Are competitive interactions influenced by spatial nutrient heterogeneity and root foraging behavior?


  • Kristin M. Bliss,

    1. Department of Biology, Virginia Tech, Blacksburg, VA 24061 USA;
    2. Present address: Biology Department, Randolf-Macon Women’s College, Lynchburg, VA 24503, USA
    Search for more papers by this author
  • Robert H. Jones,

    Corresponding author
    1. Department of Biology, Virginia Tech, Blacksburg, VA 24061 USA;
      Author for correspondence: Robert H. Jones Tel: +1 540 231 9514 Fax: +1 540 231 9307 Email:
    Search for more papers by this author
  • Robert J. Mitchell,

    1. Joseph W. Jones Ecological Research Center, Route 2, Box 2324, Newton, GA 31770, USA;
    Search for more papers by this author
  • Paul P. Mou

    1. Department of Biology, University of North Carolina at Greensboro, Greensboro, NC 27402, USA;
    Search for more papers by this author

Author for correspondence: Robert H. Jones Tel: +1 540 231 9514 Fax: +1 540 231 9307 Email:


  •  Nutrient heterogeneity, root foraging and competitive interactions were investigated for six species native to south-eastern USA.
  •  Monocultures, two- and six-species garden plots were fertilized to create spatially homogeneous or heterogeneous nutrient conditions. After 3.5 months, root proliferation in rich patches (precision), mean above-ground biomass per plant (scale) and influence of nutrient treatment on total plot biomass (sensitivity) in monocultures were measured. Competition (above-ground biomass) was assessed in two- and six-species plots.
  •  In monoculture plots, two species were relatively precise foragers, but no species showed significant sensitivity to nutrient treatment. Correlations between precision, scale and sensitivity were weak (−0.40 < r < 0.17), which contrasts with previous work showing a scale-precision trade-off. In two-species plots, competition was influenced by soil heterogeneity in two of six cases tested (anova, P < 0.05), and precise foragers grew larger in heterogeneous than in homogeneous conditions. In six-species plots, nutrient treatment had no influence on growth or competition.
  •  In our study system, heterogeneity effects on competition are context specific, generally weak and potentially mediated by the degree of root foraging precision.


Although universally acknowledged, the idea that soil resources are heterogeneous has not been fully integrated into conceptual models of plant competition. Recent models (Grime, 1977; Tilman, 1985; Wu et al., 1985; Keddy, 1989; Grace & Tilman, 1990) incorporate a number of mechanisms that control competitive outcomes and they vary considerably from descriptive to resource based, from individually to population based, and from spatially nonexplicit to explicit. However, most of them assume that soil resources are spatially homogeneous within the plant neighborhood, a condition that is not often found in nature.

Spatial heterogeneity of below-ground resources occurs in virtually all plant communities. Nitrogen, phosphorus and other key nutrients exhibit heterogeneity at scales of less than 1 m in a variety of community types, including sagebrush steppe (Jackson & Caldwell, 1993), deciduous woodland (Farley & Fitter, 1999), upland hardwood forest (Gross et al., 1995), desert (Schlesinger et al., 1996), tropical dry forest (Gonzalez & Zak, 1994) and warm-temperate conifer forest (Lister et al., 2000). Because growth of individual plants can be influenced strongly by soil heterogeneity (Cahill & Casper, 1999; Einsmann et al., 1999; Hutchings et al., 2000; but see Fransen & deKroon, 2001), it follows that interspecific competition should also be affected (Fransen et al., 2001; but see Cahill & Casper, 1999).

Root foraging behavior of individual species may influence competitive ability in heterogeneous environments. A key behavior is morphological plasticity, which is an adjustment in root system structure in response to changing environmental conditions. Two major changes in structure have been documented: precision and scale. Precision is defined as root proliferation in nutrient-rich microsites (Campbell et al., 1991; Fransen et al., 1998; Einsmann et al., 1999; Farley & Fitter, 1999; Robinson et al., 1999). Scale is total spatial extent of a root system. Rapid development of a large root system can enhance a plant’s capacity to locate nutrient-rich patches (Campbell et al., 1991; Einsmann et al., 1999). Species differ considerably in these two traits (Campbell et al., 1991; Einsmann et al., 1999); however, the extent that either trait contributes to competitive ability in heterogeneous soil is not known.

