Population Ecology: A Unified Study of Animals and Plants

C H A P T E R 4: INTERSPECIFIC COMPETITION 91

ourselves only with its current, generally accepted 1 The ecological niche of a species has been defined
meaning, originating with Hutchinson (1957). from the species' point of view. This allows it to be
contrasted with the term 'habitat', which is an objec-
If we consider a single environmental parameter tive description-in n-dimensions if need be-of the
(e.g. temperature), then a species will only be able to environment itself. Thus, although both niche and
survive and reproduce within certain temperature habitat are defined in terms of environmental param-
limits. This range of temperature is the species' eters, the term niche characterizesa species,while the
ecological niche in one dimension (Fig. 4.15a). If we term habitat characterizes an environment within
also consider the range of humidities in which the which many species may live.
species can survive and reproduce, then the niche 2 Hutchinson gave special consideration to one par-
becomes two-dimensional and can be visualized as an ticular type of environmental parameter: interspecific
area (Fig. 4.15b); and if a third dimension is added competitors. The connections between interspecific
(food particle size in Fig. 4 . 1 5 ~ )then the niche be- competition and the ecological niche should become
comes a volume. Yet it is clear that there are many much clearer below, but we can outline Hutchinson's
biotic and abiotic parameters affecting a species: the ideas at this stage. He called the niche of a species in
number of niche dimensions (n)greatly exceeds three. the absence of competitors from other species its
We cannot visualize such a situation; but we can, by fundamental niche, i.e. the niche which it could poten-
analogy with the three-dimensional model, consider a tially occupy. In the presence of competitors the
species' ecological niche to be an 'n-dimensional species is restricted to a realized niche, the precise
hypervolume' within which a positive contribution to nature of which is determined by which competing
future generations can be made. And this, in essence, species are present. In other words, Hutchinson
is Hutchinson's conception of the niche. There are, wished to stress that interspecificcompetition reduced
however, several additional, important points which contributions to future generations, and that as a
must be understood. result of interspecific competition, contributions in
certain parts of a species' fundamental niche might be
Fig. 4.15 Building up the n-dimensional hypervolume: reduced to zero. These parts of the fundamental niche
ecological niches. (a) in one dimension (temperature); (b)in are absent from the realized niche.
two dimensions (temperature and humidity);(c) in three 3 A species' niche might also be restricted in practice
dimensions (temperature, humidity and food particle size). by the habitat: parts of a species' niche-fundamental
or realized-which are simply not present at a particu-
lar location (in space and time) become, temporarily,
irrelevant. This is particularly pertinent to laboratory
investigations, in which the simplicity of the habitat
confines species to those parts of their niche which
happen to be provided by the experimenter.
4 Finally, it must be stressed that the Hutchinsonian
niche is not supposed to be a literal description of a
species' relationship with its environment. Like other
models, it is designed to help us think, in useful terms,
about an immensely complex interaction. For in-

stance, the perpendicularity of the volume in
Fig. 4 . 1 5 ~is a result of the assumption that the three
parameters are independent. Interactionof variables-
almost certainly the general rule-would lead to a less
regularly shaped niche;but in practice a perpendicular

92 PART 2 : INTERSPECIFIC INTERACTIONS

visualization may be just as useful. Similarly, the solid monospecific populations under a variety of environ-
lines around the niche in Fig. 4.15 ignore the existence mental conditions in laboratory cultures of ordinary
of variability within a species. Each individual has a flour (Fig. 4.16~).In mixed culture of this type, how-
niche, and a species' niche is, in effect, the superim- ever, T. confusum always eliminates 0. surinamensis;
position of as many niches as there are individuals: a only T. confusum has a realized niche (Fig. 4.16d). In
species' niche should have very blurred edges. Once such circumstances, T. confisum has a higher rate of
again,however, such sophisticationis rarely necessary reproduction and survival,and is also more effective in
in practice. its destruction of pre-adult individuals. (Both species
exhibit this reciprocal predation making it a ' - - ' and
4.6 The Competitive Exclusion Principle therefore competitive interaction.)However,if an extra
dimension is added to the environment,namely 'space'
We can now return to our main theme and consider (in the form of small glass tubes which are available to
two laboratory investigations of interspecific competi- 0. surinamensis but too small to be accessible to T.
tion. The first is the work of Park (1954) on the flour confusum),then 0.surinamensis gains added protection
beetles Tribolium confusum and T. castaneum. In a series against predation and there is stable coexistence. Both
of simple, sterilized cultures, Park held most environ- species now have realized niches (Fig. 4.16e).
mental variables constant, but varied a single, albeit
complex parameter: climate. In all conditions, both The standard interpretation of experiments like
species were able to survive in monospecific cultures: these has been elevated to the status of a principle: the
the fundamental niches of both species spanned the so-called 'Competitive Exclusion Principle'. This
whole climatic range (Fig. 4.16a). In mixed-species merely recasts, in terms of niches, what has already
cultures, however, the results were as outlined in been hinted at in granivorous ants and certain plants:
Table 4.4. It appears that the climatic extremes are if there is no differentiation between the realized
represented in the realized niche of only one of the
two species. In the middle of the range there is also, niches of two competing species, or if such diflerentia-
invariably, elimination of one species by the other; but
the precise outcome is probable rather than definite. tion is precluded by the limitations of the habitat, then one
The situation can therefore be visualized as in species will eliminate or exclude the other. (As the
Fig. 4.16b, with overlappingrealized niches in a single experiments described above show, such lack of
dimension. differentiation is often expressed as the total non-
existence of the realized niche of one of the species.)
The second experiment (Crombie, 1947) also con- Conversely, when differentiation of realized niches is
cerns two species of flour beetle: T. confusum and allowed by the habitat, coexistence of competitors is
Oryzaephilus surinamensis. Both species can maintain possible.

Table 4.4 Results from the work of Park (1954),see text. 4.7 Competitive exclusion in the field

Laboratory habitats tend to differ from field habitats in
having fewer dimensions, and narrower ranges of
those dimensions that they do have. It is likely, there-
fore, that, in the laboratory, habitat will frequently
preclude niche differentiation, forcing potential co-
existors to compete in a way that leads to the elimina-
tion of one of them. For this reason, field evidence of
competitive exclusion is particularly valuable.

Connell(1961)produced such evidence, working in
Scotland with two species of barnacle: Chthamalus
stellatus and Balanus balanoides. Adult Chthamalus

CHAPTER 4: INTERSPECIFIC COMPETITION 93

Fig 4.16 (a,b)Indicate, respectively,the fundamental and surinamensis and Tribolium confusum in relation to food,
realized niches of the flour beetles Tribolium confusum and while (e)indicates their realized niches when another niche
T. castaneum in relationto climate. (Based on the data of dimension, space, is added. (Based on the data of Crombie,

Park, 1954.)(c,d)Indicate, respectively,the fundamental 1947.) For further discussion, see text.
and realized niches of the flour beetles Oryzaephilus

generally occur in an intertidal zone which is above sence of Balanus. In contrast with the normal pattern,
that of adult Balanus. Yet young Chthamalus do settle Chthamalus in the absence of Balanus survived very
in the Balanus zone, so that their subsequent disap- well (Fig. 4.17). Thus, it seemed that competition from
pearance suggests either that Balanus individuals ex- Balanus, rather than increased submergence time, was
clude them, or that they are simply unable to live the usual cause of Chthamalus mortality. This was
there. Connell sought to distinguish between these confirmed by direct observation. Balanus smothered,
alternatives by monitoring the survival of young undercut or crushed Chthamalus, and the greatest
Chthamalus in the Balanus zone, taking successive Chthamalus mortality occurred during the seasons of
censuses of mapped individuals over the period of 1 most rapid Balanus growth. Moreover, the few Chtha-
year. Most important of all, he ensured at some of his malus individuals that survived a year of Balanus
sites that the Chthamalus individuals were kept free crowding were much smaller than uncrowded ones
from contact with Balanus. In other words, he carried showing, since smaller barnacles produces fewer off-
out a 'removal experiment' allowing him to compare spring, that interspecific competition was also reduc-
the responses of Chthamalus in the presence or ab- ing fecundity. It is clear, then, that the fundamental

94 PART 2: INTERSPECIFIC INTERACTIONS

niches of both species extend down into the lower Fig. 4.17 The intertidal distribution
levels of the 'barnacle belt'; but interspecific competi- of adults and newly settled larvae of
tion from Balanus excludes Chthamalus from these Balanus balanoides and Chthamalus
levels, restricting its realized niche to the upper zones stellatus, with a diagrammatic
in which it can survive by virtue of its comparatively representation of the relative effects
high resistance to desiccation. It is also clear that of desiccation and competition.
competition is markedly 'one-sided' in the lower Zones are indicated to the left; from
zones. Yet Balanus must expend some of its energy and (mean high water-spring) (MHWS)
resources in the process of smothering, undercutting down to (mean low water-spring)
and crushing invading Chthamalus, and the two spe- (MLWS).(After Connell, 1961.)
cies do undoubtedly compete for the same space. This
is, therefore, an example of interspecific competition preferred flower. Moreover, it visited more of these
bearing a superficial resemblance to amensalism. flowers than usual during each stay in a patch (such
increased stay-times being a recognized indication
Another example of competitive exclusion in the that the forager is being more successful in the patch).
field is provided by the work of Inouye (1978)on two Thus, the fundamental niches of both bee species
species of bumble bee: Bombus appositus and B. flavi- clearly include both species of flower; and, since it is
frons. In the Colorado Rocky Mountains, B. appositus the presence of one bee species that restricts the
forages primarily from larkspur, Delphinium barbeyi realized niche of the other, reciprocal competitive
and B. flavifvons from monkshood, Aconitum columbi- exclusion is strongly suggested. The probable mech-
anum. When Inouye temporarily removed one or anism is as follows: B. appositus (longer proboscis) is
other of the bee species, however, the remaining better adapted to forage from larkspur (long corolla
species quickly increased its utilization of its less- tube) than B. flavifrons, but when either bee species is
alone the difference is sufficiently slight for both
flowers to be valuable sources of nectar. When both
bee species are present, however, B. appositus depletes
the nectar in larkspur flowers to a level which makes
them unattractive to B. flavifrons (with its shorter
proboscis). As a consequence, B. flavifrons concen-
trates on monkshood, depleting the nectar there to a
level which makes them significantly less attractive (in
cost-benefit terms) to B. appositus. The two bee species
are, therefore, attracted primarily to different flowers,

C H A P T E R 4: INTERSPECIFIC COMPETITION 95

and they forage accordingly. This results in reciprocal lease'. Some of the species in the above examples of
competitive exclusion. exclusion show this phenomenon in the special con-
text of an experimental manipulation (expandingtheir
The Balanus-Chthamalus interaction and Inouye's 'range' when a competitor is removed); but the term
Bombus study are excellent illustrations of the two competitive release is more commonly applied to the
contrasting types of interspecific competition (Park, results of natural experiments, as in the following
1954). The aggressive competition by which Balanus example.
excludes Chthamalus in their joint pursuit of limited
space is termed interference competition; whereas the The New Guinea archipelago comprises one large,
reciprocal exclusion of the two Bombus species,result- several moderately sized and very many small islands.
ing from the depletion of a resource by one species to Species distribution on islands has been extensively
a level which makes it essentially valueless to the studied and analysed (see for instance MacArthur,
other species,is termed exploitation competition. Thus, 1972);but for our present purposes we need note only
with interferencethere is no consumption of a limited that in the New Guinea region, as elsewhere, small
resource, whereas with exploitation there invariably islands tend to lack species which are present on large
is. (Note that the interferencelexploitation dichotomy islands and on the mainland. One example of this
is somewhat similar to that between scramble and concerns ground doves (Diamond, 1975). As Fig. 4.18
contest (section 2.4), in that contest always involves illustrates, there are three similar species of ground
interference, while scramble usually involves only dove on New Guinea itself, and moving progressively
exploitation. Note, however, that interference and inland from the coast one encounters them in
exploitation have none of the extreme threshold or sequence: Chalcophaps indica in the coastal scmb, C.
all-or-none characteristics associated with scramble stephani in the light or second-growth forest, and
and contest.) Gallicolumba ruJigula in the rainforest. On the island of
Bagabag, however, where G. rujigula is absent, C.
These two field studies are also particularly good stephani expands its range inland into rainforest; and
examples of competitive exclusion because they are on Karkar, Tolokiwa, New Britain, and numerous
the results of experimental manipulations; the exclu- other islands where C. indica is also absent, C. stephani
sion of Glycine by Panicum in the absence of Rhizo- expands coastwards to occupy the whole habitat
bium (section 4.4.1) was equally persuasive. But gradient. Conversely, on Espiritu Santo C. indica is the
evidence is much more commonly circumstantial. only species present, and it occupies all three habitats.
Remember, for instance, that amongst the larger
species of Davidson's granivorous ants there was This, as Diamond (1975) remarks, is a particularly
never coexistence between species that shared both a neat example, but there are many other natural
size and a foraging strategy, i.e. species with appar- experiments with similar results. -Direct interspecific
ently very similar realized niches. Providing the num- competition has not been positively established, but
ber of CO-occurrencesexpected by mere coincidenceis since C. stephani and C. indica only occupy rainforest in
much greater than the number observed, reciprocal the absence of G. rufigula, C. stephani only occupies
competitive exclusion is a very plausible explanation in coastal scrub in the absence of C. indica, and C. indica
such cases; but it can never be more than a plausible only occupies light forest in the absence of C. stephani,
explanation when reliance is placed solely on observa- the conclusion is more or less unavoidable that there
tional data. is competitive exclusion on New Guinea and com-
petitive release elsewhere.
4.8 Competitive release
4.9 Coexistence: resource partitioning
A class of evidence lying between simple observation
and experimental manipulation is represented by We have seen, from both direct and circumstantial
observations of what is known as 'competitive re- evidence, that competitive exclusion can influence the

96 PART 2: INTERSPECIFIC INTERACTIONS

Fig. 4.18 Habitats occupied by three

species of ground doves on various
islands: Chalcophaps indica (vertical
bars), Chalcophaps stephani (solid
shading)and Gallicolumba rufigula

(diagonalbars). (After Diamond,
1975.)

range and distribution of species. We return now to and estimated densities at both sites at the beginning
what is, in a sense, the reverse question: when do and end of the year, noting separately (at the experi-
similar or competing species coexist? mental site) the numbers in marked and unmarked
shells. Despite the fact that many of the empty shells
Vance (1972a) studied competition and the mech- were removed by strong currents, there was a signifi-
anism of coexistence in three sympatric species of cant increase in total numbers of crabs at the experi-
intertidal hermit crabs from the San Juan Islands of mental site. However, there was no such increase at
Washington State: Pagurus hirsutiusculus, P. granosi- the control site, nor in the unmarked shells at the
manus and P. beringanus. The first priority was to experimental site. Empty shells are obviously a limited
confirm that competition was occurring.These hermit resource for P. hirsutiusculus. To establish the general-
crabs are generalized omnivorous feeders and food is ity of shell limitation, Vance gathered data on the size
apparently abundant throughout the year: a priori, the distributions of the three species of crabs, of the shells
most likely limiting resource was the empty gastropod occupied by them, and of unoccupied shells; and he
shells which the crabs inhabit. Vance chose an exper- compared these data with the preferred shell sizes of
imental and a control site, and to the former he added the crabs, determined in preference experiments.
1000 marked, appropriately sized shells once per Except for small size-classes, empty shells were rare,
month for 1year. He concentrated on P. hirsutiusculus

CHAPTER 4: INTERSPECIFIC COMPETITION 97

and all but small crabs of the three species occupied apparently specialized in winning fights in a particular
shells which were smaller than their preferred size. area. Thus, what we have called 'habitat partitioning'
Thus, it appears that empty shells constitute for these could also be considered as resource partitioning.
three species of hermit crab a common, necessary and Resource partitioning is, therefore, the means by
limiting resource for which they compete. which these three competitors coexist. As Table 4.5
shows, moreover, both types of resource partitioning
Vance went on to examine the mechanism of co- are necessary for coexistence. P. beringanus and P.
existence of these three species. He found, first of all, granosirnanus are well separated by habitat, but P.
that there was partitioning of the resource: P. hirsu- hirsutiusculus is only truly separated from the other
tiusculus prefer short-spired light shells, whereas both two in shells within its own preference range.
P. beringanus and P. granosirnanus prefer relatively
taller and heavier shells. He also found that there was We have obviously met other, similar examples
partitioning of the habitat: P. hirsutiusculus predomi- already, though in a slightly different context. Bornbus
nates in the upper intertidal and amongst Hedophyllurn appositus and B. Jlavijirons partition the nectar resource
sessile, a brown alga living on horizonal rock faces in on the basis of corolla length: they are adapted to do so
the lower intertidal; Pagurus beringanus predominates by differences in proboscis length. Balanus balanoides
in tide-pools in the lower intertidal; and P. granosi- and Chtharnalus stellatus partition the space resource
rnanus predominates under large loose stones and in on the basis of intertidal zone: Balanus is adapted to
shallow tide-pools of the mid-intertidal. Thus, there is physically oust its competitor from the lower zones
differentiationamongst the realized niches of the three and Chtharnalus to resist desiccation in the higher
species, allowing them to compete and yet coexist. zone. Glycine and Panicurn partition the 'total nitrogen'
resource: Glycine is adapted to utilize free nitrogen by
The mode of competition between these species its intimate association with Rhizobiurn. Finally,
appears, essentially, to be interference. Shell occu- granivorous ants partition the seed resource on the
pancy is determined by intra- and interspecific basis of the size, density and microdistribution of the
fighting: crabs in less-preferred shells attempting to seeds: they are adapted to do so by differences in their
displace crabs in more-preferred shells, the loser own size and in foraging strategy. In all these cases the
taking the less-preferred shell (Vance, 1972b). We basic pattern is the same: competing species appear to
have seen that the outcome of such interspecific
encounters is determined to some extent by resource coexist as a result of resource partitioning-in other
partitioning on the basis of shell shape and weight. words, by virtue of the differentiation of their realized
However, the basic resource-empty shells-is also niches. Moreover, since the same examples have
divisible into 'shells-amongst-brown-algae', 'shells-in- served to illustrate both phenomena, it is clear that
shallow-tide-pools' and so on; and each species is resource partitioning and competitive exclusion are

Table 4.5 Hermit crabs occupying Littorina sitkana shells (which fall in the Pagurus hirsutiusculus preference range) and
Searlesia dira shells (which fall in the Pagurus beringanus-P. granosimanus perference range) collected from various physical
habitats at a single site. (After Vance, 1972a.)

