22. Forey, P. L. History of the Coelacanth Fishes (Chapman & Hall, London, 1998).
23. Cloutier, R. in Devonian Fishes and Plants of Miguasha, Quebec, Canada (eds Schultze, H.-P. &
Cloutier, R.) 227–247 (Pfeil, Munich, 1996).
24. Gardiner, B. G. The relationships of the palaeoniscid fishes, a review based on new specimens of
Mimia and Moythomasia from the Upper Devonian of Western Australia. Bull. Br. Mus. Nat. Hist.
Geol. 37, 173–428 (1984).
25. Chang, M. M. Diabolepis and its bearing on the relationships between porolepiforms and dipnoans.
Bull. Mus. Natl Hist. Nat. Ser. 4 17, 235–268 (1995).
26. Schultze, H.-P. in The Biology and Evolution of Lungfishes (eds Bemis, W. E., Burggren, W. W. & Kemp,
N. E.) J. Morph. 1 (Suppl.), 39–74, (1987).
27. Ahlberg, P. E. & Johanson, Z. Osteolepiforms and the ancestry of tetrapods. Nature 395, 792–794
(1998).
28. Long, J. A new rhizodontiform fish from the Early Carboniferous of Victoria, Australia, with remarks
on the phylogenetic position of the group. J. Vert. Paleontol. 9, 1–17 (1989).
29. Johanson, Z. & Ahlberg, P. E. A complete primitive rhizodont from Australia. Nature 394, 569–572
(1998).
30. Chang, M. M. & Yu, X. Re-examination of the relationship of Middle Devonian osteolepids—fossil
characters and their interpretations. Am. Mus. Novit. 3189, 1–20 (1997).
Supplementary Information accompanies the paper on Nature’s website
(http://www.nature.com/nature).
Acknowledgements
We thank M. M. Chang and P. E. Ahlberg for advice and discussions, K. S. Thomson for
comments and corrections, M. Yang for artwork, J. Zhang for photographic work, and
X. Lu for specimen preparation. This work was supported by the Special Funds for Major
State Basic Research Projects of China, the Chinese Foundation of Natural Sciences, and
the US National Geographic Society. X.Y. thanks Kean University for faculty research and
development support.
Competing interests statement
The authors declare that they have no competing financial interests.
Correspondence and requests for materials should be addressed to M.Z.
(e-mail: zhumin@ht.rol.cn.net).
..............................................................
Increasing dominance of large lianas
in Amazonian forests
Oliver L. Phillips*, Rodolfo Va´squez Martı´nez†, Luzmila Arroyo‡§,
Timothy R. Baker*, Timothy Killeen‡§k, Simon L. Lewis*{,
Yadvinder Malhi{, Abel Monteagudo Mendoza†#, David Neillq**,
Percy Nu´n˜ez Vargas#, Miguel Alexiades††, Carlos Cero´n‡‡
Anthony Di Fiore§§, Terry Erwinkk, Anthony Jardim§, Walter Palaciosq,
Mario Saldias§ & Barbara Vinceti{
* Centre for Biodiversity and Conservation, School of Geography, University of
Leeds LS2 9JT, UK
† Jardin Bota´nico de Missouri, Jaen, Peru
‡ Missouri Botanical Garden, St Louis, Missouri 63166-0299, USA
§ Museo de Historia Natural Noel Kempff Mercado, Santa Cruz, Bolivia
kConservation International Washington DC 20036, USA
{The School of Earth Environmental and Geographical Sciences, University of
Edinburgh EH9 3JU, UK
# Herbario Vargas, Universidad San Antonio Abad del Cusco, Cusco, Peru
q Fundacio´n Jatun Sacha; ** Missouri Botanical Garden; ‡‡ Herbario QAP,
Escuela de Biologı´a de la Universidad Central del Ecuador, Quito, Ecuador
†† New York Botanical Garden, Bronx, New York 10458, USA
§§ Department of Anthropology, New York University, New York 10003, USA
kkNatural History Museum, Smithsonian Institution, Washington DC 20560,
USA
.............................................................................................................................................................................