In addition to root foraging behavior, total plant biomass response to heterogeneity may predict competitive ability. Plants that have greater biomass under patchy than under homogeneous nutrient conditions, have been described as ‘sensitive’ (Wijesinghe & Hutchings, 1997; Einsmann et al., 1999). When assessed using monocultures, sensitivity is also a measure of the impact of heterogeneity on intraspecific competition. The degree of sensitivity may or may not be correlated with degree of morphological plasticity in the root system (Einsmann et al., 1999).

Understanding how roots respond to nutrient patches is a growing interest among ecologists. However, many studies have dealt with single-plant, monoculture or potted-plant responses to nutrient heterogeneity rather than responses of mixed communities in natural soil environments. Results from this previous work are conflicting. Single plants grown in isolation may be quite sensitive to nutrient heterogeneity (Cahill & Casper, 1999; Einsmann et al., 1999), whereas plants grown in multiplant mixtures may (Fransen et al., 2001) or may not (Casper & Cahill, 1996; Cahill & Casper, 1999; Casper et al., 2000) be sensitive. Here, we explore effects of nutrient heterogeneity on plant response and on intra- and inter-specific competitive outcomes for six plant species grown in garden plots. Further, we determine whether scale and precision are correlated with competitive ability (sensitivity) in spatially heterogeneous soils. According to our previous research using individual plants and monocultures, the species used in this experiment differ in precision, scale, and sensitivity, but relationships between these traits are unclear (Mou et al., 1995, 1997; Einsmann et al., 1999). We hypothesize that some species will be sensitive to heterogeneity, and that the degree of sensitivity will be related to either precise foraging or rapid growth (i.e. large scale). We also predict that heterogeneous nutrient conditions can shift interspecific competitive interactions by enhancing the growth of either precise foragers or rapidly growing species.

Materials and Methods

Species and study site

We chose six early to mid successional species that co-occur in warm-temperate, coastal plain forests of the south-eastern USA. Three are annuals (Chamaecrista nictitans (L.) Moench, Hypericum gentianoides L., Erechtites hieracifolia (L.) Raf.), one is a perennial herb (Solidago altissima L), and two are trees (Pinus taeda L. and Liquidambar styraciflua L.). During the first year of growth after seed germination, E. hieracifolia and S. altissima grow rapidly to one or more meters high, H. gentianoides and C. nictitans develop into shorter (about 0.5 m tall), dense, shrub-like forms, and the two trees (L. styraciflua and P. taeda) produce even shorter stems (< 0.5 m tall) with narrow crowns.

In October 1997, seeds of the four herbaceous species were collected from early successional plant communities at Savannah River Site, Aiken and Barnwell counties, South Carolina, USA. Tree seeds were obtained from a seed orchard at the Flint Nursery in Byromville, GA, USA. The orchard trees have been selected for their superior timber phenotypes, but cones are open-pollinated, and thus, parentage for each seed includes orchard and nearby wild trees.

The study was conducted at the Joseph W. Jones Ecological Research Center at Ichauway, which is located in the gulf coastal plain of Georgia (Baker County), USA. In January 1998, four 15 × 20 m blocks were established in an old agricultural field. The study was blocked to control for differences in soil conditions because of a limestone outcrop in the center of the field site. Soils were Lakeland Series (sandy, thermic, coated, Typic Quartzipsamments). The site had previously been used for agriculture and contained rye (Secale cereale L.). The area had lain fallow for several years. After the blocks were disked to a depth of 20 cm and rolled to repack the soil, all vegetation and seeds were killed by fumigation with methyl bromide (98%) and chloropicrin (2%).

Plant preparation and establishment

From January 12 until March 1 seeds were germinated in a greenhouse at Virginia Polytechnic Institute and State University, Blacksburg, VA, USA. When plants had attained sufficient size to survive transplantation (4 cm high, two leaves), seedlings were planted into plugs (2.5 × 2.5 m wide and 6.3 cm deep) in a 50 : 50 (by volume) mixture of Metro Mix 200 (Scott’s Sierra Horticulture Products, Marysville, OH, USA) and sterilized sand. To inoculate with native microbes, 1.5 g of soil from the Georgia field site was added to each plug. Plants were watered twice daily, ensuring that the growing medium was kept near field capacity, and fertilized twice in March (2, 16) with a liquid 20-20-20 fertilizer solution (100 µg g−1).

