evolved as a means of escape from her-
bivory. Others argue that it is the differ-
ences in developmental phenology. result-
ing from trade-offs associated with male
versus female fiinction, that cause plants
to be differentially susceptible and/or re-
sponsive to herbivory. Too little work is
available to discriminate between these
alternatives. Studying the effects of her-
bivory in male versus female morphs of
facultatively sex-changing species may
prove useful in this regard.
Whatever the case, differences in the
developmental phenology appear crucial
for determining plants' overall responses
to herbivory (overcompensation, compen-
sation, undercompensation), as well as the
likelihood of and pattern of response to
sex-biased herbivory in dioecious plant
species. Study of the evolution, mainten-
ance and importance of patterns of devel-
opmental phenology in plants is proving

to be an exciting and productive area of re-
search, one which may give rise to predic-
tive theories in areas formerly resistant to
generalization. Already, these studies are
yielding important insights into the limits
of plant plasticity as an adaptive strategy
and the importance of developmental con-
straints in regulating plants' responses to
their environments.

Acknowledgements
The manuscript has benefitted from
the comments of Brenda Casper,
Monica Ceber. Timothy Criffith,
Katherine Preston and Jason Price.
The development of this manuscript
was funded by a grant from the
National Science Foundation.

Maxine A. Watson

Dept of Biology, Indiana University,
Bloomington, IN 47405, USA

The Keystone cops meet in Hilo

  eystone species', a term coined by
`K Painel more than 25 years ago, has
proven a powerful metaphor for the inves-
tigation of forces that organize ecological
communities. Increasingly, the concept
influences the thinking of managers and
policy makers who must set priorities in
their efforts to conserve species and habi-
tats. The meaning of the term, however, has
been blurred by overly expansive usage
and for this reason, its application poses
real dangers of misuse in decision makingz.

On the other hand, the term is too char-
ismatic, too entrenched and too useful to
be abandoned. Based on these concerns,
Hal Mooney (Stanford University, CA,
USA) and Jane Lubchenco (Oregon State
University, Corvallis, USA) organized a
workshop on the Keystone Species Con-
cept on December 8-1 1,1994 in rainy Hilo,
Hawaii, USA, as part of the Global Bie
diversity Assessment (GBA) sponsored
by the United Nations Environmental Pro-
gramme (UNEP). At the heart of the work-
shop was a belief that conservationists
and managers have too few ecological
tools to predict and prevent loss of bio-
diversity and ecosystem function, and
that a refined keystone concept would be
an Important addition. The workshop had
two goals: to seek consensus on a defi-
nition of keystone, and to explore whether
potential keystone species could be ident-
ified Q priori (before experimental confir-
mation, or their loss due to local extlnc-
tion) by any distinguishing traits. The CBA
team - Mooney, Lubchenco, Tony Janetos

(NASA, Washington, DC, USA), Osvaldo
Sala (Universidad de Buenos Aires, Argen-
tina), Hall Cushman (Sonoma State Univer-
sity, CA, USA) and Rodolfo Dirzo (UNAM,
Mexico) - met with nine ecologists who
have investigated keystone species to ad-
dress these and other questions about key-
stone species. What are keystones, how can
they be identified? How many systems have
them? How vulnerable are they to human
impacts? What ecosystem functions and
services would disappear if they were lost?
Mooney and Lubchenco launched
the workshop, challenging the group to
achieve consensus on a definition that fol-
lowed either historical (Paine's original
usage) or functional criteria (can we arrive
at an expanded definition that remains
clear and heuristic?) The need is immedi-
ate. According to Janetos, the Science and
Technology Panel of the World Bank (a
group that makes decisions about funding
projects with impacts on biodiversity) has
found no body of comprehensive state
ments in which ecologists express any
consensus on such issues
In short, ecologists need to make their
science more useful for conservation and
biodiveriity practitioners. With this man-
date, the nine workshop participants In-
itiated discussion with brief synopses of
their previous and current views on key-
stones. Bob Paine (Universlty of Wash-
ington, Seattle, USA) began by reviewing
his original definition. `Keystone' in Painel
originally referred to a species that pref-
erentially consumed and held in check

