L : {)<j~> of- eaeu.le<h'e\C, _; e.ds . P. .:r. C-eJ P.

:oaL
6- . YL. c; ~weu_ . Pu. doc._... W~v.A~....J-rk..- N~fo.u..ol s
Proc. Adv. Study Inst. Dynamics Numbers Popul. (Oosterbeek, 1970) 298-312

On the role of natural enemies in preventing competitive exclusion in some

marine animals and in rain forest trees

J. H . Connell

Department of Biological Sciences, University of California, Santa Barbara, USA

Abstract

The following question is posed and discussed. Under what environmental conditions will the
action of natural enemies prevent competitive exclusion between species of their prey? My studies
on marine animal populations and rain forest trees both suggest that fluctuations in the physical
environment and subsequent effects on the supply of resources reduce the effectiveness of natural
enemies and allow the prey to increase to sufficiently high densities that they compete for resources.
Fluctuations in weather may (1) reduce the populations of enemies more than their prey, (2) allow
the prey to get a head start after an unfavourable season and (3) i·educe the degree of specialization
of the enemies. With lesser amplitude and/or more regularity of fluctuations, natural enemies are
more effective in preventing competition. However, these conditions also lead to stronger defenses
by the prey and greater effectiveness of the secondary enemies attacking the ones in question. In
regions of very favorable physical conditions, the only defense against natural enemies seems to
be rarity and widely scattered distributions, which together preclude competitive exclusion.

Interspecific competition is defined as the striving by members of two or more species, at

the same trophic level, for the same requisite which is in short supply. It is often invoked to
explain the organization of ecological communities, the evolution of closely related spe-
cies, character displacement where ranges overlap, etc. While competitive interactions
may be observed on occasion, the competitive exculsion or displacement of one species by
another from a local area has seldom been documented clearly, as pointed out by Miller
(1967). It is usually suggested that the reason why competitive exclusion is rarely seen is
that in the past, competitors have evolved divergent strategies to share the contested
resource. However, in many ecological communities, very similar species coexist with
little obvious divergence in character. They often appear to be using completely the re-
source required; they seem to be competing, but not excluding each other.
An alternative hypothesis to explain such coexistence is that potentially competing prey
populations are kept at such low numbers by very efficient predators that competition is
•1 greatly reduced. The prey coexist because no species reaches high enough density to
exclude another in competition. There have been several experimental tests of this hypoth-
esis, both in artificial populations such as crops, heavily grazed pastures or in the labora-
tory, and under more natural conditions. Probably the first experimental test of this idea
was performed by Charles Darwin (1859, p. 67-68): 'If turf which has long been mown,
and the case would be the same with tmf closely browsed by quadrupeds, be let to grow,
the more vigorous plants gradually kill the less vigorous, though fully grown plants; thus
out of twenty species growing on a little plot of mown turf (three feet by four) nine species

298
.:.
perished, fron1 the other species being allowed to grow up freely.'
Harper (1970) has reviewed and discussed later experiments in which herbivores were
removed from or added to highly modified vegetation, such as grazed pastures, ~pwn
swards and stands of introduced plant pests such as prickly pear or St. John's wort. The
intensity and timing of grazing and the selection of prey had profound effects on the con1-
position of the vegetation. In all these studies the implication was that grazing reduced the
populations of certain species, allowing others to persist or enter.
In the laboratory, Slobodkin (1964) showed that when Hydra and Chlorohydra were
kept in the light, Hydra became extinct. He then imposed 'predation' by removing a fixed
percentage of both species. As the percentage of predation increased, so did the time to
extinction of Hydra, until at 90 % predation the two species were coexisting when the
experiment was ended.
Under more natural conditions, removal of enemies has sometimes resulted in changes
which are attributed to increased competition. Examples are the removal of rabbits from
herbaceous stands in Britain (Harper, 1970), small mammals from herbaceous vegetation
inyoungplantntions of trees (Summerhnyes, 1941) and marine grazers or predators from
rocky shores (Connell, 1961b; Paine, 1966; Paine and Vadas, 1969).
In summary, it seems clear from these experimental manipulations in the laboratory, in
simplified field situations suc!h as crops and pastures, and in natural conmmnities, that
enemies can prevent competition between some of their prey species. The question is:
Under whnt conditions will this happen in natural ecological communities? Since the most
convincing case for the existence of 'organization' in an ecological corrununity is the opera-
tion of interactions such as competition and predation, it is important to try to define the
conditions that determine how and when these interactions operate effectively under
natural conditions.
In the following case histories, I will provide some evidence concerning the conditions
which determine whether predation will or will not be effective in preventing con1petition.
These studies were made in largely undisturbed natural communities in which the species
had their full complement of con1petitors, predators, parasites and food species.

Predation inefTectivc in preventing interspccific competition in intertidal barnacles

In a study of interspecific competition and predation in Scotland, I found that one species
of barnacle excluded another species fro111 most of the fotertidal shore (Connell, 1961 b).
The losing species, Chthamalus stellatus, settled normally only in the upper half of the
intertidal zone, and was displaced from it by direct crowding during growth of the winning
species, Ba/anus bala11oides. In this upper shore region, the predatory snail Thais lapillus
attacked ma.inly Ba/anus but did so only after the prey were at least 6 months old (Connell,
1961a). By this time, much of the competition between the barnacles had already occurred,
so that the predation could not lessen the competitive exclusion of Chthamalus by Ba/anus.
At lower shore levels, Thais began to attack the barnacles when they were only about a
month old. When I transplanted the competing barnacles to these lower shore levels, the
predation was m01·e effective in preventing competition. At each monthly census I made
notes of which Chthamalus were being crowded by Ba/anus. Table 1 shows the inortality of
Chthamalus caused by this crowding. Mortality from crowding was about the same at both
levels when predators were excluded by cages, but in natural populations the predators
reduced this crowding only at lower shore levels.