In late February, square plots (90 × 90 cm and 120 × 120 cm) were established in each block for intraspecific and interspecific competition treatments, and for the mixed community experiment (Table 1, Fig. 1). During 26–28 March, seedlings were transported to the Georgia site and planted within plots using a 15 cm spacing. Cylindrical plugs of soil (8.5 cm deep by 2.5 cm wide) were removed from the plots using a sink tube to make uniform holes, and plants were placed within the holes with their root system and surrounding soil intact (i.e. to minimize transplant shock). Soil was gently compacted around the plants.

Table 1.  Size and configuration of garden plots used to test the influence of nutrient heterogeneity on competition
Competition typeTarget plantsMatrix plantsTotal plantsPlot dimensions (cm)Total number of plots
Intraspecific  940  49  90 × 9048
Two species  972  81120 × 12048
Six species3664100150 × 150  8
Figure 1.

Configurations for garden plots to test soil nutrient heterogeneity influences on competition. (a) An intraspecific competition plot. Nine target plants (T) are surrounded by border plants (X) of the same species. In two-species plots (not shown), each target plant is completely surrounded by a matrix of another species, and thus no target plant is adjacent to a conspecific plant. (b) A 10 × 10 six-species plot consisting of an inner 6 × 6 area where species (1–6) are randomized by Latin Square design, and an outer matrix consisting of two rows of border plants (X) which were a random mixture of the six study species. All intraspecific and interspecific plots are treated with uniform nutrient distribution (homogeneous treatment, not shown), or with patches of fertilizer placed such that each plant was 10.8 cm from the closest patch (heterogeneous treatment shown by dark squares in b).

Over the course of the experiment less than 3 cm of rain fell. From May until mid July, 25–30 l of water (depending on plot size) were uniformly applied by hand to each plot. Plants were watered every third day to minimize drought effects. Despite this watering regime, E. hieracifolia appeared drought-stressed, and it flowered and began senescing much earlier than expected. Therefore, the entire experiment was harvested on 15–17 July after 3.5 months of growth.

Nutrient treatments

Nutrients were distributed across both target plant and mixed community experiments in two arrays: homogeneous and heterogeneous. Each provided a mean of 2.1 g m−2 of Osmocote slow release fertilizer (17-9-12 plus minors, Scott’s Sierra Horticulture Products, Marysville, OH, USA). The nutrient release rate from this quantity of fertilizer is comparable to natural N mineralization rates found in upland, sandy, coastal plain soils (Burger & Pritchett, 1984; Bell & Binkley, 1989). In homogeneous plots, fertilizer was tilled evenly into the surface 3 cm of soil across the whole plot including a 15-cm buffer outside the plot. In heterogeneous treatments the same total fertilizer was mixed into the top 3 cm of 5 × 5 cm patches representing c. 1% of surface area. Enriched patches fell at regular spatial intervals at the intersections of a 15 × 15 cm grid (Fig. 1). The size and configuration of patches were chosen to provide each plant an equal chance of reaching a patch (i.e. each plant was 10.8 cm from the center of a nutrient patch; Fig. 1), and to create nutrient concentrations that were realistic (i.e. two orders of magnitude greater in patches than in the homogeneous treatment).

Monoculture and two-species plots

Two monoculture plots (one with heterogeneous the other with homogeneous nutrient treatment) were established in each block. In each monoculture plot, the center nine plants were designated as target plants (Table 1; Fig. 1). To limit the total number of two-species plots to a manageable size, we used data from a previous study (Einsmann et al., 1999) to choose two-species comparisons that were most likely to provide strong tests of our hypotheses. These pairs were: annual × annual (E. hieracifolia × C. nictitans), perennial × tree (S. altissima × P. taeda), and annual × tree (E. hieracifolia ×P. taeda). We used different life-history combinations to provide variety in our tests, and not to test the effects of life history categories per se. In each two-species plot, nine target plants of one species were each surrounded by neighbors (i.e. a matrix) of the other species. Both species were targets in half of the plots and used as the matrix species in the others, for a total of four plots per block (i.e. target and matrix × two nutrient treatments).

