On Bird Species Diversity

Author(s): Robert H. MacArthur and John W. MacArthur

Source: Ecology, Vol. 42, No. 3 (Jul., 1961), pp. 594-598
Published by: Ecological Society of America
Stable URL: http://www.jstor.org/stable/1932254
Accessed: 29-09-2015 07:48 UTC

REFERENCES
Linked references are available on JSTOR for this article:
http://www.jstor.org/stable/1932254?seq=1&cid=pdf-reference#references_tab_contents

You may need to log in to JSTOR to access the linked references.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact [email protected].

Ecological Society of America is collaborating with JSTOR to digitize, preserve and extend access to Ecology.

http://www.jstor.org

This content downloaded from 129.175.106.167 on Tue, 29 Sep 2015 07:48:40 UTC
All use subject to JSTOR Terms and Conditions
594 REPORTS Ecology, Vol. 42, No. 3
regeneration, and at the end of the experiment had under- 0.1 N), thereby minimizing possible undesirable effects
gone no noticeable increase in length. Variable results caused by high concentrations.
were obtained with those treated with a concentration of A suggested procedure for field work is to carry a
0.05 N; in 2 salamanders regeneration occurred whereas block of wood with holes bored in it to accommodate a
in one it was completely inhibited. Thus, the latter con- vial of beryllium nitrate and several vials of water for
centration is too weak to produce consistent results and rinsing off excess chemical after treatment. The block
therefore not well suited for use in field studies. provides a firm support and prevents spillage caused by
Although beryllium has been known to cause edema and movements of the salamanders' feet. Treatment is also
even death (Thornton op. cit.), the single death resulting facilitated if the vials have sufficiently small mouths that
from unknown causes in the present study is not believed undue movement of a salamander's foot is prevented while
to be attributable to the beryllium treatment as the animal it is immersed in the solution.
was subjected to a concentration of only 0.05 N (the low- REFERENCES
est one used). Animals treated with higher concentra- Thornton, C. S. 1949. Beryllium inhibition of regen-
tions did not show any ill effects. However, in field eration. I. Morphological effects of beryllium on
studies it would be advisable to use the lowest concentra- amputated fore limbs of larval Amnblystorna. Jour.
tion effectively preventing regeneration. (in this species, Morph. 84:459-493.

ON BIRD SPECIES DIVERSITY

ROBERT H. MACARTHUR AND JOHN W. MACARTHUR
Department of Zoology, University of Pennsylvania and Marlboro College, Marlboro, Vermont