98 P A R T 2: INTERSPECIFIC INTERACTIONS

often, though not always, alternative aspects of the 4.10 Character displacement
same process. Competitive exclusion between species
from portions of their fundamental niches leads to A particular type of resource partitioning, allowing
differentiation of their realized niches. competitors to coexist, is known as character displace-
ment. It involves the modification of the morphologi-
In fact, there are two further patterns which have cal form of a species as a result of the presence of
emerged from the examples we have considered. The interspecific competitors. Fenchel (1975) investigated
species pairs-bees, barnacles and plants-partitioned the coexistence of hydrobiid mud snails in Limfiord,
their resource along a single dimension: corolla Denmark, and paid particular attention to two species:
length, intertidal zone and total nitrogen, respectively. Hydrobia ventrosa and H. ulvae. These deposit feeders
Conversely, the three species of hermit crab required seem to ingest their substrate indiscriminately and
at least two dimensions: shell shape and weight, and utilize the attached micro-organisms, and Fenchel
shell location; while the guild of granivorous ants found that for both species there is a single linear
partitioned resources on the basis of seed size and seed relationship between shell length and food particle
density/distribution,and even these two dimensions size. Some of Fenchel's results are illustrated in
were apparently not enough to account for the co- Fig. 4.19. It is clear from this figure that when the two
existence of the smaller species. There appears, in species live alone (which they do in a range of
other words, to be a correlation between the number habitats), their sizes are more or less identical, as are
of species in a guild and the number of niche the sizes of their food particles. When the two species
dimensions involved in the partitioning of the re- coexist, however, there is character displacement. H.
source. Schoener (1974) reviewed the literature on ulvae is larger, H. ventrosa is smaller, and the sizes of
resource partitioning, and found that this correlation their food particles are similarly modified. The evi-
was, indeed, statisticallysignificant. Interestingly, this dence for interspecific competition is not direct, but
parallels the results of theoretical investigations by there is no simple alternativeto the suggestion that the
MacArthur (1965) and Levins (1968). character displacement allows the partitioning of a
potentially limiting resource, and thus allows the
In addition, Schoener (1974), like several others, coexistence of two competitors.
pointed to the niche complementarity frequently in-
volved in resource partitioning. As illustrated by the A similar situation, of course, was that desribed
examples of hermit crabs and, more especially, David- previously for the ant Veromessor pergandei (Davidson,
1978), which shows character (size) displacment in
son's ants, species which are not differentiated along response to its competitive milieu: the number and
one niche dimension tend to be separated along an- nature of its interspecific competitors. Thus, we can
other. Realized niches are, therefore, fairly evenly dis- note that, in our newly defined terms, Davidson's ants
tributed in multidimensional space, and species show competitive exclusion, coexistence through dif-
compete simultaneouslywith several species in several ferentiation of realized niches (resource partitioning)
dimensions (called difuse competition by MacArthur, and character diplacement.
1972).
4.1l Competition: its avoidance or its
Finally, however, it should be clear that we have non-existence?
avoided the most important aspect of competitive
exclusion. We have seen that when there is no niche We have seen that species can coexist when, as
differentiation, competitive exclusion occurs; and we competitors, they partition resurces between them;
have seen that when competitors coexist there is and also that such partitioning may be achieved, in
resource partitioning. But we have ignored a much special cases, by character displacement. But species
more profound question. How much niche differentia- may also coexist simply because they do not compete.
tion is necessary for the coexistence of competitors?
We return to this question in section 4.15.

CHAPTER 4: INTERSPECIFIC COMPETITION 99

Fig. 4.19 Coexistence through character displacement. adapted to its own flower that there was no (even
(a) Average lengths (plus standard deviations) of Hydrobia potentially) shared resource, and therefore no inter-
ulvae (open circles) and H. ventrosa (closed circles)at a specific competition. This explanation implies the
variety of sites at which they coexist or live alone. (b) existence of interspecific competition in the past, but
Distributions of food particle size of the same species at proposes, in addition, that selection for resource
typical sites at which they coexist or live alone. (After partitioning has been so strong and so longstanding
Fenchel, 1975.) that the partitioning is now complete and irreversible.
It might also have been suggested, however, that the
We have tried, until now, to include only those difference in proboscis length between Bombus apposi-
examples in which competition has been positively tus and B. jlavifrans merely reflects the fact that they
established. Suppose, however, that Inouye, instead of are two different species, and has nothing to do with
carrying out his experimental manipulations, had competition. There would, therefore, have been a real
observed simply that two different species of bee problem of interpretation if Inouye had relied solelyon
forage from two different species of flower, and that observational data; and to illustrate this problem we
the bee species with the longer proboscis foraged from can consider an example of the sort of information
the flower species with the longer corolla tube. The that is normally described as 'field evidence of inter-
explanation that has been suggested,involving compe- specific competition'.
tition, competitive exclusion and resource partition-
ing, would certainly have been plausible; but there are Lack (1971) described the coexistence of five spe-
two alternative explanations. It might have been cials of tit in English broadleaved woodlands: the blue
suggested that each bee species has become so well tit (Paruscaeruleus),the great tit (P. major), the marsh tit

100 P A R T 2 : INTERSPECIFIC INTERACTIONS

(P. palustris), the willow tit (P. montanus) and the coal of the seeds they take; and this separation is associated
tit (P. ater). Four of these congeneric species weigh with differences in overall size, and in the size and
between 9.3 and 11.4 g on average (great tit 20 g); all shape of the beaks. Yet, as we have seen, there are
have short beaks and hunt for food chiefly on leaves three possible interpretations of this situation. The
and twigs, but at times on the ground; all eat insects first two ('current competition' and 'competition in the
throughout the year, and also seeds in winter; and all past') are based on the assumptionthat the differences
nest in holes, normally in trees. Nevertheless, in reflect the partitioning by the tits of a potentially
Marley Wood, Oxford, all five species breed and the limited resource;but the third makes no such assump-
blue, great and marsh tits are common. tion. It states simply that the five species,in the course
of their evolution, have adapted to their environment
All five species feed their young on leaf-eating in different ways; but in ways that have nothing to do
caterpillars, and all except the willow tit feed on with interspecificcompetition. And on the basis of the
beechmast in the winters when it is plentiful; but both evidencepresented, it is impossibleto reject this interpre-
of these foods are temporarily so abundant that tation. It has not been shown that the birds would
competition for them is most unlikely. The small blue expand their niches in the absence of the other
tit feeds mainly on oak trees throughout the year, species, and it has not even been shown that food is a
concentrating on the smaller twigs and leaves of the limited resource. There is, therefore, no direct evi-
canopy to which it is suited by its agility. It also strips dence of competition, and no overriding reason for
bark to feed on the insects underneath, and generally involving it in our interpretation.
takes insects less than 2 mm in length. It eats hardly
any seeds, except those of birch which it takes from Nevertheless,the possibilityof the first two interpre-
the tree itself. The heavy great tit, by contrast, feeds tations does remain. We have seen in several
mainly on the ground, especially in winter. Most of the examples-ants, barnacles, bumble bees, etc.-that
insects it takes exceed 6 mm in length; it eats more competing species can coexist by resource partition-
acorns, sweet chestnut and wood sorrel seeds than the ing, and can retain the ability to expand their range in
other species; and it is the only species to take hazel the absence of their competitor. Thus, it may be the
nuts. The marsh tit has a feeding station intermediate case that this is also the correct interpretation of the
between the other two common species: in the shrub tits' ecology, but that the appropriate experimental
layer, in large trees on twigs and branches below 6 m, manipulations have simply not been carried out.
or in herbage. It is also intermediate in size between
the other two species, and generally takes insects Moreover, we have also seen-e.g. in the examples
between 3 and 4 mm in length. In addition it takes the of character displacement-that species can evolve
fruits and seeds of burdock, spindle, honeysuckle, morphological adaptations which allow them (largely
violet and wood sorrel. The coal tit is another small if not totally) to avoid competition. This (second)
species feeding on oak, but also, later in the winter, on interpretation, like the third interpretation (above)
ash. It generally takes insects which are less than denies the existence of current interspecific competi-
2 mm long, and which are indeed shorter, on average,
than those taken by the blue tit. Moreover, unlike the tion, but unlike the third interpretation it invokes
blue tit, the coal tit feeds mainly from branches rather interspecific competition as the evolutionary driving-
than in the canopy. Finally, the willow tit is most like force behind the differences currently observed.
the marsh tit, feeding on birch and, to a lesser extent,
elder, and in herbage. Unlike the marsh tit, however, Indeed, the first and second interpretations are
the willow tit avoids oak and takes very few seeds. based on alternative outcomes of a shared evolution-
ary process, which undoubtedly does pertain in some
As Lack concluded, the species are separated from cases. The process occurs as follows. Natural selection
each other at most times of the year by their feeding favours the survival and reproduction of those individ-
station, the size of their insect prey and the hardness uals with the greatest fitness, but interspecific com-
petition reduces fitness. Individuals that avoid
interspecific competition will therefore evolve, and

CHAPTER 4: INTERSPECIFIC COMPETITION 101

interspecific competition is avoided by resource parti- converted only one of the minuses of a ' - - ' interac-
tioning. Evolution of a realized niche which is too tion into a zero. In other cases (as in the second
small, however, will increase intraspecificcompetition, interpretation, above) it converts both.
and this, too, reduces fitness. We can, therefore,
expect each species to evolve towards a form in which 4.12 Competition and coexistence
inter- and intraspecific competition are optimally in plants
offset. Sometimes this will evolve relatively flexible
resource partitioning (first interpretation); sometimes The theory of the niche and the competitive exclusion
the partitioning will be inflexible (second interpreta- principle have their origins firmly rooted in zoological
tion). Sometimes there will be a lessening of interspe- study, and it is intuitively less easy to see how niche
cific competition; sometimes its total avoidance. Yet in differentiation can occur in autotrophic plants, when
either case, and with either interpretation, interspe- all have essentially the same basic growth require-
cific competition will be of paramount importance. ments (light, water and nutrients). It might be imag-
ined that plants have evolved specializations to
On the available evidence, however, it is impossible capture energy for photosynthesis from different
to determine whether these tit data indicate interspe- wavelengths of light, or that nutrients might be
cific competition (first interpretation),its evolutionary utilized in unique and separate ways; but comparative
avoidance (second interpretation) or its total non- physiological studies show that this is not so. Plant
existence, now and in the past (third interpretation); growth requirements ('food') are not usually discrete
and this would be true of almost all examples of packages that can be simply partitioned amongst
apparent interspecific competition in the field. There competing species. (An important exception to this
are undoubtedly cases of current resource partitioning however, is nitrogen utilization. The legumes (as we
amongst competitors; and there are undoubtedly have seen in section 4.4.1), and some other genera
cases in which species' ecologies have been moulded such as Alnus, do not place total reliance on fixed
by interspecific competition in the past. But differ- nitrogen in the soil for growth. Instead, by virtue of
ences between species are not, in themselves, indica- their symbiotic relationship with nitrogen-fixing
tions of the ways in which those species coexist; and micro-organisms, they utilize free nitrogen from the
interspecific competition cannot be studied by the air.) Moreover, the very nature of the effect of limiting
mere documentation of these interspecific differences. resources on plant growth is complex. Limitation in
water supply resulting from intense root competition,
Note, finally, that while the first interpretation is for instance, will limit leaf growth; but this may
based on a ' - - ' interaction,the second interpretation contribute in turn to a reduction in the growth of new
assumes that the interaction is essential '00'. In other roots. Thus, shortage of one limiting resource (water)
words, it is assumed that evolution, acting to avoid affects the competitive struggle to obtain both light
'minuses', converts them to 'zeros'; and this is prob- and water itself. Disentangling the web of cause and
ably also the basis of most cases of amensalism ( - 0). effect experimentally has proved very difficult (re-
The plant species that produces a toxic metabolite viewed by Harper, 1977)
causing growth reduction in a second species presum-
ably does so as an evolutionary response to the harm- In practice, most attempts to explain the coexistence
ful, competitive effects that the second species had on of plant species have rested largely on a demonstration
its growth in the past. At that time the interaction of the fact that potential competitors differ in ways
would have been competitive: ' - - ': Now, however, which might reduce competition. Attention has fo-
evolution has led to the production of toxin by the cused particularly on differences in life form, differ-
aggressive species which occurs whether or not poten- ences in the timing of various stages of growth-
tial competitors are present. The aggressive species is, particularly germination and flowering-and
therefore, unaffected by these other species, and the differences in preferred levels of abiotic factors (see
interaction is amensal. Evolution has, in this case,

102 PART 2: INTERSPECIFIC INTERACTIONS

Grubb, 1977 and Werner, 1979 for reviews). How- force behind this arrangement; but the circumstantial
ever, we have seen from the tits in Marley Wood that evidence is strengthened in this case by the fact that
such differences,taken alone, are impossible to inter- the degree of separation between species is much
pret with confidence. The conclusions that can be greater in the mature prairies than in the young
drawn from such data are severely limited. hayfield successional habitats. The position taken up
by a species along this environmental gradient is
Perhaps the most that can be obtained from the apparently the one at which it has been most success-
interpretation of these differences is illustrated by the ful in competition with the others: interspecific com-
work of Werner and Platt (1976). They studied six petition, with time, seems to realize the niche of each
species of golden rod (Solidago)that commonly occur species.
together both in old hayfields undergoing successional
change, and in mature, stable prairie communities in Intuitively, spatial heterogeneity in resource use by
North America. In both habitats, but particularly in plants is probably one of the most powerful promoters
the prairies, they were able to relate the frequency of of niche separation and coexistence between plants.
occurrence of each species to the availability of soil Species occupying distinct rooting zones within the
moisture (Fig. 4.20): differentspecies appear to 'prefer' soil may exploit nutrients and soil moisture sufficiently
different moisture levels. There is, of course, no direct independently of one another. This can be seen even
evidence that interspecific competition is the driving at the genotypic level. Ennos (1985) observed that
genotypesof Trifolium repens exhibited differentrooting
Fig. 4.20 Coexistence of competing plants? The occurrence lengths and was able to show that this character was
of six species of Solidago in relation to available soil genetically determined. He selected plants having
moisture in a hayfield undergoing succession, and in a either short or long roots and compared their perfor-
mature prairie in North America. Soil moisture percentages mance in a replacement series. Under artificially
were determined in summer. (From Werner 81Platt, 1976.) imposed drought conditions the yields of above-
ground stolons in 50 : 50 mixtures plantings of the
two genotypes significantly exceeded both of the
monocultureyields. Bearing in mind the restrictions of
this form of analysis, it does seem that the exploitation
of different soil layers may enable coexistence in
plants.