After harvest, oven-dry biomass (60°C) of the entire shoot was determined for each target plant, and for all matrix plants in the plot combined. In monoculture plots only, three root cores (10.2 cm wide and 20 cm deep) were harvested with a bucket auger from random locations in homogeneous plots, and then pooled. Three cores each were also collected and pooled from randomly selected nutrient-rich and nutrient-poor locations within heterogeneous plots. The pooled samples were rinsed through a 2-mm mesh screen, and then roots were picked out from the remaining soil and debris. Roots were dried at 60°C for 3 d and then weighed.

Six-species plots

In this experiment, a double border of plants surrounded an inner six by six grid containing the six species arranged in a Latin Square design (Table 1, Fig. 1). Each block included two such plots corresponding to the two nutrient treatments. At harvest, above-ground portions of the inner 36 plants were collected and then dried to determine total oven-dry biomass for each species (all plants of each species pooled).

Soil conditions and leaf area

Soil moisture was measured using time domain reflectometry following Topp et al. (1980). In each plot, one pair of 20-cm long stainless steel rods was placed in a 15-cm wide unplanted buffer region surrounding the seedlings. Soil moisture readings were recorded for 9 d, which were chosen to include three watering cycles. On three occasions, soil moisture was measured after watering on the day that watering occurred, and for the next two consecutive days. Soil temperature at a depth of 10 cm was measured with a soil thermometer at all plots on six consecutive dates (29 May−3 June).

Six weeks after plot establishment, anion and cation exchange membranes (3.8 × 3.8 cm) were inserted into the top 5 cm of soil in all plots. In heterogeneously fertilized plots, three pairs of membranes (one NH4 and one NO3 membrane per pair) were placed in randomly selected enriched patches and three in unenriched locations. Three pairs were also located at random within homogeneously fertilized plots. Nutrient adsorption by the membrane is a measure of nutrient availability to the plant roots (Abrams & Jarrell, 1992). Membranes were left in the plots for 72 h, and then each set of three was removed, grouped by type (NH4 and NO3), and each group was extracted with 75 ml of 2.0 m KCl. The extract was analyzed by a LACHAT QuickChem AE autoanalyser for nitrate and ammonium concentrations (Zwellinger Analytics, Inc., Milwaukee, WI, USA).

Leaf area was estimated using a Li-Cor LAI-2000 leaf area meter (Li-Cor, Lincoln, NE, USA) at all monoculture plots, 6 wk after seedlings were established and again 2 wk before harvest. This instrument estimates total leaf area by measuring light that penetrates to ground level compared with light just above the canopy.


Statistical Analysis System (SAS) software (SAS version 8.0, SAS Institute, Inc., Cary, NC, USA) was used for all statistical analyses. The experiment was a randomized complete block design, with both treatments and blocks considered as fixed effects.

Precision of root foraging was determined using monoculture, heterogeneously fertilized plots only. In one analysis, the difference between root mass in high vs low fertility soil cores in a plot was divided by the total root mass collected from the plot. This calculation is a modification of relative fine root mass difference (RFRMD) that we have used in previous experiments to indicate precision (Mou et al., 1997; Einsmann et al., 1999). We conducted an anova to test for species differences in RFRMD, and for each species, we also calculated confidence interval estimates for RFRMD, and inspected the intervals for overlap of zero (the expectation for no precision). A second analysis of precision included two factors, species and fertility patch within plots (high vs low). A split-plot anova was used with these data to detect species and nutrient patch effects on root biomass. In this analysis, species were the whole plot factor and fertility treatment was the split plot factor.

Scale and sensitivity analyses were conducted using above-ground biomass in monocultures. Differences among species in scale (size of the root system) were assessed using a one-way anova of above-ground biomass in homogeneous monoculture plots. Unpublished data from Einsmann et al. (1999) showed that below-ground mass was linearly related with above-ground mass (regression; P < 0.001; r2 = 0.95) in potted plants of these species. We also calculated scale using leaf area index at final harvest in monoculture, homogeneous plots. To test for differences between species in sensitivity, biomass in heterogeneous monoculture plots was divided by biomass in the homogeneous monoculture (separate number for each block), and these ratios were analysed using a one-way anova. In this test, differences between species would indicate that they differed in sensitivity to nutrient heterogeneity. We also evaluated sensitivity by using a two-way anova to test for species, fertility treatment and species–fertility interaction effects (significant fertility effects in this analysis would indicate sensitivity). We calculated correlation coefficients between our indices of precision (RFRMD), scale (above-ground biomass), and sensitivity (ratio of biomass in heterogeneous to homogenous plots) to determine whether species show evidence for trade-offs between traits.