It is common experience that more species of birds and other interference have very many (up to 106 at
breed in a mixed wood than in a field of comparable least in Mexico-Davis (1955)) species of breeding birds.
size. It is also well known that tropical forests seem Therefore, of course, the territories of most species are
to support more species than their temperate counter- scattered rather sparsely over the area and the terri-
parts. These facts are often explained in terms of the tories of very few species form a mosaic. It must be
number of "niches" or "ways of life" which the habitat concluded that either the species are scattered randomly
provides. In this paper, a somewhat more precise over the habitat, or else, more plausibly, that birds use
analysis is attempted. some farily subtle differences in local habitat as criteria
The actual number of species is better replaced by a for habitat selection. Returning to the bird species di-
number called the "bird species diversity," calculated as versity, it clearly can increase as the area increases.
follows: Let pi be the proportion of all of the bird in- Since there is no adequate theory of "species-area"
dividuals which belong to the ith species. Then the or "diversity-area" curves, all censuses must cover ap-
bird species diversity, is - Pi loge pi. This is a
p proximately the same area so that variations in the
bird species diversity reflect differences in habitat com-
formula used by communication engineers to calculate position rather than variations in size of census area.
the information generated, e.g., by a typist who uses Although many of the censuses reported here were taken
the different keys with frequencies pi. Thus, for in- over larger areas, the breeding bird populations of a
stance, a one species community always has zero randomly chosen 5 acres of each will be compared.
diversity; 2 species, one with 99 individuals and one The procedure of the research described here was to
with 1 individual, will have diversity of -.99 loge 99 census a wide variety of habitats, differing in (1) plant
-.01 loge .o = .046 + .010 = .056 (close to zero), species composition, (2) foliage height profiles, and
while 2 species each with 50 individuals will have di- (3) latitude, and to determine how much each of these
versity of .347 + .347 = .694. This illustrates why factors influenced the bird diversity.
diversity is a better measure than actual number of
species, for the community with 99 of one and 1 of the MEASUREMENTS
other seems closer to the community with one species. Plant species composition was measured by an index
Margalef (1957) has frequently used a similar measure of plant species diversity computed with the formula
in his plankton studies. In terms of this, the question used for bird species diversity. p. now refers to the
becomes: "What is it about the environment which total area of the leaves of the ith plant species, ex-
controls the bird species diversity ?" pressed as a proportion of the total leaf area of all of
Onu more refinement of the general problem is neces- the species on the census plot. It would be a formidable
sary. For, even if all the bird species were equally job to measure areas of individual leaves, but fortunately
abundant, a bird census of a small area (say an acre) a much easier and more accurate method bypasses this:
would have only a few of the species. Since an acre Imagine all the leaves lying on the ground as in the
sul)ports 3 or 4 pairs of birds, not more than 3 or 4 autumn. If there are, on the average 4 layers of leaves
species could be expected to have nests in the given on the ground, then there are 4 acres of leaf per acre
acre. And if the area is so homogeneous that adjacent of this habitat, and if we were to push a sticky needle
1 acre (say) territories are occupied by the same 3 or 4 through the fallen leaves at random points, then an
species, then the whole area cannot have more than 3 average of 4 leaves would adhere to it. And the areas
or 4 species. This is contradicted by the abundant of the leaves are automatically taken into consideration
evidence that stands of vegetation with the degree of so that if '4 of the leaves picked up were of some
homogeneity resulting from a uniform history of cutting large leafed species, it would indicate precisely the

This content downloaded from 129.175.106.167 on Tue, 29 Sep 2015 07:48:40 UTC
All use subject to JSTOR Terms and Conditions
July, 1961 REPORTS 595
same total leaf area of that species as of a small leafed the ground a white board, marked in squares, was
species whose leaves were picked up with the same moved horizontally away from an observer at the same
frequency. Rather than wait for the leaves to fall, height until X of its surface was obscured by leaves
it is easier to erect a stiff wire and count the leaves it from the observer's view. In forests, the board atop a
touches (since leaves on trees do not lie horizontally, high pole was erected at trial distances from the ob-
some error is introduced in this way, but it gives a server until an acceptable one was found. The distance
convenient approximation). For the canopy, rather than D between the observer and the board was then measured
using a wire, a sighting was made vertically through a and the foliage density k at that height was estimated
10 foot aluminum pipe of 1%4 inch diameter and the from e - kDr -2 or k - log,2/D.
number of leaves which a wire would intersect was esti-
This formula too is only approximate, but seemed satis-
mated.
factory. Roughly it is derived as follows. The area
For coniferous trees, and sometimes for deciduous, it
of leaf silhouetted against the board will be nA, say,
seems preferable to use fraction X of sky not obscured
where A is the area of the board and n is the leaf sil-
by foliage and estimate the number n of leaves which
houette per unit of board area. The volume of space
a wire would intersect by the formula e-n = X which
in which these leaves lie is DA, so the leaf silhouette
is the first term of the Poisson distribution.
area per unit of volume of space is
Thus, if 90% of the sky is obscured by foliage a frac-
tion equal to 1i o is unobscured and since e - 2.3 - nA n k.
1Aho,this is equivalent to an average of 2.3 randomly DA =D =k
placed leaves (of any size!) above the point from
which observation was made. In future measure- As before, n can be estimated by the first term of a
ments vertical photographs analyzed by a recording Poisson distribution so that e-n - 1? or n - log,2. The
microdensitometer will be substituted for the more method assumes that the proportion of leaf area which
subjective estimates of percentage. Since leaves may is similar in all 3 layers. The size of the board is thus.
not not be randomly placed this is obviously only an unimportant (since the A's cancel out) except that a
approximation, but in practice it seems quite accurate. large board gives an average over more vegetation'
From the foliage height profiles a number called the (10" X 18" was used inm ost of this work). In each
foliage height diversity was calculated using the same plot, at least 16 such measurements were averaged at
formula (-E pi loge pi) as before. In this case pi is each of the following heights above the ground: 6", 2',.
i 5', 10', 20', 30', 45', 60'. A profile of foliage density-
the proportion of the total foliage which lies in the ith was drawn by eye passing through each of these calcu-
of the chosen horizontal layers. The profiles were con- lated points. The profiles of the primarily deciduous.
structed as follows: At a sequence of heights above areas censused are shown in Figure 1. (The Florida

