Examination of the nitrogen limitation hypothesis in non-cyclic populations of cotton rats (Sigmodon hispidus)

Authors


Dr R.L. Lochmiller, Department of Zoology, Oklahoma State University, Stillwater, OK 74078, USA. E-mail: rllzool@okway.okstate.edu.

Abstract

1. Nitrogen-containing nutrients have long been considered a frequently limiting resource to the growth of herbivore populations (nitrogen limitation hypothesis). To explore this hypothesis, we examined the relationships between availability of essential amino acids and concentrations of phenolics in the diets of hispid cotton rats (Sigmodon hispidus) in central Oklahoma and the intrinsic characteristics of their non-cyclic populations. We posited that lower quality proteins (i.e. essential amino acid composition) and elevated phenolic levels (protein digestion inhibitors) in diets of cotton rats from low- compared to high-density populations, especially during the breeding season, would be supportive of the nitrogen limitation hypothesis. Replicated low- and high-density populations were censused by live-trapping at 3-month intervals. Samples of stomach digesta were collected from cotton rats in similar habitats adjacent to trapping grids to determine the botanical and nutrient composition of their diets.

2. During the breeding season, concentrations of essential amino acids were as much as 43% greater in diets of cotton rats from high-density populations. Dicots, typically higher in protein than monocots, were an important component of diets and were preferred forage in all seasonal collections. Seeds and arthropods were frequently utilized by cotton rats as additional high-quality sources of essential amino acids. Concentrations of total phenolics in the diet (greater in diets from low-density populations) were consistent with the nitrogen limitation hypothesis.

3. Density was consistently higher in the high-density populations throughout the study. Other demographic and body condition parameters were similar between low- and high-density populations in the non-breeding season, but reproductive activity was greater in high-density populations during the breeding season. Total number of juveniles recruited into the trappable population over the entire study was about five times greater in high- compared to low-density populations.

4. Our data did not refute the nitrogen limitation hypothesis where levels of essential amino acids and phenolic compounds in the diet during the breeding season may determine annual peak densities of cotton rats that can be supported in their habitat. However, we could not rule out the involvement of other environmental variables such as overhead cover (as well as other unmeasured variables) as contributing factors to determining annual peak densities.

Introduction

Increasing evidence supports the hypothesis that availability of high quality food can limit the growth and size of many populations of wild herbivores (Hanson 1979; Keith 1983; Doonan & Slade 1995). Studies of rodent and lagomorph populations have demonstrated that relationships exist between availability of suitable forage and onset of breeding (Cole & Batzli 1979; Keith 1987), breeding intensity (Cole & Batzli 1978; Bomford & Redhead 1987), fecundity (Cole & Batzli 1978, 1979), growth rates (Cole & Batzli 1979; Batzli & Lesieutre 1991), juvenile and adult survival (Cole & Batzli 1979; Keith 1987), and home range size or use (Ostfeld 1985; Jones 1990; Eshelman & Cameron 1996). However, most of these studies have not identified specific nutrients (e.g. specific amino acids) that might be limiting and when these limitations occur at the population level.

White (1978, 1993) and Mattson (1980) have suggested that nitrogen-containing nutrients are frequently limiting in the habitats of many populations of wild herbivores (nitrogen limitation hypothesis). Animals do not have a dietary requirement for protein per se but require specific amounts of essential amino acids from dietary protein that cannot be synthesized endogenously in adequate amounts. Therefore, diet quality relative to nitrogenous compounds is largely determined by how well dietary proteins supply an animal with a proper balance of essential amino acids (Oser 1959). Despite the potential importance of essential amino acids to populations of wild herbivores, few studies have actually examined the relationship between their availability in the habitat and population fluctuations. Quality of proteins (amino acid composition) in the diets of waterfowl (Krapu & Swanson 1975; Thomas & Prevett 1980; Sedinger 1984), willow ptarmigans (Lagopus lagopus; Steen et al. 1988), primates (Oftedal 1991), northern bobwhite quail (Colinus virginianus; Peoples et al. 1994), and eastern cottontails (Sylvilagus floridanus; Lochmiller et al. 1995) have been described. Collectively, these studies suggest that levels of essential amino acids (especially the sulphur-containing amino acids) in the diet may be seasonally deficient relative to animal requirements.

Hispid cotton rats, Sigmodon hispidus (Say & Ord; Rodentia: Muridae), are generalist herbivores that occur primarily in grass-dominated habitats in northern South America, Central America, and south-eastern and south-central North America (Cameron & Spencer 1981). In the tallgrass prairies of central Oklahoma, cotton rats are the dominant rodent species, exhibiting seasonal fluctuations in population density with peak densities occurring in late summer or early autumn and minimum densities occurring in early spring (Goertz 1964). Although this species can vary in peak density from one habitat type to another (McMurry et al. 1994), individual populations are relatively stable across years compared to more cyclic rodent species which experience dramatic multiannual changes in density. We selected the cotton rat to examine whether the nitrogen limitation hypothesis could explain differences in density among populations in the central Great Plains of North America. The primary objective of this study was to examine the dynamic relationships between availability of essential amino acids and total phenolics in the diet of cotton rats and intrinsic characteristics of their populations. Our approach was to compare seasonal changes in the botanical composition and protein quality of diets consumed by animals from replicated low- and high-density populations. Examination of relationships between diet quality and population density was chosen over direct comparisons of dietary levels of nutrients to those levels required in the diet because amino acid requirements of wild rodent species for growth and reproduction are unknown (National Research Council 1995). We posited that lower quality proteins (i.e. essential amino acid composition) and elevated phenolic levels in diets of cotton rats from low-density populations compared to high-density populations, especially during the breeding season, would be supportive of the nitrogen limitation hypothesis.