299
Table 1. The amount of mortality due solely to crowding in natural populations of C!zthamalus,
open to predation by Thais and protected from19fecfati'~ in adj~cent .en"tge·s-."Clrt.hmr;a/us stellat1J$.r· -'-' !J'
was competing with Ba/anus balanoides, the predator bemg Thars lapH/us, at M1llport, Scotland, -
1954 (Connell, 1961 b). Chthama/us dalli was competing with Ba/anus glandula and B . cariosus, the
predators being Thais emarginata and T. lamellosa at Friday Harbor, Washington, USA, 1964.
Chthamalus stellatus Clrtlwmalus dalli
upper shore lower shore 1 upper shore lower shore

open cage open cage open cage open cage

Number of sites 2 2 3 3 2 3 6 6
Number of Chthamalus in June 98 153 118 152 232 493 468 593
% mortality of Chthamalus by Sep-
tember, due to crowding by Ba/anus 31 39 13 32 6 22 10 37
1 These barnacles were on pieces of rock from the upper shore, transpanted to the lower shore.

The question then is: Why weren't the predators so effective at upper shore levels? Why
didn't they begin to attack the very young barnacles as they did at the lower levels?To
answer these questions, I will describe another example, from a similar comnmnity in
which the predators did this.

Predation effective in preventing interspecific competition in intertidal bnrnaclcs

On San Juan Island, Washington, USA, 3 species of barnacles live in the intertidal zone
with 3 species of their principal predators, snails of the genus T!tais (Connell, 1970). The
vertical distribution of the two species of Ba/anus on San Juan Island is virtually identical
with that at Millport, Scotland; adults of B. glandula were confined to a narrow band at
the top of the intertidal zone, similar to Chthamalus at Millport. B. cariosus occupies the
lower zone, like B. balanoides at Millport. But the reason why B. glandula only survived
high on the shore on San Juan Island was not because it was excluded from the lower
shore by competition with B. cariosus or any other competitor. Every year glandula settled
throughout the lower shore but was quickly eaten by predators. The predators could not
feed at the highest shore levels where adult glandula survived for several years. In situa-
tions where predators were absent, glandula survived throughout the intertidal zone. One
such habitat was at the head of quiet bays, where both predators and other barnacles were
rare (Connell, 1970).
In situations with at least moderate water movement, predators were common and fed
heavily on both species of Ba/anus when the barnacles were young. With this intense preda-
tion the barnacles were seldom allowed to get dense enough to compete for space; I have
seen crowding in B. cariosus occasionally, but never in B. glandula, or between the species.
I several times attempted to exclude Thais with cages, as I did in Millport, ~ut small Thais
invariably got inside cages and attacked B. glandu/a preferentially. I have never managed
to exclude predators for long enough to see which species of Ba/anus would win in com-
petition. Since cariosus grows sJightly faster than glandula, I would predict that cariosus
would win.
Occasionally, a heavy settlement of Ba/anus cariosus survives at high shore levels,
possibly because of unusually favorable cool weather during the settlement season. These

300
 may then crowd out the other species of barnacles because predation is much less intense
at very high levels. This ~vidently:>huppened to the 1963.settlement at the study area at San
Juan Island. When this spccies-i:eachcs the age of 2 years, it becomes invulnerable to attack '
by Thais in the intertidal zone, and so survives a long time. Since this only happens oc- , ·'
casionally, the age structure of the B. carfosus population consists of one or n1ore 'domi-
nant year classes', rather thnn the more evenly mixed age distribution of B. g/andula.
Another species of barnacle, Chthamalus dalli, occurs scattered tluoughout the inter-
tidal zone and not restricted to any particular level. In its small size and slow growth it
resembles C. stel/ntus at Mill port. I set up a series of cages to test the outcome of competi-
tion between C. dalli and the two species of Ba/anus, in the absence of predation. At each
monthly census I made notes of which Cltthamalus were being smothered, undercut or
crushed by adjacent Ba/anus. Table 1 shows the mortality from this con1petition for the
first swnmer when growth was most rapid. At both shore levels, mortality from crowding
was much higher inside the cages. The predators outside were eating the Ba/anus and thus
reducing the competition, as they did on the transplanted rocks at lower shore levels at
Millport. But on Snn Juan Island, the predators were effective at almost all shore levels in
keeping the barnacle populations below the point where they would compete for space.
Only occasionally at very high levels did B. cariosus escape predation long enough to be-
come dense and crowd out other competitors. Above this, in the narrow high band where
adult B. glandula survived in a refuge above its pl·edators, the barnacle population seldom
occupied all the space (Connell, 1970, Fig. 8). Probably the harsh physical conditions at
this high level kept the population sparse, as happened at Millport (Connell, 1961,
Table 13).
The predator responsible for most of the predation at upper shore levels was Thais
emarginata. It has a relatively thinner shell, larger opercular opening and shorter spire
'
' than either Thais lamel/osa, which lives at low levels on San Juan Island, or Thais lapillus
from Millport. As Kitching, Muntz and Ebling (1966) and Connell (1970, Table 17) have
shown, the shell characteristics of the latter two species protect them from predation by
shore crabs which feed at low shore levels. Evidently Thais emarginata has a refuge from
crab predation at the higher shore levels, and does not require a heavy shell. Since it is
reasonable to assume that constructing and carrying a thinner shell requires a lesser ex-
penditure of energy, this saving may com.pensate for the disadvantages both of taking
smaller sized prey and of having a shorter feeding period at high levels. Compared with
Thais lamellosa from lower shore levels, Thais emarginata is smaller, has a thinner shell,
1·eaches sexual maturity faster (1-! years a:s against 2! years for T. /amellosa) and lives a
shorter time (3 to 4 years as against 8 or more years for T. lamellosa).
This specialized strategy has a price; Thais emargi11ata cannot retreat to lower shore
levels if its food supply fails, because its shell structure renders it very vulnerable to preda-
tion by crabs. Therefore, it is dependent upon a very regular food supply. On San Juan
Island B. g/anc/11/a fulfills this requirement nicely; its recruitment is very regular (Connell,
1970, Fig. 4). Thus the prevention of competition at upper shore levels at San Juan Island
is probably due to the presence of a specialized predator which harvests its prey very
efficiently. The specialization is permitted by the existence of a dependable food supply,
B. g/andula.
In contrast, there is no such specialist predator at Millport, probably because the food
supply is much more variable there. B. balanoides often reaches densities of 60 to 80 per
cm 2 , but in so1ne years the settlement fails almost completely (Connell, 196la). A high

