Overlap in resource use, and interspecific competition

Oecologia (Berl.) 17, 245--256 (1974) �9 by Springer-Verlag 1974

Overlap in Resource Use, and Interspccific Competition

Peter F. Sale

School of Biological Sciences, University of Sydney, Sydney, N.S.W. 2006

Received May 20, 1974

Summary. When several species co-exist, the amount by which they overlap in their use of resources is a measure of their similarity to one another. As such, resource overlap does not measure the amount of competition among them. When the resources are not limiting to population growth, patterns of resource use may overlap to any degree. However, when the species are frequently in competition for their resources, natural selection will favor the separation of their requirements, and the amount of resource overlap will be reduced.

This paper presents a technique which permits comparison of the amount of resource overlap observed in a given case with that expected for a group of similar species co-existing in the absence of competitive interactions. From this comparison can be evaluated the likelihood of competitive processes being important in the situation under study.

Introduction

A species generally makes use of only some of the resources available to it. For example, a particular species may occur in only one of the several habitats available in its geographical range, or it may eat only a small range of sizes of the arthropods available in its habitat. Measurements of "niche breadth" along dimensions such as food size or habitat type tell us about the degree of specialisa- tion of a species.

When two species co-exist, they may or may not be similar in their ways of using resources. Measurements of "niche overlap" (----resource overlap) along dimensions like food size and habitat type will tell us the degree to which two co-existing species are similar. A number of techniques for measuring niche breadth and resource overlap have been developed (Horn, 1966; Levins , 1968; Colwell and Futuyma, 1971).

In recent years there has been considerable interest among ecologists in the measurement of resource overlap among co-existing species. The measurements obtained can be, and have been of real interest. They have been used in attempts to answer such questions as whether or not communities are "fil led" with species (Cody, 1968), whether islands contain " e m p t y " niches (Diamond, 1970), and whether tropical species are more similar to one another than temperate ones (Terborgh and Diamond, 1970). Unfortunately they have also been used in the mistaken belief that the degree of overlap between a pair of species can be used directly to assess the degree to which they are competing with one another.

This belief is based on the assumption that co-existing populations of similar species are generally limited by competition with each other for available resour- ces. Thus, the more similar two co-existing species are, the more they compete with each other. While few authors state the assumption explicitly (but see

246 P.F. Sale

Reynoldson and Davies, 1970; Hansen and Ueckert, 1970), many authors make it implicitly, assuming a direct correlation between degree of overlap and inten- sity of competition (for example, Edington and Edington, 1972; Robins, 1972; Snow and Snow, 1972; Williamson, 1971).

As Colwell and Futuyma (1971) have pointed out, there is need for consid- erable caution when interpreting measured values of resource overlap. They discus- sed the need to consider overlap in use of "equally dist inct" resources if the overlaps between different pairs of species are to be compared. And they explicitly pointed out tha t a direct equivalence of percentage overlap and amount of compe- tition does not necessarily exist. In their words, " . . . simply demonstrating an overlap in resource use by two species in nature can be evidence either for or against the existence of competition between them" . They suggested that by experimentally removing competitors, by exploiting natural situations in which competitors are absent, or by artificially increasing the supply of resources, one could measure a virtual overlap (overlap in the absence of competition) between two species. The species would be demonstrated to be competing only if the overlap observed was less than the virtual overlap measured between them. In effect, Colwell and Futuyma (1971) were proposing the detection of niche shifts-- tha t is, changes in a species' pat tern of resource use when exposed to a compe- t i t o r - a s evidence of competitive interactions.

Unfortunately, experiments of the type proposed are rarely technically feasible, while measurements of resource overlap are relatively easy to make. I t seems desirable that there be some way of using such data for formulating statements about competition when experimental methods are not possible. The approach suggested here hinges on the expected effects (including niche shifts) of competitive interactions on patterns of resource use by a group of co-existing species. This approach cannot conclusively demonstrate the present occurrence or non-occurrence of interspecific competition within a group of species. However, from data on resource use alone, it can assess whether or not interspecifie com- petition is likely to have been an important process during the course of their co- existance. The approach is illustrated with examples from the literature.