Materials and methods

Study area

After conducting extensive surveys of potential tallgrass prairie study sites in August 1993, we identified five distinct areas that supported low densities of cotton rats (peak densities of <25 animals ha−1) in western Payne Co., Oklahoma (36°3′–36°7′ N, 97°12′–97°13′ W) and two areas that supported high densities (peak densities of >80 animals ha−1) in southern Caddo Co., Oklahoma (34°53′–34°54′ N, 98°7′–98°11′ W). Herbaceous ground cover on three of the low-density (3–22 animals ha−1) sites was dominated by the grasses little bluestem (Schizachyrium scoparium (Nash)), big bluestem (Andropogon gerardii (Vitman)), tall dropseed (Sporobolus asper (Michx.)), and Indiangrass (Sorghastrum nutans (L.)), and the forbs western ragweed (Ambrosia psilostachya (DC.)), white sage (Artemisia ludoviciana (Nutt.)), goldenrod (Solidago spp.), and Sericia lespedeza (Lespedeza cuneata (G.)), an exotic legume. Woody plants occurring on these sites included smooth sumac (Rhus glabra (L.)), winged sumac (R. copallina (L.)), coral-berry (Symphoricarpos orbiculatus (Moench)), sand plum (Prunus angustifolia (Marsh.)), and eastern redcedar (Juniperus virginiana (L.)). The other two low-density sites occurred in an area that was subjected to wildfire 3 years previously and annual cattle grazing. Populations of cotton rats on these sites were at very low densities (<3 animals ha−1). Vegetation on these two sites was dominated by the grasses little bluestem, Muhlenbergia spp., purpletop (Tridens flavus (L.)), and Scribner's panicum (Panicum oligosanthes (Schultes)), the forbs western ragweed, white sage, and goldenrod, and the woody plants eastern redcedar and smooth sumac. Two high-density areas in Caddo Co. were located in heavily disturbed prairie dominated by Johnsongrass (Sorghum halapense (Pers.)), which occurred in solid stands on ≈50% of one site and ≈25% of the second site reaching heights >2 m. Dominant forbs were western ragweed, white sage, and the legume prairie acacia (Acacia angustissima (Kuntze)); smooth sumac was the most common woody plant.

Population fluctuations

A live-trapping grid consisting of 64 trap stations with one Sherman live trap (7·6 × 8·9 × 22·9 cm) was established on each of the seven study sites. We used a 15-m spacing pattern between stations on low-density grids, with each grid arrayed in an 8 × 8 design. Because areas of continuous habitat on the high-density grids were relatively small, we used a 10-m spacing pattern arrayed in an 8 × 8 and 4 × 16 design on these grids. Each population was censused in August and November 1993, and February, May, August and November 1994. All traps were opened in late afternoon (16.00–18.00 h), baited with rolled oats, provided with cotton for warmth during cold weather, and checked between 06.00 and 11.00 hours for three consecutive days during a census period. Traps were placed on the grids on the afternoon before the first day of sampling and were removed from the grids after the third day of sampling. Captured animals were toe-clipped with a unique number for identification and released immediately after data were recorded to include information on location of capture (station number), body mass (to the nearest 1 g), sex, reproductive condition (pregnant, lactating, or neither for females; scrotal or non-scrotal for males), and general condition (presence of ectoparasites or wounds). Age was determined by body mass: <60 g = juvenile, 61–99 g = subadult, >100 g = adult (Stafford & Stout 1983).

Population size of cotton rats was estimated using the program CAPTURE (Otis et al. 1978). Model M0 was selected by CAPTURE as the best estimator of population size for most data sets in this study. Because Darroch's estimator (Model Mt) is always valid when Model M0 is true and is a more robust estimator than M0 (White et al. 1982), we used Darroch's estimator for all data sets except as follows. When the number of captures and/or recaptures was insufficient to estimate population size on a given grid using CAPTURE, we used the equation: N = M/1− (1−p)t, where M is the minimum number of animals known alive, p is the mean capture probability for animals on all other grids during the same trapping period, and t is the number of trapping occasions. If trap mortality occurred and was <5% of the total animals captured on a grid during a 3-day trapping period, dead animals were not included in the data set when a population estimate was determined; the number of dead animals were added to the population estimate. If trap mortality exceeded 5% of captures (this only occurred once), the generalized removal estimate (Model Mbh) was used. We calculated density by dividing the estimated population size by the effective trapping area for each grid; effective trapping area was calculated as the area of a grid after adding to its perimeter half the mean maximum distance moved by all cotton rats on the grid (McMurry 1993). Survival rate was indexed for populations on each live-trapping grid as the proportion of animals captured during a census period that were recaptured during any subsequent census periods.

Vegetation sampling

During each census period, percentage cover of individual plant species (live plants) was estimated by the Daubenmire canopy cover method (Bonham 1989) within each of 30 randomly placed quadrats (20 × 50 cm) on each grid. Cover of standing and fallen litter also was estimated using the Daubenmire method. After cover was estimated in each quadrat, all living herbaceous vegetation was clipped at ground level and sorted into monocots and dicots. Clipped vegetation was then dried at 55°C for 5 days, and mass was recorded to the nearest 0·01 g.

Collection of stomach digesta

To collect stomach digesta for diet analysis, we used snap traps that were baited with the scent of peanut butter. Within 3 weeks of each census period, animals were removed from similar habitat >100 m from existing trapping grids (except during the last census period in November 1994 when they were collected directly from live-trapping grids). Snap traps were set near dusk, checked at night and again at dawn the following morning. Traps were checked every 3–4 h during very hot weather to ensure that trapped animals did not spoil. Variable numbers of animals were removed from each population due to fluctuating densities. It was especially difficult to obtain animals from low-density study sites despite intensive efforts during some collection periods when densities were extremely low. Cotton rats were captured from only three low-density sites in November 1993 and February 1994, two sites in May 1994, and four sites in August 1994. Cotton rats were captured from all five low-density sites in November 1994.

Trapped animals were returned to the laboratory within 1–2 h where body mass, length, age, sex and reproductive condition were recorded. Stomach contents from each animal were removed, cleared of parasites, mucosa and hair, and wet mass was determined before freezing. Animals were necropsied to measure mass of selected organs (liver, spleen, adrenal glands, seminal vesicles and testes for males, and uterus, ovaries and embryos, if present, for females).

Sample preparation and nutrient analyses

We lyophilized individual stomach contents to dryness, determined dry mass, and ground contents to even consistency with mortar and pestle. We excluded from analysis any digesta samples with a dry mass <0·1 g to minimize contributions of endogenous nitrogen (Peitz & Lochmiller 1993). To ensure adequate sample volume for all laboratory analyses (≈1 g dry mass), we composited stomach contents (1–4 rats per composite) from 222 cotton rats randomly by season and trapping site for a total of 104 composites following the procedure of Jenks et al. (1989).

We measured total phenolics in composited samples colorimetrically (absorbence read at 765 nm) with Folin-Ciocalteu reagent and gallic acid as a standard based on the procedures of Singleton & Rossi (1965). Lipid concentration was measured in composited samples using a soxlet apparatus and diethyl ether as a solvent (Williams 1984). We determined percentage nitrogen and crude protein (6·25 ×%N) of composited samples using a Perkin Elmer 2410 Series II Nitrogen Analyser calibrated with ethylenediaminetetracetic acid (EDTA). We used alfalfa (3·028% N) as an external standard (obtained from the Soil, Water and Forage Analytical Laboratory, Department of Agronomy, Oklahoma State University, Stillwater, OK, USA).