301
level specialist ......
like;11 Thais emarginata would
T\£ JllU1 lJL
oc,.c;asionally
_, _...sen. .
starve .in Millport. Apparently,
,·
the reason for the irregular barnacle recrmtment m Scotland is due to the very shoi:t t.
season of J~al-=settle~~~t. The popul~tfon of B. balanoides liberates its larvae in two
successive waves within one to one and a half months in early spring, in response to the
phytoplankton bloom (Baxnes, 1957). In some years, the first liberation perishes when the
early bloom falls due to weather conditions. Occasionally both liberations perish and
settlement is almost non-existent (Barnes, 1956; Connell, 1961a).
In contrast, the larval settlement of B. glandula on the Pacific Coast extends over six
months. It is much less likely that weather catastrophes could completely eliminate the
larval supply if settlement extends over six months, than if it occurs over a short period as
in Scotland. As a consequence, recruitment of B. glandula was very regular.
Thus, the predator-prey system in Scotland is geared to a highly seasonal, rather un-
dependable resource supply. There is a short intense spring bloom of phytoplankton,
which demands close synchrony of liberation of the barnaclelarvae,andthereare occasion-
al failures due to weather catastrophes. This system limits the degree of specialization of
the predators, and decreases their efficiency in harvesting their prey in certain habitats.
With less efficient predation and short intense recruitment of prey, strong competition
develops between the prey species.
At San Juan Island, less seasonality in primary productivity probably allows a longer
period over which larvae can be released into the plankton. This reduces the uncertainty
of recruitment of prey to the shore population and permits specialization of the predator.
It can keep up with the prey, thinning it out and preventing competition for space, as the
predator Pisaster did in Paine's (1966) study.

On the prevention of competitive exclusion in rain forest trees

In contrast to the barnacles which were kept by their predators below the maximum level
allowed by the resources available, trees in rain forests obviously compete for light and
possibly for other resources. Their natural enemies, as a group, don't prevent competition
between the trees. However, the enemies may prevent competitive exclusion of one species
of tree by another, by preventing any one from forming a single-species aggregation and
so displacing other species from the particular area.
The mechanism I suggest is that each tree species has host-specific enemies which attack
it and any of its offspring which are close to the parent. The healthy parent tree supports a
large population of these enemies without itself being killed, but the seedlings, whose
growth is suppressed in the heavy shade, succumb to the attack of insects and other ene-
mies which come from the parent tree itself or the soil below it. A similar hypothesis,
though differing in certain details, bas been proposed independently by Janzen (1970).
So the only way a species of tree can reproduce in these forests is to disperse its seed-
lings far enough from the parent to allow them to escape attack for long enough to reach the
size when they can withstand attack by enemies. If all enemies of trees were removed from
an entire forest, each species would probably form small groves, and the more rapidly
growing species would gradually spread over the habitat it is best suited for, excluding the
slower growing or less well adapted species in that area. The final result would be a lower
'pattern' diversity (less intermingling of the different species) and as a consequence, fewer
species in any local area of forest.
The following is a preliminary account ofwo;rk which I have done with J. G. Tracey and

302
 L. J. Web_~ in rain forests in Qt:c:.ensland, Australia, in an effort to test this hypothesis
(Connell, Ttatey ~nd Web1;2, in prep.). V\Te mapped.the trees in two rain forests, the la.UL<1foe" ......
ones( > ' 10 cm in diruneter at bi:east height (DBH)) on areas of about 1.7 hectares, the
smaller ones in strips within these plots.
Table 2 illustrates how the diversity increases with size and age of trees in the plot in
northern Queensland. Seedlings up to 0.1 111 high (less than about 4 years old) had a very
low 'pattern diversity' (Pielou, 1966); that is, they tended to occur in single-species clum.ps
rather than mingled with other species. (This san1pling was done by taking the first and
second nearest neighbours of each individual and recording the number of species in the
'triplets'.) In young trees of the next larger size class, 0.11 to 0.19 m high, the pattern diver-
sity increased markedly, and by the time they were over 0.8 m high, their pattern diversity
had approached that of the adult trees. Pielou's index of pattern divernity takes into
account the number of species and their relative abundance. We have given reasons
(ConneJJ, Tracey and Webb, in prep.) why a high pattern diversity should lead to a greater
species diversity and 'evenness' as shown in Table 2.