Av_P_mo____ J. MALYLA-MD

20 EI. a0 FE4D. -859

.B.5.D. -Gg839-.2-8
P.3.1 D. 9 72P FlaD. -4:48 FI. 3DI228
~~~~~
P31). 1 FED. 577 6~~~~~~~~~~~~~P..D
177o
40 ~ ~ ~ ~ ~ ~ ~ ~ ~ -6 20-
C.~~~~~~~~2
PEE...02..178
20 ~ ~ ~P.SD...........
~~~~~~~~~~~~~~
Fl-ID. 745 .~. 2.1 ~ ~ ~ ~ ~~~~~F.D N57
0 ~~ ~ ~ ~ ~ 7
02 04 08 40~~~~~~~~~7345-.1-6
20~~~~~~~~0
D. PAN......M......

2 0
02 '34 08 02 04
~~~~~~~~~~0
33)276
248 4 P30
3
PSI)~~~~~~~~~~~~~~~~C~
P1G21 022
218
3).8

60 F. 60 60 H. 60 60

0 -1)2--0--------D.i.2739...40..40
-OZ
40 p~~~~~~~~
j~~~~ 40 P~~......
..... 2503

20, FMAD. -455M0N 2PRUE 10 200 20

20 20.I-9 ro .0M
P 5. DB D5 1721 .*1
.3 ..~~~~.14.
.**- .HD 20 3
0 02 P3D 1372 02 ~~~~~~~~~~~~~~~~~~0
'a. .1*

02~~~~~~~~~~~~
are plteF
ln.heasise r .Heih. h nfe above th run sth rint. F s oig
hegh ivrst, D isbrdspcesdvesiy adP..D s latspcesdiesiy

This content downloaded from 129.175.106.167 on Tue, 29 Sep 2015 07:48:40 UTC
All use subject to JSTOR Terms and Conditions
 596 REPORTS Ecology, Vol. 42, No. 3
bird census was taken l)y Wolfenden et al. 1959, who diversity and bird species diversities are shown as well
kindly allowed measurements to be made in this area). as plant species diversity and latitude. These are plotted
The area of the profiles is divided into 3 horizontal as a graph in Figure 2, showing a close fit to the line:
layers and the proportions of the whole that each con- bird species diversity = 2.01 foliage height
stitute.s is the pi used in the formula.
diversity + .46,
REsULTS calculated by least squares. Various other subdivisions
Tlc ;actual censuses are in Table I, in which the of the profile into horizontal layers were tried, and the
mibiner of territories, or fraction of territories, is layers 0-2', 2'-25' and > 25' were chosen as those
enterc(l. layers which made the collection of points on the graph
TIhlelayers 0'-2', 2'-25' and over 25' are marked on the most orderly. It is of interest that this subdivision was
prolles in Fig. 1 and the corresponding foliage height chosen after the Vermont censuses were taken in 1959
and that it continued to be appropriate for the censuses
TABLE I. Numbers of breeding bird territories per 5 acres in 1960, elsewhere. Such subdivisions as 0-3', 3-30' and
for the regions shown in Fig. 1 > 30' were nearly as good, but more nearly equal sub-
division (e.g. 0-15', 15'-30', > 30') made a very scattered
A B C E F G 11 T J K L _
:B. 5. I)
mournri(g dov 1
yellow b)illed( . .
cuc(koo .5 3
r. t. Ihuimxloiilgb)ird......... .5 .5
flicker ...........1 . 1 5
red-bellied wood p)ecker .5
yellow-1)ellied sal)sucker.... 1
hairy wo(dpecker .......... 1 .5
downy woodpecker 1
crested flycatcher 1 1 1
acadian flycatcher 1.5 .0 *
wood pewee 2 2 1
olivesided flycatcher ....... 1.5
blue jay 1 1 .5 2.
black-capp)edchickadee 1 .5
tufted titmouse 1.5 1.5 1 1
white-breasted nuthatch .... 1
brown creeper.............1
winter wren 1
catbird ............ . . 1 1 1
robin. ... 5 .5
wood thrush 1 1.25
0~~~~~
olive-backed thrush ........ 2
veery thrush 1
gnatcatcher 1
golden-crowned kinglet....
white-eyed vireo...........
yellow-throated vireo ......1
red-eyed vireo .............1 3 1 2.5 2 4 1
black and white warbler . 5
prothonotary warbler ......2
blue-winged warbler ....... .5
Nashville warbler 1
parula warbler........ 1
magnolia warbler .5 1.5 1.5
black-throated blue warbler.
myrtle warbler 2 5
black-throated green warbler 5 1 1 ~~2
blackburnian warbler 1 2 1 F.H.D.
bay-breasted warbler 3
prairie warbler. ... 2 FIG. 2. Bird species diversity is plotted against the
ovenbird 1.5 3 3 foliage height diversity of deciduous forest plots. Point
Kentucky warbler 1 1 1 .5 1. is census D of tropical savannah and point 2 is census
mourning warbler . .75 E of pure spruce forest.
yellowthroat . . ..... 1 2.5 3 2
yellow -breasted chat ....... 2
Canada warbler 2 graph. The linearity of the cluster of points indicates that
redstart . . .1 1.5 2 the addition of a new layer of a given amount of foliage
grackle 1.5 results in the same increase in bird species diversity,