Theory Since interest here is primarily in the relation between resource overlap and

competition, consideration will be restricted to only those sorts of resources which might be potentially limiting to population growth in the species considered. Resources such as oxygen and light are rarely likely to be in short supply and influence the population processes of animals. Species can overlap completely along niche dimensions such as these with no competitive interactions occurring. On the other hand resources such as foods or habitat types could be in short supply, and competition for them could occur.

As an example, consider two species overlapping in their food requirements. Two situations may exist:

1) Food may be superabundant, or the production of food may be rapid rela- tive to the rate at which it is consumed. The populations of the two species are determined by factors other than a shortage of food, and no competition for

Resource Overlap and Competition 247

food exists. In this situation the observed amount of overlap is simply a measure of the similarity in dietary habits of the two species. They may be very similar and show a great degree of overlap, or they may be quite different overlapping only slightly. But, in either case, no competition for food is occurring.

2) Food is in short supply in the environment and the sizes of the popula- tions of the two species are limited by the amount of food available for them. In this situation competition exists, and the degree of competition and the amount of overlap are directly related.

Clearly, if one is to use measurements of resource overlap to make decisions concerning competition it is necessary to distinguish between l) and 2) above. A partial approach to this can be made in the following way. Consider a com- munity among the species of which competition for resources is a frequent event. During the development of such a community, the species successful in invading might be expected to have quite different patterns of resource use from one another. Furthermore, the occurrence of competition will mean that all species ultimately present can be expected to have been frequently subject to selection pressure for reducing still further the degree to which their resource requirements overlap. These assumptions are central to the concept of species packing (Schoe- her, 1965; MaeArthur and Levins, 1967; MacArthur, 1969, 1970).

A species carl reduce the overlap of its requirements with those of other species by narrowing its requirements--becoming more specialised, by actually shifting the range of resources used to other parts of the array available, or by both of these methods. Such adjustments of requirements might begin occurring at the time the species invade the developing community. Diamond (1970) provides several examples of such narrowing and shifting of the requirements of bird species which have colonised islands of the New Guinea archipelago. Jermsen (1973) provides a very clear example of a niche shift by a population of anoles in the presence of a close competitor.

Both the initial selection of successful invaders, and the subsequent reduc- tions in the overlap among their patterns of resource use will result in a community in which the requirements of different species are over-dispersed across the array of resources available.

By contrast, consider a community in which interspecifie competition for resources is a rare or non-existant event. The success of a species in invading this community would not be influenced by how its pat tern of resource use compared with those of the species already present. Nor would the species already present interact in ways necessarily tending to reduce the overlap in resource use among them. In this community the requirements of different species would be distributed randomly across the resources available. Because of this, overlap in resource use would be expected to be greater than if competition had influenced species packing during the community's development.

This difference in the two types of community permits asking the following question. For a particular community under investigation, is the observed overlap in resource use among the species present less than one might have expected in a community containing a similar group of species and resources, but without the ordering influence of competitive interactions. The observations of resource

248 P.F. Sale

use by the species of the community can be manipulated, as demonstrated below, so as to obtain an objective estimate of the overlap expected in an analogous community in which competition is known not to have played a role. The actual overlap can then be compared with this estimate of competition-free overlap. Data on resource overlap can be considered to provide evidence of the existance of competitive interactions only in those cases where the actual overlap is less than tha t estimated for a competition-free situation.

Methods

Measurement of Observed Overlap

Because of the nature of subsequent manipulations it is essential tha t actual overlaps be measured across an array of equally abundant resources. The tech- nique for obtaining such an array is based on an approach very similar to tha t of Colwell and Fu tuyma (1971) for obtaining one of equally distinct resources.