Table 2. Diversity of trees in rain forest at Davies Creek, North Queensland, Australia, Iat.
17° S, elevation 850 m. Saplings less than 2.5 cm in diameter (breast height )were divided into five
height classes as equal in number as possible. All the data refer to August, 1969, except those in the
second column from the right, and for trees larger than 2.5 cm in diameter. Pattern diversity was
calculated ns shown in Pielou (1966): for trees 0.11 to 6.34 m high, the 1969 figures are from the
same group of trees as in 1965, minus those which died in the interval.
Height class Number of Number of Diversity H Evenness J Pattern diversity D
(m) trees species (base e)
Aug. 65 Aug. 69
~
.
I 0 -0.10 1080 71 2.05 0.39 0.51
0.11-0.19 1038 115 3.47 0.66 0.72 0.77
0.20-0.35 983 103 3.38 0.64 0.76 0.82
0.36-0.79 1040 111 3.60 0.68 0.87 0.89
0.80-6.34 1093 131 3.80 0.72 0.93 0.93

Diameter
class (cm)

2.5-9.9 956 116 3.79 0.72 0.96

10.0 + 1502 126 4.31 0.81 0.98

The pattern diversity increases not only with size but also with age, as shown in Table 2
by the changes between 1965 and 1969. The increase in pattel'n diversity was greater in the
smaller size classes where mortality was greater. This suggests that mortality falls more
heavily on trees which are next to others of the same species, than on those intermingled
with other species. Since we had mapped every individual in 1965, we could test this hy-
pothesis by tabulating for each tree whether its nearest neighbour was the same or a differ-
ent species. As shown in Table 3, the small saplings having the same species as a nearest
neighbour had a higher mortality rate than if they had a different species as neighbour.
Larger saplings had a lower mortality and the species of the neighbour had no effect.
Adult trees also have a deleterious effect on smaller trees of the same species. We planted

I 303
.. ~
~" . Table 3. Mortality of tree saplings in rain forest in relation to the species of their nearest neigh-
bour within the samer-s1ze c_lass. August 1965 to August 1969, Davies Creek, l-lorth Queens1~ 1 It
Australia, lat. 17° S .. elevat1on 850 m. ·
Height Species of nearest neighbour
class (m)
same species different species

0.11-0.19 Number living in 1965 64 148

Number dying by 1969 24 27
% mortality in 4 years 37.5 18.2
0.20-0.35 Number living in 1965 153 435
Number dying by 1969 33 49
% mortality in 4 years 21.6 11.2
0.36-0.79 Number living in 1965 100 548
Number dying by 1969 11 36
% mortality in 4 years 11.0 6.6
0.80-6.34 Number living in 1965 59 604
Number dying by 1969 2 18
% mortality in 4 years 3.4 3.0

seeds of Planchonella sp. nov. in four plots of 2 x 1 m each under two adjacent adult trees,
and in four other plots nearby under adult trees of other species. In each place we cut the
roots around 2 of the plots to eliminate competition for nutrients with the adult trees.
Over 80 % of the seedlings became established in all plots. The results in Table 4 indicate
that mortality of the young trees during the next 3 years was much greater under adults of
the same species than under those of other species. Removal of roots had no consistent
effect; it certainly did not improve the survival under adults of the same species.
In other experiments we placed the seeds of other species on the surface instead of
burying them, both under adults of the same species and of different species; some were
protected by cages of 1 cm mesh chicken wire. To test the effect of density, we placed
seeds one meter apart in a line through the forest as a 'sparse' treatment for comparison
with the 'dense' treatments described above where the seeds were 100 per n1 2 • An insecti-
cide spraying treatment wasn't effective; indications of insect attacks were as prevalent in

Table 4. Mortality between 2 February 1967 and 4 April 1970 of seedlings of Planchonella sp. nov.
at Davies Creek, North Queensland, Australia, Iat. 17°S, elevation 850 m. Seeds were planted un-
der mature trees of the same species and under trees of other species. Each of the eight plots was
2 x 1 m and contained 98 seeds. Half of the plots were trenched, all roots being cut to a depth of
0.3 m.
Number dying out of 98
under adults of under adults of a
the same species different species
plot 1 plot 2 plot 1 plot 2

Trenched 65 69 14 26
Not trenched 50 54 25 42

304
these plots as in the other treatments. Seeds of Cryptocarya cormgata, from the plot itself,
and EugeJtia. brachya11dra, brought from the other plot at 27 ° S, were used. None was used
.. ,h r \,.~---1-. •lfr!I/)£
jf it had already l;>een attacked b Y insects. As shown in Table 5, almost every seed was. · .
killed with~iii a year, regardless of its position or density. Since seeds were killed insidenaa
cages and since many seeds had holes caused by insect attack, it is probable that insects
were the main cause of death. For these species, at least, the increase in patten1 diversity
evidently occurs after germination, not in the seed stage. That is, all seeds are probably
killed most of the time, but in occasional years a seed crop may escape when, for some
reason, the seed eaters are reduced. Then a great nwnber of new seedlings becomes estab-
lished in a patch, and these are attacked and thinned out over the succeeding years. The
fact that both the evenness component of diversity and the pattern diversity were very low
in the young trees which entered the population between 1965 and 1969 (Table 2, height
class 0 to 0.1 m), is evidence that recruitment followed this pattern on the study site. This
model differs from that pr oposed by Janzen (1970). He suggests that the probability of a
fallen seed surviving is a: function either of its distance from the parent or its population
density. In the species I studied, neither was related to survival. However, we agree on the
fundamental point that overall survival should be greater at greater distances from the
parent. My data indicate that the mol'tality which increases pattern diversity occurs mainly
after germination during the seedling and sapling stages.
Evidently, mortality is greater when individuals of the sarne species grow together than
when they are intermingled with other species. At least three possible mechanisms could
produce this effect. Firstly, each species might secrete toxins into the environment which
poison individuals of its own species more than those of different species. Although this

