Tropical Forest Community Ecology

150 Stephen P. Hubbell

0 50 100 150 200

0.95

1.00

1.05

1.10

1.15

1.20

Per capita birth rate/death rate

c = 1

c = 100

b(n) = d(n)

c = 10

Density (population size), n

a = 40

x = 0.99

Figure 9.1 Theoretically expected curves of the ratio of the per capita birth rate to the per capita death rate in the
density dependence version of neutral theory. The dotted line is the line of population replacement, when birth and
death rates are equal. Parameterccontrols the strength of density dependence, and the size of the population that
experiences a rare-species advantage increases with the value ofc. The two other parameters of the theory are the
biodiversity numberθ(Fisher’sα) andx.

CONFRONTING THE THEORY WITH

RELATIVE ABUNDANCE DATA ON

TROPICAL TREE COMMUNITIES

Since my book was published, an old controversy
has re-emerged over which distribution fits rel-
ative abundance data better, Fisher’s logseries,
Preston’s lognormal, or now, the distribution
predicted by neutral theory. McGill (2003) and
McGillet al.(2006) contend that Preston’s lognor-
mal fits most available data on relative abundance
better than the distributions from neutral the-
ory. I have many problems with their analysis in
addition to my earlier-stated objections to meta-
analyses.The first problem is based on the number
of free parameters. Under the dispersal limitation
version of neutral theory (Hubbell 2001), the dis-
tribution has two free parameters (onlyθandm–
the number of species is a prediction), whereas
the lognormal has three (mean, variance, and

the modal number of species), so from an AIC

perspective one needs to devalue the lognor-

mal hypothesis relative to neutral theory for

having more parameters. Second, even without

this devaluation, the fit of the lognormal to

the Barro Colorado Island (BCI) data is actually

slightly worse than neutral theory in the case

of BCI (Volkovet al.2003) which was the basis

of McGill’s original assertion (McGill 2003). In

a subsequent paper, we showed that both ver-

sions of neutral theory – the dispersal limitation

version and the newer version with symmetric

density dependence (Volkov et al.2005) – fit

the static relative tree species abundance data

from six tropical forests very well (Figure 9.2).

The forests in question have very different evo-

lutionary histories and ecology, but despite this,

they are all fit quite well by the same neutral

model with different values of the free parameters

(Table 9.1).