In a few plant examples, however, it is possible to do
more than merely implicate interspecific competition
as an important interaction between coexisting spe-
cies. Sharitz and McCormick (1973), for instance,
studied the population dynamics of pairs of annual
plant species (Sedum smallii and either Minuartia
uniflora or M. glabra) which dominate the vegetation
growing on granite outcrops in the south-eastern USA.
As Fig. 4.21 shows, there is very strict zonation of the
adults of the two species associated with the soil depth
around the outcrops; and soil depth itself is strongly
correlated with soil moisture. The experimental re-
sults in Fig. 4.21b and 4.21c, however, indicate that
this zonation is not simply a reflection of the tolerance
ranges of the species. In fact, their fundamental niches
cover the same range of experimental conditions.

CHAPTER 4: INTERSPECIFIC COMPETITION 103

( W Sedum smallii

lntraspecific competition

High Low

Sedum smallii

+m-Loo Interspecificcompetition
Hiah
Low

Seedling Soil depth (cm)
Rosette
0 2 4 6 8 10 12 1 4 1FMlaotwuerering (C) Minuartia glabra
15 f lntraspecificcompetition
Soil depth (cm)
High Low
Minuartia uniflora 15 f

100 Seedling lnterspecificcompetition
0 Rosette
Flowering
Soil depth (cm)

Fig. 4.21 (a)The zonation of individuals, according to soil Soil depth (cm)
depth, of two annual plants Sedum smallii and Minuartia
uniJlora at four stages of the life cycle. (b,c)The initial seed sown) is shown when grown alone and with the
consequences of competitive interaction between Sedum other species. The experiment was conducted at three soil
depths and three relative moisture levels: S, saturation; FC,
smallii and Minuartia glabra, respectively. For each species field capacity; D, one-third of field capacity. (From Sharitz
the final density at plant maturity (as a percentage of the & McCormick, 1973.)

Nevertheless, while Sedum is clearly more capable realized niches (in practice, 'zones') that are signifi-
than Minuartia of tolerating the lack of moisture at the cantly smaller than their fundamental niches. Further
low end of the range, Minuartia is obviously much less evidence of this is provided by the very incomplete
affected than Sedum by interspecific competition at the zonation at the seedling stage, prior to any substantial
high end. It is apparent, in other words, that interspe- competitive interaction (Fig. 4.21a). The parallel with
cific competition in nature restricts these species to the barnacles in section 4.7 is quite striking.

104 PART 2: INTERSPECIFIC INTERACTIONS

In at least some cases, therefore, plant species Fig. 4.22 (a)Competition for one limiting resource in
which are potential competitors for a limited resource relation to an abiotic factor. (b)The outcome of competition
can coexist by virtue of a differentiation of their for silicate for two species of planktonic algae over a wide
realized niches. Yet it must be recognized that there temperature range. A$, Asterionella formosa; S.U.,Synedra
are very few instances in which this has been posi- ulna. (Modified from Tilman et al., 1982.)
tively established(see Werner, 1979). It is very easy, in
plants, to demonstrate the reductions in fitness which SiO, and outcompetes Synedra ulna, but the reverse is
can result from interspecific competition; but it has true above 20°C. Although there is no temperature
proved very difficult so far to demonstrate the mech- range in which both species have the same demand
anisms which allow potential competitors to coexist. for silicate, we can envisage coexistence if there are
There are two reasons for this. Grubb (1977) has fluctuations in temperature over the range in which
suggested that this difficulty may stem from an alternative competitive displacement can occur. In-
overemphasis on adult plants. Irrespective of any deed, diurnal temperature fluctuations between 16
niche differentiation, it is almost impossible for a
seedling of one species to outcompete an established
adult of another species. The most important compe-
tition between plants, therefore, may be pre-emptive
competition for the regeneration niche, i.e. competition
amongst seedlings to become established in a part of
the environment which has recently become vacant. It
seems certain that in future, the study of pre-emptive
competition will teach us a great deal about the
coexistence in nature of competing plants. A very
similar phenomenon, restricted to single species and
referred to as 'space capture', is discussed in section
6.11.

The second reason is a neglect of the fact that
environmental parameters indirectly interact to deter-
mine the intensity of competition. Utilization of re-
sources in limited supply may be determined by the
species' response to other (non-limiting) environmen-
tal factors which govern the competitive outcome.
Figure 4.22a portrays the requirements of two hypo-
thetical species for a limiting resource, determined by
an abiotic factor such as temperature or pH. At the
lower part of the range, species A, by virtue of a lower
requirement for the limiting resource will tend to
outcompete species B. The converse will be true in the
upper part of the range. In mid-range, all other factors
being equal, growth is limited to the same extent and
coexistence occurs. Partial support for this view comes
from a further experiment using the planktonic algal
species Asterionella formosa and Synedra ulna (Fig.
4.22b). Over much of the temperature range (4-20°C),
Asterionella formosa has the lower requirement for

CHAPTER 4: INTERSPECIFIC COMPETITION 105

and 24°C are very likely in the surface waters of the than species 1has on itself.) We now simply need to
lakes in which these species live. The only proviso that replace N, in the bracket of our logistic equation with
we must add in final explanation of this mechanism is a term which signifies: 'N, plus N,-equivalents', i.e.
that it clearly depends on the time scale over which
population growth responses can occur. -dt = rlN1 K1
or
4.13 A logistic model of
two-species competition -dt = rlN1

Having examined what is known about interspecific WL-:e can, of course, write a similar equation for species
competition, it will be valuable to turn (as we did with
single-species pop~ationst)o some simple models, to This is our basic (Lotka-Volterra) model.
see whether they can improve our undershnding of To describe the properties of this model, we must
the interaction.
ask the following questions: when (under what cir-
The conventional starting point for such models is cumstances) does species 1 increase in numbers?
the differential logistic equation (followingLotka, 1925; When does it decrease? And, when does species 2
Volterra, 1926). Obviously,therefore, in conforming to increase and decrease? In order to answer these
this convention, we will be incorporating into our questions we construct what are, in essence, the
model all of the logistic's shortcomings. Nevertheless, equivalents of maps. Thus, while in maps there are
as will become apparent below, a useful model can be areas of land and areas of sea, with a coastline(neither
constructed. land nor sea) dividing them; in our case we will have

The logistic equation: areas of N, (or N,)increase.and areas of N, (or N,)

K-N decrease, with a zero isocline (neither increase nor
decrease) dividing them. Moreover, if we begin by
= rN ( X ) drawing a zero isocline (coastline), we will know that
there is increase (land) on one side of it .and decrease
contains, within the brackets, a term which is respon- (sea) on the other. As Fig. 4.25 shows, the axes of our
sible for the incorporation of intraspecificcompetition. 'map' will be N, and N,: the bottom left-hand corners
We can proceed by replacing this term with one which are areas where there are low numbers of species 1
incorporatesnot only intra- but also interspecificcom- and 2, and the top right-hand corners areas where
petition. We will denote the numbers of our original there are high numbers of species 1and 2.
species by N, (carrying-capacity, K,; intrinsic rate of
increase, r,), and those of a second species by N,. In order to draw the NI--isocline we will use the
fact that on it dN,/dt = 0, i.e.
Suppose that, together, 10 individuals of species 2
have the same competitive,inhibitory effecton species )=o.
1 as does a single species 1 individual. The total
competitive effect on species 1(inter-and intraspecific) This is tme for two trivial cases (when rl or are
zero), but also for an important case:
will then be equivalent to (N, + {N,/lO)) species 1
K, - N, - a12N2= 0
individuals. We call the constant-1/10 in the present
case-a coeficient of competition, and denote it by a,,
since it measures the competitive effect on species 1of
species 2. In other words, multiplying N, by a,,
converts it to a number of W,-equivalents'. (Note that
a,, < l means that species 2 has less inhibitory effect
on species 1than species 1has on itself, while a,, > 1
means that species 2 has a greater inhibitory effect

106 PART 2: INTERSPECIFIC INTERACTIONS

Indeed, the straight line represented by this equation N2-isoclinestogether on a single figure. In so doing, it
is our isocline, and since it is a straight line we can should be noted that the arrows in Fig. 4.23 are
draw it by finding two points on it and joining them. actually vectors-with a strength as well as a direc-
Thus when tion-and that, to determine the behaviour of a joint
NI-N2 population, the normal rules of vector addition
N1 = 0 , N2 = -K1 (point A, Fig. 4.23a) should be applied (see Fig. 4.24). It is clear from
a12 Fig. 4.25 that there are four different ways in which
the two isoclines can be arranged. In Fig. 4.25a and
and when 4.25b, one isocline lies entirely beyond the other, and
the vectors indicate that, as a consequence,the species
N2=0,N1=K, (point B, Fig. 4.23a). with the inner isocline becomes extinct, while the
other species attains its own carrying-capacity.
The line in Fig. 4.23a is, therefore, the NI-isocline.
Below and to the left of it, numbers are low, competi- Such situations can be defined by the intercepts of
tion is comparatively weak and species 1increases in the isoclines. In Fig. 4.25a, for instance:
abundance (arrows from left to right, N, on the
horizontal axis); above and to the right of it, numbers K, > K2a12and K, < K1a2,.
are high, competition is comparatively strong and
species 1decreases in abundance (arrows from right Species 1 exerts more effect on itself than species 2
to left). exerts on it, but also exerts more effect on species 2
than species 2 does on itself. In other words, species 1
Based on an equivalent derivation, Fig. 4.23b has is a strong interspecificcompetitor, species 2 is a weak
areas of species 2 increase and decrease separated by interspecificcompetitor, and species 1drives species 2
the (straight)N2-isocline;arrows, like the N2-axis, are to extinction. The situation is reversed in Fig. 4.25b.
vertical.

All that is required now is to put the N,- and

Fig. 4.23 (a)The N,-isocline
generated by the Lotka-Volterra
competition equations. Species 1
increases below and to the left of

the isocline (arrows left to right),
and decreases above and to the
right of the isocline (arrows right to
left). (b)The equivalent NZ-isocline.

CHAPTER 4: INTERSPECIFIC COMPETITION 107

In Fig. 4.25~:

K, < K2a12and K, < K1a2,. Fig. 4.24 Vector addition. When species 1and 2 increase in

Interspecificeffects are more important than intraspe- the manner indicated by the N, and N, arrows (vectors),
cific effects: both species are strong interspecific com-
petitors. There are two stable points (N, = K,, N , = 0 the joint population increase is given by the vector along
and N2 = K,, N, = 0) and an unstable equilibrium
combination of N, and N,. In other words, one species the diagonal of the rectangle, generated as shown by the N,
always drives the other to extinction, but the precise
outcome depends on the initial densities. and N,-vectors.

Finally, in Fig. 4.25d:

K, > K2a2, and K, > K1a2,. Fig. 4.25a and 4.25b successfully describes situations
in which the second species lacks a realized niche in
Intraspecific effects are now more important than competition with the first.
interspecific effects, both species are weak interspe-
cific competitors, and there is stable coexistence at a In Fig. 4.25c, both species are stronger competitors
particular, equilibrium combination of N, and N,. on the other species than they are on themselves:
there is reciprocal interference competition. This will
4.13.1 The model's utility occur when each species produces a substance that is
toxic to the other species but harmless to itself, or
We can now proceed to examine this simple model's when there is reciprocal predation. In fact, this latter
utility. Clearly, it can produce the full range of situation is the mechanism by which Park's flour
outcomes of interspecific competition: stable coexist- beetles compete (section 4.6), and it is satisfying,
ence, predictable exclusion of one species by another, therefore,to see that the model's predictions are borne
and exclusion between the two species with an out by Park's data: there is competitive exclusion, but
indeterminate outcome. It should be recognized,how- the precise outcome is indeterminate. Whichever
ever, that these are the only conceivable outcomes; to species starts with (or, at some point, attains) a more
be useful, the model must produce the right outcome favourable density will 'outpredate' (or outpoison)the
at the right time. other.

The model indicates (Figs 4.25a & 4.25b) that one Finally, Fig. 4.25d indicates, quite reasonably, that
species will outcompete and exclude a second if the stable coexistence is only possible when, for both
first species is a stronger competitor on the second species, intraspecific competition is more inhibitory
than the second is on itself, and the effect is not than interspecific competition, i.e. when there is niche
reciprocated. We saw in section 4.7 that this was, differentiation. We have seen repeatedly that this is
indeed, the case for each bumble bee on its own flower the case. However, the model avoids the more pro-
(exploitation competition), and for Balanus excluding found question of how much niche differentiation is
Chthamalus (interference competition).In other words, necessary for stable coexistence.

It is clear, then, that in broad terms the model is
successful in spite of its limitations. Indeed, it is

108 PART 2: INTERSPECIFIC INTERACTIONS

Fig. 4.25 The outcomes of competition generated by the the original inspiration of the Competitive Exclusion
Lotka-Volterra competition equations for the four possible Principle; investigation of the relevance of the princi-
arrangements of the N,- and N,-isoclines. Vectors, ple, and its applicability in the real world, came later.
generally, refer to joint populations and are derived as Yet, the most important aspect of the model is that it
indicated in (a). For further discussion, see text. makes exact, quantitative predictions about coexist-
ence, based on the numerical values of the Ks and as.
important to realize that this model (given 'experimen- Ultimately the model's utility must be tested in these
tal teeth' by the laboratory work of Gause, 1934)was quantitative terms.

CHAPTER 4: INTERSPECIFIC COMPETITION 109

4.13.2 A test of the model: fruit fly competition Fig. 4.26 Carrying-capacities(open circles) and stable two-
species equilibrium points (closed circles) for eight
There have been several studies in which experimen- combinations of two species of Drosophila. Each division
tal results have tended to support the model's predic- along the coordinates corresponds to 250 flies. In all but
tions. It will be more instructive,however, to consider the last case, the point of stable two-species equilibrium falls
an example in which results and predictions disagree, below the straight line joining the carrying-capacities. (After
because this will illustrate a point of general Ayala et al., 1973.)
importance: models are of their greatest utility when
their predictions are not supported by real data, as long which would generate the appropriately curved iso-
as the reason for the discrepancy can subsequently be clines. They judged these on a number of criteria,
discovered. Confirmation of a model's predictions rep- including simplicity, biological relevance of the para-
resents consolidation; refutation with subsequent ex- meters and quality of fit to the data, and found that
planation represents progress. one equation was particularly satisfactory. This
equation:
Ayala et al. (1973)reviewed Ayala's own findings on
laboratory competition between pairs of Drosophila is a modification of the original, simple model, but it
species. Their results are illustrated in Fig. 4.26. contains, in addition, a parameter, 8,' which modifies
Ayala's basic procedure was to maintain either one- or the underlying logistic equation, such that the func-
two-species populations in culture bottles using a tion relating growth-rate to density need no longer be
'serial transfer' technique-transferring adults to new
food at regular intervals-and to monitor the popula-
tions for several generations until an approximate
equilibrium was reached. This allowed him to esti-
mate the carrying-capacities (K, and K,) in single-
species populations, and the numbers for stable

coexistence (N, and N,) in two-species populations.

Figure 4.25d shows that our simple model predicts
that stable coexistence should only occur at a point
above and to the right of the line joining K, and K,. It
is clear from Fig. 4.26, however, that in seven out of
eight cases Ayala found coexistence below and to the
left of this line. Our simple model is, therefore, unable
to account for Ayala's results.

The simplest explanation for such a discrepancy is
illustrated in Fig. 4.27. If the isoclines are concave,

rather than straight, the values of K,, K,, G, and N,

immediately become compatible. In fact, Ayala et al..
(1973) were able to support this solution, in a partic-
ular species-pair, by following a number of popula-
tions for a single generation and obtaining an actual
series of vectors. These, too, are shown in Fig. 4.27, in
which we can see that the curved isoclines are,
indeed, satisfactory lines of demarcation between
areas of increase and decrease.