For amino acid analysis, fat-extracted composited samples of digesta (≈40 mg of protein) were placed in 25 × 150 mm glass tubes with Teflon caps, purged with N gas for 4 min, and hydrolysed in 15 mL 6N HCl at 110 C for 24 h. We filtered hydrolysed samples through a 0·45-μm syringe filter (Acrodisc CRPTF, Fisher Scientific, Plano, TX, USA) and added 50 mL internal standard (4 mm Norleucine in 0·1N HCl) to 150 mL of filtered hydrolysate prior to derivitization. Amino acids were derivitized (precolumn) with phenylisothiocynate to produce phenylthiocarbamyl amino acids using the procedure of Cohen, Meys & Tarvin (1988). Samples were then refiltered through a 0·45-μm syringe filter. We determined concentrations of nine essential (histidine, arginine, threonine, valine, methionine, isoleucine, leucine, phenylalanine, and lysine) and eight non-essential (aspartic acid, glutamic acid, serine, glycine, alanine, proline, tyrosine, and cystine) amino acids using high-performance liquid chromatography (Waters Model 820 System Controller and Model 501 Pumps, Millipore Corporation, Milford, MA, USA).

We used the following chromatographic conditions: Waters Pico-Tag Silica/C18 (150 mm × 3·9 mm) column and guard column (20 mm × 3·9 mm); column temperature, 37°C; flow rate, 1·0 mL min−1 with pump back pressure of 16 095 kg cm−2; system sensitivity, 489 mv s−1 (recorder) and 0·1 absorbance units full scale (Waters Model 484 UV detector, set at 254 nm); injection volume, 12 μL; and run time, 27·5 min. We used solvents Eluent A and Eluent B (catalogue no. 88208 and 88112, respectively, Millipore Corp., Milford, MA, USA) under conditions and gradients described for separation of amino acids by Cohen et al. (1988). Two ground, ether-extracted feeds (A & M Complete Rabbit Pellets and A & M Quail Starter, Stillwater Milling Company, Stillwater, OK, USA) of known amino acid composition (determined by a certified laboratory, University of Missouri Experimental Station Chemical Laboratory, Columbia, MO, USA) were hydrolysed and analysed with samples for comparison of amino acid recovery. Concentrations obtained for methionine and cystine were combined, as were phenylalanine and tyrosine, because cystine and tyrosine may supply up to 50% of the requirement for their respective amino acid (National Research Council 1995). Tryptophan was destroyed during acid hydrolysis and therefore was not measured (Cohen et al. 1988). Non-protein nitrogen was calculated as (crude protein nitrogen – amino acid nitrogen)/crude protein nitrogen.

Food habit analysis

Food items in composited digesta samples were identified through microhistological analysis. Dried, composited samples were cleared of pigments with 95% ethyl alcohol, bleached, stained using lactophenol blue solution, and permanently mounted on microscope slides using glycerin gel (Davitt & Nelson 1980).

For each composited sample, botanical composition of the diet was determined by randomly locating 25 microscope fields on each of three slides, identifying the centre-most fragment in each field at 100× magnification, and counting the 0·25-mm2 squares on a 10 × 10-mm ocular grid that were occupied by each fragment (McMurry et al. 1993). Identification of plant fragments was facilitated with reference slides of plant tissues prepared as above. Relative composition of plant species in the diet was estimated for each composite by dividing the total coverage of each plant species by the total coverage of all fragments.

For statistical purposes items in the diet were placed into the following categories: monocot foliage (stems and leaves), dicot foliage (stems and leaves), non-legume seeds (monocots, woody and herbaceous dicots, gymnosperms), legume seeds (including pods), arthropods, and other (including fungi and unidentified fragments). A relative food preference index (PI = proportion of plant in diet/proportion of total plant cover in habitat; Lindroth & Batzli 1984a) was calculated for categories of plants (monocots, forbs, legumes, woody plants) in diets for each census period. PI values >1·0 indicated preference while values <1·0 indicated avoidance relative to availability of plants in the habitat (for true avoidance, PI = 0).

Statistical analyses

An analysis of variance procedure (Proc MIXED, SAS Institute, Inc. 1996) was used to test for differences in habitat composition, animal condition, diet composition, and diet quality between low- and high-density study sites. The model used was a repeated measures model with density as the main unit variable and season as the subunit variable. Both season and density were assumed to be fixed effects, with grid as the random experimental unit. Because many of the response variables showed significant density by season interaction, we analysed the simple effects of season (effect of density given season) using the SLICE option from an LSMEANS statement in Proc MIXED (SAS Institute, Inc. 1996). Satterthwait's approximation was used to estimate the denominator degrees of freedom for the pooled variance estimate used to test simple effects. This often resulted in a non-integer value for denominator degrees of freedom (see Steel & Torrie 1980). All of the response variables were analysed in this way, regardless of the presence of interaction in the overall model, in order to maintain consistency in the analyses. To correct for non-normality and heterogeneity of variance within the data sets, data were transformed prior to analysis when necessary (arcsine transformation for percentage data, log transformation for other data; Steel & Torrie 1980).

Population characteristics (percentage juveniles, subadults, and adults; percentage reproductive adult–subadult females; sex ratios; survival rates) were analysed within season between low- and high-density populations using chi-squared analysis (2 × 2 contingency tables; Proc FREQ, SAS Institute, Inc. 1996). Fisher's exact test (Proc FREQ, SAS Institute, Inc. 1996) was used to analyse any contingency tables where expected values were <5 (Zar 1984). Because of small numbers of animals captured from low-density populations, these demographic parameters were analysed within seasons after pooling animals from all low-density sites vs. all animals from high-density sites.

Stepwise discriminant analysis (Proc STEPDISC, SAS Institute, Inc. 1996) was used within season to select a reduced set of discriminator variables (out of nine essential amino acids) that provided the best model for classifying individual digesta composites according to population density. Variables were chosen to enter or leave the model based on a significance level of 0·15. Discriminant function analysis (Proc DISCRIM, SAS Institute, Inc. 1996) was used to determine classification accuracy among individual digesta composites from low- and high-density populations using the reduced set of variables selected by stepwise discriminant analysis. Canonical discriminant analysis (Proc CANDISC, SAS Institute, Inc. 1996) was used within season to derive canonical variate scores for individual digesta composites based on a linear combination of nine essential amino acids (the canonical variate). Data were rank transformed prior to multivariate analysis to correct for non-normality and heterogeneity of variance within the multivariate data set.

Results

Habitat composition

All sites showed considerable seasonal variation with standing crop biomass of monocots and dicots more than 10-fold greater in May and August 1994 than in other months(Fig. 1). Cover of forbs was significantly higher on low- compared to high-density sites in May 1994 (Fig. 2), a period when dicot biomass also tended to be higher on low-density sites. Biomass of both monocots and dicots was greater on high- compared to low-density sites in August 1994 (Fig. 1). Cover of legumes (absent in November 1993 and February 1994), woody plants, and warm-season grasses did not differ significantly between sites during any sampling period (Fig. 2).

