The Influence of Functional Diversity and
Composition on Ecosystem Processes
David Tilman,* Johannes Knops, David Wedin, Peter Reich,
Mark Ritchie, Evan Siemann
Humans are modifying both the identities and numbers of species in ecosystems, but
the impacts of such changes on ecosystem processes are controversial. Plant species
diversity, functional diversity, and functional composition were experimentally varied in
grassland plots. Each factor by itself had significant effects on many ecosystem pro-
cesses, but functional composition and functional diversity were the principal factors
explaining plant productivity, plant percent nitrogen, plant total nitrogen, and light pen-
etration. Thus, habitat modifications and management practices that change functional
diversity and functional composition are likely to have large impacts on ecosystem
processes.
Although the organisms living in an eco-
system control its functioning (1–4), it has
not been clear how much of this control is
determined by the identities of the species
present (4, 5), by the number of species
present (2, 4, 6, 7), by the number of
different functional roles that these species
represent (1, 2, 8), or by which functional
roles are represented (4, 9). The effects of
species or functional diversity are expected
to increase with the magnitude of the dif-
ferences among species or functional groups
(10). These differences are also expected to
influence the magnitude of the effects
caused by compositional differences. How-
ever, the relative effects attributable to di-
versity versus composition are unclear.
We performed a field experiment in
which plant species diversity (defined as
number of plant species added to plots),
functional diversity (defined as number of
functional groups added to plots), and
functional composition (defined as which
functional groups were added to plots)
were directly controlled (11). Our 289
plots, each 169 m
2
, were planted and
weeded to have either 0, 1, 2, 4, 8, 16, or
32 perennial savanna-grassland species
representing 0, 1, 2, 3, 4, or 5 plant func-
tional groups. Grassland-savanna plants
were classified into functional groups on
the basis of intrinsic physiological and
morphological differences, which influ-
ence differences in resource requirements,
seasonality of growth, and life history. Le-
gumes fix nitrogen, the major limiting nu-
trient at our site (7). Grasses with the
three-carbon photosynthetic pathway
(C
3
) grow best during the cool seasons and
have higher tissue N than do grasses with
the C
4
pathway, which grow best during
the warm season. Woody plants have high
allocation to perennial stem and low
growth rates, and forbs do not fix N and
often have high allocation to seed.
When analyzed in separate univariate
regressions, species diversity had significant
effects on plant productivity (Fig. 1A) and
on three of five other response variables
measured in the third year of study (12, 13,
14). Functional diversity significantly influ-
enced plant productivity (Fig. 1B) and all
other variables (13, 14). Species diversity
had a highly significant effect (P , 0.001)
in a one-way multivariate analysis of vari-
ance (MANOVA) that included all six re-
sponse variables, as did functional diversity
in a similar MANOVA.
In multiple regressions of each of the
six response variables on both species and
functional diversity, functional diversity
was significant in all six cases, but species
diversity was not (Table 1) (14). Plant
productivity and plant total N significant-
ly increased, and soil NO
3
, soil NH
4
, plant
percent N (% N), and light penetration
significantly decreased as functional diver-
sity increased. A two-way MANOVA that
included all six response variables showed
highly significant effects of functional di-
versity (Wilk’s lambda F 5 7.58; df 5 6,
277; P , 0.0001) but no significant effects
of species diversity (Wilk’s lambda F 5
0.12; df 5 6, 277; P 5 0.99). Similar
results were obtained in alternative anal-
yses (14), including a two-way MANOVA
that used observed species and functional
diversities from 1996 (15). Thus, the func-
tional group component of diversity is a
greater determinant of ecosystem processes
than the species component of diversity.
The independent effects of functional
composition can be tested by ANOVAs in
which each of the 32 possible functional
compositions (16) is nested within the
appropriate level of functional diversity.
There were highly significant effects of
both functional diversity (Fig. 1B) and
functional composition (Fig. 2) on plant
productivity, plant % N, plant total N,
and light penetration (Table 2). Soil NH
4
and soil NO
3
depended on functional di-
versity but not on functional composition.
Thus, for four of the six variables, both
functional composition and functional di-
versity had significant impacts. A two-way
MANOVA that included all six variables
found highly significant effects of both
functional diversity and functional com-
position (14, 17).
