap in these body masses.

Bioenergetics researchers have historically used mass-specific metabolic rate (dividing metabolism by total body mass or, when it is available, lean body mass) as a method for correcting for differences in body size. However, this approach has some statistical problems, because the relationship between metabolism and body mass is allometric, not isometric (Packard & Boardman, 1987, and references therein). Simply dividing by body mass (or lean mass) forces the metabolism-body size curve through the origin, and does not accurately describe metabolic differences among animals of differing size within the same species. To circumvent this problem, researchers now generally use statistical methods to correct for body mass effects [either analysis of covariance (ancova) or analysis of variance (anova) of residuals from the regression of metabolic rate on body mass], or express metabolic rate as energy per gram of body mass raised to an exponent where the exponent empirically describes the relationship between metabolism and body mass for the species being studied. For any of these three methods, using lean body mass may be preferable to whole body mass because fat is not very metabolically active (Martin & Fuhrman, 1955), and can be a substantial component of whole body mass for some individuals (especially for older ad lib fed laboratory mice).

We also tried to standardize mouse age as closely as possible, because older mice (at least in captivity) typically have lower lean-to-fat mass ratios than younger mice. In deciding which age to use we considered that wild mice are young by laboratory standards. For laboratory mice, the median lifespan ranges from about 600 to 900 days, and 90% mortality typically occurs by about 900–1100 days. A variety of field studies have indicated that the median lifespan of wild mice is about 130 days with 90% mortality occurring by about 280 days (Berry & Jakobson, 1971; Phelan & Austad, 1989). Therefore, a random animal sampled from the field will be a youngster by laboratory standards.


Body mass was significantly greater for laboratory mice than either wild or wild-derived mice (anova, F2,18 = 12.0, P = 0.001; Tukey's HSD test for multiple comparisons; Fig. 1), but did not differ statistically between wild and wild-derived mice. Similarly, absolute fat mass (F2,11 = 7.5, P = 0.015) as well as per cent body fat (F2,11 = 4.5, P = 0.049; Fig. 1; Table 2) were greater in laboratory mice compared with wild and wild-derived mice, with the latter two groups not differing statistically from one another. However, fat mass adjusted for body mass using ancova did not show significant differences among groups (F2,11 = 0.6, P = 0.588).

Figure 1.

Body mass and composition (lean vs. fat mass) for laboratory, wild-derived and wild house mice. Per cent body fat is given for each bar. Values are means ± 1 standard error (sample sizes: laboratory, n = 4; wild-derived, n = 5; wild, n=2).

Table 2.  Body composition of laboratory mice, wild-derived mice in captivity (four or fewer generations captive born) and wild mice in nature
GenotypeMass (g)Dry fat mass (g)Fat (%)SexAge (months)Temperature (°C)Reference
  • a

    Age estimated based on median age of adult wild house mice.

  • b

    Two male and two female mice were collected between 28 March 2002 and 17 April 2002 in Moscow, ID, when average temperature was 5 °C (range of averages: 3–6 °C).

  • c

    Modified from data published in Kristan & Hammond (2000).

Mice in captivity
 Wild derived142.215F1.324Bronson et al. (1991)
 Wild derived213.315F7.523Kristan & Hammond (2003)
 Wild derived192.010F523D. M. Kristan & K. Hammond (unpubl.)
 Wild derived171.6 9F2.523D. M. Kristan (unpubl.)
 Wild derived201.7 9M2.523D. M. Kristan (unpubl.)
 C57BL/6J306.622F1.222Harrison et al. (1984)
 Swiss Webster346.019F2.523Kristan & Hammond (2000a)
 Swiss Webster262.911F223Kristan (2002b)
 Swiss Webster313.110M223Kristan (2002b)
Mice in field
 Wild190.7 4M4a 5bS. N. Austad & D. M. Kristan (unpubl.)
 Wild150.8 5F4a 5bS. N. Austad & D. M. Kristan (unpubl.)
In demanding conditions
 Wild derived (lactation)260.8 3F623D. M. Kristan & K. Hammond (unpubl.)
 Swiss Webster (lactation)384.311F323Kristan (2002a)
 Wild derived (cold)212.612F3 5Kristan & Hammond (2003)
 Swiss Webster (cold)326.821F2.5 5Kristan & Hammond (2000)c

Because of differences in body mass, we examined both absolute energy use, and energy use standardized by body mass, either by analysis of covariance (ancova) with a body mass covariate (and present least squares means ± 1 SEM), or by dividing absolute energy use by body mass raised to 0.568 (the slope of the relationship between metabolism and body mass for small mammals; Koteja, 1991). This was statistically evaluated by analysis of variance (anova), followed by a Tukey HSD post-hoc comparison (and present arithmetic means ± 1 SEM).