Ecological orthodoxy suggests that old-growth forests should be
close to dynamic equilibrium, but this view has been challenged
by recent findings that neotropical forests are accumulating
carbon1,2 and biomass3,4, possibly in response to the increasing
atmospheric concentrations of carbon dioxide5,6. However, it is
unclear whether the recent increase in tree biomass has been
accompanied by a shift in community composition. Such changes
could reduce or enhance the carbon storage potential of old-
growth forests in the long term. Here we show that non-frag-
mented Amazon forests are experiencing a concerted increase in
the density, basal area and mean size of woody climbing plants
(lianas). Over the last two decades of the twentieth century the
dominance of large lianas relative to trees has increased by 1.7–
4.6% a year. Lianas enhance tree mortality and suppress tree
growth7, so their rapid increase implies that the tropical terres-
trial carbon sink may shut down sooner than current models
suggest8–10. Predictions of future tropical carbon fluxes will need
to account for the changing composition and dynamics of
supposedly undisturbed forests.
Recent field studies1,2,3 indicate that old-growth tropical forests
are absorbing 1–2 Gt C yr21, but the mechanisms and stability of the
tropical carbon sink, and its implications for the ecology of tropical
vegetation, are highly uncertain. Shifts in functional composition
and biodiversity are expected as a result of climate changes and
increased CO2 (refs 11, 12) but so far there is no evidence of
widespread compositional change in old-growth forests. This
absence of evidence might imply evidence of absence—or it could
simply reflect our failure to monitor adequately forest behaviour, or
even to examine existing data across sufficient spatial and temporal
scales. Lianas in particular are ignored in forest inventories and
models alike, in spite of their key functional roles. As structural
parasites, lianas exert a much greater ecological effect than their size
suggests, representing less than 5% of tropical forest biomass but up
to 40% of leaf productivity13. They also suppress tree growth and
encourage tree mortality, and affect the competitive balance among
trees by disproportionately infesting some taxa and suppressing the
regeneration and growth of non-pioneers7. Climbers respond
strongly to increased CO2 concentrations14,15 and benefit from
disturbance7,16,17, and a biome-wide trend to increased tree turnover
rates has been detected in old-growth forests18 so increases in liana
densities might be anticipated19. Here we assemble several unique,
Figure 1 Structural importance of lianas over 10 cm in diameter in each neotropical site
as a function of date of first inventory. a, Liana stem density in stems ha21; b, liana basal
area in m2 ha21. ‘Central America’ is Panama and tropical countries to the north;
‘Northwest South America’ is the Choco´ bioregion, west of the Andes; ‘Amazonia’ is the
Amazon river basin and contiguous forested zones of Guyana and eastern Brazil. Linear
regressions are fitted to the Amazonian data.
letters to nature
NATURE | VOL 418 | 15 AUGUST 2002 | www.nature.com/nature770 © 2002 Nature Publishing Group

long-term, multi-regional data sets of liana and tree populations.
We use them to test both the general hypothesis that the compo-
sition of old-growth tropical forests is changing over large scales,
and the specific prediction that lianas are benefiting. We analyse 47
interior-forest sites in four Amazonian regions (North Peru, South
Peru, Bolivia and Ecuador) where we are monitoring all woody
plants of over 10 cm in diameter in 1-ha plots, and include
published data from a further 37 neotropical sites. We find that
the density and the basal area of large lianas have increased
substantially over the last two decades of the twentieth century.
The same trends are observed however liana populations are
analysed.
First, we plot liana dominance as a function of each site’s first
inventory data (Fig. 1). There is broad scatter, reflecting forest
variation and large sampling error at the plot scale17, but a
significant trend for late-censused neotropical sites to have greater
liana dominance than early-censused sites. Analysis of covariance
(ANCOVA) shows that this is not an artefact of spatial changes in
sampling intensity—in models with lianas as the dependent vari-
able, and region and year as independent variables, the year
contributes significantly to both neotropical and Amazon liana
stem density (F 1,69 ˆ 15.30, P , 0.001; F1,53 ˆ 14.37, P , 0.01)
and to neotropical and Amazon liana basal area (F 1,49 ˆ 8.96,
P , 0.01; F 1,43 ˆ 6.99, P , 0.02). The increases in lianas as a
function of first inventory dates are also unlikely to be artefacts of
a change through time in the environmental characteristics of the
forests sampled, because the date of first inventory contributes
significantly to statistical models of liana density and basal area even
after accounting for edaphic and climatic effects (Table 1).
Second, we analysed a different data set: the changes that
Table 1 Modelled changes in liana density and basal area
Dependent variable Source of variation
A. Environment only B. Environment plus time
Adjusted R2 (%) F-value, d.f. D Adjusted R2 (%) F-value, d.f. T-value and significance for T in model
...................................................................................................................................................................................................................................................................................................................................................................