On average, across the six ANOVAs of
Table 1, species and functional diversity
together explained 8% of the variance in
response variables, whereas functional com-
position and diversity together explained
37% (Table 2), suggesting that composition
is the greater determinant of ecosystem
processes.
To determine if particular functional
groups were responsible for the effects of
D. Tilman, J. Knops, E. Siemann, Department of Ecology,
Evolution and Behavior, University of Minnesota, St. Paul,
MN 55108, USA.
D. Wedin, Department of Botany, University of Toronto,
Toronto, Ontario, M5S 3B2, Canada.
P. Reich, Department of Forest Resources, University of
Minnesota, St. Paul, MN 55108, USA.
M. Ritchie, Department of Fisheries and Wildlife, Utah
State University, Logan, UT 84322, USA.
*To whom correspondence should be addressed. E-mail:
tilman@lter.umn.edu
Fig. 1. (A) Dependence of 1996 aboveground
plant biomass (that is, productivity) (mean and SE)
on the number of plant species seeded into the
289 plots. (B) Dependence of 1996 aboveground
plant biomass on the number of functional groups
seeded into each plot. Curves shown are simple
asymptotic functions fitted to treatment means.
More complex curves did not provide significantly
better fits.
SCIENCE
z
VOL. 277
z
29 AUGUST 1997
z
www.sciencemag.org1300
functional diversity, we repeated the mul-
tiple regressions of Table 1, but replaced
functional diversity with five dummy
variables, each describing a functional
group as either absent from a plot or
represented by at least one species. For
each of the six ecosystem variables, there
were significant (P , 0.05) effects of the
presence of particular functional groups
and no significant effects of species diver-
sity. Only C
4
grasses and legumes signifi-
cantly affected productivity (Fig. 2) and
light penetration (P , 0.001 for each,
overall r
2
5 0.19 for C
4
grasses and 0.27
for legumes). Plant % N depended on all
five functional groups (P , 0.05 for all, r
2
5 0.57). The other ecosystem variables
were significantly dependent only on ei-
ther legumes (plant total N) or C
4
grasses
(soil NH
4
, soil NO
3
). On average, across
plots containing two, four, or eight spe-
cies, the presence of one or more C
4
grass
species led to a 40% increase in produc-
tivity, and the presence of one or more
legume species led to a 59% increase. The
greater biomass from legumes is consistent
with their ability to fix N. The greater
biomass from C
4
grasses is consistent with
their lower tissue N concentrations.
Another multiway MANOVA, in which
the five independent variables were the
species diversity within each functional
group (number of plant species within a
functional group planted in a plot) and the
dependent variables were the six ecosystem
responses, showed significant (P , 0.01)
effects of species diversity within each func-
tional group except woody plants. Thus,
both the presence of some functional groups
and the number of species within most
functional groups had significant effects on
ecosystem processes.
The increase in productivity with di-
versity was partially caused by overyield-
ing of species, especially C
4
grasses, in
high-diversity plots. Specifically, a regres-
sion for each species of log(percent cover)
on log(species richness) revealed signifi-
cant (P , 0.05) overyielding at high spe-
cies diversity (that is, slopes significantly
less negative than –1) for 14 of the 34
species, but significant underyielding at
high diversity for only four species. All
eight C
4
grasses significantly overyielded
(Andropogon gerardi, Bouteloua curtipen-
dula, B. gracilis, Buchloe dactyloides, Pani-
cum virgatum, Schizachyrium scoparium,
Sorghastrum nutans, and Sporobolus
cryptandrus), as did the C
3
grass Elymus
canadensis, the legumes Lespedeza capitata
and Petalostemum villosum, the forb Aster
azureus, and the woody plants Quercus ellip-
soidalis and Q. macrocarpa. Thus, many spe-
cies inhibited themselves in monoculture
and low-diversity plots more than they were
inhibited by other species in high-diversity
plots. This is consistent with several mech-
anisms of niche differentiation and coexist-
ence (18), suggesting that such mechanisms
may explain the increase in productivity
with diversity (10).