Let s species make use of an array of r resources. A resource matr ix can be constructed with s rows and r columns in which each value, Pij, is the proportion of all resources used by the ith species which is made up of the jth resource. Sums across rows of this matrix, Pi" = 1.0 in each case.

Values of Pii are readily obtained from original field data. In addition to observation~ of resource use, measurements are made of the relative abundance of each of the r resources. A i is the proportion of all resources available to the s species as the ~-th resource. Note tha t in cases where these measurements have not been made, it is essential tha t estimates of them be obtained. Methods of determining Aj, the estimated proportion of resources available as the ]th resource will be considered later.

An expanded array is now generated in which the s species make use of m equally available resources (m ~ r). In generating this array, each actual resource, j, is split into Afro parts, and each of these parts becomes one of the m equally available resources. The proportion of species i 's resource use made up of resource J, Pij', is similarly split into Aj.m equal parts apportioned one to each of the equally available resources derived from j. Thus in the expanded array, the summed resource use by the i th species, pi. --~ 1.0 as before.

Resource overlap between each pair of species can now be calculated using any appropriate formula. Here (following Colwell and Futuyma, 1971) the formula used is:

C i h = l " O - - i / 2 ~ (1) k = l

where Cia is the resource overlap between species i and species h, and Pi~ and Phk are, respectively, the proportions of resources used by species i and species h which are derived from the k th equally available resource. Oih varies from 0.0 when the resources used by species i and h are completely different, to 1.0 when their pat tenls of resource use are identical. The values of Cih obtained from the expanded array are identical to what would be obtained using the data in the original s by r array. The mean of the Cih is the average observed resource overlap, Co, for the community of s species.

Resource Overlap and Competition 249

Est imate of Competition-Free Overlap

A set of synthetic communities is now generated. Each contains m equally available resources, and s species, and in each synthetic community, species make use of resources without regard to the resource use of other species. The procedure is as follows.

Each of the s species in the original community has a set of m values, Pik, of proportional resource use which range from some maximum (Pik ~ 1.0) down to some minimum w~lue (Pi~ ~ 0.0). The particular values of Pi,r belonging to the ith species determine tha t species' niche breadth along this particular resource dimension regardless of which of the m equally available resources is associated with each value of Pik. The particular sequence of Pik belonging to the ith species determines tha t species' niche shape in this resource dimension.

In generating a synthetic community, the particular pa t tern of resource use, both niche breadth and shape, is retained, however the position of this pa t tern of resource use on the array of resources available is randomly chosen. This is done for the ith SP ecies by randomly selecting resource j, one of the m equally available resources. The value, Pil, of resource use is then re-allocated to resource j, and succeeding valaes, Pi~ .. . Plm are allocated to each of the following resources, ] ~- 1, ?'-~ 2 . . . . m, 1 . . . . ?" - - 1, so tha t all m again have a value of p ~ associated with them. This process is repeated for each of the ~ species so tha t a new, randomised s • m array is generated. In this new array row totals have not been changed (Pi' = 1.0), but column totals, P.k" have been permit ted to vary.

Resource overlaps for each pair of species, and the average resource overlap among all pairs of species in this synthetic community, Cs, are calculated as before. Following generation and analysis of all synthetic communities, the mean and variance of the C 8 are calculated. This mean, Cop is the estimated average competition-free overlap among the s species examined. For the sets of data which I have analysed, I have found the generation of 100 or fewer synthetic communities produces values of Cc! with quite small associated standard errors.

I t is essential, as discussed later, tha t the above procedure be carried out on an array of equally available resources. When the abundances of the actual resources have not been measured, they must be estimated. This can be done on the assumption tha t they are used in approximately the proportions in which they occur. Thus the estimated relative abundance of the ~th resource is:

8

Aj -= X p i j .A i (2) i=t

where Ai is the relative abundance of the i th species a M Pi] is as before. I f values of A i are also estimated, or assumed equal, the method becomes of only slight value.