Ignoring body size, not surprisingly the much heavier laboratory mice consumed more energy than either wild-derived or wild mice (F2,18 = 16.6, P < 0.0001; Fig. 2). We found an overall significant difference between groups of mice. When we corrected for size using body mass0.568, we found a significant difference between groups (F2,18 = 4.1, P = 0.038; Fig. 2); when correcting using ancova, there was a marginal statistical difference between the groups (F2,18 = 3.6, P = 0.054; Fig. 2). Using post hoc analyses, we found that laboratory mice consumed more energy than wild-derived mice in captivity (Tukey test of body mass0.568: P = 0.04; non-overlapping 95% confidence intervals for the ancova), but not more than wild mice (Fig. 2). Furthermore, wild-derived mice did not consume significantly less energy than wild mice (P = 0.117).

Figure 2.

Energy use for laboratory, wild-derived and wild house mice evaluated on an absolute basis and by adjusting for body mass using ancova and body mass as a function of the relationship between body mass and metabolic rate. Values are means ± 1 standard error (sample sizes: laboratory, n = 12; wild-derived, n = 4; wild, n = 2).


After accounting for body mass, laboratory mice do not appear to eat more than wild mice (Fig. 2). Although our sample of energy consumption from wild mice comes from only two studies, somewhat surprisingly we did find some statistically significant differences between mice in the field and the laboratory groups. These two field studies did reach remarkably similar results for field metabolic rate, however. Furthermore, one of the studies –Mutze et al. (1991) – was exceptionally thorough, sampling a total of 59 animals at different times of the year, during both breeding and non-breeding seasons. During some sampling periods, energy use was as high as 12.4 kJ day−1 (g0.568)−1 or higher than laboratory mice under any conditions measured. A third field study performed in a colder climate reported approximately the same level of energy use as the highest level found in the long-term study (Table 1). So field data from varying localities under varying conditions seem to lie within a surprisingly consistent range. Those data clearly do not support the laboratory glutton hypothesis.

The unexpectedly high values for food consumption in nature are probably due to the high energetic demands of foraging, thermoregulation, predator avoidance and reproduction. For example, wild mice measured at an average of 5 °C had energy consumption that was 43% greater than the average energy consumption of laboratory mice kept at room temperature. Furthermore, wild-derived mice had 140% greater energy consumption during peak lactation than laboratory mice (Table 1; for effects of cold exposure and lactation in resting metabolism see also Hammond & Diamond, 1992; Hammond et al., 1996; Kristan & Hammond, 2000).

Inadvertent selection for increased food intake

One issue that can be addressed by the data in hand is the extent to which laboratory mice have been selected to eat more than wild-derived mice when both are fed ad lib. J. Harper and S. N. Austad (unpublished data) measured the ad lib food consumption of 3-month-old male wild-derived mice from Idaho and D. M. Kristan and K. Hammond (unpublished data) measured ad lib food consumption for 5-month-old female wild-derived mice from Arizona (Table 1). Our analysis shows that after accounting for body mass effects, laboratory mice consume almost 20% more energy than wild-derived mice maintained under similar captive conditions. Indeed, if we delete one study of C57BL/6J mice (Harrison et al., 1984), in which food consumption for some reason was about one-third lower than in comparable studies with the same genotype, laboratory mice consume more than 30% more energy on a mass-corrected basis than wild-derived mice. Therefore, we propose that increased food intake by laboratory mice is a correlated response to the well-documented selection for rapid growth and large litter size in laboratory mice (Berry, 1969; Clark & Price, 1981; Miller et al., 2000; 2002).

Relationship of body composition with metabolism and caloric restriction

It may be preferable to use lean body mass, rather than whole body mass, to standardize metabolism measures because body fat is not very metabolically active (Martin & Fuhrman, 1955). However, lean body mass was not available for the wild mouse studies, in which energy use was measured. It is likely that using lean body mass as a covariate in our analyses would not alter our conclusions. For example, effects of cold exposure on resting metabolism in laboratory mice were the same regardless of whether whole body mass or lean body mass was used in the analysis, despite the fact that body composition differed between cold-exposed and warm-adapted mice (Kristan & Hammond, 2000). Similarly, the effects of short-term caloric restriction on resting metabolism of laboratory mice was similar when either whole body mass or lean body mass was used as a covariate, even though the amount of body fat was less for calorically restricted than for ad lib fed mice (Kristan & Hammond, 2001). It should be noted that because laboratory mice are still gaining mass at 4 months of age and do not become exceptionally fat until later in life, the differences in using whole vs. lean body mass to standardize metabolism measures may become important if older laboratory mice are used.

An obvious additional consideration is whether body composition plays a role in the caloric restriction effect directly (as opposed to potential indirect effects associated with metabolism as described above). This topic has not yet been clarified. Some evidence suggests that changes in body fat are of minor importance at best in the caloric restriction effect. For example, body mass and per cent body fat have been manipulated in laboratory rats by providing them access to exercise wheels, but these treatments have far less impact on longevity than does altering caloric intake itself (Holloszy & Schechtman, 1991; McCarter et al., 1997). In addition, among calorically restricted rats, individuals with more body fat have been reported to live longer (Bertrand et al., 1980). In another study suggesting that per cent body fat was not involved in the CR effect, genetically obese (ob/ob) mice subjected to CR exhibited 48% body fat, yet lived longer than ad lib fed C57BL/6J controls (22% body fat). Harrison et al. (1984) concluded that reduced caloric intake, not reduced body fat, was the primary component of increased longevity associated with CR. However, this conclusion may have been premature. Recent work with mice genetically engineered to lack insulin-receptor in adipose tissue (FIRKO mice) produced animals that ate as much as control animals yet were considerably leaner and lighter and lived 18% longer than controls (Blüher et al., 2003). Although the authors of the FIRKO mouse paper interpreted their results as demonstrating ‘the beneficial effects of reduced adiposity on the extension of life-span . . .’, clearly there would be many metabolic consequences associated with knocking out this receptor. Whether this result is attributable to reduced adiposity remains to be determined.