Other studies have shown that the
number of species (2, 6, 7, 19), the num-
ber of functional groups (8), or ecosystem
species composition (20, 21) influence
various ecosystem processes. Our results
show that composition and diversity are
significant determinants of ecosystem pro-
cesses in our grasslands. Given our classi-
fication of species into functional groups,
functional diversity had greater impact on
ecosystem processes than did species di-
versity. This suggests that the number of
functionally different roles represented in
an ecosystem may be a stronger determi-
nant of ecosystem processes than the total
number of species, per se. However, spe-
cies diversity and functional diversity are
correlated; each was significant by itself,
as was species diversity within functional
groups; and either species or functional
diversity may provide a useful gauge of
ecosystem functioning.
Our results show a large impact of com-
position on ecosystem processes. This
means that factors that change ecosystem
composition, such as invasion by novel or-
ganisms, nitrogen deposition, disturbance
frequency, fragmentation, predator decima-
tion, species extinctions, and alternative
management practices (20, 21), are likely
to strongly affect ecosystem processes. Our
results demonstrate that all species are not
equal. The loss or addition of species with
certain functional traits may have a great
impact, and others have little impact, on a
particular ecosystem process, but different
processes are likely to be affected by differ-
ent species and functional groups.
Fig. 2. Effects of functional composition on 1996
aboveground plant biomass (productivity) in plots
containing at least one legume species (Legume),
at least one C
4
grass species (C
4
grass), at least
one of each (C
4
grass plus legume), or only spe-
cies from other functional groups (Other). Mean
and SE are shown, using all plots containing 1, 2,
or 3 functional groups.
Table 1. Dependence of ecosystem variables on diversity treatments as determined by multiple
regression. Values shown are regression parameters. A separate regression was performed for each
ecosystem variable. Regressions have df 5 2, 283 to 2, 286. NS, P . 0.05; *, 0.05 $ P . 0.01; **,
0.01 $ P . 0.001; and ***, P , 0.001 for tests of significant difference of parameter value from 0.
Response
variable
Regression parameters
Overall
r
2
Overall
F value
Intercept
Species
diversity
Functional
diversity
Productivity 81.1*** 20.19NS 20.0*** 0.09 14.0***
Plant % N 1.24*** 20.0003NS 20.072*** 0.11 17.15***
Plant total N 104.3*** 20.193NS 12.06* 0.02 3.61*
Soil NH
4
1.07*** 0.003NS 20.082** 0.04 5.60**
Soil NO
3
0.37*** 0.001NS 20.041*** 0.09 13.4***
Light penetration 0.75*** 0.0001NS 20.040*** 0.11 18.3***
Table 2. Dependence of response variables on functional diversity treatments and functional compo-
sition based on ANOVAs. Functional composition was nested within each level of functional diversity. A
separate analysis was performed for each ecosystem response variable.
Response
variable
F values
Overall r
2
Functional
diversity
(df 5 5, 254)
Functional
composition
(df 5 26, 254)
Overall model
(df 5 31, 254)
Productivity 9.36*** 2.87*** 4.02** 0.33
Plant % N 22.2*** 17.3*** 17.4*** 0.68
Plant total N 4.23** 3.92*** 4.18*** 0.34
Soil NH
4
2.40* 1.23NS 1.40NS 0.14
Soil NO
3
22.3*** 1.17NS 4.57*** 0.36
Light penetration 12.1*** 3.21*** 4.57*** 0.36
REPORTS
www.sciencemag.org
z
SCIENCE
z
VOL. 277
z
29 AUGUST 1997 1301
REFERENCES AND NOTES
___________________________
1. J. H. Lawton and V. K. Brown, in Biodiversity and
Ecosystem Function, E.-D. Schulze and H. A.
Mooney, Eds. (Springer-Verlag, Berlin, 1993), pp.
255–270.
2. P. M. Vitousek and D. U. Hooper, ibid., pp. 3–14.
3. B. H. Walker, Conserv. Biol. 6, 18 (1991).
4. F. S. Chapin III, J. Lubchenco, H. L. Reynolds, in
Global Biodiversity Assessment, V. H. Heywood, Ed.
(Cambridge Univ. Press, Cambridge, 1995), pp.
289–301.
5. T. J. Givnish, Nature 371, 113 (1994).
6. S. J. McNaughton, in (1) , pp. 361–383; S. Naeem,
L. J. Thompson, S. P. Lawler, J. H. Lawton, R. M.
Woodfin, Nature 375, 561 (1995).
7. D. Tilman, D. Wedin, J. Knops, Nature 379, 718
(1996).