Table 5. Survival of seeds of two species of rainforest trees, placed on the surface of the ground at
the Davies Creek plot, North Queensland, 17°S. lat., 850 m. Seeds of Cryptocarya corrugata were
from this plot. Eugenia bracltyandra seeds were from South Queensland, 27 °S. !at., 850 m. The
figures are numbers of seeds at the start of the experiment in September, 1969, and of germinated
seedlings in June, 1970.
Under adults of Under adults of
same species other species
Sep. 69 June 70 Sep. 69 June 70

C. COl'l'ugata, 'dense', 100/m2

Control, no treatment 134 1 130 0
Under wire mesh roof 75 0 75 0
Inside wire mesh cage 75 0 75 0
Insecticide, 0 .1 % Dieldrin 141 0 129 2

Sep. 69 June 70
C. corrugata
'sparse', one seed every meter in a line through
the forest 100 0

Eugenia braclzyandra
'dense', IOO/m 2 100 0
'sparse', one every meter in a line through the
forest 100 1

305
 remains a possibility, it seems less likely than the others. T he second possibility is that each
t!JJI
'.'pecies has such specialized requirements for nutrients tnat competition is more sever0 "!:>e- "
twee~ :Pi~nts of the same species than with plants of different species. However, the resu!ts ,,....;:, .. ~
of the experiment in Table 4 seem to exclude the possibility that intraspecies root com-
petition with adults was more deleterious than interspecies competition. Competition
between adults and young trees in these dense rain forests is probably primarily for light,
not for soil nutrients or water. Competition for light should not be more severe within than
between species, unless different species produce qualitatively different sorts of shade. On
the basis of present evidence this mechanism seems less likely than the following one.
The third possibility is that attack by natural enemies is heaviest when trees live in single-
species aggregations, rather than mixed with other species. There has been, to my knowl-
edge, no direct experimental test of this hypothesis, but there is much indirect evidence
(Gi11ett, 1962). For exan1ple, single-species stands such as rubber plantations were un-
successful in their native Brazil where they were heavily attacked by disease. But, when
introduced into Asia without their native enemies, they could be grown in plantations. In
our own work (Table 4) there was much more grazing on the seedHngs planted under
adults of the same species than on those under other species.
For this mechanism to be effective the enemies must be both specialized in their choice
of prey and not particularly mobile. Populations of insects or fungi centering their atten-
tion on an adult tree would probably be the most effective at producing the high pattern
diversity we observed. Obviously, other factors, particularly heterogeneity in space and
time, provide other opportunities for the maintenance of diversity in rain forest trees. I
suggest that the role of predation is important in preventing trees from forming single-
species groves, which is probably the only way in which one species of tree could ex.elude
others by interspecies competition.

Conditions determining the effectiveness of natural enemies in preventing competitive