References

1 Belsky, A.J. (1986) Am. Not 127.870-892
2 Palge, K.N. (1994) Am. Not. 143,739-749
3 Malhews, J.N.A. (1994) Am. Nut. 144,528-533
4 Trumble. J.T.. Kolodny-Hlrsch, D.M. and

Ting, I.P. (1993)Annu. Reu. hfomd. 38,93-119
5 Watson, M.A. (1990) In Clonal CIocurn in planis:
Regulation and Function (van Crmendael, 1.
and de Kroon, H., eds), pp. 43-55, SPB Academic

6 Boecklen. W.J., Price. P.W. and Mopper. S.

7 Boecklen. W.J. and Hoffman, M.T. (1993)
Oecologia 96.49-55
8 Boecklen, W.J., Mopper, S. and Price, P.W.
(1994) Oikos 71,267-272

10 Cox, PA (1981) Am. Nut. 117,295-307
11 Lovett Doust, L and Lovett bust. 1. (1987)

(1990) €CO/W 71,581-588

9 JL~& S.W. and colw. P.D. (1990) olpassa, 369-377

€co/~ 68,2056-2058
12 Delph, LF. (1990) €CO/OS~ 71,1342-1351

Bot. 80,607-615

13 Delph, LF., Lu. Y. and Jape. LD. (1993) Am. J.

14 Mutlkalnen, P.. Walls, M. and Ojala. h Oikos
(In press)

another species that would otherwise
dominate the system. This suppressed
target species could be a competitive domi-
nant (e.g. the mussels suppressed by a
starfish in Paine's rocky intertidal system)
or coupled to a trophic cascade (e.g. sea
urchins, suppressed by sea otters3). The
interaction had to be of sufficient strength
to mediate observable indirect effects in
the community. Traits suggesting a poten-
tial keystone role for a consumer were pref-
erence for prey of high competitive ability,
and the ability to control this prey at all
sizes (so the prey had no escape in size).
David Tilman (University of Minnesota,
Minneapolis, USA) defined keystone in
terms of the effect of a change in one
species on some characteristic of its com-
munity or ecosystem, such as species
richness, productivity or nutrient avail-
ability. He stressed that the measure
should reflect per capita or per unit bio-
mass effects of the species, and the total,
not the partial, derivative of the change in
the community, characteristic with change
in the species, because the total deriv-
ative reflects indirect as well as direct ef-
fects. The change should be measured
after all Important feedbacks have had
time to occur. These feedbacks may be
more rapid in aquatic systems, but can
take longer in terrestrial systems such as
the deserts in which Jim Brown and col-
leagues found more unexpected results as
the experiment aged`. Tilman also argued
that the term keystone should not be re-
stricted to animals or a particular trophic
level, although consumers at hlgher fro-
phic levels would, a priori, seem more
likely to have larger per capita effects than
plants, which commonly have larger bio-
mass.

182

cb 1995. Elsevier Sclence Ltd 01645347/95/S09.50

TREE ool. IO. no. 5 May 1995

Jim Estes (University of California,
Santa Cruz, USA) drew upon his classic
studies of sea otters and their role in tro-
phic cascades that can completely alter
the structure of subtidal and even adjac-
ent Intertidal communities. These studies
have led to three questions of broad rel-
evance to other systems containing key-
stones: (1) How general are the direct
and indirect interactions? (2) What is the
breadth of effects on other members of the
community? (3) What are the evolution-
ary consequences? Since the vast majority
of systems in nature either have not been
or cannot be studied in comparable detail,
it is important to evaluate whether or not
keystone species and interactions can be
predicted a priori based on some aspect
of the predator, prey or ecosystem. In the
sea otter-kelp forest system, the keystone
predator is much larger and more mobile
than its primary prey. Estes argued for in-
creased use of 'natural experiments' and
'adaptive management' in evaluating the
potential importance of other large, mobile
consumers. His comparative surveys of
otter-urchin interactions throughout the
northeast Pacific have shown that prey
properties (e.g. whether urchins graze at-
tached or drift kelp) can also influence the
strength of these interactions and the
breadth of the cascading effects that they
generat e.
The theme of prey influence on key-
stones was continued by Juan Carlos
Castilla (Pontificia Universidad Cat6lica de
Chile, Santiago, Chile), who reviewed his
studies of the keystone carnivorous gas-
tropod, Concholepas, on the rocky shores
of Chile. In his studies, the snail's keystone
status depended on whether its prey could
be both eaten and bulldozed off the sub
strate. Mussels and barnacles were sus-
ceptible, hence Concholepas plays a key-
stone role where these prey are domi-
nants. Other intertidal communities are
dominated by a tunicate, which cannot be
bulldozed, and In these, the snail Is not a
keystone. Castilla also reviewed results
from studies of Homo sapiens, which feeds
heady on Concholepas, with consequences
that tipple through the intertidal food web.