Liana stem density (ha21) 35.3 3.83*, 26 ‡13.2 5.89***, 26 ‡2.93**
Liana basal area (m2 ha21) 27.2 2.94*, 26 ‡21.1 7.06***, 26 ‡4.42***
Number of lianas/number of trees 34.1 4.36*, 26 ‡12.4 6.66***, 26 ‡3.93***
Basal area of lianas/basal area of trees 34.6 3.60*, 26 ‡21.7 6.59***, 26 ‡4.80***
Mean liana basal area (cm2) 0.3 1.04, 26 ‡5.5 1.54, 26 ‡1.55
...................................................................................................................................................................................................................................................................................................................................................................
*, 0.05 . P > 0.01; **, 0.01 . P > 0.001; ***, P , 0.001.
Multiple regression of large liana density and basal area (BA) at each initial census, modelled as a function of the environment, and the environment plus time. Environmental variables measured at North
Peru, South Peru and Bolivian sites include climate (mean annual rainfall, seasonality), soil chemistry (pH, Ca, K, Mg, Na, P, Al), soil particle size distribution (sand, silt, clay), and hydrology (drainage and risk
of water-logging). Soil variables were converted to principal components before multiple regression; time variable (T, in years) is the decimal date in which the liana parameter was first recorded in the
plot. See Methods for details.
Table 2 Linear trends in the structural importance of large lianas
a Amazonian sites with a monitoring period of over 5 yr
...................................................................................................................................................................................................................................................................................................................................................................
Parameter Annual rate of change in
parameter (mean ^ 95% CI)
Annual rate of change (proportion
of site initial value)
Annual rate of change (proportion
of site final value)
...................................................................................................................................................................................................................................................................................................................................................................
Liana stem density (ha21) ‡0.22 ^ 0.11 (n ˆ 28) ‡4.03 ^ 2.56% ‡1.78 ^ 0.82%
Liana basal area (m2 ha21) ‡3.72 ^ 1.16 £ 1023 (n ˆ 28) ‡4.58 ^ 2.60% ‡2.40 ^ 0.62%
Liana stems as a fraction
of tree stems
‡3.45 ^ 2.10 £ 1024 (n ˆ 28) ‡3.27 ^ 2.10% ‡1.70 ^ 0.87%
Liana basal area as a
fraction of tree basal area
‡1.19 ^ 0.53 £ 1024 (n ˆ 28) ‡4.05 ^ 2.30% ‡2.07 ^ 0.71%
Mean basal area per liana
stem (cm2)
‡1.03 ^ 0.97 (n ˆ 27) ‡0.96 ^ 0.66% ‡0.65 ^ 0.65%
b Amazonian sites with a monitoring period of over 5 yr analysed by region
...................................................................................................................................................................................................................................................................................................................................................................
Parameter ANOVA, test of hypothesis of
a regional cluster effect
Annual rate of change in
parameter (mean ^ 95% CI) for each region
...................................................................................................................................................................................................................................................................................................................................................................
Liana stem density (ha21) F ˆ 0.84, P ˆ 0.48 N. Peru ‡0.34 ^ 0.25 (n ˆ 7)
S. Peru ‡0.19 ^ 0.22 (n ˆ 12)
Bolivia ‡0.06 ^ 0.41 (n ˆ 5)
Ecuador ‡0.28 ^ 0.32 (n ˆ 4)
Liana basal area (m2 ha21) F ˆ 0.64, P ˆ 0.60 N. Peru ‡0.0049 ^ 0.0022 (n ˆ 7)
S. Peru ‡0.0029 ^ 0.0025 (n ˆ 12)
Bolivia ‡0.0042 ^ 0.0029 (n ˆ 5)
Ecuador ‡0.0033 ^ 0.0033 (n ˆ 4)
Liana stems as a fraction F ˆ 0.95, P ˆ 0.43 N. Peru ‡6.34 ^ 4.75 £ 1024 (n ˆ 7)
of tree stems S. Peru ‡2.20 ^ 3.91 £ 1024 (n ˆ 12)
Bolivia ‡2.10 ^ 7.89 £ 1024 (n ˆ 5)
Ecuador ‡4.30 ^ 5.70 £ 1024 (n ˆ 4)
Liana basal area, as a F ˆ 0.62, P ˆ 0.61 N. Peru ‡1.79 ^ 0.82 £ 1024 (n ˆ 7)
fraction of tree basal area S. Peru ‡0.97 ^ 1.21 £ 1024 (n ˆ 11)
Bolivia ‡0.80 ^ 1.81 £ 1024 (n ˆ 5)
Ecuador ‡1.11 ^ 1.46 £ 1024 (n ˆ 4)
Mean basal area per liana F ˆ 0.15, P ˆ 0.93 N. Peru ‡0.66 ^ 2.92 (n ˆ 7)
stem (cm2) S. Peru ‡0.96 ^ 1.83 (n ˆ 11)
Bolivia ‡1.61 ^ 1.90 (n ˆ 5)
Ecuador ‡1.20 ^ 2.49 (n ˆ 4)
...................................................................................................................................................................................................................................................................................................................................................................