Figure 1.

Mean biomass of monocots and non-woody dicots in habitats supporting low- and high-density populations of cotton rats in central Oklahoma from November 1993 to November 1994. Within each season, statistically significant differences between low- and high-density sites are indicated by +P < 0·10, *P < 0·05, **P < 0·01.

Figure 2.

Mean percentage cover of categories of vegetation in habitats supporting low- and high-density populations of cotton rats in central Oklahoma from November 1993 to November 1994. Within each season, statistically significant differences between low- and high-density sites are indicated by +P < 0·10, *P < 0·05, **P < 0·01. Data were arcsine transformed prior to statistical analysis.

Cool-season annual brome grasses (Bromus spp.) dominated high-density sites in November and February sampling periods after warm-season grasses matured and died, but brome grasses were less common on low-density sites. Cover of cool-season grasses was three times greater on high- compared to low-density sites in November 1994 (Fig. 2). Fallen litter was more abundant on high- than low-density study sites throughout the study (particularly in February 1994), but standing dead litter did not differ between sites (Fig. 2).

Population characteristics

Population densities of cotton rats were consistently greater on high-density census grids compared to the low-density grids, especially during August peaks (over 13-fold higher in August 1993 and 8-fold higher in August 1994; Fig. 3). In general, the demographic parameters we measured were not different between low-and high-density populations, except during the August 1993 breeding season (Table 1). The total number of juvenile recruits per study site that entered the trappable population from August 1993 to November 1994 was about 5-fold greater in high-(62·0 ± 19·8 juveniles site−1) compared to low-density (11·6 ± 17·6 juveniles site−1) populations (U = 10, n = 2 + 5, P = 0·050). In August 1993, the proportion of juveniles was 2-fold greater and the proportion of reproductively active (pregnant or lactating) females in combined live-trapped and snap-trapped census data was about 7 times greater in high- compared to low-density populations (Table 1). A similar trend was observed for snap-trapped females in August 1994 when there was a greater proportion of reproductively active females from high- (75·0%, n = 12) compared to low-density (33·3%, n = 12) populations (χ2 = 4·20, 1 d_f., P = 0·041). There was no difference in the number of embryos per pregnant female (among snap-trapped animals) between low- and high-density populations in May 1994 (6·8 ± 0·8 for low-density, 5·8 ± 0·5 for high-density; F1,8·8 = 1·38, P = 0·271) or August 1994 (5·3 ± 0·8 for low-density, 5·7 ± 0·6 for high-density; F1,15·3 = 0·18, P = 0·678). The per capita recruitment rate (total number juveniles divided by the total number of adult females) across the entire study was not different between low- and high-density sites (U = 9, n = 2 + 5, P = 0·100), averaging 1·89 (range from 0·5 to 3·50) per site. Sex ratios were similar between low- and high-density populations, but we observed a disproportionately high ratio of males to females in all populations during February and May 1994 (Table 1). During the breeding season (May [N0] to August [N1]), mean rate of population increase (r) was similar between low- (0·596 ± 0·239) and high-density (0·579 ± 0·060) populations (r = ln N1 − ln N0; Zejda & Nesvadbova 1996).

Figure 3.

Seasonal fluctuations in density (±SE) of five low- and two high-density populations of cotton rats in central Oklahoma from August 1993 to November 1994.

Table 1.  Seasonal changes in selected demographic attributes of low-and high-density cotton rat populations in central Oklahoma from August 1993 to November 1994 (N in parentheses, P in brackets). Attributes were analysed within season using 2 × 2 contingency tables (χ2 analysis or Fisher's exact test).
Demographic
parameter
Density
L = low;
H = high
August
1993
November
February
1994
May
August
November
  • *

    Only females >60 g.

  • Survival rate (3 months) was indexed as the proportion of animals alive at the start of a census period that remained alive to the start of the next census period, as indicated by subsequent recapture in later census periods. Survival rate could not be indexed for November 1994.

Juveniles (%)L17·9 (67)33·3 (45)46·4 (28)11·8 (34)22·5 (89)29·3 (75)
H33·1 (178)37·3 (83)24·5 (53)2·8 (72)25·0 (116)29·0 (69)
   [0·019] [0·651] [0·045] [0·062] [0·674] [0·963]
Subadults (%)L32·8 (67)33·3 (45)39·3 (28)17·6 (34)30·3 (89)34·7 (75)
H38·8 (178)41·0 (83)67·9 (53)12·5 (72)27·6 (116)37·7 (69)
   [0·897] [0·734] [0·013] [0·478] [0·666] [0·707]
Adults (%)L49·3 (67)33·3 (45)14·3 (28)70·6 (34)47·2 (89)36·0 (75)
H28·1 (178)21·7 (83)7·5 (53)84·7 (72)47·4 (116)33·3 (69)
   [0·002] [0·230] [0·334] [0·088] [0·917] [0·737]
Females inL3·8 (26)6·7 (15)0 (7)66·7 (9)32·1 (28)0 (23)
reproductiveH27·8 (54)0 (20)0 (14)55·6 (27)34·2 (38)0 (23)
condition (%)*  [0·012] [0·429]  [0·705] [0·860] 
Survival rate (%)L
H
20·0 (50)
20·7 (135)
 [0·912]
20·0 (35)
41·7 (60)
 [0·031]
30·0 (10)
43·8 (32)
 [0·490]
14·3 (21)
21·4 (42)
 [0·735]
21·8 (55)
12·8 (86)
 [0·157]
 
Sex ratio (% female)L50·7 (67)48·9 (45)39·3 (28)35·3 (34)40·2 (87)50·7 (75)
H48·9 (178)47·0 (83)39·6 (53)40·8 (72)47·4 (116)52·2 (69)
   [0·794] [0·837] [0·976] [0·623] [0·308] [0·857]

Survival rates from one census period to the next (3 months) were consistently low (≈20%) for all populations, indicating that turnover rates of populations were high (Table 1). A shorter time interval between census periods (probably monthly sampling) would have provided greater sensitivity for detecting differences between high- and low-density populations. Despite this shortcoming, we did observe a significantly greater survival rate in high- compared to low-density populations from November 1993 to February 1994.

Animal condition

No consistent patterns in body mass were evident from seasonal comparisons between low- and high-density populations (Table 2). Mean body mass of juveniles, reproductive females, and adult–subadult males was similar across all seasons of capture. The only difference in body mass that we observed was for adult–subadult females (>60 g) in November 1993, which was greater in low-density populations (Table 2). Similar to body mass, no consistent trends in organ mass were apparent between low- and high-density populations, although spleen mass was greater in animals from high- compared to low-density populations in August 1994 (Table 2).