In two-species plots, nutrient heterogeneity effects on interspecific competition were tested using a one-way anova for each target and matrix plant combination. A two-way anova that included species as a factor was not possible because different target species were not exposed to the same matrix species (i.e. they did not have a common ‘treatment’ applied to them). In each test, only target plant biomass was used as the response variable. In six-species plots, we used a two-way anova with nutrient heterogeneity, species, heterogeneity × species and block effects.

The effects of spatial nutrient treatments on the six-species communities were also analyzed using Goldberg’s competition index (D = Ryim − RYix), a measure of each species’ competitive ability within the community (Goldberg, 1994). RYim is the relative yield of species i in monoculture (calculated by the yield for a species in monoculture divided by the sum of the monoculture yields for all species). RYix, the relative yield of species i in mixture, is calculated similarly, with the yield for a species in mixture divided by the sum of all monoculture yields for all species. If D is zero, interspecific and intraspecific competition are equivalent for that species. One D-value was calculated for each species in each nutrient treatment within each block. D-values were analysed using a two-way anova with species and nutrient treatment (heterogeneous vs homogeneous) as the main effects. We also calculated the absolute value of D for each species and summed these values for each plot. This is an overall index of community level change describing the effect of interspecific competition (Goldberg, 1994). A one-way anova was used to test for the influence of nutrient arrangement on this sum of D index.


Soil conditions

Moisture in the top 20 cm of soil ranged from 13 to 33% across all plots on all sampling dates. Mean soil moisture was 20% at the driest point of the watering cycle, and 24% at the wettest point. Moisture was uniform across fertility treatments (Table 2) and across all types of plots including monoculture, two-species and six-species plots (data not shown). Soil temperature averaged 32.2°C, and no significant differences between nutrient treatments (Table 2) or between different types of plots were detected (data not shown).

Table 2.  Soil measures (n = 104 plots) and leaf area index (n = 48 monoculture plots only) measured in garden plots, showing contrasts between heterogeneous and homogeneous nutrient treatments, and between high and low fertility patches within heterogeneously fertilized plots
VariableMeasurement date or locationHeterogeneousHomogeneous
  1. Within variables, measures with different superscripts are significantly different at P < 0.05.

Soil moisture (%)Mean of nine dates20.86a  3.8021.43a  3.50
Soil temperature (°C)Mean of six dates32.15a  0.4832.16a  0.49
Ammonium (µg g−1)High fertility85.37a65.79  8.09b  8.05
Low fertility  5.24b10.35  
Nitrate (µg g−1)High fertility60.23a49.7512.68b11.92
Low fertility  5.05b  4.84  
Leaf area index (m2 foliage m−2 ground)28–30 May  0.86a  0.52  0.97a  0.56
11–13 June  1.27a  0.54  1.42a  0.67

Mean ammonium and nitrate values 6 wk after planting indicated that patches within heterogeneous plots were distinctly different (Table 2). Nutrient levels in the homogeneous treatment were intermediate, but were not significantly different from the nutrient poor patches in heterogeneous plots (Table 2). Species and species × patch type (heterogeneous-high, heterogeneous-low and homogeneous) interactions within monoculture plots were not significant (anova; P > 0.49), and neither were plot type (one-, two- and six-species) or plot type–patch type interactions (P > 0.73).

General plant responses

Mean plant survival was 99.2% 4 wk after transplantation and 97% by the experiment’s end. Erechtites hieracifolia had 11% mortality, whereas all other species had less than 5%.

Plant heights differed by species (P < 0.0001) but not by treatment. At harvest, E. hieracifolia was the tallest, with a mean height of 54.0 cm, followed by S. altissima (43.7 cm), H. gentianoides (24.5 cm), C. nictitans (24.5 cm), P. taeda (17.5 cm), and L. styraciflua (15.4 cm). Leaf area index (m2 foliage m−2 ground) was not significantly affected by fertility treatment (Table 2).