Assessing Likelihood of Competition

The observed average resource overlap, Co, is compared to Q! , the average competition-free overlap, using the following test for comparing a single measure- ment with the mean of a sample (Simpson et al., 1960, p. 183) :

250 P.F. Sale

y 2V t = [C~Z-0~ ~V7 1 (3)

d

with (N-- I ) degrees of freedom, where N is the number of synthetic communi- ties generated in estimating Cel, and d is the standard deviation associated with Col. The test is one-tailed since interest is limited to whether Co is less than Co/. When Co is not less than Col there is no evidence available in the observed patterns of resource use that competitive intereactions have occurred among the species.

Discussion of the Method The major assumption on which the above method is based is that species in

competition for resources will be under selective pressure to maintain, and to increase still further those differences existing in their patterns of resource utili- sation. This assumption is well based in competition theory although the excep- tion is postulated that two very similar species will tend to converge still further with the ultimate extinction of one of them (MacArthur and Levins, 1967; MacArthur, 1970, 1972).

The only evidence of competition available from measurements of resource overlap is that such an ordering of patterns of resource use has occurred. This is an important restraint on the use of observations of overlap in identifying the presence of competitive interactions. In the present case the possibility of such an ordering having happened is assessed by comparing the observed average- overlap in resource use with that estimated for an analogous set of species and resources in a competition-free situation. Here the procedure for making that estimate must be justified, and possible limitations in applicability of the method must be considered.

When several species make use of an array of resources, they will overlap with one another to an amount dependant upon 1) the number of resoures available to them, 2 ) t he degree to which each species is currently specialised in resource use (its present niche breadth), 3) the degree to which such specialisation constrains each species to use a specific part of the total range of resources, and 4) the degree to which competition between the species has caused them to adjust their resource use to different portions of the total range. I t is recognised that this adjustment due to competition may have involved a reduction in niche breadth- - tha t is, an increase in specialisation from an earlier condition.

The technique described for estimating competition-free overlap determines the mean overlap among all pairs of species in each of a family of syathesised com- munities. These communities have in common with the actual community the number of resources available, the number of species using them, and the current niche breadths and shapes of each of these species. They differ from the actual community in that, in each synthetic community, each species' use of resources is known to be unaffected by that of the other species.

In generating these synthetic communities, while competition between species is not permitted, the degree to which specialists are unable to use certain of the range of resources (3 above), even in the absence of putative competitors is also

Resource Overlap and Competition 251

ignored. All s species are considered potentially capable of using any of the m resources. This is done on the expectation that if the family of communities actually existed, species in different ones could have evolved the same degrees of speeialisation on any part of the range of resources available. Note also that in these synthetic communities, niche breadths of the species present are those currently exhibited by the species of the actual community. There is no basis in the original observations for deciding whether or not the species in the actual community are now more specialised than at some earlier time. To assume that they are is invalid.

The synthetic communities contain species with similar characteristics to those currently in the actual community. Thus the mean amount of overlap observed among the species in them will differ from that in the actual community only when species in the actual community have resource requirements non-randomly distributed across the resource array. An overdispersal of requirements is likely when competitive processes have been important during the community's develop- ment. Under-dispersal could result from convergence of species onto abundant resources.

All calculations of resource overlap are made from an array of many (approxi- mately 100) equally available resources derived by splitting up the differentially abundant resources actually used. This procedure has two important results. I t minimises any bias due to the assumption mentioned above, of equal capability in use of all resources, because species will seldom be highly specialised in use of the expanded array of resources. Secondly it makes the process of shifting of patterns of resource requirements along the resource array more realistic. Specific requirements of a species are shifted from one to another equally available resource, rather than from abundant, to a rare, or from a rare to an abundant resource.

Colwell and Futuyma (1971) emphasise the importance of measuring resource overlap over a range of equally distinct resources if values obtained are to be compared with those from other studies. I t is important to note that the develop- ment of equally available resources carried out here is not the same thing. In fact the values of overlap obtained for the observed distribution of species' require- ments are identical whether measured over the observed, or the equally available categories of resource. Since, for the present purposes, these observed values are only going to be compared with values synthesised from the same data, this lack of equal distinctiveness seems unlikely to matter.