Table 2.  Seasonal changes in body and organ masses (mean ± SE) in low- and high-density cotton rat populations in central Oklahoma from August 1993 to November 1994. For the entire study, statistically significant effects of season were found for all body/organ masses given below (P < 0·05). Within each season, P is given in brackets for comparisons of body/organ mass between low- and high-density populations
Condition
parameter
Density
H = high
L = low
August
1993
NL=5/NH=2
November

NL=3/NH=2
February
1994
NL=3/NH=2
May

NL=2/NH=2
August

NL=4/NH=2
November

NL=5/NH=2
Body mass (g)
JuvenilesL34·2 ± 5·145·1 ± 4·548·9 ± 4·6 34·2 ± 4·147·8 ± 4·1
H36·3 ± 4·045·7 ± 4·351·5 ± 4·9 36·7 ± 4·246·1 ± 4·8
   [0·872] [0·974] [0·796]  [0·957] [0·711]
Adult andL105·6 ± 5·0109·4 ± 6·788·8 ± 10·6122·4 ± 8·2108·4 ± 4·995·5 ± 5·4
subadultH101·8 ± 3·582·3 ± 5·878·6 ± 7·0120·3 ± 5·0113·6 ± 4·295·3 ± 5·3
females  [0·651] [0·001] [0·350] [0·965] [0·334] [0·967]
Adult andL112·2 ± 5·691·2 ± 7·575·0 ± 9·4127·9 ± 6·6112·7 ± 4·799·3 ± 5·5
subadultH105·1 ± 4·3101·3 ± 5·589·6 ± 6·1123·9 ± 5·0121·8 ± 4·8100·2 ± 6·1
males  [0·235] [0·174] [0·067] [0·646] [0·406] [0·846]
Organ mass
Liver (g)L3·58 ± 0·343·59 ± 0·422·76 ± 0·355·51 ± 0·373·82 ± 0·253·57 ± 0·25
H3·21 ± 0·283·10 ± 0·313·08 ± 0·355·77 ± 0·284·39 ± 0·273·83 ± 0·28
   [0·409] [0·351] [0·521] [0·570] [0·152] [0·491]
Spleen (mg)L178·8 ± 23·9159·3 ± 30·260·7 ± 27·6295·6 ± 28·8176·4 ± 17·4121·3 ± 15·9
H191·8 ± 19·1121·7 ± 19·5108·1 ± 23·9285·7 ± 17·7247·8 ± 18·1159·9 ± 18·4
   [0·671] [0·297] [0·195] [0·771] [0·005] [0·114]
Adrenals (mg)L28·5 ± 3·832·4 ± 4·820·3 ± 3·949·2 ± 4·234·0 ± 2·724·7 ± 2·6
H30·3 ± 3·021·4 ± 3·223·3 ± 3·752·4 ± 3·139·6 ± 2·931·4 ± 3·0
   [0·704] [0·059] [0·588] [0·542] [0·169] [0·100]
Testes (g)L1·97 ± 0·170·15 ± 0·210·20 ± 0·182·43 ± 0·192·19 ± 0·130·09 ± 0·13
H1·79 ± 0·160·10 ± 0·150·23 ± 0·152·31 ± 0·151·89 ± 0·140·17 ± 0·16
   [0·444] [0·855] [0·911] [0·623] [0·150] [0·725]

Diet composition

Diets of cotton rats from both low- and high-density sites showed considerable seasonal variation (Fig. 4). Animals from high-density sites consumed greater amounts of monocot foliage in November 1993 (F1,12 = 10·58, P = 0·007) and November 1994 (F1,9·1 = 7·90, P = 0·020) but consumed less in August 1994 (F1,8·9 = 6·16, P = 0·036) compared to those from low-density sites. We observed no differences in the consumption of other categories of food items (dicot foliage, seeds, arthropods, other) between low- and high-density populations, although animals from high-density sites tended to consume more non-legume seeds in August 1994 compared to those from low-density sites (F1,10·5 = 4·10, P = 0·069).

Figure 4.

General botanical composition of composited digesta samples from cotton rats collected seasonally from low- and high-density populations in central Oklahoma from November 1993 to November 1994.

During the non-breeding season (February and November sampling periods), animals from high-density sites ate primarily monocot and dicot foliage (stems and leaves), which occupied > 76% of the diet. During the breeding season (May and August), animals from high-density sites shifted their diet to include a much larger percentage of seeds (>46% of the diet).

Consumption of seeds on low-density sites was relatively low in February 1994 (23%) but increased in November (68% and 32% of the diet in 1993 and 1994, respectively). During November, animals from low-density sites ate more legume seeds than during any other season (22% and 18% of the diet in 1993 and 1994, respectively). This coincided with an increase in the availability of seeds from Sericia lespedeza, an exotic legume which was common on low-density sites but absent from high-density sites. In the early breeding season (May 1994), seeds made up 58% of diets from low-density sites, while foliage made up only 37%. In August 1994, however, consumption of seeds dropped to 26% while consumption of foliage increased to 68%.

The preference index calculated for selected categories of plants (monocots, forbs, legumes, woody plants) indicated that cotton rats on both low- and high-density sites generally preferred forbs and avoided monocots and woody plants relative to availability, while legumes were preferred when seeds were available (May 1994 on high-density sites, November 1993–94 on low-density sites; Table 3). Preference for individual plant species in diets of cotton rats changed seasonally on all study sites.

Table 3.  Percentage relative cover* (±SE), percentage of diet (±SE), and preference index (±SE; Lindroth & Batzli 1984a) for categories of plants identified in diets of cotton rats collected seasonally from low- and high-density populations in central Oklahoma from November 1993 to November 1994
 Low-densityHigh-density
 % Cover% DietPI% Cover% DietPI
  • *

    Relative cover for categories of plants was calculated as the proportion of each category out of all living cover available during each season.

  • Percentage cover of legumes from August 1994 was used to estimate available cover of legumes in November (when seeds were consumed) because cover of living legumes in November could not be used to estimate the abundance of seeds from senescent plants.