ANOVA, analysis of variance. CI, confidence interval.
letters to nature
NATURE | VOL 418 | 15 AUGUST 2002 | www.nature.com/nature 771© 2002 Nature Publishing Group

occurred through time within multi-census sites. Here we find
consistently strong positive changes in measures of liana dominance
(1.7–4.6% per year) and stem size (0.6–1.0% per year), evident
across all regions (Table 2).
Finally, we assembled all available data from our four west
Amazonian regions, that is single-census and multi-census data,
to create running means across sites, and find similarly consistent
patterns over the last two decades (Fig. 2). The year-on-year increase
in mean values of large lianas is not driven by a few atypical sites or
by the intrinsic liana richness of any one region, but is rather a
general phenomenon across all five ecological parameters and all
four regions (ANCOVA, with year as the continuous variable and
region as fixed factor, shows that year contributes (P , 0.01) for all
20 combinations except for mean stem size in North Peru
(P . 0.05)).
We examined the underlying dynamics of lianas in all plots
censused three or more times, and find that the rates of both large
liana growth and large liana loss have increased (comparing annual
rate of basal area gain for lianas interval 1 versus interval 2,
t ˆ 23.10, n ˆ 21, P , 0.01; annual rate of basal area loss for
lianas interval 1 versus interval 2, t ˆ 22.66, n ˆ 21, P , 0.02),
whereas growth rates have consistently exceeded loss rates (annual
rate of basal area gain versus loss for lianas > 10 cm in diameter for
interval 1: t ˆ 4.19, n ˆ 21, P , 0.001; and for interval 2: t ˆ 2.38,
n ˆ 21, P , 0.05). Thus the net increase in liana basal area was
driven by high liana growth rates that have increased through time,
and occurred in spite of an acceleration in the rate of liana mortality.
Tree basal area increased by 0.34 ^ 0.20% a year in the 1980s and
1990s in Amazonia3 but the increase in liana values has been much
more rapid: the relative importance of large lianas has approxi-
mately doubled over a similar period across all sites (Fig. 2) and the
annual rates of increase in liana density and BAwithin sites exceeded
the rate of increase in tree BA by an order of magnitude (Table 2a).
Mature Amazonian forest plots have therefore undergone sub-
stantial change in functional composition. The increase in lianas has
been concerted in the sense that it has occurred simultaneously over
a wide spatial, climatic and edaphic range. We have shown that this
is not an effect of changes in the kinds of forests sampled through
time, and other potential artefactual explanations can also be ruled
out (Supplementary Information). We can explore the generality of
the changes indicated by the permanent plot data set with an
independent data set from 70 old-growth lowland forests across
the Neotropics. These are single-census 0.1-ha forest samples
surveyed between 1971 and 1997, in which every tree and
liana > 2.5 cm in diameter was inventoried once (see Methods),
thus sampling plants with basal area as low as about 6% of the
smallest plants in our permanent plots. After controlling for climate
and soil effects, the neotropical 0.1-ha data set confirms the
temporal trends seen in the permanent plot data set: liana popu-
Figure 3 Relative dominance of lianas in neotropical 0.1-ha plots as a function of
inventory date. Liana basal area is shown as a fraction of tree basal area for all stems over
2.5 cm in diameter. A polynomial curve is fitted for the Amazonian sites.
Figure 2 Changes through time of the importance of lianas over 10 cm in diameter in
western Amazonia. a, Liana stem density in stems ha21; b, liana basal area in m2 ha21;
c, relative liana density as a percentage of tree stems; d, relative liana basal area as a
percentage of tree basal area; e, mean basal area of each liana stem in cm2. Graphs show
5-yr running means with 95% confidence intervals, with values plotted separately for
North Peru, South Peru, Bolivia and Ecuador.
letters to nature
NATURE | VOL 418 | 15 AUGUST 2002 | www.nature.com/nature772 © 2002 Nature Publishing Group

lations have become more dense, liana basal area has increased, the
relative dominance of lianas has increased, and the size of individual
lianas has increased (Fig. 3, and Supplementary Information
Table 3). The degree of internal consistency within and between
data sets across differing sample unit sizes, target variables, mini-
mum plant sizes, climatic regimes, edaphic conditions, regional
locations and spatial scales is a critical factor in assessing confidence
in the changes. The results demonstrate a substantial increase in the
density and relative dominance of lianas in western Amazonia, and
available evidence suggests that this change in the structural and
functional composition of forests has been even more widespread.