scarlet tanager .5 (not however the same increase in number of bird
summer tanager 2 species) no matter which layer (0-2', 2-25' or > 25') is
cardinal .................. 1 1 1
added, and no matter which other layers are present to
indigo bunting ............ 1.5 1.5
goldfinch . . . . 1 .5
begin with. Thus, we can say that the layers 0-2', 2'-25'
towhee .................. .25 1.5 1 1 1 and > 25' are roughly equally important to the birds.
slate-colored junco ..........5 (The reasons for this will be discussed later.) Looked
field sparrow . .......... 2 1 3 at from this point of view, we can see the trouble with
white throated sparrow 1 the other subdivisions. For definiteness, consider 0-15',
song sparrow . ... .. .5 1 15'-30', > 30'. Adding a 0-15' layer to a habitat without
it causes a much greater increase in bird species diversity

This content downloaded from 129.175.106.167 on Tue, 29 Sep 2015 07:48:40 UTC
All use subject to JSTOR Terms and Conditions
July, 1961 REPORTS 597
than the addition of the layer > 30'. There is nothing on Mount Desert Island in Maine which the authors
biological about the number of layers chosen. Four or 5 studied earlier (MacArthur 1957) had as great a bird
layers in a roughly similar subdivision would be more species diversity as the, forests of very mixed composition
cumbersome to analyse but would presumably be even in Pennsylvania and Maryland.
more accurate. In particular, the layers O-'2', '2'-6', The lack of effect of latitude is puzzling but is only
6'-15' and > 15' suggested by Elton and Miller (1954) tentative since a rather poor sample of latitudes have
allow a rather good prediction of the bird species been examined. Furthermore, a brief trip to Panama
diversity. yielded a tentative bird species diversity in a savannah
area of 1.906 and foliage height diversity of .455 (see
Fig. 2). Thus, the tropical bird species diversity is
much greater than the temperate one for habitats of
comparable profile. The excess of tropical savannah
.4 over temperate fields is about equal to the excess of
the bird species diversity in the tropical forest over
temperate forests. These facts will be discussed further
in a separate paper.
DIscussION
These results are rather statistical in nature. What
.2. is their meaning in terms of individual birds or species?
The simplest explanation which seems to account for
the observations, describes the "shape" of a bird's niche.
Let us return to the picture of many territories distrib-
.1.
uted over an area and consider the following evolu-
*
tionary argument. A large number of species can be
0 accommodated in an environment in a variety of ways
of which there are 2 extremes. Each species may have
P.52D. 1 3 different habitat preference and feed throughout this
habitat on all kinds of food, or, all species may share the
entire habitat, each species feeding on a different variety
of food or in a different situation within the habitat.
-.1~~~~~ The first extreme violates what might be called the
"jack of all trades-master of none" principle that nat-
- .2. ural selection favors the increased efficiency resulting
from a certain amount of specialization. In the other
. extreme, specialization has proceeded so far that time
-.3 and energy are wasted in travelling between spots for
which the specializations are adapted. It is hard to
say just where the balance of these opposing require-
ments would be reached, but it is clear that greater
- .4 specialization resulting in increased efficiency would al-
Fi. 3. The ordinate is the residual scatter, after ways be favored as long as no time or energy are wasted.
partial regression of bird species diversity on foliage And no, time or energy will be wasted if niches are
height diversity. This is plotted against plant species "convex" in the sense that between any 2 fairly distant
diversity, showing that knowledge of plant species feeding places there will be a fairly natural route also
diversity does not provide additional knowledge of bird consisting of feeding places. A specialization to a