November 1993
Monocots45·6 ± 15·142·8 ± 11·30·98 ± 0·1255·1 ± 7·951·0 ± 1·40·95 ± 0·16
Forbs19·6 ± 4·131·7 ± 16·71·48 ± 0·598·8 ± 0·545·3 ± 2·45·17 ± 0·54
Legumes*3·6 ± 1·521·9 ± 21·914·62 ± 14·62 4·5 ± 1·7 3·8 ± 3·80·61 ± 0·61
Shrubs and trees33·9 ± 12·43·6 ± 2·50·31 ± 0·2836·1 ± 8·30 ± 00 ± 0
February 1994
Monocots39·3 ± 2·751·3 ± 4·91·33 ± 0·2379·2 ± 8·845·3 ± 12·80·56 ± 0·10
Forbs16·0 ± 1·226·7 ± 4·61·48 ± 0·597·0 ± 0·552·2 ± 12·37·41 ± 1·29
Legumes0 ± 01·8 ± 1·8 0 ± 02·4 ± 0·7 
Shrubs and trees44·7 ± 2·920·2 ± 11·00·47 ± 0·2413·9 ± 8·30·1 ± 0·10·02 ± 0·02
May 1994
Monocots60·4 ± 15·427·2 ± 15·10·55 ± 0·2355·7 ± 0·347·2 ± 13·10·85 ± 0·24
Forbs9·2 ± 0·467·8 ± 16·26·49 ± 2·2311·5 ± 1·840·3 ± 5·43·53 ± 0·09
Legumes4·6 ± 3·30·2 ± 0·20·03 ± 0·032·7 ± 2·19·9 ± 8·53·14 ± 0·67
Shrubs and trees24·1 ± 10·70 ± 00 ± 030·1 ± 4·30 ± 00 ± 0
August 1994
Monocots59·3 ± 8·740·9 ± 13·70·61 ± 0·2055·8 ± 1·139·4 ± 6·70·71 ± 0·14
Forbs15·6 ± 1·752·7 ± 15·83·86 ± 1·6714·0 ± 0·955·7 ± 8·14·03 ± 0·83
Legumes3·5 ± 1·10·1 ± 0·10·05 ± 0·054·5 ± 1·70·9 ± 0·10·25 ± 0·10
Shrubs and trees21·7 ± 10·40·3 ± 0·30·02 ± 0·0225·7 ± 1·50 ± 00 ± 0
November 1994
Monocots41·7 ± 8·525·2 ± 10·51·14 ± 0·7776·1 ± 2·339·8 ± 00·53 ± 0·02
Forbs31·4 ± 9·752·4 ± 16·12·64 ± 0·8511·3 ± 0·653·4 ± 4·44·77 ± 0·66
Legumes2·9 ± 1·018·9 ± 12·25·85 ± 3·544·5 ± 1·70·6 ± 0·60·09 ± 0·09
Shrubs and trees28·1 ± 11·22·3 ± 2·30·28 ± 0·2811·9 ± 3·75·1 ± 5·10·62 ± 0·62

Arthropods were present in diets of cotton rats from low- and high-density sites in all seasons except November 1993 (all sites) and February (low-density sites; Fig. 4). Consumption of arthropods was < 5% of the diet during all seasonal collections and did not differ between sites.

Diet quality

Wet mass of stomach digesta was 2-fold greater in cotton rats collected from high- compared to low-density populations in February 1994 (Table 4). This was also reflected in the moisture content of diets, which was 21% greater in February 1994 on high-density sites. No differences in dry mass of digesta or in fat content were observed between low- and high-density populations. Crude protein was 45% greater in diets of animals from high-density populations in August 1994, but was similar between low- and high-density populations in other seasons. Non-protein nitrogen was 2- to 3-fold greater in diets from high-density populations during the non-breeding season but tended to be lower during the breeding season compared to diets from low-density populations.

Table 4.  Seasonal changes in measures of diet quality (mean ± SE) from low- and high-density cotton rat populations in central Oklahoma from November 1993 to November 1994. Within each season, P is given in brackets for comparisons of parameters of diet quality between low-and high-density populations. Data were arcsine transformed prior to statistical analysis(no transformation was used for wet mass)
Nutrient
category
Density
H = high
L = low
November
1993
NL = 3/NH = 2
February
1994
NL = 3/NH = 2
May

NL = 2/NH = 2
August

NL = 4/NH = 2
November

NL = 5/NH = 2
  1. abc For the entire study, statistically significant effects of density, season, and density × season are indicated by a, b, and c, respectively (P < 0·05).

Wet mass (g)bcL3·12 ± 0·793·36 ± 0·673·23 ± 0·732·55 ± 0·512·58 ± 0·51
H3·00 ± 0·607·39 ± 0·673·84 ± 0·582·64 ± 0·572·72 ± 0·57
  [0·905][0·001][0·522][0·913][0·863]
MoistureL73·4 ± 2·674·9 ± 2·377·0 ± 2·677·0 ± 2·677·7 ± 1·9
content (%)abcH74·4 ± 2·390·5 ± 2·481·4 ± 2·372·7 ± 2·378·5 ± 2·3
  [0·832][<0·001][0·204][0·139][0·417]
Fat contentL5·0 ± 3·67·7 ± 3·05·2 ± 3·511·1 ± 2·714·1 ± 2·7
(% dry mass)H5·4 ± 2·92·0 ± 3·18·2 ± 2·916·3 ± 2·98·4 ± 2·9
  [0·963][0·124][0·586][0·348][0·285]
Crude proteinL22·8 ± 2·917·5 ± 2·618·3 ± 3·119·1 ± 2·320·1 ± 2·2
(% dry mass)H19·5 ± 2·816·2 ± 2·924·1 ± 2·827·2 ± 2·815·8 ± 2·8
  [0·486][0·740][0·193][0·057][0·235]
Non-essential amino acids (% dry mass)
Aspartic acidL2·52 ± 0·272·02 ± 0·231·87 ± 0·261·97 ± 0·202·18 ± 0·20
H2·18 ± 0·222·07 ± 0·232·60 ± 0·222·53 ± 0·221·79 ± 0·22
  [0·413][0·804][0·072][0·108][0·181]
Glutamic acidbcL3·88 ± 0·452·39 ± 0·372·50 ± 0·422·41 ± 0·333·25 ± 0·33
H3·17 ± 0·341·65 ± 0·373·64 ± 0·344·77 ± 0·342·11 ± 0·34
  [0·298][0·134][0·078][0·003][0·051]
SerinebL1·01 ± 0·150·72 ± 0·130·91 ± 0·160·88 ± 0·120·88 ± 0·12
H0·79 ± 0·150·55 ± 0·151·17 ± 0·151·27 ± 0·150·65 ± 0·15
  [0·424][0·316][0·300][0·091][0·234]
GlycinebcL1·15 ± 0·140·87 ± 0·111·01 ± 0·131·02 ± 0·101·09 ± 0·10
H0·90 ± 0·110·73 ± 0·111·42 ± 0·111·49 ± 0·110·81 ± 0·11
  [0·073][0·356][0·059][0·027][0·081]
AlaninebL1·10 ± 0·190·96 ± 0·171·08 ± 0·201·19 ± 0·151·03 ± 0·14
H1·09 ± 0·190·76 ± 0·191·45 ± 0·191·77 ± 0·190·91 ± 0·19
  [0·994][0·393][0·227][0·052][0·630]
ProlinebcL1·40 ± 0·180·96 ± 0·061·29 ± 0·321·13 ± 0·121·22 ± 0·12
H1·23 ± 0·130·76 ± 0·141·40 ± 0·131·78 ± 0·130·92 ± 0·13
  [0·310][0·282][0·377][0·012][0·090]
Total essentialL10·49 ± 1·587·97 ± 1·438·67 ± 1·689·05 ± 1·269·37 ± 1·22
amino acidsH8·51 ± 1·565·72 ± 1·6112·25 ± 1·5613·23 ± 1·566·99 ± 1·56
(% dry mass) [0·472][0·243][0·165][0·087][0·245]
Non-proteinL5·5 ± 2·77·0 ± 2·25·3 ± 2·56·4 ± 1·94·0 ± 1·9
nitrogenH11·9 ± 1·926·0 ± 2·12·9 ± 1·93·9 ± 1·911·1 ± 1·9
(% of total [0·012][<0·001][0·180][0·070][0·001]
N pool)abc