We asked what is driving this change? If regional climates have
changed in western Amazonia or across the neotropics that could
provide an explanation, but we failed to find convincing evidence
for this (Supplementary Information). The documented increase in
CO2 concentrations is another possibility, because lianas respond
strongly to CO2 fertilization over the historical range of concen-
trations15. The direct impacts of CO2 on photosynthesis may drive
nonlinear compositional responses—for example, the relative rate of
stimulation on liana growth may be particularly strong in deep
shade15, enhancing the likelihood of lianas reaching the sunlit
canopy. Effects of increasing CO2 on climber growth14,15 could
also be magnified in a positive feedback loop with the simultaneous
increase in tree turnover18 (more rapid turnover favours gap-
specialist lianas which in turn accelerate tree mortality).
The increased liana load in trees may have a major impact on the
Amazon carbon sink. The biomass of lianas themselves is usually
small7,20, but lianas can substantially suppress tree biomass7,17. We
examined the relationship between liana infestation and subsequent
tree mortality in the 13 sites where we have accurate records of
liana–host relationships. Here, liana infestation of tree biomass is
associated with a 39.6 ^ 31.3% excess risk of tree mortality. The
expected annual rate of increase in the amount of tree biomass
mortality is estimated as 1.64 ^ 1.38%, the product of the liana-
associated excess tree mortality risk and the annual rate of increase
in large lianas per unit of tree biomass in the sites. However climate
models suggest that in east and south Amazonia moisture supply
will become more seasonal9,10, conditions which may favour lianas
(Supplementary Information), so synergisms between climate
change and increasing liana densities could magnify the impact of
either process alone. Better understanding of these risks will require
intensive field research to improve the liana-on-tree mortality
functions and to begin including lianas within full tropical forest
vegetation models and coupled carbon cycle/climate models.
The increase in liana density and biomass is the first evidence for a
widespread functional shift in old-growth tropical forests. Regard-
less of the impacts on the carbon cycle, this has important
implications for the biodiversity of tropical forests. First, if the
increase is driven by increasing CO2 concentrations it implies that
the extensive tropical forests of the Cretaceous and Tertiary periods
when CO2 concentrations peaked at .2,000 p.p.m. (ref. 21) may
have differed radically from today’s in structure and function.
Second, lianas and trees are differentiated phylogenetically22 and
by distinctive pollination and dispersal ecologies23, so changed
relative densities has knock-on consequences for conservation of
plants and animals. Third, increased liana density has the potential
to alter tree species composition because climber impacts on trees
vary with host phylogeny and ecology7,16. Finally, the change in
composition has direct societal and economic impacts, because
lianas are valued less than trees by forest communities24 and are
major silvicultural pests for the tropical timber industry7. A
Methods
Study sites of 1 ha
We censused large liana populations in 47 sites spanning the climatic and edaphic
gradients of western Amazonia25 (Supplementary Information Appendix). 1-ha
permanent plots were sited since as early as 1979 in old-growth forest and recensused every
2–5 yr, most recently in 2002 (eight sites) and 2001 (19 sites), yielding up to 19 yr of growth
and dynamics data. Plot locations were constrained by the need for reasonable access
(,10 km to nearest road or navigable river) and long-term protection, but are otherwise
sited randomly or haphazardly within landform strata and are unbiased by sylvigenetic
state3,4. The straight-line distance between all pairs of plot centroids ranges by four orders
of magnitude (0.2–2,380 km), but intersite distance had no effect on liana change metrics
(Supplementary Information). Distances to edges are 0.1 to ,3.6 km, where ‘edge’ is
defined as a previously forested location where there has been an anthropogenic impact
creating a canopy gap of >0.5 ha the effects of which are still apparent at the time of the
first census. Edges were formed by farmers, research stations, tourist lodges and logging
activities. We also searched the literature for published single-census large liana
inventories in neotropical forests with >1,500 mm rain, including all except montane/
cloud forests, small fragments (,1,000 ha), or with known human disturbance to forest
structure.