.  Mary Power (University of California,
Berkeley, USA) emphasized that keystones
have been distinguished from other types
of strong interactors by having effects that
are much greater than would be expected
from their relative abundance. She also
pointed out that the original metaphor has
had a second useful connotation: keystone
specles, If experimentally perturbed, can
be keys to understanding the forces that
organize communities. Although it would
be useful to discover traits that identify
keystones apriori, she was skeptical that a
'field guide to strong interactors', as Steve
Carpenter @en. commun.) has called it,

could be compiled until we have a better
understanding of the context dependen-
cies of interaction strength. She illustrated
this point by reviewing her studies in
northern California rivers, in which the
keystone role of fish as top predators de-
pends on whether or not scouring floods
have occurred.
Bruce Menge (Oregon State University,
Corvallis, USA) supported the argument
that keystone species are context depen-
dent, citing a variety of studies from mar-
ine habitats including his own recent work.
He further noted that detecting keystone
species can be difficult, particularly when
several similar species are candidates for
potentially controlling a system. Remov-
ing a single suspected keystone from a
multispecies group of consumers could
lead to no change in the community for one
of three reasons: (1) the removed species
had no impact; (2) control was imposed
not by a single species but by the entire
group, the remaining members of which
compensated for the absent species; or
(3) none of the consumers had an effect.
To distinguish among these, Menge advo-
cated, where feasible and appropriate, a
combination of experiments that would

estimate both the collective trophic impact
of all consumers, and the isolated impacts
of single species (or individual trophic
groups). As yet, no species or system traits
reliably distinguish keystones from non-
keystones. Examples froni a wide variety
of habitats are available, however, which
could be mined for comparative guidelines
in efforts to identify the most likely strong
interactors in unstudied systems or sys-
tems being considered for management.
Scott Mills (University of Idaho,
Moscow, USA) raised the question of
whether keystone characteristics are di-
chotomous (some species are keystones,
the rest are wimps), or whether the ef-
fects of species on their Communities were
in fact more continuously distributed.
Presumably, a community with keystone
species would have 'community import-
ance' or interaction strength values dis-
tributed very differently from communities
that are traditionally modeled. He warned
of the dangers of the lackof an operational,
repeatable criterion for designating key-
stone status if this term is to be used by
decision makers. He stressed, however,
that the keystone phenomenon is real and
the need for application is great, so that

I/

Proportional biomass of species

Fig. l. Total (collective) impact of species versus its proportmd abundance. A point representing a species
whose total impact is proportional to its abundance would fall along Ute diagonal line x= y. Keystones have
effects that exceed their proportional abundances by some large factor. They also have a total effect whose
magnitude exceeds some absolute threshold. Therefore, although a rhinovirus that made wildebeests
sneeze (V,) might have a total effect that far exceeded that expected from its very low biomass, it would
not be a keystone if the total effect fell below the threshold. On the other hand, a distemper virus (V,) that
killed lions or wild dogs might have a collective effect of sufficient magnitude for keystone designation.
Pisaster (P), sea otters (0). the predatory snail Concholepas (C) and freshwater bass (B) have
disproportionately large impacts on their communities. Trees (T), giant kelp (K). prairie grasses (G) and
reef-building corals (C,), which dominate community biomass, have total impacts that are large, but not
disproportionate to their biomass. Quantitative values that should be prescribed for thresholds of absolute
total collective impact (vertical position) and factors by which keystone effects should exceed a species'
proportional abundance (distance left of the line x= y required for keystone status) may vary with the
community trait (e.g. species richness, biomass of other species or guilds. primary prwiuctivity. nutrient ol
soil retention, albedo) under consideration. In addition, interaction strengths. and therefore status as key
stones or dominants. may change for given species under different environmental circumstances. Impacts
of succession, changes in productivity. Or deletions of other species on the status of a species could be
represented by families of points representing the same species in different contexts.