Precision, scale and sensitivity

According to comparisons of confidence interval estimates, only S. altissima had a RFRMD greater than zero (Table 3), an indication that this species was a precise forager. The split-plot anova of species and soil fertility effects, however, detected significant species (df = 5,15; P = 0.004), fertility (df = 1,18; P = 0.004) and species–fertility interactions (df = 5,18; P < 0.02; Fig. 2). Thus, at least some species were precise (indicated by the significant fertility treatment effect) and species differed in the degree of precision (significant species × fertility effect). According to RFRMD values, the most precise species were S. altissima and L. styraciflua, and the least precise were E. hieracifolia and H. gentianoides (Table 3).

Table 3.  Plant responses measured in monoculture plots
RFRMD Mean (CI95%)RankAbove-ground biomass (g)Biomass rankLeaf area index (LAI)LAI rankBiomass ratioRank
  1. RFRMD (root mass in high fertility − root mass in low fertility cores)/(sum of mass in both cores); Sensitivity is above-ground biomass in heterogeneous plot within a block divided by biomass in homogeneous plot in same block; Within a column, superscript letters that are different indicate significance (anova followed by the Ryan–Einot–Gabriel–Welsh (REGW) Multiple range test, P < 0.05).

Solidago altissima    0.419a (0.081−0.756)18.25a12.05a10.97a4
Liquidambar styraciflua    0.322a (−0.003−0.647)21.41c51.12ab40.92a5
Chamaecrista nictitans    0.220a (−0.074–0.514)34.66b41.90a20.98a3
Pinus taeda    0.046a (−0.352−0.445)41.31c60.58b61.11a1
Erechtites hieracifolia −0.003a (−0.613–0.607)56.73ab21.72a30.91a6
Hypericum gentianoides −0.017a (−0.273–0.239)64.82b31.08ab51.02a2
Figure 2.

Root dry mass in soil cores collected from high (filled bars) and low (emty bars) fertility patches in monoculture listed left to right in descending relative fine root mass difference (RFRMD): Sa, Solidago altissima; Ls, Liquidambar styraciflua; Cn, Chamaecrista nictitans; Pt, Pinus taeda; Eh, Erechtites hieracifolia; Hg, Hypericum gentianoides. Error bars are standard errors based on n = 4 blocks.

Species differed significantly in scale. Based on both above-ground biomass and leaf area, S. altissima ranked first (largest) in scale and P. taeda ranked last (Table 3). Ranks for the remaining species varied according to measure used, but trees were always small relative to herbs (Table 3).

No species were sensitive according to our one-way and two-way anovas (P > 0.25). Mean ratios of biomass in heterogeneous/homogeneous plots, our index of sensitivity, had a narrow range of only 0.92–1.11 (Table 3). When analyzed across species, correlations were weak and not significant between precision and scale (r = 0.17; P = 0.75), precision and sensitivity (r = −0.34; P = 0.51), and scale and sensitivity (r = −0.40; P = 0.43).

Competition in two-species plots

The spatial arrangement of nutrients had a significant effect on competitive interactions in two of the two-species mixtures. Both S. altissima and P. taeda target plants grew larger in the heterogeneous treatment than in the homogeneous treatment (anova; df = 1,3; P < 0.05). Precision was related more to these results than were scale or sensitivity. In both cases, the target species was more precise than the matrix species. Furthermore, in three of the remaining four cases tested (all but E. hieracifolia grown as a target with P. taeda) the more precise forager had larger biomass in the heterogeneous treatment, while the less precise forager had larger biomass in the homogeneous treatment (Fig. 3), although these results were not statistically significant. When these results were compared with scale ratings (Table 3), the two significant cases conflicted (i.e. the larger-scale plant gained by heterogeneity in the S. altissimaP. taeda pair, but the opposite occurred for the P. taedaE. hieracifolia comparison). Furthermore, in the remaining four tests, the larger-scale species gained biomass in two cases, and lost biomass in the other two cases. Finally, the two species with relatively low sensitivity rank in monoculture (S. altissima and E. hieracifolia) were at opposite ends of the spectrum in terms of response to heterogeneity in two-species mixtures.

Figure 3.

Above-ground biomass for target species in competition with a matrix species. Notation is target species (matrix species). Error bars are standard errors based on n = 4 blocks. Asterisks indicate a significant difference in biomass between heterogeneous (filled bars) and homogeneous (empty bars) treatments. Sa, Solidago altissima; Ls, Liquidambar styraciflua; Cn, Chamaecrista nictitans; Pt, Pinus taeda; Eh, Erechtites hieracifolia; Hg, Hypericum gentianoides.