Liana and tree measurements
Tree measurements and analysis follow RAINFOR protocols25 (http://
www.geog.leeds.ac.uk/projects/rainfor/). For lianas the diameter of each independently
rooted climbing stem rooted within each plot and potentially over 10 cm wide was
measured at a height of 1.3 m (d1.3: all sites except 14, 15 and 25) and at the widest point
within 2.5 m of the ground (d max: all sites except 23, 24, 34 and 35), or at both points for
most censuses (40 sites). We used the d 1.3:d max ratio of individual lianas to determine d max
of lianas at censuses where only d 1.3 was recorded. This procedure gives unbiased
estimates, because the ratio of d 1.3 to d max is independent of d max (mean ^ 95%
confidence interval, CI, of r, the correlation between d 1.3/d max and d max ˆ 0.007 ^ 0.080,
n ˆ 28). For both measurement methods we calculated the number of lianas and total
basal area (the sum of cross-sectional stem areas) at each census. Lianas support greater
biomass and productivity than trees of equal diameter7,13,26, so stem density and basal area
are not equivalent measures of functional importance between life-forms, but for each life-
form basal area provides a close approximation to biomass3,26,27. For reasons of brevity we
only report results based on d 1.3 measurements. Results based on dmax are presented in the
Supplementary Information Tables 1 and 2 and are essentially equivalent.
Climate and soil data
Climate data were sourced directly from local records, or indirectly from interpolated
maps25,28, to derive mean annual rainfall and seasonality (consecutive months averaging
,100 mm rain) for every neotropical site. At Bolivia and South and North Peru sites, soils
were sampled from up to ten randomly chosen locations within each plot at 0–15 or 0–
20 cm depth; samples were bulked, dried, and subsampled. Analysis followed standard
ISRIC procedures29. Most sites have been visited at least five times, allowing assessment of
hydrologic conditions on a scale of one (permanently water-logged) to ten (excessively
draining).
Change in permanent plots
For within-site and within-region change analyses (Table 2) we included all interior-forest
old-growth sites with at least two full censuses of trees and lianas over 10 cm in diameter
more than 5 yr apart. For between-site change analyses (Fig. 1, Table 1) we included
interior-forest old-growth plots with at least one full census of trees and lianas over 10 cm
in diameter. To minimize effects of any asymmetric sampling of environments through
time, we first excluded permanent swamp and white sand sites and then used principal
components ordination analysis (PCA) to describe the major gradients in normalized and
standardized soil variables in Peruvian and Bolivian sites, and then applied multiple
regression to test the effects of the PCA factors, climate variables and the time variable on
forest structure (Table 1), repeating the procedure with neotropical 0.1-ha sites
(Supplementary Information Table 3). We report multiple-regression models with
greatest adjusted-r 2 values and control for climate and edaphic effects by computing the
variance explained by inventory date after accounting for the variance explained by
environmental variables.
For the displays of change within- and between-sites (Fig. 2), we included interior-
forest old-growth plots with at least one full census of trees and lianas. We used linear
interpolation between each census to estimate structural values within sites and then
derived cross-site 5-yr running means to smooth the effects of site-switching.
For comparisons of liana growth and loss within plots we split each monitoring period
into intervals of similar length (first interval, 5.9 ^ 0.7 yr; second interval, 6.0 ^ 0.9 yr),
so that turnover rates could be compared directly while controlling for possible effects of
interval length on estimated mortality and growth rates.
Study sites of 0.1 ha
Inventories were completed by A. Gentry (58 sites) (http://www.mobot.org/MOBOT/
Research/gentry/transects.html), R.V.M. and O.L.P. (6), and T.K., L.A. and M.S. (6). All
scandent lianas and hemiepiphytes with d max over 2.5 cm and non-climbing stems with
d 1.3 over 2.5 cm were measured. See ref. 30 for detailed descriptions. All available
neotropical old-growth forest samples with more than 1,500 mm rain were included,
except montane/cloud forests, small fragments (less than 1,000 ha), or sites with known
human disturbance to forest structure (extra-Amazonian sites also exclude island
hurricane-impacted forests).
Received 25 March; accepted 30 May 2002; doi:10.1038/nature00926.
1. Grace, J. et al. Carbon dioxide uptake by an undisturbed tropical rain-forest in Southwest Amazonia,
1992-1993. Science 270, 778–780 (1995).
2. Malhi, Y. et al. Carbon dioxide transfer over a Central Amazonian rain forest. J. Geophys. Res. Atmos.
103, 31593–31612 (1998).
letters to nature
NATURE | VOL 418 | 15 AUGUST 2002 | www.nature.com/nature 773© 2002 Nature Publishing Group

3. Phillips, O. L. et al. Changes in the carbon balance of tropical forest: evidence from long-term plots.
Science 282, 439–442 (1998).