Ayala et al. took their analysis a stage further by
considering a range of possible alternative models

110 PART 2: INTERSPECIFIC INTERACTIONS

Thus, with a notation similar to that used in
equation 3.5 in section 3.2.2 the mean yield iG1 of a
plant of species 1 in a mixed population containing

species 1and 2, at densities N,and N , is

in which y,, is a 'competition coefficient' or 'equiva-
lence value' converting the density of species 2 into
numerical terms equivalent to species 1.Likewise, the
effect of species 2 on the mortality of 1is

Fig. 4.27 Isoclines for two species of Drosophila fitted by where dI2 is the appropriate equivalence value deter-
visual inspection of the vectors, which were derived mining the density-dependent mortality in species 1
empirically and are reduced to one-third of their actual attributable to species 2. This pair of equations can
length for clarity. Each division along the coordinates also be written for species 2 with (of course) different
corresponds to 200 flies. Closed circles indicate the carrying equivalence coefficients.
capacities and the point of stable two-species equilibrium,
which were also derived empirically.(After Ayala et al., We can appreciate the value of this model by
1973.) re-examining the data of Marshal1 and Jain (1969)on
oats (see Fig. 4.10). Whereas from our earlier interpre-
a straight line symmetrical about Kl2. Thus, the work tation we could qualitativelyconclude that Avena fatua
of Ayala et al. confirms our original misgivings regard- was the superior competitor in mixture with A, barbata
ing the incorporation of the.logistic equation into our we can now quantify this. Fitting equation 4.1 to the
model. It also points to an alternative model which, in results (Firbank & Watkinson, 1985) gave the follow-
quantitative terms, is certain to be more generally ing pair of equations:
applicable than the original. Note, however, that the
more qualitative conclusions concerning exclusion and
and coexistence remain unaffected.
The subscripts B and F refer to A. barbata and A. fatua,
4.14 Analysis of competition in plants respectively, whilst S is the mean number of seeds
produced per plant and N is the density of each species
Where competitive outcomes are assessed by additive in the mixture. This pair of equations neatly encapsu-
designs (see Fig. 4.7) the analysis of competition is lates the absolute differences between the two species
achieved by the use of yield-density models. These and their competitive interactions. On average an
come in a variety of forms depending on the design isolated plant of A. fatua produced 47 seeds (188-141)
used and the ability of the model to fit the observed less than A. barbata but required more space to do so
data empirically (Cousens, 1985). since its ecological neighbour area (1.02) is larger than
for A. barbata (0.41). This numerical superiority in
To analyse additive designs we must reconsider the seed output is offset, however, by the fact that on a
model developed earlier for intraspecific competition one-to-one basis A. fatua is much more competitive as
(equations 3.5 and 3.6). Hassell and Comins (1976) judged by the competition coefficients. In mixture, A.
and Watkinson (1980) have pointed out that this barbata 'perceives' each A. fatua as equivalent to 1.44
model may be easily extended to include the influence of its own individuals. Each A. fatua plant on the other
of a second species.

CHAPTER 4 : INTERSPECIFIC COMPETITION 111

hand 'perceives' each A. barbata plant as about a fifth Fig. 4.28 An abundance diagram for Phleum arenanum and
(0.19) of one of its own. (Note that this analysis does Vulpiafasiculata showing the isoclines (derived from the data
not assume a reciprocity in competition coefficients as shown in Fig. 4.11) and the trajectories(arrowed)
in the case of the replacement series.) We can imme- illustrating the long-term outcome of competitionover
diately see then that this approach extends our
qualitative conclusions about the nature of interspe- succeeding generations. (After Law & Watkinson, 1987.)
cific interactions.
The 'competition coefficient' in equation 4.1 is
density-independent-a 'fixed' value applying to all
densities of the second species. In analysing the
response surface shown in Fig. 4.11,Law and Watkin-
son (1987)showed, by careful statistical analysis, that
a better fit to the observed data was achieved by a
model that allowed competition coefficients to vary
both with frequency and density.
The form of model used was

where, following previous conventions, y iis the mean ary equilibrium density for Vulpia of 227 seeds, with
the eventual eradication of Phleum. Note that this
seed yield per plant of species i, and ymi is the happens despite the fact that isolated plants of Phleum
maximum yield of an isolated plant of species i, and Nl are much more prolific seed producers than Vulpia,
and N2 are the initial densities of each species. The 1160 as opposed to 187 seeds per plant.
power terms b, and b, enable competitive abilities to
be described in a density-dependent manner, but it is The outcome of competition described by yield-
not possible to attach biological meaning to them in density relationships in the form of equation 4.1 may
the same way as b in equation 4.1. The estimated be analysed in a similar manner on joint abundance
parameter values are indicated in the equations 4.3 diagrams but in actual fact there is an alternative and
and the outcome of competition can be determined by easier method of analysis. Consider the pair of recur-
iterating them from any initial pair of population rence equations(4.4)for two species X and Y in which
densities, N,,, (Phleum) and N,,, (Vulpia).Note that this competitive effects are described by an 'equivalence
model assumes discrete generations of population coefficient'.
growth and the seed yield of each species after the
action of competition is reduced by 50% to mimic seed Hassell and Comins (1976)have shown that difference
losses during the dormant stage of the life cycle. equations of this form are analytically tractable. That
is to say, that it is possible by the mathematical
Figure 4.28 gives in a joint abundance diagram of technique of phase plane analysis to identify four
the two species the density combinations that give regions in the parameter space of the coefficients a
zero population growth for each species-the zero
growth isoclines-and some illustrative population and p in which diierent outcomes of competition will
trajectories. All of the population trajectories move
(usually in a single generation) to the space between occur. In two regions the model predicts that one
the isoclinesand then consistentlytowards the bound- species will be driven to extinction by the other; in
another region stable coexistence of both species will
result, whilst in the fourth unstable equilibria may
result (Fig.4.29). Thus from knowledge of the entire

112 PART 2: INTERSPECIFIC INTERACTIONS

Fig. 4.29 Regions in the ap-plane defining the dynamical
behaviour of two species interactions governed by equation
4.9. (After Hassell & Comins. 1976.) The boundaries of the
four regions are determined by Q where

Q = - l])/(al[(hy)1/b2- l]).

See text for details.

set of parameter values, we might make predictions of Fig. 4.30 Resource utilization curves for three species
the long-term outcome. As mentioned above, yield- coexisting in a one-dimensional niche system; d is the
density responses have long been explored by agron- distance between curve maxima, which occur at the centre
omists seeking models to describe the relationships of the curve; W is the standard deviation of the curves.
between crop yield and weed densities in additive (a)Narrow niches with little overlap (d > W).(b)Broad
competition experiments. Spitters (1983) reviews the niches with considerable overlap (d < W).
range of reciprocal yield models and Mortimer et al.
(1989) describe their relationship to equations 4.4. initiated by MacArthur and Levins (1967) and devel-
oped by May (1973).
4.15 Nicheoverlap
Imagine three species competing for a single, unidi-
We return to the question first posed in section 4.10. mensional resource which is distributed continuously;
How much niche differentiation is necessary for the food size and food at different heights in a forest
coexistence of interspecific competitors? We shall see canopy both conform to this description. Each species
that this question is intimately related to a problem has its own niche in this single dimension within
which, in evolutionary terms, confronts all species: which it will (in our examples) consume food. More-
that of offsetting interspecific competition against over, its consumption-rate is highest at the centre of
intraspecific competition (section 4.12). There have its niche and tails off to zero at either end. Its niche
been several theoretical approaches to the solution of
this problem, but since, for the most part, they reach
similar conclusions, we can concentrate on the one

CHAPTER 4: INTERSPECIFIC COMPETITION 113

can, therefore, be visualized as a resource utilization
curve (Fig.4.30). Clearly, the more adjacent species'
utilization curves overlap, the more they compete.
Indeed, if we make the (fairly restrictive) assumptions
that the curves are 'normal' (in the statistical sense),
and that the different species have similar curves,then
the competitioncoefficient(applicableto both adjacent
species)can be related to the standard deviation of the
curves, W,and the difference between their peaks, d,
by the following formula:

-d2

Thus a is very small when there is considerable Fig. 4.31 Niche overlap and coexistence. The range of
separationofadjacentcurves(dlw >> 1,Fig. 4.30a)and resource utilization (indicated by the carrying capacities, K,
approaches unity as the curves themselves approach and K,, where K, = K,) which permits a three-species
one another (dlw < 1,Fig. 4.30b). equilibrium community with various degrees of niche
overlap (dlw). (After May, 1973.)
In terms of this model, we can ask the following
question. How much overlap of adjacent utilization the conditions for coexistence are extremely restric-
curves is compatible with stable coexistence? Obvi- tive, but these restrictions lift rapidly as dlw ap-
ously, if there is very little overlap, as in Fig. 4.30, proaches and exceeds unity. In other words, stable
there is very little interspecific competition and com- coexistence is possible under the restrictive conditions
petitors can coexist. Conversely, in such a case the imposed by low values of dlw, but as May (1973)points
species have rather narrow niches. This means, since
all conspecifics are consuming very similar food, that out, because conditions are so restrictive '...it may be
there is intense intraspecific competition. Moreover,
plausibly argued that environmental vagaries in the
the food items in those positions along the resource real world will upset such an equilibrium'.
spectrum where the curves overlap are being almost
totally ignored by the consumers. It is, therefore, likely So we have seen that high values of dlw there is
that natural selection will favour an increased con- intense intraspecific competition and underexploita-
sumption of these neglected food items, an increase in tion of resources, and at low values of dlw the
niche breadth, a lessening of intraspecificcompetition, equilibrium is too fragile to be maintained in the real
and thus an increase in niche overlap. The question is: world. Theory, therefore, suggests that the coexistence
how much? of competitors (using a unidimensional resource) will
be based on niche differentiation in which &Wis
MacArthur and Levins (1967) and May (1973) approximately equal to, or slightly greater than unity.
answered this question by what was, in essence, an Unfortunately, the testing of this suggestion is hin-
extension of the search for stable coexistence pursued dered by two major problems. The first is that it
in section 4.13. They assumed that the two peripheral applies only to situations in which there is a simple,
species had similar carrying capacities (K,, propor- unidimensional resource (probably quite rare), and in
tional to the area under the utilization curve), and which the utilization curves are at least approximately
considered the coexistence between them of an inter- the same as in the model. Competition in several
mediate species (carrying capacity K,). Their results dimensions, and certain alternative utilization curves
are illustrated in Fig. 4.31, which indicates the values (Abrarns,1976)would both lead to lower values of dlw
of K,lK, that are compatible with stable coexistence being compatible with robust, stable coexistence. The
for various values of &W.At low values of dlw (high a) second problem is the collection of the appropriate

114 P A R T 2 : INTERSPECIFIC INTERACTIONS

data. These are needed in order to establish the petitors. However, for many competing species, habi-
dimensionality of competition, and the form of the tats are often less predictable and shorter lived. Many
utilization curve; and also to determine not only d but species exploit resources which are divided into small,
W as well. Thus, the field evidence persuasively used discrete patches, such as carrion, dung, fungi or fruit.
by May (1973) in support of this model's suggestions In these patches there is often time for just one
was even more persuasively criticized by Abrams generation before the resource unit (patch) becomes
(1976). In particular, there is a grave danger, when unusable. Here, species' competitive abilities are not
attempting to 'test' the predictions of this model, that the only determinants of coexistence or competitive
those field examples that support it will be selected, exclusion.
while those that do not support it are ignored. From
our own examples, for instance, we could select the Atkinson and Shorrocks (1981) explored by com-
ant Veromessor pergandei (see Fig. 4.6) and the two puter simulation, and Hanski (1981) by a simple
hydrobiid snails (see Fig. 4.19) as providing empirical analytical model, ways of incorporating habitat heter-
support for the model's prediction (dfw= 1). Yet, while ogeneity into competition models. They examined the
such support is gratifying, it most certainly does not effect of resource subdivision on two species competi-
prove that the model is correct. What is needed is tion and found that their coexistence could be facili-
evidence that such 'supportive' patterns occur much tated by dividing the resource into more and smaller
more frequently than would be expected by chance breeding patches, assuming that the distribution of
alone; and as yet we have insufficient data to provide individuals was aggregated over those patches. The
this evidence. opposite of an aggregated distribution is an even
distribution when each patch contains an equal num-
Nevertheless, the model has shown us that there is ber of competitors. If both competitors were evenly
likely to be some limit to the similarity of competing distributed then the inferior competitor would come
species; and that this limit represents a balance into conflict with the superior competitor on every
between, on the one hand, the evolutionary avoidance patch and would be eliminated. This is not the case
of intraspecific competition and the underexploitation with aggregated distributions, when, depending on
of resources, and on the other hand, the evolutionary the level of aggregation, many individuals of the
avoidance of equilibria which are too fragile to with- inferior competitor meet few, if any, individuals of the
stand the vagaries of the real world. Once again, superior competitor. In particular, the models of
therefore, a model, without being 'correct', has been Atkinson and Shorrocks and of Hanski demonstrate
immensely instructive. the possibility of continued coexistence of an inferior
with a superior competitor, when the two species have
4.16 Competitionand heterogeneity independently aggregated distributions over the re-
source patches. Coexistence occurs because the com-
Most of this chapter has been concerned with compet- petitive pressure of the superior competitor is largely
itive situations in which the habitat has remained directed-as a result of aggregation-at members of its
more or less constant and environmental conditions own species,those present in the high density patches.
have remained more or less stable. Under these In other patches, the superior competitor will be
circumstancesthe models we have explored, based on scarce or absent, and here the inferior competitor is
intrinsic rates of increase, carrying capacities and able to escape interspecific competition. In other
competitive abilities, can be expected to give a reason- words, coexistence is facilitated becasue the variation
able description of interspecific competition. Individu- in population densities among the habitat patches
als have had time to distribute themselves relatively shifts the balance from interspecific towards more
evenly in the suitable habitats and, as assumed by the intraspecificcompetition. It should be noted that if two
classical theory, each individual encounters about the species do not have relatively independent distribu-
same number of conspecific and heterospecific com- tions but have a tendency to aggregate in the same

CHAPTER 4: INTERSPECIFIC COMPETITION 115

patches, then coexistence is less likely. Thus habitat per patch was only 2.7. AIl nine species showed highly
patchiness itself is not critical: the important question aggregateddistributions and particular pairs of species
is how that heterogeneity is perceived by individuals. came into contact only rarely, thus interspecific com-
petition was much less intense than might have been
Rosewell et al. (1990) and Shorrocks et al. (1990) expected on the basis of the number of species and
have tested the assumptions of the aggregation mode1 number of patches present. Ives (1991), by a mixture
of coexistence using 360 data sets, largely for droso- of experiment similar to that described above, and
philid flies-the group that inspired the original model. theory, has estimated that the effect of aggregation on
The assumptions tested were that: (i) the competing coexistence for the five carrion fly species present in
individuals (larvae) are aggregated; (ii) the degree of his study was equivalent to reducing the amount of
aggregation can be represented by a parameter (k of larval competition between pairs of species by an
the negative binomial distribution)which remains the average of at least 57%.
same at different overall densities; and (iii) there is no
strong association between species. The vast majority The aggregation model predicts that coexistence of
(930) of the data sets showed larval aggregation, so many species is often dependent on their indepen-
the first assumptionwas upheld: competing stages are dently aggregated spatial distributions. Hanski (1987)
strongly aggregated. Of the data sets 79% were ade- has tested this prediction with a field experiment with
quately described by a negative binomial distribution. blowfiies. Female flies were allowed to oviposit for 3
The assumption that aggregation (as measured by the days on small carcasses placed at 15m intervals in a
parameter k of the negative binomial distribution)does homogeneous field. After the oviposition period, the
not change with density was less well supported.
Aggregation declined at higher densities and thus Fig. 4.32 Comparison between the control and
coexistence may be less likely in larger patches (see experimental rearings in species number. The open
also Shorrocks & Rosewell, 1987). Even if this second histogram is the null hypothesis, derived by pooling flies
assumption is relaxed, however, the predictions of the from each combination of two and four of the 20 control
model remain essentially unchanged. The final as-
sumption, that species utilizing patchy resources are rearings (n = 190 and 4845, respectively). The number of
distributed independently of each other was poten-
tially a complicating factor for the two-species model, experimental rearings was 11(two pieces) and nine (four
but in only 5%of possible comparisonsexamined were pieces). Removing resource patchiness increases the
significant associationsfound. Thus coexistence of two dominance of the best competitor and decreases species
species with independently aggregated distributions is
likely in situationsin which one or other of the species richness. (From Hanski, 1987.)
would have been eliminated in a homogeneous envi-
ronment.