NEWS & COMMENT

steps towards a clearer operational defi-
nition would be welcome.
Gretchen Daily (University of California,
Berkeley, USA) described how determin-
ing the consequences of a species deletion
requires overcoming some of the greatest
challenges in ecology: starting with a flash
of naturalists' intuition; bridging across
spatial scales to detect the influence of
individual behavior on characteristics of
communities and ecosystems; working
over potentially formidable timescales and
with a daunting diversity of taxa: and all
the while persevering with knowledge that
similar consequences may not manifest
themselves in seemingly similar systems.
She argued that, nonetheless, it is impera-
tive to incorporate the potentially dra-
matic role of subtle, indirect species inter-
actions when depicting communities in
conservation and management models.
This point was emphasized by William
Bond (University of Cape Town, Cape
Town, South Africa), who decried the lack
of contribution by ecologists of protocols
that would allow conservation priorities
to be set on the basis of interactions. Key-
stone, he pointed out, was the only single
name widely associated with interactions
that can conform to database formats
(`species i is a keystone, and it's here'). A
keystone can be pragmatically defined as
a species whose loss from or addition to
a system would change community com-
position, structure or function enough to
arouse concern. Keystones are best dem-
onstrated by experiments, which, unfortu-
nately, are slow, particularly in terrestrial

systems. Consequently, we do not yet know
whether deletions of possible keystone
mutualists (pollinators, seed dispersers)
would in fact alter population dynamics to
produce large effects on the rest of their
communities. In general, we lack and need
a protocol for evaluating probable key-
stone interactions for biodiversity conser-
vation, so that we can flag those species
involved and the species or key processes
that would be vulnerable to their loss.
During the subsequent days, the group
hammered out a verbal, and a potentially
operational, definition of keystone. The
verbal definition, which expands the orig-
inal definition to include interactions other
than trophic ones, was: `A keystone species
is a species whose impacts on its community
or ecosystem are large, and much larger than
would be expected from its abundance '. This
verbal definition distinguishes keystones
from other strong interactors whose effects
are due to their dominance of ecosystem
biomass. The potentially operational defi-
nition takes a step towards quantifying
`large, and larger than expected' by lor-
mally defining the community importance
of a species, deriving from this species'
total (collective) impact, and relating this
impact to its proportional biomass in the
community (Fig. 1). This definition dis-
tinguishes keystones from dominants and
other species. Its potential for examining
the context dependence of species inter-
action strengths and keystone status (for
example, how these change with the gain
or loss of other members of interaction
webs, or over the course of succession)

Behavioural brain research in natural and

semi-natural settings

will be explored in a paper by this group in
preparation for BioScience.
By developing a more operational defi-
nition, the workshop participants hope
to refocus the term keystone for ecologi-
cal research, and to make it more useful
for policy makers concerned with pre-
serving biodiversity. All agreed that it was
premature to prescribe `magic numbers'
(specified quantitative thresholds above
which species would be designated as
keystones). Nevertheless, there was grow-
ing excitement over research questions
that emerged or were clarified during dis-
cussion, and a general optimism that the
keystone concept was evolving towards a
form that could take an important place in
the conservation biologist's tool box.

Acknowledgements
We thank Adrian Sun for help in
preparing Fig. 1.

Mary E. Power

Dept of Integrative Biology, University of
California, Berkeley, CA 94720, USA

L. Scott Mills

Depf of Fish and Wildlifr, tJnioeni& of Idaho,
Moscow, ID 83844, USA

References

1 Paine, R.T. (1969)Am. Nat. 100,91-93
2 Mills. L.S. et ol. (1993) BioScience 43,219-224
3 Estes, J.A. and Palmisano, J.F. (1974) Science

4 Brown, H.H. and Heske, E.J. (1990) Science 250,

185, 1058-1060

1705-1 707

(3) the limit and complementarity of labor&
tory and field approaches, and (4) the func-
tions of the hippocampus. Neuroanatomy
methods were taught by specialists with
the aim of helping researchers to do in
one week what might take a year in the
absence of supervision. Behavioural tech-
niques were demonstrated, from homing

-. ' * ---.:&l. -nmhi.tipatpd