During the non-breeding season, we observed no significant differences in the concentrations of non-essential amino acids between diets from low- and high-density study sites (although they tended to be higher on low-density sites; Table 4). In the August 1994 breeding season, however, concentrations of three out of six non-essential amino acids were significantly greater in diets of animals from high-density populations. A similar pattern was observed for the essential amino acids; concentrations of total essential amino acids tended to be higher on high-density sites during the breeding season but were similar between low- and high-density sites during other seasons (Table 4). Concentrations of arginine and leucine were >40% higher in high-density populations during the August 1994 breeding season (Fig. 5a,b); concentration of phenylalanine + tyrosine and methionine + cystine also tended to be higher in high-density populations in August 1994 (Fig. 5c). Concentrations of essential amino acids were similar between sites during the non-breeding seasons, except for arginine, which was higher on low-density sites (Fig. 5).

Figure 5.

Mean (±SE) concentrations of essential amino acids in composited digesta samples from cotton rats collected seasonally from low- and high-density populations in central Oklahoma from November 1993 to November 1994. Within each season, P is given for comparisons of amino acid concentrations between low- and high-density populations. Data were arcsine transformed prior to statistical analysis.

Concentrations of total phenolics in the diets of cotton rats were 2-fold greater in low-density populations in November 1993 and August 1994 compared to those from high-density populations (Fig. 6). There were no significant differences in concentrations of total phenolics between low- and high-density populations during other census periods.

Figure 6.

Mean (±SE) concentrations of total phenolics in composited digesta samples from cotton rats collected seasonally from low- and high-density populations in central Oklahoma from November 1993 to November 1994. Within each season, P is given for comparisons of concentrations of total phenolics between low- and high-density populations. Data were log-transformed prior to statistical analysis.

Multivariate statistical analysis demonstrated that the overall essential amino acid composition of diets was different between individual digesta composites from low- compared to high-density sites during most seasons. Within each season arginine, methionine ± cystine, and/or leucine were selected as important discriminators of population density by stepwise discriminant analysis. Using these variables, discriminant function analysis classified digesta composites into their appropriate low- or high-density categories with an overall accuracy rate of 79% in November 1993, 92% in August 1994, and 100% in February, May and November 1994. Canonical discriminant analysis provided canonical variate scores for individual digesta composites based on a linear combination of nine essential amino acids (Table 5). A graphical representation of the canonical variate scores for digesta composites shows a distinct separation between low- and high-density populations (Fig. 7).

Table 5.  Coefficients used to calculate canonical variate scores for digesta composites of cotton rats collected seasonally from low- and high-density populations in central Oklahoma from November 1993 to November 1994
 Canonical variate
VariableNovember
1993
February
1994
May
August
November
Lysine−0·232−0·030 0·154−0·120−0·216
Phenylalanine ± tyrosine−0·229 0·701−0·153 0·075−0·241
Methionine ± cystine 0·600−0·256 0·272 0·057−0·200
Valine−0·574 0·750−0·266 0·552−0·177
Isoleucine 0·495−0·506−0·184−0·175 0·628
Leucine−0·088−1·000 0·716 0·093 0·071
Histidine−0·122−0·494−0·192 0·080−0·144
Arginine−0·114 1·412 0·130−0·040 0·674
Threonine 0·158−0·111−0·199−0·370−0·170
Figure 7.

Plot of canonical variate scores for individual digesta composites from cotton rats collected seasonally from low- and high-density populations in central Oklahoma from November 1993 to November 1994. The canonical variate is a linear combination of nine essential amino acids.

Discussion

Although other environmental factors may be involved in driving rhythmic fluctuations or cycles in herbivore populations, the abundance of quality food resources often dictates potential peak densities. Perhaps the most important factor in determining food quality for herbivores is the availability of nitrogen-containing nutrients (protein; White 1993). Previous studies of cotton rat populations have shown increases in density (Doonan & Slade 1995) and habitat affinity (Eshelman & Cameron 1996) in response to habitats supplemented with protein or foods rich in protein. Prairie voles (Microtus ochrogaster), which are considered an ecological equivalent to cotton rats in the northern prairies of North America, have also responded to greater food availability with increases in population density (Cole & Batzli 1978, 1979; Desy & Batzli 1989).

Seasonal changes and interpopulational differences in the botanical composition of diets consumed by cotton rats in this study clearly demonstrate the generalist foraging strategies utilized by this grassland herbivore (Randolph, Cameron & Wrazen 1991). These observations also demonstrate that cotton rats, like other small herbivores (Belovsky 1986), do not make dietary choices randomly. When availability in the habitat is considered, they portray a reliance on typically high-protein dicots (Mattson 1980) compared to monocots. As noted in several studies by Cameron and his associates (Kincaid & Cameron 1985; Randolph et al. 1991; Cameron & Eshelman 1996), cotton rats appear to rely heavily on dicots in their diet to obtain sufficient protein for life processes. Indeed, cotton rats from our study sites included a large proportion of dicots in their diet during most seasons. We observed that seeds (especially legume seeds) were also an important item in diets of cotton rats, and that dietary protein was >30% dry mass when legume seeds were consumed. As has been shown by Kincaid & Cameron (1982), this herbivore consistently consumed a variety of arthropods (≈5%), which may be an extremely valuable source of essential amino acids (Peoples et al. 1994). The significance of this is readily apparent in the work of Campbell & MacArthur (1996), which showed that the herbivorous muskrat (Ondatra zibethicus) can meet its maintenance requirement for nitrogen by consuming diets containing as little as 3% animal tissue. Animal tissue is also a notable component of vole and lemming diets (Batzli 1985).