Studies on the coexistence of competitors on
ephemeral resources have not been restricted to fruit
flies. Hanski and Kuusela (1977), Hanski (1987) and
Ives (1991) have looked at aggregation and coexist-
ence in carrion fly communities (the community of
flies whose larvae feed on dead animals, usually
vertebrates). In Hanski and Kuusela's study, 50 small
pieces of carrion were laid out in a small area and
natural colonization was permitted to take place. A
total of nine species of carrion fly was recovered from
the patches, but the mean number of species emerging

116 PART 2: INTERSPECIFIC INTERACTIONS

carcasses were removed from the field and placed in carcasses maintains species richness in this commu-
containers to rear the flies out. In the control rearings, nity.
each carcass was placed in its own container, whereas
in the experimental rearings two or four carcasses There seems little doubt that aggregation plays a
were placed next to each other in the same container. major role in explaining the coexistence of competitors
In the experimental rearings fly maggots could easily on divided and ephemeral resources that would ex-
move from one carcass to another, and thereby clude one another on resources that were more
density variation among the carcasses placed in the homogeneous in space and time. Although insects
same container was effectively removed. Figure 4.32 living in carrion, dung, fruit or fungi may be extreme
shows that decreasing larval aggregation (the experi-
mental rearings) decreased the number of fly species examples of species exploiting patchy habitats, they
emerging from the rearing, in comparison with the should not be regarded as exceptions to a rule of
expected results calculated from the control rearings.
This result directly demonstrates how the aggregated habitat homogeneity, rather as one end of a contin-
spatial distribution of fly maggots among replicate
uum. All habitats are patchy and ephemeral to a

degree, hence the theory outlined in this section can
be expected to be applicable to many communities of
competitors.

Chapter 5

5.1 Introduction are much more like parasites; aphids, for instance, live
in close association with plant hosts, gain sustenance
Although this chapter is entitled 'Predation', we shall from them, and reduce their vitality. Many other
be dealing in it with a variety of interactions.In all the herbivores, however, fall into neither category. They
cases we shall be considering, however, animals will do not live in close association with any one plant
totallyconsume,orpartlyconsumeand alsoharm,other host; but by consuming parts of plants, they are often
animals or plants. This broad umbrella is designed to the ultimate, if not the immediate, cause of plant
include each of the following categories of 'predator'. mortality. Thus, they certainly do have a detrimental
1 True predators that consume other animals-their effect on their plant 'prey'.
prey-and thus gain sustenance for their own survival
and reproduction. These four categories are all quite distinct and
2 Parasitoids. These are insects, mainly from the order special in their own way, but, as their inclusion within
Hymenoptera, but also including many Diptera, a single definition shows, they share important com-
which are free-living in their adult stage, but lay their mon features. For this reason there will be many
eggs in, on or near other insects (or, more rarely, statements in this chapter that apply to all categories.
spiders or woodlice). The larval parasitoids then de- It would be tedious to refer individually in every such
velop on or within their 'host' (usually, itself, a case to 'true predators and their prey, parasitoids and
pre-adult), initially doing little harm, but eventually their hosts, parasites and their hosts, and herbivores
almost totally consuming, and therefore killing, the and their food plants'. Instead, for simplicity, and
host prior to pupation. Often just one parasitoid where the context precludes ambiguity, the four
develops from each host, but in some species several categories will be referred to, together, as predators
individuals share a host. Nevertheless, in either case, (or, rarely, consumers) and their prey.
the number of hosts attacked in one generation closely
defines the number of parasitoids produced in the We can, therefore, apply a single definition and a
next. This, along with the fact that the act of 'preda- single,all-embracinglabel to these four categories.But
tion' is confined to a particular phase of the life history we can also apply a single naive expectation regarding
(adult females attacking hosts), means that parasitoids their population dynamics. If we imagine a single spe-
are especially suitable for study. They have provided a ciesof predator and a singlespeciesof prey,then we can
wealth of information relevant to predation generally. expect predators to increase in abundance when there
3 Parasites. These are organisms (animals or plants) are large numbers of prey. However, there should then
that live in an obligatory, close association, usually be an increase in predation-pressure on the prey,
with a single host individual for a large portion of their leading to a decrease in prey abundance. This will
lives. They gain sustenance from their host and do ultimately lead to food-shortage for the predators, a
their host harm, but parasite-induced host mortality is decrease in predator abundance, a concomitant dropin
by no means the general rule, and is often rather rare. predation-pressure, an increase in prey abundance
4 Herbivores. These are animals that eat plants. Some and so on. In other words, there appears, superficially,
appear to act as true predators, since they totally to be an inherent tendencyfor predators and their prey
consume other organisms for their own sustenance (and parasitoids and their hosts, parasites and their
(seed-eaters are a particularly good example). Others hosts, and herbivores and their food plants)to undergo
coupled oscillations in abundance, predator numbers
'tracking' those of the prey.

118 PART 2: INTERSPECIFIC INTERACTIONS

5.2 Patternsof abundance exhibit periodic outbreaks (Fig. 5.le), and cases in

Some actual examples of abundance patterns are which the 'predators' have no noticeable, simple effect
shown in Fig. 5.1. Certainly, some do appear to show
coupled oscillations (Fig. 5. l a & 5.lb), but there are on the prey since the fluctuations in the sizes of the
also cases in which the 'prey' are kept at a constant two populations are apparently unconnected (Fig. 5.lf
low level (Fig. 5. l c & 5.ld), cases in which the prey & 5.lg).

It is clear, in other words, even from this limited
range of examples, that actual predators, parasitoids,

(f 1 I Wood mice and bank voles

0' 5 10 15 20 25
Generation
5 t(c) g 60
Beetle introduced

.L L'II \ Estimated 80 2

II . house mouse
II

0 50 I

150
Days

Fig. 5.1 Patterns of abundance in predator-prey systems. beetle Laemophloeas minutus of being infected by the
(a)The lynx Lynx canadensis and the snowshoe hare Lepus protozoan Mattesia dispora; parasite-host. (After Finlayson,
americanus; true predator-prey. (After MacLulick, 1937.) 1949.)(e)The psyllid Cardiaspina albitextura and mortality
(b)The wasp Heterospilus prosopidis and the bean weevil caused by Syrphus species; true predator-prey. (After Clark,
Callosobruchus chinensis; parasitoid-host. (After Utida, 1963.)(f)Tawny owls Strix aluco and wood mice and bank
voles Apodemus sylvaticus and Clethrionomys glareolus: true
1957.)(c) Changes in the abundance of the Klamath weed
Hypericum perjoratum following the introduction of the leaf- predator-prey. (After Southern, 1970.)(g) The house mouse
Mus musculus and barley Hordeum vulgare; herbivore-plant.
eating beetle Chpjsolina quadrigemina; herbivore-plant.
(After Newsome, 1969a.)
(After Huffaker & Kennett, 1959.) (d)The effect on the

CHAPTER 5 : PREDATION 119

parasites, herbivores and their 'prey' all exhibit a wide of prey, paying special attention to the effects of
variety of patterns of abundance, and we can imme- herbivores on plants (section 5.5). Then, in section
diately see two rather obvious reasons for this. The 5.6, consideration is given to the ways in which prey
first is that predators and prey do not normally exist as 'thresholds' and food quality complicate the relation-
simple, two-species systems. To understand the abun- ship between predation-rate and the beneficial effects
dance patterns exhibited by two interacting species, to the predator. Section 5.7 covers the way in which
these must be viewed in a realistic, multi-species predation-rate is influenced by the availability (espe-
context. The second reason is that our conception of cially the density) of prey items; while a long section
even the simplest, abstracted, two-species system is 5.8 considers some of the consequences of environ-
itself excessively naive. Before multi-species systems mental heterogeneity (particular attention being paid
are even considered, we must abandon our expecta- to the unequal way in which many predators distrib-
tion of universal prey-predator oscillations, and look ute their harmful effects amongst individual prey). In
instead, much more closely, at the ways in which section 5.9, the consequences of mutual interference
predators and their prey interact in practice. amongst predators are examined, and in section 5.10,
the similarities between the effects produced by the
We shall examine, in turn, the individual compo- processes in sections 5.8 and 5.9 are discussed. Then,
nents of the predator-prey relationship (an approach in section 5.11 attention is drawn to the tendency of
employed successfully by Hassell, 1976, 1978); but predators to maximize their 'profits' by 'foraging
the journey covering these components will be a long optimally'; while section 5.12 is a resume of the
ancl fairly complicated one. Moreover, it is only after all preceding nine sections.
the topics have been examined in detail that we shall
be able to reassemble them into an integrated whole. Throughout these sections we will be especially
At this stage, therefore, we provide an itinerary which concerned with the effects that these individual com-
can be referred to now or part-way through the ponents have on the dynamics of the predator and
journey. This is set out in Table 5.1, and in the prey populations; and in particular, with the regula-
folllowing outline of the rest of this chapter. tory, stabilizing effects of density-dependentprocesses,
and the destabilizing effects of inversely density-
In section 5.3, the patterns of food preference dependent processes (sections 2.3 and 2.6). Then, in
shown by predators are examined, along with some section 5.13, we attempt to incorporate many of the
possible determinants of these patterns. Then, in behaviourally complex components into models of
section 5.4, a number of effects of time scales and predator-prey interactions, hoping to explore further
timing are considered. This is followed by an exami- their effects on population dynamics and stability;and
nation of the detailed effects of predators on the fitness this allows us, in section 5.14, to reconsider the
abundance patterns of the present section in the light
Table 5.1 Summary of the components of the predator- of sections 5.3-5.13. Finally, section 5.15 examines
prey interaction which are examined in this chapter. problems emanating from the contrived human pre-
dation involved in the process of harvesting.

5.3 Coevolution, and specialization
amongst predators

Predators of all types can be classified, according to
their diet width, as monophagous (feeding on a single
prey type),oligophagous(fewprey types)or polyphaous
(many prey types),and the degree of specializationcan
have important effects on predator-prey dynamics.

120 PART 2: INTERSPECIFIC INTERACTIONS

The abundance of a monophagous predator, for in- fitness; fitness is increased by increasing the profit-
stance, is likely to be closely linked to the distribution ability of food-acquisition; evolution, therefore, fa-
and abundance of its prey; while a polyphage is very vours predators that choose profitable prey). It is also
unlikely to have its abundance determined by any one borne out by the facts. Figure 5.2, for instance,
of its prey types. If we are to understand the patterns illustrates examples of predators actively selecting
of diet width amongst predators, however, we must those prey items which are most profitable, i.e. prey
begin by establishingtwo basic points. The first is that items for which the gain (in terms of energy intake per
predators choose profitable prey. This follows, by an act unit time spent handling prey) is greatest.
of faith, from a consideration of natural selection
(evolution favours those individuals with the highest The second point stems from the fact that all
animals and plants have evolved in response to
Fig. 5.2 Predators eating 'profitable'prey: (a)crabs eating selection pressures originating, to a large extent, from
mussels (Elner & Hughes, 1978); (b) pied wagtails eating the other animals and plants in their environment.
flies (Davies, 1977). (After Krebs, 1978.) 1 calorie (non-S1 The point, then, is that predators and their prey are likely
unit)= 4.186 joules. to have coevolved. There is a continuous selection
pressure on prey to avoid death (or, more generally,
fitness-reduction)at the hands of their predators, and
a reciprocal, continuous pressure on predators to
increase their fitness by exploiting their prey more
effectively. At perhaps its most trivial, this evolution-
ary arms race consists of prey that can run quickly
from their predators, and predators that can run
quickly after their prey, both being favoured by
natural selection. However, we can see the results of
analogous pressures on prey in the distasteful or
poisonous chemicals in the leaves of many plants, in
the spines of hedgehogs, the camouflage coloration of
many insects, and the immunological responses of
hosts to parasite infection; while in predators these
pressures results, for instance, in the long, stout,
penetrative ovipositors of wood wasps, the hooks and
suckers on the heads of tapeworms, and the silent
approach and sensory excellence of owls. It is clear, in
short, that no natural predator-prey interaction can
be properly understood unless it is realized that each
protagonist has played an essential role in the evolu-
tion of the other.

It is equally clear, however, irrespective of any
coevolution, that no predator can possibly be capable
of consuming all types of prey; simple design con-
straints prevent shrews from eating owls, and
humming-birds from eating seeds. Nevertheless, co-
evolution provides an added force in the restriction of
diet width. Each prey species responds (in an evolu-
tionary sense and on an evolutionary time scale) in a
differentway to the pressures imposed by its predators,

CHAPTER 5: PREDATION 121

but a predator cannot 'coevolve' in a wide range of voured by the exclusivity(and, perhaps, predictability)
directions simultaneously. Predators, therefore, tend of the Senita, and the consequent lack of interspecific
to specialize to a greater or lesser extent, and the competition. To a certain extent, similar arguments
better adapted a predator is for the exploitation of a can be advanced to explain diet width generally.Thus,
particular prey species, the less likely it is to profit where prey exert pressures which demand specialized
from the exploitation of a wide variety of prey. morphological adaptation by the predator, there is a
Parasites, in particular, because they live in such tendency for the predator to have a narrow range of
intimate association with their hosts, tend to specia- diet; and where predators feed on an unpredictable
lize, evolutionarily, on a single host species. Their resource, there is a tendency for them not to be
whole life style and life cycle is finely tuned to that of specialists.
their host, and this precludes their being finely tuned
to any other host species. They are, therefore, com- 5.3.1 One explanation for the
monly monophagous. For similar reasons, the same is degrees of specialization
true of many parasitoids, although in their case,
possibly because they are free-living as adults, polyph- In many cases, however, some further explanation is
agy (or at least oligophagy) is rather more common. needed, and this may be provided by the ideas of
MacArthur and Pianka (1966; see also MacArthur
Conversely,with herbivores and predators the situ- 1972). In order to obtain food, any predator must
ation is much less clear-cut. At one extreme there are expend time and energy, first in searching for its prey,
examples of complete monophagy. The fruit fly, Droso- and then in handling it (i.e. pursuing, subduing and
phila pachea, for example, indigenous to the Sonoran consuming it). Searching will tend to be directed, to
Desert of the south-west USA, is the only species some degree, towards particular prey types; but, while
capable of consuming the rotten tissues of the Senita searching,a predator is nevertheless likely to encoun-
cactus (and its associated micro-organisms), because ter a wide variety of food items. MacArthur and
all other species are poisoned by an alkaloid- Pianka, therefore, saw diet width as being determined
pilocereine-which the Senita produces (Heed et al., by the choices made by predators once they had
1976). Moreover, natural selection has favoured D. encountered prey. Generalist predators are those that
pachea's exploitation of this exclusive niche to such an choose to pursue (and, hopefully, subdue and con-
extent that the fly has an absolute requirement for sume) a large proportion of the prey they encounter;
certain unusual sterols which are also only produced specialists are those that continue searching except
by the Senita. Many plants produce 'secondary', when they encounter prey of their specifically pre-
protective chemicals and, probably as a consequence, ferred type.
many herbivorous insects are specialists (Lawton &
McNeill, 1979; Price, 1980). However, most true The basic conclusion MacArthur and Pianka drew
predators and many herbivores feed on a variety of from these considerations can be stated as follows.
prey items. Predator choices are determined, ultimately, by the
forces of natural selection, and are driven by these
A partial explanation of the range of diet widths is forces towards the maximization of profitability for the
provided by a simple consideration of the fact that, predator. Natural selection, therefore, favours a pre-
although there is pressure from coevolution towards
predator specialization, there is a counterbalancing dator that chooses to pursue a particular prey item if,
evolutionary pressure discouraging the reliance of
predators on an unpredictable or heavily exploited during the time it takes to handle that prey item, the
resource. For many parasites, the necessity for special- predator cannot expect to search for and handle a
ization clearly provides a pressure that outweighs any more profitable prey item. On this basis, predators
disadvantages stemming from resource unpredictabil- with handling times that are generally short compared
ity; for D. pachea, specialization is presumably fa- to their search times should be catholic in their tastes,
because in the short time it takes them to handle a

122 PART 2: INTERSPECIFIC INTERACTIONS

prey item which has already been found, they can and white spruce. As Table 5.2 shows, the deer, with
barely begin to search for another prey item. This is free and equal access to all four species, exhibited a
MacArthur and Pianka' s explanation for the broad fairly consistent preference for jack pine followed by
diets of many insectivorous birds that 'glean' foliage. white pine, with red pine being only lightly browsed
Search is always moderately time-consuming; but and white spruce ignored. Such preference amongst
handling the minute, stationary insects takes neglig- polyphagous animals is, in fact, the general rule; but
ible time and is almost always successful. Ignoring as Murdoch and Oaten (1975)have shown, there are
such prey items (i.e. narrowing diet width) would, two distinct forms that this preference can take.
therefore, decrease overall profitability, and the birds
tend to be generalists. An example of the simpler and much commoner
form is shown in Fig. 5.3 (Murdoch, 1969). Two types
By contrast, many other predators have search of predatory shore snails, Thais and Acanthina, were
times that are short relative to their handling times. In presented with two species of mussel (Mytilus edulis
such cases, specialization will be favoured, because and M. californianus) as prey, at a range of prey
the predators can expect to find a more profitable food proportions. When the porportions were equal, the
item very soon after ignoring a less profitable one. snails showed a marked preference for the thinner-
Lions, for instance live more or less constantly in sight shelled M. edulis. The line in Fig. 5.3 has been drawn
of their prey so that search time is negligible; con- on the assumption that they retained this same
versely, handling time and particularly pursuit time, preference at other (unequal) proportions, and this
can be very long (and energy-consuming). Lions con- assumption is clearlyjustified: irrespectiveof availabil-
sequentlyspecialize on those prey that can be pursued ity, the snails showed the same marked preference for
most profitably: the immature, the lame and the old. the less protected prey, which they could exploit most
Thus, on the basis of MacArthur and Pianka's ideas effectively.
we can expect pursuers or handlers (like lions)to have
relatively narrow diets, and searchers to have rela- This can be contrasted with data of a rarer sort in
tively broad ones. Along similar lines, Recher (in Fig. 5.4a, obtained from an experiment in which the
MacArthur, 1972)found that great blue herons in the predatory water bug, Notonecta glauca, was presented
productive waters of Florida (in which search times
were consequently short) had a much narrower range
of food size in their diet than those inhabiting the
unproductive lakes of the Adirondacks (where search
times were relatively long).