4. Phillips, O. L. et al. Changes in the biomass of tropical forests: evaluating potential biases. Ecol. Appl.
12, 576–587 (2002).
5. Prentice, I. C. et al. in Intergovernmental Panel on Climate Change Third Assessment Report, Climate
Change 2001: The Scientific Basis Ch. 3 (Cambridge Univ. Press, Cambridge, UK, 2001).
6. Malhi, Y. & Grace, J. Tropical forests and atmospheric carbon dioxide. Trends Ecol. Evol. 15, 332–337
(2000).
7. Schnitzer, S. A. & Bongers, F. The ecology of lianas and their role in forests. Trends Ecol. Evol. 17,
223–230 (2002).
8. Chambers, J. Q., Higuchi, N. & Tribuzy, E. S. Carbon sink for a century. Nature 410, 429–429 (2001).
9. Cox, P. M. et al. Acceleration of global warming due to carbon-cycle feedbacks in a coupled climate
model. Nature 408, 184–187 (2000).
10. White, A., Cannell, M. G. R. & Friend, A. D. CO2 stabilisation, climate change and the terrestrial
carbon sink. Glob. Change Biol. 6, 817–833 (2000).
11. Condit, R., Hubbell, S. P. & Foster, R. B. Assessing the response of plant functional types to climatic
change in tropical forests. J. Vegn. Sci. 7, 405–416 (1996).
12. Ko¨rner, C. Biosphere responses to CO2 enrichment. Ecol. Appl. 10, 1590–1619 (2000).
13. Hegarty, E. E. & Caballe´, G. in The Biology of Vines (eds Putz, F. E. & Mooney, H. A.) 313–336
(Cambridge Univ. Press, Cambridge, UK, 1991).
14. Condon, M. A., Sasek, T. W. & Strain, B. R. Allocation patterns in two tropical vines in response to
increased atmospheric CO2. Funct. Ecol. 6, 680–685 (1992).
15. Granados, J. & Korner, C. In deep shade, elevated CO2 increases the vigour of tropical climbing plants.
Glob. Change Biol. (in the press).
16. Pe´rez-Salicrup, D. R., Sork, V. L. & Putz, F. E. Lianas and trees in Amazonian Bolivia. Biotropica 33,
34–37 (2001).
17. Laurance, W. F. et al. Rain forest fragmentation and the structure of Amazonian liana communities.
Ecology 82, 105–116 (2001).
18. Phillips, O. L. & Gentry, A. H. Increasing turnover through time in tropical forests. Science 263,
954–958 (1994).
19. Phillips, O. L. The changing ecology of tropical forests. Biodivers. Cons. 6, 291–311 (1997).
20. Putz, F. E. Liana biomass and leaf-area of a tierra firme forest in the Rio Negro basin, Venezuela.
Biotropica 15, 185–189 (1983).
21. Retallack, G. J. A 300 million year record of atmospheric carbon dioxide from fossil plant cuticles.
Nature 411, 287–290 (2001).
22. Gentry, A. H. in The Biology of Vines (eds Putz, F. E. & Mooney, H. A.) 3–49 (Cambridge Univ. Press,
Cambridge, UK, 1991).
23. Gentry, A. H. in The Biology of Vines (eds Putz, F. E. & Mooney, H. A.) 393–423 (Cambridge Univ.
Press, Cambridge, UK, 1991).
24. Phillips, O. L. & Gentry, A. H. The useful plants of Tambopata, Peru. II: Additional hypothesis testing
in quantitative ethnobotany. Econ. Bot. 47, 33–43 (1993).
25. Malhi, Y. et al. An international network to monitor the structure, composition and dynamics of
Amazonian forests (RAINFOR). J. Vegn. Sci. (in the press).
26. Gerwing, J. J. & Lopes Farias, D. Integrating liana abundance and forest stature into an estimate of
total aboveground biomass for an eastern Amazonian forest. J. Trop. Ecol. 16, 327–335 (2000).
27. Brown, S. Estimating Biomass and Biomass Change of Tropical Forests: a Primer (Food and Agriculture
Organisation Forestry Paper 134, Rome, 1997).
28. Sombroek, W. G. Spatial and temporal patterns of Amazon rainfall: consequences for the planning of
agricultural occupation and the protection of primary forests. Ambio 30, 388–396 (2001).
29. van Reeuwijk, L. P. (ed.) Procedures for Soil Analysis, Tech. Pap. 9, 5th edn (International Soil Reference
and Information Centre, FAO, Rome, 1995).