single tree species in a mixed forest would clearly vio-
species diversity. The least squares equation was B.S.D.
= 2.546 F.H.D. - .152 P.S.D. -.250, late this since, in passing from one suitable tree to
and the residual
another, the bird would go through many unsuitable ones.
scatter was computed from the formula B.S.D. - 2.546
Thus, natural selection would tend to eliminate a situa-
F.H.D. - .048.
tion in which bird species diversity depended upon tree
species diversity, unless, as in some fruit eating spe-
The next question is: How much of the remaining cies, a very remarkable improvement in efficiency is
scatter, i.e. how much of the variability in bird species achieved along with the restriction in feeding position.
diversity not accounted for by the variation in foliage Thus, one principal result of these censuses can be
height diversity, can be accounted for in variations of
predicted on assuming that niches are convex.
plant species diversity and latitude ? Remarkably
Next, we may ask "why are the layers 0-2', 2-25',
enough, the answer is "None." This is seen most easily
> 25' equally important? Is it because birds respond
by glancing at Figure 3 which shows that the residual
scatter after partial regression is more or less inde- to different heights, or is it because they respond to
pendent of plant species diversity. Thus, although plant different configurations of vegetation in different
species diversity alone is a good predictor of bird species layers?" In the latter case, herbs, bushes and trees pre-
diversity it is because plant species diversity is high sumably correspond to the layers 0-2', 2-25' and > 25'
when foliage height diversity is high, and, when this is respectively, although small trees count as bushes etc.
taken account of, plant species diversity can contribute There is good evidence for this latter explanation. For,
nothing further. In other words, habitats of the same although deciduous forests vary principally with height
profile have the same bird species diversity whether above the ground and hence have a bird diversity pre-
composed of few or many plant species. As a striking dictable from the height profile, conifers (especially
example of this, the almost pure stand of white spruce spruce) have a marked "inside" and "outside" for which

This content downloaded from 129.175.106.167 on Tue, 29 Sep 2015 07:48:40 UTC
All use subject to JSTOR Terms and Conditions
598 REPORTS Ecology, Vol. 42, No. 3
species are specialized (MacArthur 1957). Hence bird aspects of this research at length and it has benefited
species diversity would be high in a mature spruce forest from their ideas. James Preer helped make the measure-
even if few layers were present. This is precisely what ments.
happens in the Maine white spruce wood mentioned SUMMARY
earlier, with bird species diversity of 1.712 and foliage
height diversity of .287 which is seriously off the graph 1. Bird censuses on a wide variety of areas are com-
of deciduous forests (Fig. 1). pared in order to see what aspects of environmental
A different way of looking at the data gives additional variation control bird species diversity.
insight. Watt (1947) has pointed out that plants are 2. Indeciduous forests, bird species diversity can be
distributed in patches. Hutchinson and MacArthur predicted in terms of the height profile of foliage
(1959) attempted to explain the sizes of coexisting density. Plant species diversity, except by influencing
organisms in terms of an environment composed of a this profile, has nothing to do with bird species
mosaic of kinds of patches. Different combinations of diversity. The layers 0-2', 2'-25', > 25' seem equally
patches formed the habitats selected by different species. important in determining bird species diversity;
The present research can be easily interpreted in terms these layers presumably correspond to different con-
of this picture of the environment. In fact, our results figurations of foliage. This should not be interpreted