8. D. U. Hooper, Ecol. Monogr., in press.
9. P. M. Vitousek, Oikos 57, 7 (1990); F. S. Chapin III,
H. L. Reynolds, C. D9Antonio, V. Eckhart, in Global
Change in Terrestrial Ecosystems, B. Walker, Ed., in
press.
10. D. Tilman, C. L. Lehman, K. T. Thomson, Proc. Natl.
Acad. Sci. U.S.A. 94, 1857 (1997).
11. To prepare for planting, a field at Cedar Creek Nat-
ural History Area, in Minnesota, was treated with
herbicide and burned in August 1993, and had the
upper 6 to 8 cm of soil removed to reduce the seed
bank, was plowed and repeatedly harrowed, and
divided into 342 plots, each 13 m by 13 m (only the
inner 11 m by 11 m was sampled). Plots were seed-
ed in May 1994 and again in May 1995. To test for
effects of species diversity, we determined compo-
sition of each of 167 plots by random draws of 1, 2,
4, 8, or 16 species from a core pool of 18 species
(four species each of C
3
grasses, C
4
grasses, le-
gumes, and forbs; two woody species), with 29 to 35
replicates at each level of species diversity. To better
distinguish between effects of species and functional
diversity, we assigned combinations of 1, 2, or 3
functional groups containing 2, 4, or 8 species to 76
more plots, with compositions chosen by random
draws of functional groups followed by species.
When needed, we used a pool of 16 additional spe-
cies (four in each of the nonwoody functional
groups). Another 46 plots were created with 32 of
these 34 species. Four plots were kept bare. These
289 plots uncouple species diversity, functional di-
versity, and functional composition, but have a weak
correlation between these and species composition.
There is no such correlation in the 167-plot random
species subexperiment. The 289 plots have the fol-
lowing numbers of plots assigned to species and
functional diversity classes:
12. Unless noted otherwise, all analyses use treatment
species diversity, treatment functional diversity, and
treatment functional composition. In each plot we
estimated the percent cover of each species in four
subplots (0.5 m by 1 m each). We measured peak
aboveground living plant standing crop (an estimate
of plant productivity) by clipping, drying, and weigh-
ing four 0.1 m by 3.0 m strips per plot. We measured
% N in this aboveground biomass (plant % N), its
total N (plant total N), soil NH
4
and soil NO
3
extract-
able in 0.01 KCl [four soil cores (2.5 cm by 20 cm
depth) per plot], and the proportion of incident light
(PAR) that penetrated to the soil surface. In 1996,
plots contained mature, flowering plants, but the rel-
ative abundances of species may still be changing.
13. Linear regressions for effects of species diversity:
productivity, r 5 0.20, P , 0.01, n 5 289; plant % N,
r 5 –0.24, P , 0.001, n 5 286; plant total N, r 5
0.10, P 5 0.08, n 5 286; soil NH
4
,r5–0.11, P 5
0.06, n 5 289; soil NO
3
,r5–0.18, P , 0.01, n 5
289, light penetration, r 5 –0.24, P , 0.001, n 5
288. For effects of functional diversity: productivity,
r 5 0.30, P , 0.001, n 5 289; plant % N, r 5 – 0.33,
P , 0.001, n 5 286; plant total N, r 5 0.16, P ,
0.01, n 5 286; soil NH
4
,r5–0.19, P 5 0.01, n 5
289; soil NO
3
,r5–0.29, P , 0.001, n 5 289, light
penetration, r 5 –0.34, P , 0.001, n 5 288.
14. Regressions (as in 13), multiple regressions (as
in Table 1), ANOVAs (as in Table 2), and MANOVAs
that used only the 167 plots of the random species
subexperiment (11) had similar results and gener-
ally higher r
2
values, indicating that results are
not caused by the weak correlation between diver-
sity and species composition in the full 289-plot
experiment.
15. The 1996 average percent cover of each species or
functional group in each plot was used to calculate
its effective species or functional diversity as e
H9
,
where H9 is the Shannon-Wiener diversity index for
species or functional groups. Trends found using
treatment diversity variables also occurred when us-
ing 1996 effective diversity.
16. There were 32 different combinations of five func-
tional groups drawn 0, 1, 2, 3, 4 or 5 at a time. All 32
combinations were represented in the experiment.
For the nested ANOVAs, each plot with a given level
of functional diversity was further classified by which
of the 32 combinations it contained. Similar results
occurred when plots with bare soil or with 32 species
were excluded.