exclusion in their prey

What circumstances determine whether natural enemies can prevent competitive exclusion
between prey species? From the examples discussed above and from the studies of other
workers, it seems clear that fluctuations in the physical environment decrease the effective-
ness of natural enemies in several ways.
Firstly, prey populations aye apparently less vulnerable to the d irect effect of extreme
fluctuations in weather, than are their enemies. For example, Michelbacher and Leighly
(1940), Lord and MacPhee (1953) and DeBach, Fisher and Landi (1955) all found that
extremes of weather killed a much higher proportion of the natural enemies than of their
prey, herbivorous insects. Why this should be so isn't known, although it is probable that
the enemy would be more ~xp~sed to the effects of weather when it was searching than
would its prey, sheltered by the vegetation. However, this is not the whole story since
much of the mortality occurred during winter when the parasite was a larva inside the
hibernating host (Lord and MacPhee, 1953). On the intertidal shore, barnacles extend
higher than their predators, evidently being better able to withstand the extremes of
weather at the very high levels.
Secondly, fluctuations of the weather can reduce the effectiveness of natui-al enemies in
less direct ways. After the reduction in numbers or activity of both predator and prey
4.ur4ig an u_pfavorab~e S!iason, the_prey . PqP1:Jl~~i~ns µiay recover. a~d grow quickly to
306
densities at which they begin to compete with each other before the predator populations
have-increased enough to control. them:+lfttis+evfdently happens""''a"t'1iigh latitudes w.Qer.e _·-
phytoplankton blooms or huge swarms of insects occur in the spring. The favorable season
may be so short that the natural enemies don't have time to increase enough to control
their prey. In this respect the terrestrial high arctic resembles the high intertidal zone where
there is too short a time for the predators to feed effectively. At the lower edge of the inter-
tidal zone, the predators can feed almost continuously, as they can in those parts of the
terrestrial wet tropics with little seasonal change. Even though the natural enemies have
evolved to respond rapidly to such spring increases in their prey - as for example the re-
lease of barnacle larvae in response to an increase in phytoplankton, or the spring m..igra-
tion of insectivorous birds into the arctic - the prey may still outstrip them for a while.
Thirdly, irregular or unpredictable fluctuations in the physical environment may cause
such unpredictable variations in the supply of some critical resource, such as the numbers
of a species of prey, that the natural enemy cannot afford to specialize on it. This is prob-
ably the reason why Thais lapillus has not specialized to the extent that Thais emarginata
has, with the consequence that T. lapillus cannot attack the barnacles on the upper shore
early enough to prevent competition. Whether specialist predators are usually n1ore effec-
tive in reducing their prey than generalists is hard to say at this time. In Paine's (1966)
example, the large predatory starfish, P;saster ochraceus, kept the nrnssel populations
from competitively excluding other species. Pisaster is specialized in feeding if given several
species of prey in equal abundance, showing a preference for mussels (Landenberger,
1968). Thus one would predict that when Pisaster first moved into a new area mussels
would be conunon, and it would feed as a specialist on them, reducing their competitive
effect. Thereafter, as shown by Paine (1966), it would feed more generally on a greater
variety of the species of sedentary animals available. Thus when Pisaster was reducing the
dominant prey it was acting as a specialist.
Other independent evidence that specialists are more effective than generalists in re-
ducing their prey is, unfortunately, lacking. Although it is often stated (DeBach et al.,
1964) that the best biological control of pests is done by highly specialized ene1nies, there
has never, to my knowledge, been a test of this idea. Specialists are used in biological con-
trol, not because they are known from independent evidence to be more effective than
generalists, but because they are not likely to attack beneficial plants or animals after they
are introduced to a new country.
Another exrunple of differences in degree of specialization is the common observation
that species nearer the 'top' of food webs tend to be less specialized in their choice of food.
For example, Paine (1963, 1966) has found that whereas small predatory snails in the
rocky intertidal habitat eat only a few species of suspension feeding or herbivorous grazing
animals, the larger snails and starfish eat more species, including both the smaller pred-
atory snails and their prey. Brues (1946) points out that herbivorous insects often feed
on only one or a few species of plants, whereas many predatory insects attack many species
of herbivorous insects. Also, he states that entomophagous parasitoid insects usually have
fewer species of hosts than the hyperparasites, which often attack both the herbivorous
host and the entomophagous parasitoid. If, as discussed earlier, unfavorable weather
causes less fluctuation of populations at lower than at higher trophic levels, this would
permit greater specialization of animals feeding on organisms at lower levels. Thus for
several reasons, one might expect greater specialization and presumably greater effective-
ness of natural enemies at lower trophic levels.

307
 In summary in the sort of environment with fewer irregular occurrences of extreme
weather
. ,
:
and I~wer amplltu~ofvariations
. . ~. "
in temperallu-e, rainfall, etc.,.natural enemies
wili be most effective in preventing competitive exclusion in their prey. In such places the
enemies should be more specialized, especially at lower trophic levels. There will be less
chance for the prey to increase suddenly until it is beyond control of the predator, since
the unfavorable seasons will be reduced in length and severity.
However, these same environmental advantages also work to diminish the effectiveness
of natural enemies in at least two ways. Firstly, there will be strong selection on the prey to
evolve defences which not only protect them from existing enemies, but also have the
effect of reducing the number of species attacking them. Defences of the prey, for example
plants with poisonous chemical substances or spines, or animals with armor, protective
coloration, etc., require specialized methods of attack by the predator. Such a requirement
for specialization must reduce the number of species of predator which can attack a
particular species of prey. With fewer species of natural enemies there is a greater chance
of the prey escaping. There is evidence for this from work in biological control; a pest of
the olive, which was somewhat reduced during the cooler seasons by one species of enemy,
was almost eliminated when another enemy species was introduced which attacked it in the
summer (Huffaker and Kennett, 1966). Holling (1959) indicated that control of the Euro-
pean pine sawfly in Canada was more likely with seveJ·al species of manunalian predators,
each of which was most effective at a different prey density.
Secondly, such favorable environments also increase the effectiveness of the species of
predators or parasites which attack the natural enemies in question. Since even large
carnivores have parasites, and general predators feed on hyperparasites, no species is
without its complement of natural enemies. For this reason, as natural enemies become
more effective in inore favourable climates, it becomes more difficult to predict whether, for
example, plants will be reduced more significantly by herbivores than the herbivores by
their predators or parasites. In some instances, such as in the rain forests I have studied,
herbivores seem to be reducing competitive exclusion of the plants. But in the second
growth of dry savannah woodlands in Central America, ants living on swollen-thorn
Acacia trees kill both the herbivorous insects and the nearby competing vegetation
(Janzen, 1966). In any event, in places with a more predictable or less fluctuating physical
environment, attacks by natural enemies at all trophic levels would be expected to be more
intense and continuous. The great increase in incidence and intensity of parasitism in
humans as one approaches the equator is probably the best available evidence of this trend
(LaPage, 1963).
In contrast, in areas with extrerne and irregular fluctuation in climate the prey often
excape their natural enemies and reach densities at which they compete for resources. As
the amplitude and irregularity are reduced, natural enemies becon1e more effective. In a
completely constant benign physical environment, natural enemies should be very special-
ized and effective in finding almost all prey. In this situation, the only way in which any
species could exist would be to be so rare and widely scattered that some individuals
escape being found by their specialized natural enemies, at least until they are well estab-
lished and reproductively mature. Because no environment is constantly benign, no
ecological community is so organized. However, as one approaches such conditions in a
tropical rain forest or coral reef, species certainly do become scarcer and more widely
scattered amongst individuals of other species. I believe that increased effectiveness of
natural enemies accounts for this trend.