suggest that the patches forming the birds' environmental as evidence that a forest is made up of discrete
mosaic are sections of canopy C (over 25'), patches of layers. These 3 layers are constructed by the
bushes B from 2-25', and the herbaceous and other observer.
cover H less than 2' from the ground. And the sequence 3. An evolutionary argument is given which predicts
of patches encountered in moving through the habitat the observations and at the same time suggests that
(or in taking ever larger samples) is then represented niches should be "convex." Supporting evidence is
by a sequence of letters e.g. C, B, H, H, B, C, ,,, with provided.
certain random properties but also subject to the con- 4. These results provide no evidence about the real
dition that the long term frequency of C's, B's, and H's causes of tropical diversity (i.e. whether the tem-
should conform to their respective densities (pi) in the perate regions, given enough time, can support as
particular habitat. If the sequence is ergodic, which great a diversity as the tropics now have) or about
defines what we call a homogeneous habitat, then it is the diversity which could be expected in a composite
well known that the uncertainty of the next letters in census of 2 habitats. These are essentially different
the sequence is appropriately measured by the formula problems and are under investigation now.
- E pi loge pi which we used. If, instead of considering
i REFERENCE S

the uncertainty of future single letters in the sequence, Davis, L. I. 1955. Census 27. Audubon Field Notes
we ask for the uncertainty of future pairs of letters, 9: 425-426.
the formula becomes - 2 E pi loge pi which is 2 X foliage Elton, C. and R. Miller. 1954. The ecological survey
i of animal communities. J. Ecol. 42: 460-496.
height diversity, which is essentially the predicted value Hutchinson, G. E. and R. H. MacArthur. 1959. A
of the bird species diversity. Thus we can say that theoretical ecological model of size distributions
bird species diversity is determined as if the birds recog- among species of animals. Amer. Nat. 93: 117-125.
nized suitable habitats by pairs of foliage types (> 25', MacArthur, R. H. 1957. Population ecology of some
2-25', 0-2'). The species area curve could then be warblers of northeastern coniferous forests. Ecol-
predicted from this. ogy 39: 599-619.
Margalef, R. 1957. La teoria de la informacion en
ACKNOWLEDGMENTS ecologia. Barcelona. Memorias de la real academia
This work was supported (in different stages) by de ciencias y artes.
grants from the University of Pennsylvania, The Amer- Watt, A. S. (1947). Pattern and process in the plant
ican Academy of Arts and Sciences, and The National community. J. Ecol. 35: 1-22.
Science Foundation (No. G11575). Drs. G. E. Hutchin- Woolfenden, G. 1959. Census 14. Audubon Field
son, Peter Klopfer and Monte Lloyd have discussed Notes 13: 466.

HABITAT OF JUVENILE SHRIMP (FAMILY PENAEIDAE)l

GORDONGUNTER2
Texas Game and Fish Commission, Field Laboratory, Seabrook, Texas

For the past 50 years penaeid shrimp have been studied other higher Crustacea in that the eggs are shed free into
along the shores of the Gulf of Mexico and the South the water and develop into larval stages ranging from
Atlantic coast of the United States. These shrimp are of nauplii (very similar to early copepod developmental
intrinsic interest to zoologists. The group differs from stages) to post-larvae, with as many as 14 probable molts
' Contribution No. 53 from the Marine Laboratory, intervening, the precise number still being a matter of
Texas Game and Fish Commission. argument. A 2nd reason for the extensive study of the
2 Present address, Gulf Coast Research Laboratory, Penaeidae is that the shrimp fishery is the most valuable
Ocean Springs, Mississippi. one in the United States (cf. the various annual fishery

This content downloaded from 129.175.106.167 on Tue, 29 Sep 2015 07:48:40 UTC
All use subject to JSTOR Terms and Conditions