Clearly, we can go some way at least towards an
understanding of the degrees of specialization shown
by predators.

5.3.2 Food preference and predator switching Fig. 5.3 Snails exhibiting a consistent preference amongst

Irrespective of this range of diet width, however, othfethmeiursrseellastiMveytailbuus neddualniscean(md eMa.nsca+li2foSrnEi)a.n(AusfitrerreMspuercdtiovceh
polyphagy is very common; especially amongst pred-
ators and herbivores. Yet polyphagous animals are & Oaten, 1975).
rarely indiscriminate in the various types of food they
eat. Horton (1964),for example, presents the results of
an accidental field experiment in which deer broke
into a plantation containing four species of tree
arranged at random; white pine, red pine, jack pine

CHAPTER 5 : PREDATION 123

Table 5.2 Food preferences exhibited by deer. Percentage
browsing incidence of deer on planted trees. (After Horton,
1964.)

with two types of prey-the isopod, Asellus aquaticus Fig. 5.4 (a) The percentage of Asellus in the diet of Notonecta
and mayfly larvae (Cloeon dipterum)--with overall prey as a function of their relative abundance;the thinner l i e
density held constant (Lawtonet al., 1974). In this case indicates the function expected on the basis of a consistent
the preference exhibited when the two prey types preference. (b) The effect of 'experience'on the success of
were equally available, if extrapolated to other avail- Notonecta in attacking Asellus. (After Lawton et al., 1974.)
abilities (thin line in Fig. 5.4a), is obviously not a good (Means and total ranges are indicated.)
indication of the overall response. Instead, Notonecta
took a disproportionately small number of Asellus tubificid worms as prey (Fig. 5.5). Figure 5.5a shows
when they were scarce, and a disproportionately high that there was, indeed, switching by the guppies. But
number when they were common. This is known as more interesting than this is Fig. 5.5b, which shows
predator switching, since it suggests that predators that although the population of predators showed little
switch their preference to whichever prey is most preference when offered equal proportions of the two
common. In the case of Notonecta, the explanation is prey, the individual guppies showed considerable spe-
apparently illustrated in Fig. 5.4b: the more previous cialization. This does not contradict the idea of a
experience Notonecta has of Asellus, the more likely it is search image, but does suggest that predator switch-
to make a successful attack. It appears, in other ing in a population does not result from individual
words, as if predator switching is based upon a learnt predators gradually changing their preference, but
ability to specialize. This, essentially,is the view taken from the proportion of specialists changing. Murdoch and
by Tinbergen (1960), who proposed that certain pre- Oaten provide evidence for the occurrence of this type
dators, particularly vertebrates, develop a 'specific of switching in other vertebrate predators, in some
searching image'. This enables them to search more
successfully (since they effectively 'know what they
are looking for'), and results in them concentratingon
their 'image' prey to the relative exclusion of their
non-image prey. Moreover, since the searching image
develops as a result of previous experience, and since
the predators (or herbivores) are most likely to experi-
ence common prey, we can expect predators to
concentrate on a prey type when that type is common
andl switch to another prey type when it is rare.

The basis for predator switching has been discussed
in more detail by Murdoch and Oaten (1975). One of
their most instructive examples is the work of Mur-
doch, Avery and Smyth (in Murdoch & Oaten, 1975)
on guppies offered a choice between fruit flies and

124 PART 2: INTERSPECIFIC INTERACTIONS

Fig. 5.6 Seasonal changes in diet. The percentage
contributionsof different food types to the diet of the wood
pigon Columba palumbus in Cambridgeshire,UK. (After
Murton et al., 1964.)

Fig. 5.5 (a) Switching in guppies fed on Drosophila and 5.4 Time and timing
tubificids (with total ranges indicated). (b) A frequency
histogram showing the number of individual guppies with In any discussion of predation there are several points
particular types of diet when offered equal numbers of the that must be made regarding the effects of time and
two prey types. All showed a preference (< 0.4 or > 0.6) timing. The first is that levels of specialization need not
even though the population as a whole consumed involve morphological adaptation, but may, in many
approximately equal amounts of the two types. (After cases, be the result of differing degrees of 'temporal'
Murdoch & Oaten, 1975.) coincidence. Thus, the European rabbit flea (Spilopsyl-
lus cuniculi), like very many parasites, has a life cycle
invertebrate predators, and in a herbivore: the feral that coincides more or less exactly with that of its host:
pigeon feeding on peas and beans (Murton, 1971). maturation of the flea can only occur on a doe in the
Nevertheless, it is fair to conclude that this more latter part of pregnancy, and eggs are laid only on the
complex type of food preference is most common in newborn rabbit young (Mead-Briggs & Rudge, 1960).
vertebrates, where the ability to learn from experience The wood pigeon (Columba palumbus), by contrast,
is most highly developed. The consequences of it are switches its food preference seasonally, depending on
discussed in section 5.7.4. availability (Fig. 5.6: Murton et al., 1964). Thus, while
it may, at any one time, be fairly specialized in its
feeding habits, it is temporally (as well as morphdo-

CHAPTER 5 : PREDATION 125

gically) a generalist; and, unlike the parasite, its dealingwith systemsin which both consumer and food
population dynamics are not strongly dependent on contribute to the effects of any time-lags. There is,
the availability of any one type of prey. consequently, a tendency for populations of predators
and their prey to fluctuate out of phase with one
Related to the effects of temporal coincidence are another, and it is essentially this that leads to the
the effects of differing lengths of life cycle amongst 'naive expectation' of coupled predator-prey oscilla-
predators and their prey. Aphids, for instance, gener- tions outlined in section 5.l.Varley (1947) coined a
ally pass through several generations for each time special term-delayed density-dependence-to de-
their host plant passes through a seasonal cycle (never scribe such a relationship between predation-rate and
mind a generation). They can, therefore, be expected prey density.
to react quickly towards, and fairly accurately reflect,
the quantity and quality of their food. Conversely, the 5.5 Effects on prey fitness
dynamics of a sycamore population is unlikely to be
greatly influenced by the intraseasonal fluctuations in Turning to this next component, we can note that
the abundance of its aphids. Similarly, small mam- when predators and prey interact, the fitness of prey
mals, with a fairly high intrinsic rate of increase and a individuals is obviously affected by the predators, but
life span never exceeding a year, exhibit a pattern of it is also influenced by the prey themselves through
abundance that reflects the yearly changes in environ- the density-dependent process of intraspecific compe-
mental quality; while tawny owls in the same habitat, tition. As Chapters 1-3 made clear, this will tend to
often living 6 years and failing to breed when food is regulate the size of a single-species prey population,
scarce, maintain a comparative constancy of abun- but in so doing it will also tend to stabilize the
dance irrespective of these environmentalfluctuations interaction between the prey and their predators. Prey
(Southern, 1970). populations reduced by their predators will experience
a compensatory decline in the depressant effects of
Iin part such 'discrepancies' are the result of making intraspecific competition; while those that grow large
connparisons on an artificial, yearly basis, when it through the rarity of predators will suffer the conse-
might be more appropriate to consider the fluctua- quences of intraspecific competition all the more
tions from generation to generation.Nevertheless,it is intensely.
quite clear that generation times, and in particular the
relative generation times of a 'predator' and its 'prey', The most important effects on prey fitness in the
can have important effects on predator-prey dynam- present context, however, are attributable to preda-
ics through their influence on the speeds at which tors. These can be most easily described in the case of
species respond to changes in the environment. parasitoids and their hosts: a host which is success-
fully attacked dies-its fitness is reduced to zero. It
Another aspect of timing relevant to predator-prey might appear, moreover, that the effects of true
dynamics is the effect of time-lags. We have already predators on their prey are equally straightforward-
seen (in section 3.4.1) that time-lags tend to cause and they often are. But consider, by contrast, the work
increased levels of fluctuation in dynamic systems(i.e. of Errington (1946). Errington made a long and
they tend to destabilize systems), and we can expect intensive study of populations of the musk-rat (Onda-
this to apply to populations of predators and their tra zibethicus) in the north-central USA. He took
prey. Once again, populations with discrete genera- censuses, recorded mortalities and movements, fol-
tions will tend to oscillate in numbers more than lowed the fates of individual litters, and was particu-
populations breeding continuously; but there is, in a larly concerned with predation on the musk-rat by the
sense, an additional reinforcement of such time-lags mink (Mustela vison). He found that adult musk-rats
in predator-prey systems. In section 3.4.1 we were that were well established in a breeding territory were
considering systems in which the food resource re- largely free from mink predation; but those that were
mains constant in size (as evidenced by the constancy
of the single species' carrying capacity). Now we are

126 PART 2: INTERSPECIFIC INTERACTIONS

Fig. 5.7 The relationship between the rate of infection of tion. Conversely with parasites it is well known that
the water bug Hydrometra myrae (host) with the mite the effects on the host are often not drastic (though,of
Hydrgphantes tenuabilis (parasite) and (a)host survival, course, to conform with the definition in section 5.l,
(b)host life expectancy, (c) host maturity, (d) host fecundity, and thus warrant inclusion in this chapter, the para-
and (e) host rate of increase. (After Lanciani, 1975.) site must have some adverse effect on its host's
fitness). 'Predation-rate' in parasites, therefore, can-
wandering without a territory, or were exposed by not be equated with host death-rate. Instead, it can be
drought, or injured in intraspecific fights were very taken as the rate at which host tissue and energy is
frequently preyed upon. Certainly, those that were diverted from hosts to parasites.
killed had their fitness reduced to zero. Yet, because
these were individuals that were unlikely to ever Predation-rate in parasites, then, is obviously lilkely
produce offspring, they had low or zero fitnesses to increase with increases in the mean number of
anyway. Similar results have been obtained, in fact, parasites per host; and the effect this can have on host
for predation on other vertebrates. Those most likely fitness is depicted in Fig. 5.7, illustrating the work of
to succumb are the young, the homeless, the sick and Lanciani (1975) on the ectoparasitic mite, Hydry-
the decrepit-the very individuals whose immediate phantes tenuabilis, and its host, the water bug Hy-
prospects of producing offspring are worst. The hann- drometra myrae. It is clear from Fig. 5.7a & 5.7b that
ful effects on the prey population are clearly not as the mite affects the survival of its host, and from
drastic as they might be; and while the effects of true Fig. 5 . 7 ~& 5.7d that it affects its host's fecundity.
predators on their prey are often straightforward, Taken all in all, therefore, the mite affects the host's
there are obviously cases, particularly amongst verte- reproductive potential (Fig. 5.7e), and this effect in-
brates, in which the superficial simplicity can be creases with increasing intensity of infection. In other
misleading. In particular, a predator that effectively words, the greater the proportion of parasites to hosts,
ignores the potential contributors to the next prey the greater their depressant effect will be; there is
generation will have very little effect on prey abun- inverse density-dependence tending to destabilize the
dance. host-parasite interaction.

Nevertheless, it is possible in both true predators 5.5.1 The effects of herbivores on plant fitness
and parasitoids to equate the predation-rate in a
population with the prey death-rate; the difficulty is Being primary producers, plants provide food re-
that prey death does not always lead to simple sources and are prey to the attentions of herbivores.
reductions in the overall vitality of the prey popula- Careful examination of natural populations of plants

CHAPTER 5 : PREDATION 127

frequently reveals individuals bearing leaves or stems tored the fates of experimentally sown Douglas fir

that have been trimmed and seeds that are bored or seeds, both in open plots and in plots supposedly
cracked open. Such evidence, moreover, may only be screened from the attacks of vertebrate herbivores.
conspicuous at particular phases in the life cycle of a Their results (Table 5.3) indicate that their screening
plant species. In considering the effects on plants of was effective in that vertebrate granivoreswere largely
animals that eat them it is useful to recognize a excluded; but they also illustrate two further points.
continuum. At one end, the effect may be the imme- The first is that these herbivores do, indeed, appear to
diate death of the individual plant or seed because of have an effect on parental output, since the recruit-
(almost) entire consumption; whilst at the other the ment of seedlings, 1 year after germination, was
effect may be the removal of plant parts with the significantly increased in their absence. The second
donor remaining alive and apparently unaffected in point, however, is that this effect is rather less than
terms of its sgrvival and fecundity. In a simple might be expected from a consideration of vertebrate
classification we can thus distinguish between, on the granivores alone. Other sources of mortality, particu-
one hand, true predators such as granivorous ants larly fungal attack at various stages, appear to act in a
(Chapter 2) and frugivorous bats, and on the other density-dependent, compensatory fashion. Thus the
hand grazers such as caterpillars that characteristi-
cally tend to leave plant meristems from which plant Table 5.3 Compensatory mortality when predation is
regrowth can occur. Between these end-points lies an prevented. The fates, in percentage terms, of Douglas fir
array of herbivore effects that may precipitate the seeds sown in open and screened plots. (After Lawrence &
earlier death of an individual than would otherwise Rediske, 1962.)
occur. We might, for instance envisage a plant patho-
gen reducing the photosynthetic area of an individual
to such a degree that its ability to compete with
neighbours is lowered and its death hastened.

To seed-eating animals, the crops of seed represent
a source of highly nutritious food, which is packaged
in a discrete, compact way; and it would be easy to
argue that each seed eaten represents a measurable
reduction in fitness, since it constitutesthe death of a
whole individual. It is perhaps more valid, however, to
consider not the fitness of the seeds themselves, but
the fitness of the parent that produced them. We can
see, for instance, in the related case of fruit-eating
herbivores, that although a significant quantity of
plant material is lost with every fruit eaten, and many
seeds (i.e. individuals)are destroyed in the process, the
herbivore by acting as an essential agent of seed
digpersal is, from them parent plant's point of view, a
net contributor to fitness. (Similar comments apply to
animal pollinators.) And while the situation is not
quite so clear-cut in the case of specialist seed-eaters,
there is certainly some evidence to suggest that a
parent plant's fitness is not incrementally reduced
each time one of its seeds is eaten.

Lawrence and Rediske (1962). for instance. moni-

128 PART 2: INTERSPECIFIC INTERACTIONS

herbivores, like the minks preying on Errington's Table 5.4 Some examples of plant responses to herbivory
musk-rats, are, to some extent, removing individuals
that are already doomed. plants classified according to 'fundamental botanical
attributes, the nature of their present herbivore fauna,
We have, of course, already met a similar example and the apparent nature of their long-term coevolu-
of granivory in section 4.3: Brown and Davidson's tionary relationships with herbivores'. One group,
(1977) granivorous ants tended to consume seeds that comprising mainly the monocotyledons exhibit
would otherwise have been taken by rodents, and vice growth forms that are likely evolutionary responses to
versa. Yet the two guilds, together, caused a measur- grazing by mammals and orthopterans; plant species
able reduction in the size of the seed population; and may possess short-lived repeating modular units with
since germination and subsequent establishment de- basal meristems, have high capacity for clonal repro-
pends on the occupation of a favourable microsite duction and possess general physical defence mecha-
(which is largely a matter of chance), this, in turn, nisms. The other group (consisting of gymnosperms,
must have caused a reduction in the size of the dicotyledons and non-graminoid monocotyledons)
populations of adult plants in subsequent generations. possess more elaborate growth forms, develop from
Granivores seem, therefore, to adversely affect plants, apical rather than basal meristems and commonly
but not necessarily to the extent suggested by superfi- display toxic specialist chemical defences, recent evo-
cial examination. The attractive hypothesis that dis- lution having been in association with herbivorous
persal of seeds by ants to improved soil close to ant insect taxa. In evolutionary terms then, present
nests has little support from field studies (Rice & observable plant responses may reflect fitness
Westoby, 1988).