308
Acknowledgments -
.......
Many people helped in these observations and exp-eriments and criticized the manuscript.
Together with those who arranged the m.eeting in Holland and who discussed the paper
after I had delivered it, I would like to thank J. Choat, M. Hopkins, R. Knowlton, B.
Menge, W. Murdoch, C. Peterson, D. Potts, T. Southwood, J. Tracey, L. Webb and J.
White. The work was supported by grants from the John Simon Guggenheim Foundation,
the Office of Naval Research and the National Science Foundation.

References

Barnes, H., 1956. Ba/anus balanoides in the Firth of Clyde: the development and annual variation
in the larval population and the causative factors. J. Anim. Ecol. 25: 72- 84.
Barnes, H., 1957. Processes of restoration and synchronization in marine ecology. The spring dia-
tom increase and the 'spawning' of the common barnacle, Ba/anus balanoides (L.). Ann. Biol.
33 : 67-85.
Brues, C. T., 1946. Insect dietary. Harvard University Press, Cambridge (Mass.), pp. 466.
Connell, J. H., 1961a. Effects of competition, predation by Thais lapillus and other factors on
natural populations of the barnacle Ba/anus balanoides. Ecol. Monogr. 31: 61- 104.
Co11nell. J. H ., 1961 b. The influence of interspecific competition and other factors on the distribu-
tion of the barnacle Clzthamalus stel/atus. Ecology 42: 710-723.
Connell, J. H., 1970. A predator-prey system in the marine intertidal region. I. Ba/anus glandula
and several predatory species of Thais. Ecol. Monogr. 40: 49-78.
Darwin, Charles, 1859. On the origin of species. First ed. J . Murray, London.
DeBach, Paul (Ed.), 1964. Biological control of insect pests and weeds. Reinhold Puhl. Co., New
York. pp. 844.
DeBach, P ., T. W. Fisher & J. Landi, 1955. Some effects of meteorological factors on all stages of
Aphytis lingnanens/s, a parasite of the California Red Scale. Ecology 36: 743-753.
Gillett, J.B., 1962. Pest pressure, an underestimated factor in evolution. In : Taxonomy and geo-
graphy. System. Assoc. Puhl. No 4: 37-46.
Harper, John L., 1970. The role of predation in vegetational diversity. Brookhaven Natn. Lab.
Symp. Biol. No 22: 48- 62.
Holling, C. S., 1959. The components of predation as revealed by a study of the small-mammal
predation of the European Pine Saw.fly. Can. Ent. 91: 293-320.
Huffaker, C. B. & C. E. Kennett, 1966. Biological control of Parlatoria oleae (CoLVEE) through the
compensatory action of two introduced parasites. Hilgardia 37: 283-335.
Janzen, D . H., 1966. Coevolution of mutualism between ants and Acacias in Central America.
Evolution 20: 249- 275.
Janzen, D. H ., 1970. The role of herbivores in tropical tree species diversity. Am. Nat. (in press).
Kitching, J . A., L. Muntz & F. J. Ebling, 1966. The ecology of Lough In.e. XV. The ecological
significance of shell and body forms in Nucella. J. Anim. Ecol. 35: 192-301.
LaPage, Geoffrey, 1963. Animals parasitic in man. Dover Press, New York.
Lord, F. T. & A . W . MacPhee, 1953. The influence of spray programs on the fauna of apple or-
chards in Nova Scotia. YI. Low temperatures and the natural control of the Oystershell scale,
Lepidosaphes ulmi (L.) (Homoptera: Coccidae). Can. Entom. 85: 282- 291.
Landenberger, D. E., 1968. Studies on selective feeding in the Pacific starfish Pisaster in southern
California. Ecology 49: 1062-1075.
Michelbacher, A. E. & J. Leighly, 1940. The apparent climatic limitations of the alfalfa weevil in
California. Hilgardia 13: 103-139.
Miller, R . S., 1967. Pattern and process in competition. Adv. Ecol. Res. 4: 1-74.
Paine, R. T., 1963. Trophic relationships of 8 sympatric predatory gastropods. Ecology 44:
63-73.
Paine, R . T., 1966. Food web complexity and species diversity. Am. Nat. 100: 65-75.
Paine, R. T. & R. L. Vadas, 1969. The effects of grazing by sea urchins, Strongylocentrotus spp.,
on bentbic algal populations. Limnol. Oceanogr. 14: 710-719.

309
Pielou, E. C., 1966. Species-diversity and pattern-diversity in the study of ecological succession..., .., -·
J. Theor. Biol. 10: 370-383. J ;,,,
Slobodkin, L. B., 1964. Experimental pc pulations of Hydrida. J. Eco/. 52 (suppl.): 131-148.
0

Summerhayes, V. S., 1941. The effect of voles (Microtus agrestis) on vegetation. J. Eco/. 29: 14-48.