30. Phillips, O. L. & Miller, J. Global Patterns of Plant Diversity: Alwyn H. Gentry’s Forest Transect Data Set
(Missouri Botanical Garden, St Louis, in the press).
Supplementary Information accompanies the paper on Nature’s website
(http://www.nature.com/nature).
Acknowledgements
We acknowledge the contributions of more than 50 field assistants in Peru, Ecuador and
Bolivia, the residents of Constancia, Infierno, La Torre, Mishana and Florida, as well as
logistical support from Instituto Nacional de Recursos Naturales (INRENA), Amazon
Center for Environmental Education and Research (ACEER), Cuzco Amazo´nico Lodge,
Explorama Tours SA, Instituto de Investigaciones de la Amazonı´a Peruana (IIAP), Parque
Nacional Noel Kempff, Peruvian Safaris SA, Universidad Nacional de la Amazonı´a
Peruana, and Universidad Nacional de San Antonio Abad del Cusco. Field research was
supported by the EU Fifth Framework Programme (RAINFOR), the UK Natural
Environment Research Council, the National Geographic Society, the American
Philosophical Society, the National Science Foundation, the WWF-U.S./ Garden Club of
America, Conservation International, the MacArthur and Mellon Foundations, US-AID,
the Max-Planck Institute for Biogeochemistry and the Royal Society (Y.M.). The
manuscript benefited from comments by C. Ko¨rner and N. Pitman. We are indebted to the
late A.H. Gentry for helping to make this work possible.
Competing interests statement
The authors declare that they have no competing financial interests.
Correspondence and requests for materials should be addressed to O.P.
(e-mail: o.phillips@geog.leeds.ac.uk).
..............................................................
Involvement of DARPP-32
phosphorylation in the
stimulant action of caffeine
Maria Lindskog*, Per Svenningsson†, Laura Pozzi*‡, Yong Kim†,
Allen A. Fienberg†, James A. Bibb†‡, Bertil B. Fredholm§,
Angus C. Nairn†, Paul Greengard† & Gilberto Fisone*
* Department of Neuroscience, Karolinska Institutet, 17177 Stockholm, Sweden
† Laboratory of Molecular and Cellular Neuroscience, The Rockefeller University,
New York, New York 10021, USA
§ Department of Physiology and Pharmacology, Karolinska Institutet, 17177
Stockholm, Sweden
‡ Present addresses: Department of Neuroscience, “Mario Negri” Institute for
Pharmacological Research, Milan, Italy (L.P.); Department of Psychiatry,
UT Southwestern Medical Center, Dallas, Texas, USA (J.A.B)
.............................................................................................................................................................................
Caffeine has been imbibed since ancient times in tea and coffee,
and more recently in colas. Caffeine owes its psychostimulant
action to a blockade of adenosine A2A receptors1, but little is
known about its intracellular mechanism of action. Here we show
that the stimulatory effect of caffeine on motor activity in mice
was greatly reduced following genetic deletion of DARPP-32
(dopamine- and cyclic AMP-regulated phosphoprotein of relative
molecular mass 32,000)2. Results virtually identical to those seen
with caffeine were obtained with the selective A2A antagonist
SCH 58261. The depressant effect of the A2A receptor agonist,
CGS 21680, on motor activity was also greatly attenuated in
DARPP-32 knockout mice. In support of a role for DARPP-32 in
the action of caffeine, we found that, in striata of intact mice,
caffeine increased the state of phosphorylation of DARPP-32 at
Thr 75. Caffeine increased Thr 75 phosphorylation through
inhibition of PP-2A-catalysed dephosphorylation, rather than
through stimulation of cyclin-dependent kinase 5 (Cdk5)-cata-
lysed phosphorylation, of this residue. Together, these studies
demonstrate the involvement of DARPP-32 and its phosphory-
lation/dephosphorylation in the stimulant action of caffeine.
Striatal medium spiny neurons have an important role in the
control of voluntary movements. A large subpopulation of these
neurons project to the substantia nigra pars reticulata, the major
Figure 1 Emulsion autoradiogram illustrating co-expression of adenosine A2A receptor
mRNA (silver grains) and DARPP-32 mRNA (dark cells). Shown are subpopulations of
medium spiny neurons in mouse (a) and rat (b) striatum. Single arrows indicate neurons
that only express DARPP-32 mRNA. Double arrows indicate neurons that express both
DARPP-32 mRNA and adenosine A2A receptor mRNA.
letters to nature
NATURE | VOL 418 | 15 AUGUST 2002 | www.nature.com/nature774 © 2002 Nature Publishing Group