When we turn from seeds and fruits to seedlings
and adult plants, the assessment of herbivore damage
becomes, if anything, more difficult. The range of
responses of growing plants (Table 5.4) in general to
the effects of herbivores is considerable and has been
the subject of substantialresearch interest not least for
weed biocontrol (Crawley, 1989).

At first sight it might be argued that all of the
responses illustrated in Table 5.4 will automatically
lead to fitness reductions. For instance metabolic
energy and resources diverted towards chemical or
mechanical defence might otherwise be used for
growth and enhanced survival, and changes in rate of
development and re-allocation of resources may lead
to reduced fecundity. However, we need to view such
responses in context of (i) the coevolution of a plant
species and its herbivores; and (ii) knowledge of the
key factors acting at each stage in the plant's life cycle
that operate in a density-dependent,regulatory man-
ner.

Coevolution might be expected to select for more
tolerant plants, less damaging herbivores and more
specialist relationships. McNaughton (1983) has pro-
posed that the long-term evolution of plants and
herbivores has led to two major groups of terrestrial

CHAPTER 5: PREDATION 129

advantages accrued in response to past selection by

herbivores.

On ecological time scales there is evidence for

indisputable fitness reduction in some instances.

Haines (19 7 9 , for instance, obtained frequencydistri-

butions of seedling populations in different areas

whilst investigating the activities of leaf-cutting ants

(Atta colombica tonsipes). (These ants cut leaf, flower

and fmit material, which is then transported to their

nests where the harvested plant material is decom-

posed by a fungus. The fungus is eaten by the ants,

and the degraded plant remains dumped at 'refuse

.tips1.)The survivorship and size of seedlings of forest

plants are inversely related to the freqeuncy of visita-

tion by the ants (Fig. 5.8). Seedlings in parts of the

forest away from nests and dumps are present in all

size-classes, but on the nests and dumps themselves

most of the seedlings present are very small. Thus, the

chance of a seedling maturing into the forest canopy

appears to be zero if it arises in areas regularly

frequented by ants. It is, however, unusual for insect

herbivores to cause the death of perennial established

plants except in the cases of the deliberate introduc-

tion of biological control (Crawley, 1989).

Much more common than the destruction of whole

plants is loss of stems or leaves. The consequences of

partial defoliation to the subsequent survivorship of a

plant are generally difficult to unravel, while the

demonstration of proximate death by defoliation is

often elusive.

This is because a frequent effect of partial defolia-

tion is to disturb the integrated nature of the whole

physiology of the plant. Loss of leaf means not only the

loss of photosynthetic area, but also the mobilization

of stored reserves for leaf replacement through bud

growth; new growth demands protein, carbohydrate

and minerals-the high quality components required

by predators. Thus, persistent defoliation of young,

actively growing leaves may constitute a rapid drain of

stored reserves. Moreover, such changes in physiology fig. 5.8 S e e d h s do not survive to a large size in the

may result in an alteration in the rate of root growth, presence of ants. The size-structure of seedling populations
in tropical rainforests on nests and refuse dumps of leaf-
*and in some cases in root dieback. grazing cutting- ants Atta colombica fonsipes and on forest soil. The
particular feature of 'pecies that density of seedling height classes are presented. (Data from
is the ability to quickly restore the root to shoot ratios Haines, 1975.)

(in biomass) present before grazing started (Piper &

130 PART 2: INTERSPECIFIC INTERACTIONS

Weiss, 1993). Some species are able to increase vegetative growth in perennial plants tends to reduce
photosynthesis in the remaining plant in response to flowering, but the time at which defoliation occurs is
defoliation (Gold & Caldwell, 1989) and defoliated critical in determining the actual response of the
plants therefore show a higher relative growth-rate plant. If the inflorescence is formed prior to defolia-
than undefoliated ones. In some cases the increased tion, the response to leaf removal is seed abortion, or
net primary production can exceed that of control for individual seeds to be smaller; whereas defoliation
plants, in which case the response has been described before inflorescence production is likely to halt or
as 'overcompensatory', in contrast to those where the severely constrain flower formation. Flower loss itself
net rate is decreased ('undercompensatory'-Archer & by insect attack or very high aphid infestations may
Tieszen, 1983). When such production results in the severely reduce seed production.
development of new potentially autonomous plants
(e.g. shoots in perennials) or overall larger plant size Some plant predators do not defoliate their host, but
the response has been described as 'herbivore in- extract food requirements from within the plant. The
creased fitness'. However there is (i) no clear consen- detailed work of D i o n (1971a, b), for instance, has
sus on the spectrum of plant regrowth patterns that clearly demonstrated that aphid infestations on lime
may occur in response to grazing and certainly such saplings cause at least a 10-fold reduction in the rate
responses are dependent on both defoliation history of total dry weight increase (from 7.7 down to 0.6 g
and environment; and (ii) whilst there has been week-'), even though average and total leaf area per
considerable debate that herbivory in principle could sapling remained unaltered. Yet, examination of the
increase plant fitness the available evidence suggests root systems of the infested plants revealed that no
that it does not (Belsky, 1986). What is clear is that growth had occurred below ground subsequent to
reductions in photosynthetic area, root volume and infestation; and the energetics of this interaction
stored reserves necessarily place a defoliated individ- suggest that, on average, 30 aphids per leaf during the
ual at a disadvantage in any competitive struggle-a growing season is sufficient to completely drain the
disadvantage that would obviously be intensified still annual net production of the tree. This average is
further if buds, the regenerative organs, were subject obviously an oversimplification, but the figure agrees
to predation as well. Competitive disadvantage will reasonably well with observed natural aphid densities
result in reduced plant size (see below) which in (Dixon, 1971b),implying that although they cause no
return will be reflected in fecundity (Rees & Crawley, immediate visual damage, aphids may substantially
1989). The low levels of defoliation that have been limit the growth of lime trees, and they may, in
observed in natural woodlands has often been as- consequence,affect survivorship.
sumed to have negligible effects on plant fecundity
because of plant compensatory growth processes. To Evidence from exclusion experiments point strongly
test this, Crawley (1985) examined matched pairs of to the indirect effects of herbivory leading to altered
oak (Quercus robur) trees over 4 years; in each pair one competitive abilities in plant communities. From a
tree was regularly sprayed with insecticide to kill all range of field studies Crawley (1989)concluded that in
defoliating and sap-sucking insects and the other was almost all cases the dominant plant species in a
sprayed with water. He showed that whilst tree natural community was changed after the deliberate
growth (girth)was unaffected by insect exclusion and exclusion of vertebrate herbivores accompaniedl by
unsprayed trees lost only 8-12% of their leaf area, detectable changes in relative plant abundance; but
sprayed trees consistently produced more seeds (2.5- much less so in studies of insect exclusion.
4.5-fold) than unsprayed ones.
However, very few critical experiments have been
Defoliation may thus have important repercussions conducted to examine the change in competitive
on seed production but the time and frequency at interactions amongst plants experiencing herbivory.
which it occurs is also important. Regular removal of In one such study Weiner (1993) deliberately manip-
ulated both herbivore (the snail, Helix aspersa) and
plant (Hypocharis radicata) density in an experimental

CHAPTER 5 : PREDATION 131

system. Predictably, mean plant size diminished with factor in plant population dynamics remains to be
increasing plant density in the absence of herbivory, unequivocally demonstrated.
indicating intraspecificcompetition in the plant popu-
lation. Moreover, mean plant size was further reduced 5.6 The effects of predation-rate on
by snail herbivory, the more so the higher the density predator fitness
of snails, but to the same proportional extent at each
plant density (i.e. there was no interaction between 5.6.1 Thresholds
herbivore and plant population density). Increasing The effects of predation-rate on predator fitness-like
herbivore density did, however, reduce the number of the effects of predation-rate on prey fitness-appear,
plants surviving to the end of the experiment, mortal- superficially, to be straightforward;and in the case of
ity being most marked at high herbivore and plant parasitoids they certainly are: every host successfully
densities. The most conspicuous effects were, how- attacked by a parasitoid represents an incremental
ever, seen in the size distribution of plants. At low increase in parasitoid fitness. In all other cases,
plant density, herbivory reduced the size of many however, there is an added complication. Predators,
individuals without influencing the larger plants, herbivores and parasites all require a certain quantity
causing an increase in the inequality of plant sizes in of 'prey' tissue for basic maintenance; and it is only
the population. But at high plant density, the main when their intake exceeds this threshold that in-
effect of snails was a reduction in surviving plant creases in predation-rate lead to measurably increased
density with relatively little effect on plant size varia- benefits to the predator. This is illustrated with typical
tion. This one example illustrates some of the subtle examples in Fig. 5.9a (predator growth-rate) and
effects of herbivory on plants competing in monocul- Fig. 5.9b (predator fecundity). The consequences of
tures. If intraplant competition is intense and asym- this threshold for the stability of predator-prey inter-
metric (Chapter 2) herbivory may reduce competition actions are fairly obvious. There is a tendency at low
and restrict (or decrease) the resulting size variability prey densities for predation-rate to fall below the
that asymmetric competition generates. However, if threshold, causing predator fitness to slump to zero.
herbivore attack is size-dependent, then it may am- The adverse effects of low prey density on predator
plify asymmetric competitive interactions amongst
plants surviving herbivory. (Weiner observed that Fig. 5.9 Prey thresholds for predators. (a) Growth in the
snails did not feed on large plants grown in isolation spider Linyphia triangularis (Turnbull, 1962).
reflecting their aversion to feeding in a low cover (b)Reproduction in the water flea Daphnia pulex var.
environment.) pulicaria (Richman, 1958). (After Hassell, 1978.) 1 calorie
(non-SI unit)= 4.186 joules.
At the present time it remains impossible to gener-
alize on the effects of herbivores on competitive
interactions in plant populations due to the lack of
detailed experimental study; empirical observations
on community changes (Harper, 1977)and theoretical
studies (Crawley& Pacala, 1991)strongly suggest that
changes in competitive ability may well be mediated
by herbivores but the precise processes remain hid-
den.

Overall, therefore, we can see that there is ample
evidence to suggest that the effects of herbivores on
both seeds and growing plants may be substantial.
The extent to which plant fitness is altered by herbi-
vores, and in turn herbivory, acts as a regulating

132 PART 2: INTERSPECIFIC INTERACTIONS

fitness are, therefore, exaggerated, and the interaction
generally is destabilized.

5.6.2 Food quality Fig. 5.10 (a)The quality of food measured as crude protein
available to (0)and eaten by (0)wildebeest in the Serengeti
There is another factor complicating the relationship during 1971. During the dry season, food quality fell below
between predation-rate and the fecundity and survi- the level for nitrogen balance (5-6% crude protein) despite
vorship of predators, however, which itself is of selection. (b) The fat content of the bone marrow of the live
greater importance, namely food quality. It is not the male population (0)and those found dead from natural
case that each item (or even each gram) of food
consumed by a predator or herbivore is equivalent. cases (a).Vertical lines, where present, are 95% confidence
The chemical composition of food, and its accessibility
via digestion to the predator, both have a considerable limits. (After Sinclair, 1975.)
bearing on the way in which food consumption affects
predators. This is particularly apparent amongst her- or four times the normal (AgriculturalResearch Coun-
bivores (see White, 1978; Lawton & McNeill, 1979). cil, 1965), it becomes obvious that shortage of high
quality food can have drastic effects on herbivores.
In particular, herbivores are greatly affected by the
nitrogen content of their food. One has only to A similar conclusioncan be drawn from the work of
consider the honeydew (excesscarbohydrate) excreted McNeill (in McNeill & Southwood, 1978).As Fig. 5.l1
by aphids, to realize that many herbivores must ingest shows, seasonal peaks in the densities of insects
vast quantities of plant tissue in order to consume feeding on the grass Hotcus mollis are related to peaks
sufficient amounts of amino acids. Moreover, there is in food quality (measured as soluble nitrogen) in the
good evidence that herbivore abundance can be lim- leaves and stems-the latter approximating phloem
ited by nitrogen content (i.e. food quality). Many flows to the flowers and developing seeds in the
herbivores only have access to low quality food, and middle of the summer. Once again, food quality
they have insufficient time and energy to digest
enough of this to provide them with protein for
maintenance, let alone growth and reproduction.This
was shown, for instance, by Sinclair (1975)who noted
the protein content of the food available to the
wildebeest in Serengeti (Tanzania) during 1971, and
compared this with the protein content of the food
they ate (Fig. 5.10a). He also monitored (Fig. 5.lob)
the fat reserves in the bone marrow of live males, and
of males that died of natural causes (these reserves
being the last to be utilized).

It is clear from Sinclair's results that, despite select-
ing nitrogen-rich plants and plant-parts, the wilde-
beests consumed food in the dry season which was
below the level necessary even for maintenance (5-6%
crude protein); and, to judge by the depleted fat
reserves of dead males, this was an important cause of
mortality. Moreover, when we consider that during
late-pregnancy and lactation (December-May in the
wildebeest) the food requirements of females are three

CHAPTER 5: PREDATION 133

Fig. 5.12 The functional response of tenth-instar damselfly
larvae to Daphnia of approximately constant size. (After

Thompson, 1975.)

Fig. 5.11 Seasonal changes in the mean numbers of insects tenth-instar larvae of the damselfly Ischnura elegans
on the grass Hobs mollis related to changes in food quality
in the leaves and stems (McNeill & Southwood, 1978). (Thompson, 1975). Figure 5.12 clearly shows that, as
(After Lawton & McNeill, 1979.)
prey density increases, the predation-rate responds

appears to be having an extremely important effect on less and less and approaches a plateau (approximately

predators. 16 Daphnia per 24-hour period). A similar result is
show" in Fig. 5.13 for a herbivore: slugs feeding on
Moreover, plant quality does not only affect herbi-
L0liurn Perenne (Hatto & Harper* 1969). Such func-
vores because of what plants lack (in terms of nutri-
tional of predators to changes in Prey
ents), but also because of the toxic or digestibility-
density were described first by Sobmon (1949), but
reducing compounds~that many plants
discussed more extensively by Holling (l959)~who
contain by way of protection. However, while this may
attributed the form taken by curves like the ones in
have an impofimt evolutionaly effecton herbivores,

causing them to specialize, it presumably has rela-

tively little effect on those herbivores that are specifi-

calky adapted to feed on the plants producing these

compounds (Lawton & McNeill, 1979).

5.7 The functional response of predators Fig. 5-13 The functional response of single slugs to changes
to prey availability
in the amounts of the grass Lolium perenne available to be
5.7,lThe 'type 2' response eaten. (Data from Hatto & Harper, 1969.)

We now turn, for our next component, to the way in
which the predation-rate of predators, herbivores or
parasitoids is influenced by prey availability, and we
begin with the simplest aspect of availability: prey

A is in
Fig"5-12,which depicts the numbers of D a ~ h n i a(of a
particular size) eaten during a 24-hour period by

134 PART 2: INTERSPECIFIC INTERACTIONS

Figs 5.12 and 5.13 (which he called 'type 2' responses) rizes some of the results obtained by Griffiths (1969)in
to the existenceof the predator's handling time. (This,as his work on the ichneumonid Pleolophus basizonus
we have seen,is the time the predator spends pursuing, parasitizing the cocoons of the European pine savvfly,
subduing and consuming each prey item it finds, and Neodiprion sertifer. Taking parasitoids of different ages,
then preparing itself for further search.)Holling argued GrifFiths plotted the number of ovipositions per para-
that as prey density increases, search becomes trivial, sitoid over a range of host densities, but he also
and handling takes up an increasing proportion of the calculated the actual maximum oviposition-rate by
predator's time. Thus, at high densities the predator presenting other parasitoids of the same ages with a
effectively spends all of its time handling prey, and the superabundant supply of host cocoons. It is clear from
predation-ratereaches a maximum, determinedby the Fig. 5.14 that the type 2 functional response curves
maximum number of handling times that can be fitted
into the total time available. did indeed approach their appropriate maxima. Yet,
while these maxima (of around 3.5 ovipositions day-')
This view of the type 2 functional response is suggest a handling time of around 7 hours, further
confirmed and illustrated in Fig. 5.14, which summa- direct observation indicated that oviposition takes, on
average, only 0.36 hours. The discrepancy is ac-
Fig. 5.14 The functional responses of the ichneumonid counted for, however, by the existence of a 'refractory
parasitoid Pleolophus basizonus to changes in the density of period' following oviposition during which there are
its host Neodiprion sertifer; arrows indicate maxima observed no eggs ready to be laid. 'Handling time', therefore,
in the presence of excess hosts. (a)Parasitoids on their third includes not only the time actually taken in oviposi-
day, and (b)parasitoids on their seventh day. (After tion, but also the time taken 'preparing' for the next
Griffiths, 1969.) oviposition. Similarly, the handling times suggested by
the plateaux in Figs 5.12 and 5.l3 will almost cer-
tainly include time devoted to activities,peculiar to the
darnselflies and slugs, other than the direct manipula-
tion of food items. It is in this (non-literal) sense that
handling time must be understood.