Additionally, Barzilai & Gupta (1999) argue that previous studies of the relationship between CR, fat mass and longevity are flawed and unable to distinguish the effects of changes in fat mass during CR on longevity. They posit that many of the systemic changes that occur with CR can be attributed to changes in circulating levels of a variety of adipocyte-derived factors (e.g. peptides, cytokines, complement factors) and that an examination of fat distribution and gene expression for factors derived from fat tissue will provide valuable information on how CR can modulate life expectancy in laboratory rodents. Therefore, although the effects of body composition may have a minor impact on basal or resting metabolism, and by extension on daily metabolic requirements, its role in whole animal response to CR remains unresolved. We are only beginning to understand the very important endocrine functions of fat tissue as they relate to physiological changes during CR. Further investigation of whether absolute fat levels are important as well as potential differences in function among different fat depots in the body that change mass differentially during CR will provide useful information towards understanding whole animal effects of CR.

Is absolute caloric intake more important than mass-adjusted intake?

One provocative aspect of the findings presented here concerns total energy consumption per animal, i.e. energy consumption uncorrected for differences in body mass. The range of values for non-reproductive wild animals in natural populations occurring in warmer climates is approximately 40–45 kJ per animal per day whereas those for ad lib fed laboratory animals of approximately the same age ranged from roughly 42 to 75 kJ per animal per day (or 53–75 kJ deleting the same anomalous C57BL/6 study as previously). Reducing the ad lib level of laboratory consumption by 40% (a typical experimental protocol) results in restricted mice consuming between 25 and 45 kJ (or 32–45 kJ after deleting the same study) per animal per day. Therefore, ignoring body size, CR laboratory mice consume somewhat fewer, or in some cases nearly equal, calories as are expended by house mice in nature.

During long-term 40% CR, the peak body mass of restricted laboratory mice ranges from less than 40% to more than 70% smaller than ad lib fed animals (Turturro et al., 1999). Body mass of CR laboratory mice overlaps the body masses of only the largest populations of wild mice, but is still heavier than the average mass of most wild house mice. Therefore, when body mass corrections are used, the CR laboratory mice are consuming much less than wild mice because of their relatively large body mass. Therefore, traditional CR studies are indeed limiting the restricted mice in both absolute and relative caloric intake compared with wild mice.

Experimental procedures

We chose 4 months (about the median age of wild mice) as a reasonable age for mice in our comparisons based on the median age of wild mice in nature. In addition, whereas mice in the laboratory continue to gain weight with age, their food consumption after 4 month does not increase dramatically (Fig. 1; Turturro et al., 1999). Therefore, 4 months probably also represents an age when ad lib food intake (our proxy for energy expenditure) per gram of body mass for laboratory mice is near its maximum.

We compared energetics data from two studies of wild house mice in nature, three studies of wild-derived or wild-caught mice and five studies that examined eight laboratory mouse genotypes (both inbred and outbred). We also compared these with several other studies of mice under extreme energy stress. Energy consumption for wild mice was measured directly with doubly labelled water. Energy consumption for captive mice was based either on 24-h energy expenditure measured by indirect calorimetry (measurement of respiratory gas exchange) or on digestible food intake (a measure of metabolizable energy). We calculated digestible food intake from total food intake multiplied by published digestibility values for each diet, converting all energy measures to metabolizable energy (kJ day−1). Where we could obtain information on diet composition (i.e. percentage fat, protein and carbohydrate), we assumed that metabolizable energy was 6% less than total calories. We determined this value based on a comparison of the relationship between total calories and metabolizable energy for seven published diets formulated by Purina Mills Inc. (average difference ± 1 standard error 6.4 ± 0.9%). Because lean body mass was not available for wild mice in the published papers, we tested energy-use data using whole body mass corrections both by ancova and with metabolic rate divided by body mass raised to an appropriate exponent for intraspecific studies (Koteja, 1991) analysed by anova and Tukey HSD post hoc test for multiple comparisons.

We examined body composition for wild mice captured near Moscow, Idaho, by ether extraction of the entire carcass, and compared these values with those from four published studies of wild-derived house mice, and of three studies that examined two laboratory mouse genotypes (both inbred and outbred) (Table 2).


This work supported in part by grants from NIH (R01 AG13711) and the Ellison Medical Foundation. We thank Jim Harper, Kim Hammond, Roger McCarter and an anonymous reviewer for data and/or helpful comments on this topic.