Discussion

Participants: Connell (Author), den Boer {P. J.), Cavers, Clough, Jacobs, Jain, Kuenen,
Murdoch, Pimentel, Rosenzweig, Solomon, Turnock and Watt

This work seems to be in keeping with a generalization, that in ecosystems which have
evolved in stable environments for a long time, species maintenance is more by subtle
behavioral or biochemical mechanisms than is the case in immature north and south tem-
perate zone ecosystems in unpredictable environments. The fact that there are a great
number of poisonous fish species in shallow coral seas, whereas these are absent in e.g.
the Great Lakes, which are highly unstable, may bear on this (WATT). However, many
poisonous species of fungi (mushrooms) are found in the subarctic Scandinavian forests;
what is the position in the tropics (CLOUGH)?
The tropic zone may well be less variable than the temperate and arctic zones, but it is
not clear that it is a more predictable environment (MURDOCH). The level of differences
between seasons in arctic and boreal continental l·egions may be highly predictable.
Moreover, in connection with the variability of tropic and temperate zones the influence
of natural selection may have been such that animals in some tropic regions are about
equally.susceptible to the rather subtle weather changes prevailing there as are animals in
temperate regions to more pronounced weather changes (DEN BOER). I agree with this
(AUTHOR).

If one starts with a clumped distribution (as was the case with the small rain forest trees),
and then allows mortality work at random, would not the spatial pattern diversity index
increase towards 1? This would seem to mean that the supposed neighbourhood (intra-
specific) competition is not needed to get an increase of spatial diversity with increase of
tree size (JA COBS). This could easily be simulated. But random mortality should not kill trees
with nearest neighbours of the same species more than those with nearest neighbours of
different species, as was shown to be the case. And if mortality is random, pattern diver-
sity should stay the same (AUTHOR).
It shown that species diversity and pattern diversity increase with tree height classes.
But what is the possible effect of patterns of diversity between different tree heights on the
changes in overall diversity measures (JAIN)? The neighbours used were all within the same
height class. But the two nearest neighbours which were smaller than the central individual
could be used to get some idea of the influence of larger trees on the pattern of smaller
ones (AUTHOR).
How do you explain there being a higher number of species in the taller group than in
the lowest height class, if only random mortality is involved (JAIN)? The height classes are
not age classes, and the larger classes probably represent much longer periods of time.
The smallest class are offspring only of those species which reproduced successfully over
the past few years (AurHOR).

Healthy phytophagous sawfly larvae on larch trees do not drop from trees, even during

310
rather violent storms, unless they have completed feeding or are depleting the foliage.
Therefore, it seems unlikely that small trees under large ones, which support low denslties
of phytophagous ~fpecies will be destroyed by this means (TURNOCK). In the tropics tr~es
might 'declare a territority' by raining insects on their surroundings, by abscission of
twigs and leaves having insects on them (ROSENZWEIG).
To explain a presumed non-random mortality of seedlings under a tree of the san1e spe-
cies it is not necessary to assume the presence of specialized feeders. Each phytophagous
insect coming down from the tree overhead will find seedlings of the same species more
available than those of other species; moreover, it is very probable that even extreme gen-
•
eralists will show some preferences. There may be another complication; in temperate
regions many phytophagous insects show a definite preference for small and young trees;
to give an example: whereas nearly all leaves of the available young birches J.nay be rolled
up by Deporaus betulae (or other Rhynchitini), mature trees are hardly infested. How does
this compare with the tropics? (DEN BOER). The problem is the meaning of 'more availa-
ble'. Several species of young trees occur under a single large tree; if they are all of about
the same size, they are equally available to any phytophagous insect. If the insects eat them
without preference, i.e. in the same proportion in which they occur~ those of the same
species as the adult tree above need not be removed proportionately faster there than
under adults of a different species. To produce the results shown in Table 4, the insects
must specia1ize to some degree on young trees of the same species as the adult. Since most
insects do show some preferences, this will probably happen. In regard to the second point,
a greater rate of attack on young trees would intensify the removal of seedlings; I have no
data on whether this tendency also occurs in the tropics, but I expect that it does. AUTHOR).
If something tends to prevent the growth of seedlings under a mature tree of the same
species, this may be of advantage to the species, by giving preference to its seedlings a few
meters away where the species has not depleted the soil of trace elements, etc., correspond-
ing to its particular needs (SOLOMON). The experiment shown in Table 4 indicates that
competition for soil nutrients between seedling and parent is not important in the species
studied (AUTHOR).
There are examples of progeny arising only from the exact site of the parent plant.
Koller and Roth (Am . J. Bot. 51, 1964) working with Gymnarrhena in the desert areas of
Israel found that two kinds of seeds are produced (amphicarpy). One kind is light, aerial
and wind dfapersed. The other kind is heavy and is produced below ground; heJ·e the off-
spring actually grows up through the dead tissues of the parent plant. This site would be
the most favourable for the species in a very unfavourable environment (CAVERs).

Is it not possible for two competitors to coexfat although they are feeding on the srune
living resource if the fobct~ ost can evolve differentially to the two competitors (PIMENTEL)?
Perhaps, but this is not"" possible in plants whose food does not evolve. Probably it will
occur in son1e animals, but the barnacles studied were competing for a resource which
does not evolve either, namely space (AUTHOR). See also the discussion after the paper of
Pimentel.
A few Chthalamus individuals survive and reproduce in the region of competition with
Balanus, but they are smaller and so contain fewer larvae than uncrowded ones (AUTHOR
to JAJN).
Could there be any selective pressure toward a greater heterogeneity of dispersion; com-
pare Birch: Cactoblastis-Opuntia (JAIN). After a period in the plankton the offspring do not

311
nece~sru:ily return to the habitat of the adults. Most of the offspring come from the up.Q~lf:.c:s
shore zone where there is no competition with Balanus, and where this presumf"d selective
advantage of aggregation would be less (AUTHOR).
Studying more cases and more species, experimentally, on rocky coasts may indeed
upset some of the conclusions so far reached (AUTHOR to KUENEN).

312

- - ------- ----~