Competition in six species plots

In six-species plots, species had different total biomass (anova; df = 5,33; P < 0.001), but the heterogeneity treatment was not significant (df = 1,33; P = 0.744) and neither was the heterogeneity–species interaction (df = 1,33; P = 0.948). Competitive ability measured by Goldberg’s D index differed between species (anova; df = 5,31; P < 0.001; two missing data points). However, again there were no heterogeneity treatment effects (df = 1,31; P = 0.379), nor were there treatment–species interaction effects (df = 5,31; P = 0.928). The two largest species (i.e. with greatest scale) performed better in mixture than predicted based on monoculture yields (Fig. 4). Therefore, mean D-values calculated for each species (heterogeneous and homogeneous plots combined) were positively correlated with scale (r = 0.79; P = 0.06). They also had a positive correlation with precision (r = 0.48; P = 0.34) and a negative one with sensitivity (r = −0.36; P = 0.48). Heterogeneity treatments had no significant impact on the sum of D-values within plots, an index of mean community change (df = 1,3; P = 0.765; D for homogeneous = 0.406  0.092 SE and for heterogeneous = 0.362  0.063 SE).

Figure 4.

Goldberg’s D index (relative yield of plants in mixture minus relative yield in monoculture) with species listed in order of decreasing scale (left to right). Sa, Solidago altissima; Eh, Erechtites hieracifolia; Hg, Hypericum gentianoides; Cn, Chamaecrista nictitans; Ls, Liquidambar styraciflua; Pt, Pinus taeda. Error bars are standard errors based on n = 4 blocks; filled bars, heterogeneous treatment, empty bars, homogeneous treatment. Nutrient arrangement was not significant (anova, P > 0.05).


Our first hypothesis, that some species will be sensitive to heterogeneity and that the degree of sensitivity will be related to precision or scale of root foraging, was not supported. No significant sensitivity was detected in any species and our index of sensitivity had a very small spread around the expected value for no sensitivity (i.e. 1.0; Table 3). Furthermore, the index of sensitivity was only weakly correlated with scale and precision. Our results therefore are consistent with other studies that have shown little or no effect of soil nutrient heterogeneity on mean biomass in multiplant monocultures (Casper et al., 2000).

Our finding that precision and scale were positively correlated (r = 0.17) contrasts with previous work suggesting a trade-off (negative correlation) between these two traits (Campbell et al., 1991), but is consistent with a study that included five of our six species (Einsmann et al., 1999). Our measure of scale, however, is imprecise and may not fully capture the concept of scale. We used above-ground biomass and leaf area as surrogate measures of below-ground biomass, and then we assumed that larger below-ground biomass would equate to greater exploitation of soil volume or horizontal area by roots (i.e. greater scale). A previous study using potted plants showed that species with high root-mass density also had high root-length density (Einsmann et al., 1999); however, we have no data on spatial extent of root systems.

The monocultures showed that plants responded to nutrient heterogeneity with proliferation in patches (precision), and that the degree of proliferation differed between species. Other studies have also shown interspecific differences in degree of precision (Caldwell et al., 1991; Mou et al., 1995; Einsmann et al., 1999). Differences in precision measured in the multiplant garden plots of our study, however, were not as striking as in an earlier greenhouse study of the same species using a single plant per pot (Einsmann et al., 1999). This suggests that presence of conspecific neighbors may decrease root foraging precision, possibly because of antagonisms between root systems of different plants.

Our second hypothesis, that under heterogeneous nutrient conditions interspecific competitive interactions would shift so that either precise foragers or rapidly growing species would benefit, was partly supported. Success in a competitive environment was determined by average above-ground biomass in the heterogeneous treatment compared with above-ground biomass in the homogeneous treatment. In two-species plots, two (of three) cases showed the more precise species of the pair gained in relative competitive ability (increased above-ground biomass) under heterogeneous conditions. In the remaining test, the same trend in means was observed, although the results were not statistically significant. In pairs where the target species was less precise, biomass tended to be higher in the homogeneous treatment (in two of three cases), although the results were not statistically significant. The gain in biomass under heterogeneous conditions was apparently associated more with precision of foraging than with rapid growth, because even species with small stature relative to their competitors (e.g. P. taeda targets growing with E. hieracifolia matrix) were larger in heterogeneous conditions than in homogeneous conditions (Fig. 3).