It is estimated that minimum protein levels in the diet necessary for reproduction in cotton rats is about 11% (dry mass basis) when consuming natural forages (Randolph, Cameron & McClure 1995; Hellgren & Lochmiller 1997); requirements for growth are unknown (see Cameron & Eshelman 1996). Based on this estimate, cotton rats in our study from both low- and high-density populations appeared to consume sufficient dietary protein to support reproduction throughout the year (see Table 3). However, the 11% estimate assumes that animals are consuming a protein source that contains a proper balance of the required essential amino acids. Proteins in natural forages are known to be highly variable and frequently deficient in selected essential amino acids, especially the sulphur-containing nutrients methionine + cystine, relative to an animal's requirements for maximum growth and reproduction (Thomas & Prevett 1980; Sedinger 1984; Peoples et al. 1994). This also may be true of forages consumed by cotton rats in central Oklahoma.

Based on our earlier observations with cottontail rabbits (Lochmiller et al. 1995), we hypothesized that diets of cotton rats may be limited in selected essential amino acids, particularly the sulphur-containing amino acids methionine + cystine. It was our contention that an observation of greater levels of essential amino acids in the diets of cotton rats from high-density populations compared to those from low-density populations would support the nitrogen limitation hypothesis. Cotton rats in central Oklahoma reproduce primarily between late spring and early autumn and rarely breed at other times of year (McMurry 1993; McMurry et al. 1994; R.L. Lochmiller, unpublished data). During the period of peak breeding activity in our study (the August census period), dietary concentrations of two essential amino acids were greater by as much as 43% in high-density populations compared to low-density populations. Two other essential amino acids also tended to follow this pattern, including the sulphur amino acids (methionine + cystine), which have been found to be the most limiting essential amino acids in diets of eastern cottontails (Lochmiller et al. 1995) and a variety of avian species (Murphy 1994)

Also supporting the nitrogen limitation hypothesis were the higher concentrations of total phenolic compounds in the diets of animals from low- compared to high-density populations. Phenolics act either as toxins or as protein digestive inhibitors and thus decrease the quality of forage consumed by herbivores (Lindroth & Batzli 1984b). Phenolics influence forage selection (Jung & Batzli 1981; Bucyanayandi & Bergeron 1990) and have the potential to negatively influence growth and survival in voles (Jung & Batzli 1981; Lindroth & Batzli 1984b); they probably have the same effect on cotton rats. In our study, concentrations of total phenolics in the diets of cotton rats from low-density sites were over 2-fold greater compared to high-density sites during the breeding season; a period when high-quality diets are especially important to meet the increased amino acid requirements for reproduction.

Protein restrictions (4% crude protein diet) imposed on female cotton rats in the laboratory have been shown to influence age at first oestrus and pregnancy rate (Cameron & McClure 1988; Cameron & Eshelman 1996). Phenotypic plasticity in litter size of cotton rats across its range has been well documented (Cameron & McClure 1988). Reproductive performance has also been shown to vary with primary productivity of habitats (Doonan & Slade 1995; Slade, McMurry & Lochmiller 1996). Although we did not observe large differences in reproductive parameters between populations in all census periods, indicators of fecundity were greater in high- than in low-density populations in the August 1993 breeding season. Additionally, total numbers of juveniles recruited into high-density populations were nearly 5-fold greater than low-density populations over the duration of the study. Differences in body mass of females in November 1993 could have been a reflection of differences in reproductive activity over the previous months; females on low-density sites bred less and conserved body reserves. However, another possible explanation is that an older age structure existed on low-density sites in November 1993 as a result of reproductive patterns the previous August (Cameron & Eshelman 1996).

Survival estimates were consistently low over the 3-month interval between census periods, so caution is warranted in interpreting the significance of the differences we did observe. With this caveat in mind, the lower 3-month survival rate between the August 1993 breeding season and November 1993 census period on low- compared to high-density sites may have reflected a combination of increased subadult dispersal and mortality of animals after the annual peak. Population densities may be maintained at levels that are in balance with the availability of nitrogen-containing nutrients in the habitat by triggering dispersal before nutrients such as sulphur-containing amino acids become severely limiting, similar to the presaturation dispersal hypothesis described by Lidicker (1975).

Cotton rats prefer habitats with dense overhead cover (Goertz 1964); numerous studies have found that availability of suitable cover influences population dynamics of cotton rats (Goertz 1964; Fleharty & Mares 1973; Kincaid, Cameron & Carnes 1983). We observed more fallen litter on high-density sites than on low-density sites (particularly in February 1994), but the importance of fallen litter as overhead cover is difficult to evaluate. Standing crop biomass of monocots and dicots, which provides important overhead cover, was also greater on our high-density sites in August 1994 during peak population densities. Although this could be interpreted as a measure of greater nutrient availability in support of the nitrogen limitation hypothesis, it remains possible that this affords greater protection from predation. Eshelman & Cameron (1996) investigated the interaction between the effects of cover and protein availability on habitat use by cotton rats and found that both factors were important.

It is important to stress that the populations of cotton rats we studied are largely stable over time and only show major seasonal fluctuations, which peak during late summer. As with most small mammal populations in the central Great Plains, they do not show classic multi-annual increases and decreases in density. Thus, long-term population growth rates (r = annual intrinsic rate of increase) are not relevant in these situations, and the primary question becomes, what determines the peak density or amplitude of seasonal fluctuations? Perhaps we could best address this question by borrowing the concept of habitat carrying capacity as used by resource managers. As demonstrated by Hobbs & Swift (1985) using an algorithm for estimating maximum supportable densities of herbivores in a habitat, the interaction of both forage quality and quantity in the habitat is an important determinant. We would further add that the timing of nutrient limitations relative to life-history events is also important; for example, nutrient levels during the breeding season may be most important in determining peak density in a seasonally fluctuating herbivore population that shows little multi-annual changes in density. Our observation of no difference in rate of per capita recruitment between low- and high-density sites would be expected if populations are basically stable over time.

Acknowledgements

We greatly appreciate the particularly helpful comments of G. N. Cameron and two anonymous referees on earlier drafts of this manuscript. We also acknowledge the excellent assistance that was provided by the many undergraduate and graduate student volunteers in both the field and laboratory. This research was supported in part with funds from the National Science Foundation (IBN-9318066, BSR-8657043), Oklahoma Agricultural Experiment Station, Oklahoma Cooperative Fish and Wildlife Research Unit (Oklahoma State University, Oklahoma Department of Wildlife Conservation, US Geological Survey, Wildlife Management Institute Cooperating), and Department of Zoology, Oklahoma State University.

Received 12 March 1997;revision received 21 December 1997

Ancillary