17. In the MANOVA, P , 0.0001 for both functional
diversity and functional composition using Wilks9
Lamba, Pillai’s Trace, Hotelling-Lawley Trace, or
Roy’s Greatest Root.
18. J. L. Harper, Population Biology of Plants (Academic
Press, London, 1977); D. Tilman, Resource Compe-
tition and Community Structure, Monographs in
Population Biology (Princeton Univ. Press, Prince-
ton, NJ, 1982).
19. J. J. Ewel, M. J. Mazzarino, C. W. Berish, Ecol. Appl.
1, 289 (1991).
20. R. T. Paine, Am. Nat. 100, 65 (1966); J. H. Brown,
D. W. Davidson, J. C. Munger, R. S. Inouye, in
Community Ecology, J. Diamond and T. Case, Eds.
(Harper and Row, New York, 1986), pp. 41–61;
S. R. Carpenter et al., Ecology 68, 1863 (1987); J.
Pastor, J. D. Aber, C. A. McClaugherty, J. M. Me-
lillo, ibid. 65, 256 (1984); G. C. Daily, P. R. Ehrlich,
N. M. Haddad, Proc. Natl. Acad. Sci. U.S.A. 90,
592 (1993).
21. P. M. Vitousek, L. R. Walker, L. D. Whiteaker, D.
Mueller-Dombois, P. A. Matson, Science 238, 802
(1987).
22. We thank C. Lehman, C. Bristow, N. Larson, and our
research interns for assistance and C. Bristow, C.
Lehman, C. Mitchell, S. Naeem, and A. Symstad for
comments. Supported by NSF and the Andrew Mel-
lon Foundation.
21 April 1997; accepted 16 July 1997
The Effects of Plant Composition and Diversity
on Ecosystem Processes
David U. Hooper* and Peter M. Vitousek
The relative effects of plant richness (the number of plant functional groups) and com-
position (the identity of the plant functional groups) on primary productivity and soil
nitrogen pools were tested experimentally. Differences in plant composition explained
more of the variation in production and nitrogen dynamics than did the number of
functional groups present. Thus, it is possible to identify and differentiate among po-
tential mechanisms underlying patterns of ecosystem response to variation in plant
diversity, with implications for resource management.
Recent experiments have shown increas-
ing net primary productivity (NPP) and
nutrient retention in ecosystems as the
number of plant species increases (1, 2).
Ecosystem response to plant richness could
occur via complementary resource use if
plant species differ in the ways they harvest
nutrients, light, and water (3, 4). Comple-
mentarity could happen in space, for exam-
ple, because of differences in rooting
depths; in time, for example, because of
differences in phenology of plant resource
demand; or in nutrient preference, for ex-
ample, nitrate versus ammonium versus dis-
solved organic N. Greater plant diversity
would then allow access to a greater propor-
tion of available resources, leading to in-
creased total resource uptake by plants,
lower nutrient losses from the ecosystem,
and increased NPP, if the resources in
question are limiting growth. However,
differences in plant composition (the
identity of the species present) may have
large effects on ecosystem processes if the
traits of one or a few species dominate (5).
For example, if one species or group of
species reduces soil nutrients to a lower
level than do other species, then this spe-
cies (or group) may dominate pools of
available soil nutrients in mixtures (6).
Such effects of composition could also
lead to lower soil nutrient pools and great-
er nutrient retention as diversity increases
because of an increasing probability of
including the dominant species at higher
levels of richness. In this case, however,
increased ecosystem nutrient retention re-
sults from the presence of only one species
rather than from niche differentiation and
complementary resource use among many.
Department of Biological Sciences, Stanford University,
Stanford, CA 94305–5020, USA.
*
To whom correspondence should be addressed at: Depart-
ment of Integrative Biology, Room 3060, Valley Life Sciences
Building, University of California, Berkeley, CA 94720–3140,
USA. E-mail: hooper@socrates.berkeley.edu
Species per plot
012481632
04––––––
1 – 34 11 12 14 – –
2 – – 331314 – –
3 – – – 20 14 – –
4 – – – 10 18 1 16
5 – – – – 113430
Functional
groups
per plot
SCIENCE
z
VOL. 277
z
29 AUGUST 1997
z
www.sciencemag.org1302