A further point to note from-Fig.5.14 is that, while
the plateau level (and thus the handling time) remains
approximately constant with increasing age, the rate
of approach to that plateau is much more gradual in
the younger parasitoids. In other words, it is apparent
that the younger parasitoids search less efficiently, or
attack at a slower rate. Thus, at low host densities
they oviposit less often than their older conspecifics;
but at high densities there is such. a ready supply of
hosts that even they are limited only by their handling
time. Hence, the form taken by a type 2 functional
response curve can be characterized simply in terns
of a handling time and a searching efficiency (or attack
rate); and Hassell (1978) discusses the methods by
which these parameters can be obtained from the
data. The value of estimating these parameters is
illustrated by the work of Thompson (1975) who fed
Daphnia of a variety of sizes to tenth-instar damselfly
larvae (Fig. 5.15). Figure 5.15b shows that the various

CHAPTER 5 : PREDATION 135

Fig. 5.16 The functional response of Daphnia magna to

different concentrationsof the yeast Saccharomyces
cerevisiae. (After Rigler, 1961.)

Fig. 5.15 (a) Functional responses of tenth-instar damselfly parasitoid-host interactions. Much less common, by
larvae to Daphnia of various sizes. Size A prey (0);B (0)C; comparison, is his 'type 1' response, an example of
(H); ID (0);E (A). Standard errors are fitted to the top line. which is shown in Fig. 5.16. This figure describes
(b)The attack-rates and handling times of these functional work carried out by Rigler (1961) on the feeding-rate
responses. (After Thompson, 1975.) of Daphnia magna with the yeast Saccharomyces cerevi-
siae as it 'prey'. Daphnia magna is a filter feeder,
functional responses in Fig. 5.l5a are the result of the extracting yeast cells at low density from a constant
ways in which both handling time and attack-rate volume of water washed over its filtering apparatus.
change with prey size. It appears, quite reasonably, Below 105 yeast cells ml-l the predation-rate is
that damselfly larvae need more time to handle larger directly proportional to the food concentration. But
prey, and that they are most efficient and effective at the Daphnia must also swallow (i.e. handle) their food.
catching Daphnia of size D, with attack-rate declining At low concentrationsthis does not interfere with the
rapidly as prey size increases. predation-rate, because it happens sufficiently quickly
to remove all the food accumulated by filtration.
5.7.2 The 'type 1' response Above 105cells ml-l however, the Daphnia are unable
to swallow all the food they filter. At all such concen-
A Holling 'type 2' functional response is commonly trations, therefore, they ingest food at a maximal rate,
observed in herbivore-plant, predator-prey and limited by their 'handling' time. The type 1response
is, therefore, an extreme form of the type 2 response in
which the handling time exerts its effect not gradually
but suddenly.

It is important to note, in both the type 1 and
particularly the type 2 response, that the rate of
predation on the prey declines as prey density in-
creases. This is, in other words, a case of inverse
density-dependence (section 2.6), in which large prey
populations suffer proportionately less mortality than
small. The consequent effect on the predator-prey
interaction is clearly destabilizing.

136 PART 2:\INTERSPECIFIC INTERACTIONS

5.7.3 Variation in handling time and searching Fig. 5.17 (a)The relationship between the time spent
efficiency: 'type 3' responses probing by Venturia canescens (as a percentage of total
observation time) and the density of its host larvae Plodia
It is instructive, at this point, to note explicitly the interpunctella. (b)The sigmoid functional responses of
various components of the handling time and search-
ing efficiency (following Holling, 1965, 1966). Han- Venturia canescens parasitizing Cadra larvae of second (a),
dling time is determined by:
1 the time spent pursuing and subduing an individual third (0)and fourth (A) instars (Takahashi,1968).(After
Hassell et al., 1977.) For further discussion, see text.
prey;
2 the time spent eating (or ovipositing in) each prey; constant, leading to the decrease in slope (region B)
and which is typical of a type 2 response. Overall, the
3 the time spent resting or cleaning or fulfilling any resulting functional response is S-shaped or sigmoitdal,
other essential function (like digestion) prior to the act and is, in Holling's (1959) terminology, 'type 3'.
of predation itself. Clearly it will occur whenever attack-rate increases or
handling time decreases with increasing prey density.
Searching efficiency (attack-rate)will depend on:
1 the maximum distance at which a predator can 5.7.4 Switching and 'type 3' responses
initiate an attack on a prey;
2 the proportion of these attacks that are successful; There are other (related) circumstances, however,
3 the speed of movement of the predator and prey
(and thus their rate of encounter); and
4 the 'interest' taken by a predator in obtaining prey
as opposed to fulfilling other essential activities.

We have already seen that at least some of these
components are likely to change with predator age
(Griffiths, 1969; see Fig. 5.14) and prey size (Thomp-
son, 1975; see Fig. 5.15);and the length of time since
the predator's last meal (its hunger) is also likely to
modify its response (i.e. its 'interest' in food, and thus
its attack-rate, will be altered). Of particular impor-
tance, however, is the way in which these components
vary with prey density, or, more generally, relative
and absolute prey availability.

An example which effectively shows attack-rate
varying with prey density is illustrated in Fig. 5.l7a
(Hassell et al., 1977):when host densities are low, the
parasitoid Venturia canescens spends a relatively large
proportion of its time in activities other than probing
for larvae (see component 4 above). The consequence
of this is shown in Fig. 5.17b (Takahashi, 1968). At
low host densities, there is an upward sweep in the
functional response curve (Fig. 5.17b, region A), be-
cause an increase in density elicits an increased
amount of probing and thus an increased effective rate
(Fig. 5.17a). Conversely, at higher host densities the
attack-rate (as well as the handling time) is relatively

CHAPTER 5: PREDATION 137

which will also lead to a type 3 functional response. action. The actual importance, however, depends on:
These are the cases of predator switching considered 1 the concavity of the curve in region A; and
in section 5.3-2. The resemblance between Fig. 5.l7b 2 the relevance of the prey densities in region A to a
and Fig. 5.4a and 5.5a is obvious. The major differ- particular field (or laboratory) situation.
ence is that in the latter case there were two types of
prey (rather than one), so that the numbers of prey 5.8 Aggregated effects
eaten varied with relative prey availability; we cannot
be sure that a type 3 functional response would have 5.8.1 Parasite-host distributions
resulted had the alternative prey been absent. Never-
theless, the effective result of predator switching is a It was established in section 5.3.2 that polyphagous
type 3 functional response: the numbres of prey predators, by exhibiting preferences, distribute their
consumed varies with prey density in a sigmoidal ill-effects unevenly between prey species. We turn
fashion. now to the distribution of these ill-effects within a
single species of prey. Consider, to begin with the
Type 3 functional responses, then, resulting from distributions of parasites on their hosts shown in
switching or from changes in handling time or attack- Fig. 5.18. In both examples, the observed patterns
ing efficiency, can occur in predators, parasitoids or have been compared with the patterns that would
herbivores; althogh it remains to be seen how com- have arisen if the parasites had been distributed at
mon and widespread they are. Whenever and for random. Random distributions are the simplest imag-
whatever reason they occur, however, the effect on inable arrangements in that they occur when all hosts
the stability of the interaction will be essentially the and all parasites are equivalent and independent. (An
same. Throughout region A (Fig. 5.17btthe upward- 'even' distribution would only occur if the parasites
sweeping part of the curve-increases in prey density positively avoided one another.) Yet the observed
lead to increases in the predation-pressureon the prey: patterns are far from random. Instead they are dis-
the process is density-dependent. It can, therefore, have tinctly 'clumped'. There are more hosts than expected
a potentially important stabilizing effect on the inter-

Fig. 5.18 The aggregated
distributions of parasites on hosts

(points and curve), compared to a

random distribution (histograms).
(a)Ticks lxodes trianguliceps on the

wood mouse Apodemus sylvaticus

(data from Randolph, 1975).
(b)Tapeworms Caryophyllaeus
laticeps in the bream Abramis brama.
(Data from Anderson, 1974.)

138 PART 2: INTERSPECIFIC INTERACTIONS

supporting large numbers of parasites, but also more only extend its ovipositor a certain distance into the
than expected with no parasites at all. In effect, there flour medium. A proportion of the caterpillars-those
is a partial refuge for the host: the pattern of distribu- lying deeply enough to be beyond the ovipositor's
tion ensures that an 'unexpectedly' large number of reach-are therefore protected in an effective refuge
hosts escape parasitization. (Hassell, 1978).

Such patterns are extremely common amongst In the second type of total refuge,by contrast, afixed
parasites, and, although proposals for underlying number of prey are protected. This is illustrated by
mechanisms are usually speculative(Anderson & May, Connell's (1970) work on the barnacle, Balanus glan-
1978),the effects these patterns have on host-parasite dula, in which he found that adults are restricted to
dynamics are fairly obvious. By leaving a large num- the upper zones of the shore, while juveniles are
ber of hosts unparasitized, even at high levels of distributed throughout a much broader region
infection, the distributions ensure that the host popu- (Fig. 5.19). This occurs because in the upper zones
lation~are buffered from the most drastic effects of the there are two exposures to the air at low tide alrnost
parasites. This, in turn, tends to ensure that the every day, which means that the whelks that consume
parasite population has a population of hosts to live the juvenile barnacles in the lower regions-even Thais
on. The basic effect of this partial refuge, then, is to emarginata (Fig. 5.19thave only two short, high-tide
stabilize the interaction. periods to feed (as opposed to a single long one at
lower levels). And this is never -long enough to find
5.8.2 Refuges and eat an adult Balanus glandula in the upper regions.
Thus, the B. glandula individuals which the upper
Partial refuges, as we shall see, are of considerable zones can support are protected from predation,
importance in a wide variety of predator-prey inter- irrespectiveof Thais numbers. There is a fixed-number
actions. Yet there are, in some cases, not partial but refuge.
total refuges, and these can be of two types.
From what has already been noted about partial
The first is a refuge for afixedproportion of prey. The refuges, it is clear that total refuges will tend to
parasitoid Venturia canescens, for instance, when at- stabilize predator-prey interactions; but it should be
tacking caterpillars of flour moths (Ephestia spp.) can equally clear that the (density-independent) fixed-

Fig. 5.19 Vertical distributions of barnacle
prey (Balanus glandula) and whelk
predators (Thais emarginata, Th. canaliculata
and Th. lamellosa) in relation to typical tidal
fluctuations of large and small amplitude.
(After Connell, 1970.) For further
discussion, see text.

CHAPTER 5: PREDATION 139

proportion refuge will be much less potent than the of aphids. Moreover, the uninfested leaves of the

(density-dependent)fixed-number refuge, since in the 'one-leaf colonized plants' were healthy at the end of

latter case the proportion protected increases as prey the experiment, while the 'four-leaf colonized plants'

density decreases. were virtually dead. The aphids' behaviour, therefore,

does more than increase their own productivity: it also

5.8.3 Partial refuges: aggregative responses provides a partial refuge for the cabbages. Thus, as a
by-product of the aphids' behaviour, many more

In the living world as a whole, total refuges are leaves escape destruction than would do so if the

probably rather rare. Considerably more common, aphids were randomly distributed.The cabbage popu-

however, is a tendency for prey to be affected by their lation is partially buffered from the aphids' ill-effects,

herbivores, parasitoids or predators in much the same and the interaction is (relatively)stabilized.

way as hosts are affected by their parasites (see Of course, the aggregated cabbage aphids are them-

Fig. 5.18): the ill-effects are aggregated so that the selves the prey of other animals and, indeed, their

prey have a partial refuge. We can illustrate this, distribution is typical of the 'patchiness' of prey

initially, in a herbivore-the cabbage aphid, Brevico- animals generally. The responses of a predator and a

gne brassicae. This species, an important pest of parasitoid to their distribution are shown in Fig. 5.20.

cabbage and its relatives, has a marked tendency to In both cases (over at least part of the density range)

form aggregates at two separate levels: nymphs, when the time spent on a leaf increases with the density of

isolated experimentally on the surface of a single leaf, aphids on that leaf. Thus, if we make the reasonable

quickly form large groups; while populations on a assumption that an increased searching time leads to

single plant tend to be restricted to particular leaves an increased proportion of the leaf being searched,

(Way & Cammell, 1970).The effects are illustrated by which leads to an increase in the proportion of aphids

an experiment (Way& Cammell, 1970)in which eight attacked, then clearly the aphids at the lower densities

small cabbage plants were each artificially infested by have a smaller probability of being attacked. Once

16 ;aphidnymphs. Four of the eight plants had all 16 again, therefore,there is a partial refuge, but this time

nymphs on a single leaf (the normal situation), while it is the aphids themselves that are protected: aphids

the other four had four nymphs on each of the four in low-density aggregates are most likely to be ig-

leaves. Aphid-free leaves were protected from cross- nored. (It is important to note, however, that such

infestation, and colonization by 'outside' aphids was effects apply only to those parts of the density axis in

prevented. Productivity was measured by the num- which 'time spent' increases. The importance of the

bers and weights of adult aphids subsequently pro- effects depends on the relevance of these densities in

duced. Although the differences are small (Table 5.5) nature.)

it is clear that the normal, aggregated situation, ~ t a h Most herbivores, predators and parasitoids appear

single leaf colonized, is the more productive in terms capable of exhibiting an 'aggregative response', con-

centrating their ill-effects on a particular portion or

Table 5.5 Weights of winged adult cabbage aphids patch of their prey population; and by providing a
produced on small cabbage plants. (From Way & Cammell, partial refuge for the prey this tends to stabilize the

1970.) interaction. But it must be realized that in each case

this is essentially a by-product of the response (albeit

an extremely important one): there is no question of

evolution favouring stable populations. This funda-

mental point is illustrated in Table 5.5 and Fig. 5.20.

The predator and the parasitoid concentrate on par-

ticular patches (leaves) and make their individual

search more profitable; the aphids aggregate in

140 PART 2: INTERSPECIFIC INTERACTIONS

Fig. 5.20 Aggregative responses by
a parasitoid and a predator of the

cabbage aphid Brevicoryne brassicae.
(a)Searching time of the braconid
Diaeretielh rapae at different densites
of its host (Akinlosotu, 1973).(b) As
(a)but using coccinellid larvae
CoccineIIa septempunctata.(After
Hassell, 1978.)

patches supporting other aphids and increase their live at an apparently stable low level of abundance,
individual productivity. In all cases, the aggregative which can be explained in terms of aggregation by
behaviour is advantageous to the consumer. considering some of the results obtained by Monro
(1967) for C. cactorum and Opuntia inermis (Table 5.6).
These data also provide further support for the The distribution of Cactoblastis cactorum egg-sticks
assertion that predators choose profitable prey (in this (each containing 70-90 eggs) on prickly pears is
case, profitable patches). Experimental illustration of distinctly clumped,and the very limited mobility of the
this is provided for a two-patch situation by the work larvae means that the 'unexpectedly' high number of
of Krebs et al. (1978).Great tits were offered rewards
(food) at two different perches (patches), but the
profitabilities of the two perches were unequal. After a
period of learning, the birds concentrated (almost
exclusively)on the more profitable patch (Fig. 5.21).

Of course, by blandly asserting that consumers con- Fig. 5.21 When great tits are faced with a choice of two!
centrate on profitable patches we beg the question of
what actually constitutes a patch. In the examples perches to obtain food,they go for the one with the higher
shown in Fig. 5.20, for instance, the aphids' predators reward rate. The ''dinate shows the percentage of
and parasitoids appear to treat each leaf as a patch; responses on one of the perches, and the abscissa is the
while in the further examples in Fig. 5.22 (which once percentage of rewards available (Krebs et al., 1978). (After
again show concentration on profitable 'patches') the Krebs, 1978.)(Geometric means and 95010 confidence limits
term 'patch' applies to a whole plant in the case of the are shown.)
braconid Apanteles glomeratus (Fig. 5.22a), and simply
to a unit area in the case of the ichneumonid
Dial -romus pulchellus (Fig. 5.22b). Similarly, while the
cabbage aphid appears to treat a leaf (or part of a leaf)
as a patch, the appropriate scale for many herbivores

is the whole plant. One example of this is the moth

Cactob'astis cactorum' which has been used in
lia to control the prickly pear cacti O ~ u n t i ainermis and
0.stricta. Cactoblastis cactorurn and its food plants now