The method used by Colwell and Futuyma (1971) for obtaining equally distinct resources is not applicable in the present case. The overlap in use of resources by all pairs of species is to be considered, so use by some species cannot also be used to measure degree of distinctiveness. No other objective measure of relative distinctiveness seems available, lkTeedless to say an at tempt should be made to divide categories of resource as evenly as possible when collecting field data.

How should results be interpreted, and what restrictions should be applied to the use of the method ? When Co, the mean resource overlap observed among all pairs of species, is significantly less than Ocl, that estimated for a similar set of species in a competition-free situation, it is an indication that competitive interactions during the course of their coexistance have caused an ordering of

17 Oecologia (Berl.), u 17

252 P.F. SMe

resource requirements of the species present. Their patterns of resource use differ more than one might have expected them to in the absence of competition. Note that it may still be possible, particularly if the species are taxonomically widely separated, that they have always been very different in their pattern of resource use. That is, that they have each been constrained by their particular specialisa- tions and have never come into competition for resources.

When C o is found not to be less than Col a stronger conclusion can be made- - t ha t there is no evidence from patterns of resource use that competition has been important among the species examined.

Clearly, use of this method should be restricted to the study of groups of taxonomically close species, which are believed to have been co-existing for some time. Species which have come together very recently through range extension may not have had time to diverge, even if competition between them is severe.

Application to Examples Use of the technique is demonstrated using two examples from the literature.

MacArthur (1958) examined the patterns of resource use among five species of warbler. He considered that a major category of resource was the foraging sites used, and he presented data showing the proportion of foraging time spent by each species in each of 16 foraging zones. He presented no data on the relative availability of these zones, nor more particularly of food items within zones, but he did provide data on the relative abundance of four of the species of warbler on his study site (site E of MaeArthur's Fig. 10).

I considered only the four species for which relative abundance was known, and used their abundances to estimate the relative abundances of the 16 foraging site resources. Using MacArthur's (1958) data on proportion of sightings in each zone, I calculated overlaps observed between each pair of species. These observed overlaps, and abundances of species and resources are given in Table 1. Note that the calculated relative abundances of the 16 foraging site resources bear little rela- tion to the relative amount of space in each of those zones of a tree. Insect food is evidently not randomly distributed throughout the trees.

The mean overlap observed among the four species was Co = 0.654. The esti- mate of competition-free overlap, based on generation of 100 communities, was C~i=0 .678 i0 .0143 S.D., and Co is significantly less than Col (P <0 .0 5 , t test). That is, there is evidence in the patterns of use of these resources alone suggesting that interspecific competition has been an important process during the co-exis- tance of this group of warblers. The birds are partitioning the food resources available to them.

A quite different result was obtained when I applied the technique to the data of Ueckert and Hansen (1972). They provided extensive information on the dietary overlap of 14 species of grasshopper co-existing in Colorado. They had measured the abundance (as above-ground biomass) of the major food plants, and had determined the proportion each made of the contents from crops of each species of grasshopper.

For convenience I reduced the 49 categories of food plant recognhsed by the authors to 20 categories. This was achieved by pooling those species of plant

l~esource Overlap and Competition 253

Table 1. Analysis of overlap in foraging sites used by four species of warbler. Data are derived from MaeArthur (1958)

a) Proportional abundances of birds

Species Proportion of population

Myrtle warbler 0.1559 Black-throated green warbler 0.4498 Blaekburnian warbler 0.0870 Bay breasted warbler 9.3074