In the six-species experiment, no significant nutrient distribution effects were found. Scale contributed disproportionately to overall community response, while heterogeneity (treatment) effects were dwarfed by comparison. A species’ interspecific competitive ability (D) was more strongly related to scale than to any other factor we measured. The lack of any soil nutrient arrangement effects was surprising, since heterogeneity influenced competition in the two species plots.

Why were effects of below-ground root foraging evident in two species plots, and unimportant in intraspecific competition (monoculture) and six species (mixed community) plots? First, we speculate that root foraging may be less important in monocultures than in mixed-species plots. In single-species plots, an individual plant that proliferates in rich patches may not gain biomass because its neighbors employ the same strategy, thus offsetting any possible advantage. According to Casper et al. (2000), roots of neighboring plants that proliferate in patches reduce nutrient availability in the patch so that it is similar to background nutrient levels. Second, results of the study were possibly confounded by unintended variation in size asymmetry and density of individual species, two factors that influence competitive interactions (Rees et al., 1996). Although all plots had the same total density (15 × 15 cm spacing, or 44 plants m−2), size asymmetry increased, and density for individual species decreased as more species were added to a plot. Target plant density for any one species was 44 m−2 in monocultures, 16 m−2 in two-species plots, and only 6 m−2 in six-species plots. For some of the larger species, the low density of conspecific neighbors and the likelihood that neighbor plants were smaller may have led to weaker competition in the six-species plots than in either the monocultures or two-species plots. Even when plots with the same number of species are compared, the larger size of fast-growing herbaceous species may well have created more intense competitive pressure than occurred in plots dominated by trees. Thus, caution must be taken in interpreting results of our experiments and further work is needed to tease out the independent influences of species, density, and intraspecific vs interspecific effects.

We have shown that precise foraging may influence competitive interactions when soils are heterogeneous. There is indirect evidence that other foraging mechanisms may operate to enhance uptake and competitive ability in patchy environments. One species, P. taeda had greater above-ground biomass in the heterogeneous treatment when grown in competition with E. hieracifolia; however, it was not very precise, even though it was more so than E. hieracifolia. Physiological plasticity might have provided P. taeda with an advantage in patchy soil. Fransen et al. (2001) found that physiological, rather than morphological plasticity influenced relative competitive ability of two perennial grass species grown in soils with heterogeneous nutrient arrangement. Grime et al. (1986) argued that physiological plasticity should be more common in unproductive environments, while morphological plasticity should predominate in productive environments. It follows that species from relatively unproductive environments need to be examined for other root foraging traits including physiological (uptake) and demographic plasticity (turnover). In addition to physiological plasticity, there are many other facets of nutrient foraging that were not addressed by this study, yet are potentially important for explaining our results. These include within-species genetic variability, root–shoot signaling, mycorrhizal symbioses, and soil fauna (Zhang & Ford, 1998; Beveridge, 2000).

Low sample size in our study precluded a powerful test of relationships between root foraging and other plant traits. However, as we have seen in another study (Einsmann et al., 1999), precision of foraging appeared unrelated to either growth form or scale. Within the annual plants, for example, C. nictitans showed evidence of precise foraging whereas E. hieracifolia did not. Only a weak positive correlation (r = 0.17) was found between precision and scale. By contrast, a study by Campbell et al. (1991) documented an inverse relationship between scale and precision for eight herbaceous species.

In this paper, we found that individual plant responses to patchy resources can influence competitive interactions between neighboring species. Although our data suggest that morphological plasticity of root systems may be related to competitive ability, it remains to be determined how important this root foraging characteristic is relative to others.


We thank the following for help with field and laboratory assistance: Stephen Pecot, Stacy Hurst, Glen Stevens, Erica Mahar, Anna Maria Escherich, Anne Kottman, Kasey Simmons, Ginger Piper and Dwan Williams. We also thank Erik Nilsen and anonymous reviewers for advice on earlier drafts. Funding was provided by USDA-NRICGP Grants 96-35101-3452 and 99-35101-7872.