b) Estimated proportional abundances of foraging sites

Site Proportion

Zone 1, T 0.1203 M 0.0519 B 0.0059

Zone 2, T 0.1349 M 0.0904 B 0.0288

Zone 3, T 0.1471 M 0.1327 B 0.0662

Zone 4, T 0.0544 M 0.0474 B 0.0550

Zone 5, T 0.0103 l~I 0.0052 B 0.0294

Zone 6 0.0201

e) Observed overlap in use of foraging sites, Cih

Blaekthroated Blackburnian Bay breasted Green

Myrtle 0.661 0.567 0.670 Blaekthroated Green 0.670 0.732 Blackburnian 0.626

which both a) represented no more than 4 % of the diet of any one grasshopper, and b) occurred in measurable amounts in the diets of fewer than 4 grasshoppers. Table 2 shows the relative abundances of the food categories used in m y analysis of their data. The est imated average competition-free d ie tary overlap among the fourteen species was Co! = 0.538 • 0.0057 S.D., and the average observed overlap was Co----0.615, clearly no t less t han Col. I n fact, formula (2) can be used to demonst ra te t ha t Co is significantly greater than Col ( P ~ 0 . 0 0 1 , t test). My analysis therefore suggests t ha t ra ther than competi t ion for food having led to an overdispersal of the separate species' requirements across the a r ray of resources

17"

254 P.F. Sale

Table 2. Relative abundances, as above-ground biomass, of 20 food resources for grasshoppers. Data derived from Ueekert and Hansen (1972)

Resource Proportion

Grasses and grasslike Agropygon smithii a 0.0160 Andropogon hallii 0.0055 Bouteloua gracilis a 0.4180 Calamovil/a longi/olla a 0.1850 Carex heliophila 0.0055 Festuca octo/lora a 0.0040 Sporobolus cryptandrus 0.0055 Stipa comata a 0.1550 All other grasses 0.0055

All grasses a 0.8000

Forbs Amaranthus retroflexus 0.0086 Ambrosia psilostachya 0.0086 Artemisia /ili/olia a 0.1120 Crisium undulatum 0.0086 Evolvulus nuttallianus 0.0086 Psoralea tenui/olia 0.0086 Tradescantia occidentalis 0.0086 All other forbs 0.0086

All forbs a 0.1724

Other Fungus Moss Arthropod parts, hair

All other

a Category for which data on abundance available in values estimated.

0.0093 0.0093 0.0093

0.0279

Ueckert and Hansen (1972). Other

available, there has probably been convergence of food habits. Thus data on patterns of resource use do not indicate that the co-existance of these 14 species of grasshopper has been influenced by interspecific competition for foods. This conclusion is in contradiction to the authors' interpretation of some of their results.

On Detecting Competition The technique outlined, and demonstrated above, and the experimental

procedure suggested by Colwell and Futuyma (1971) are both quite limited. Both detect separation in patterns of resource use. The procedure of Colwell and Futuyma identifies this as due to niche shifts brought about by competition for those resources. The present technique, being non-experimental, cannot discrimi- nate separation brought about in this way from separation due to different evolutionary backgrounds, but this seems unimportant if taxonomically close groups are being considered.

Resource Overlap and Competition 255

Bu t separat ion of different species' resource requirements is not competition, only a possible, and probable result of its action over time. I t seems unlikely tha t the degree of such separat ion can be correlated with the intensi ty of compe- tition. Nor can its existance be used as evidence for the present existence of competi t ive interactions.

The chief value of the present technique is tha t by a simple manipulat ion of da ta on resource use alone, one can determine the likelihood of competi t ive processes being impor tan t in the communi ty studied. When Co is less than Co! compet i t ion is likely to have been impor tan t th roughout the period of the com- mun i ty ' s existance, and is presumably still important . When Co is no t less than Cr competi t ive interactions are very unlikely to have influenced tha t community .

However, this is by no means a final answer. For example, consider the case of the grasshoppers examined above (Ueckert and Hansen, 1972). While da ta on resource use indicate t h a t competi t ion for food is unlikely to have been an impor tan t process in this community , it remains possible t ha t the grasshoppers are limited by competi t ion for other kinds of resources, or even by a competi t ion for food which has not led to a part i t ioning of the food resources. Whenever it is possible the direct measurement of changes in populat ion size following experimen- tal changes in the amoun t of resources, or in the numbers of presumed compe- ti tors, is the preferred method for assessing competi t ion.

Acknowledgements. I wish to thank the ecology group at the University of Sydney, and in particular, Dr. G. J. Caughley, Dr. A. C. Hodson, Dr. A. W. Meats, and Dr. A. J. Underwood for their helpful comments on the ideas and on the manuscript. Mr. R. Dybdahl assisted in the development of the technique. Comments of two referees on an earlier version of the manuscript assisted me greatly. I enjoyed support from a Crown of Thorns grant, and a grant from the Australian Research Grants Committee while developing the ideas in this paper.

References

Cody, M. L.: On the methods of resource division in grassland bird communities. Amer. Natur. 102, 107-147 (1968)

Colwell, R. K., Futnyma, D. J. : On the measurement of niche breadth and overlap. Ecology 52, 567-576 (1971)

Diamond, J. M. : Ecological consequences of island colonisation by Southwest Pacific birds. I. Types of niche shifts. Proc. nat. Acad. Sci. (Wash.) 67, 529-536 (1970)

Edington, J.M., Edingten, M. A.: Spatial patterns and habitat patterns in the breeding birds of an upland wood. J. anita. Ecol. 41, 331-359 (1972)

Hansen, R. M., Ueckert, D. N. : Dietary similarity of some primary consumers. Ecology 51, 640-648 (1970)

Horn, H. S.: Measurement of overlap in comparative ecological studies. Amer. Natur. 106, 419-424 (1966)

Jenssen, T. A. : Shift in the structural habitat of Anolis opalinus due to congeneric compe- tition. Ecology 54, 863-869 (1973)

Levins, R.: Evolution in changing environments. Princeton, N. J.: Princeton Univ. Press 1968

MacArthur, R. It. : Population ecology of some warblers of northeastern coniferous forests. Ecology 39, 599-619 (1958)

MacArthur, R. H. : Species packing, and what interspecies competition minimizes. Proc. nat. Acad. Sei. (Wash.) 64, 1369-1371 (1969)

MaeArthur, R. H. : Species packing and competitive equilibrium for many species. Theoret. Popular. Biol. 1, 1-11 (1970)

256 P .F . Sale

MacArthur, R. H.: Geographical ecology: patterns in the distribution of species. New York: Harper and Row 1972

MacArthur, R. H., Levins, R." The limiting similarity, convergence and divergence of coexisting species. Amer. Hatur. 191, 377-385 (1967)

Reynoldson, T. B., Davies, R .W. : Food-niche and coexistence in lake-dwelling triclads. J. anita. Ecol. 89, 599-617 (1970)

Robins, J. D. : Differential niche utilization in a grassland sparrow. Ecology 52, 1065-1070 (1972)

Schoener, T. W.: The evolution of bill size differences among sympatric congeneric species of birds. Evolution 19, 189-213 (1965)

Simpson, G. G., Roe, A., Lewontin, R. C." Quantitative zoology, revised edit. New York: Harcourt, Brace and Co. 1960

Snow, B. K., Snow, D. W. : Feeding niches of hummingbirds in a Trinidad valley. J. anita. Ecol. 41, 471-486 (1972)

Terborgh, J., Diamond, J. M.: Niche overlap in feeding assemblages of New Guinea birds. Wilson Bull. 81, 29-52 (1970)

Ueckert, D. N., ttansen, R. M.: Dietary overlap of grasshoppers on sandhill rangeland in northeastern Colorado. Oecologia 8, 276-295 (1972)

Williamson, P.: Feeding ecology of the Red-eyed Vireo (Vireo olivaeeus) and associated foliage-gleening birds. Ecol. Monogr. 41, 129-152 (1971)

Dr. Peter F. Sale School of Biological Sciences University of Sydney Sydney, N.S.W. 2006 Australia