Testing Explanations for Species Diversity using FDPs 107of seedlings near to the parent trees, while a rela-
tively small number of seedlings will escape the
parental neighborhood with a high density of
conspecifics. As a consequence intraspecific com-
petition between siblings dispersed furtheraway
will be less. This mechanism, then, promotes the
survival of a variety of species in any location,
and therefore forest diversity is increased. Because
it is a stabilizing effect (Chesson 2000), it offers a
direct contrast to neutral theory.
The development of the Janzen–Connell
hypothesis, and the various ways it has been
studied in tropical forests, is treated in more
detail by Adler and Muller-Landau (2005) and
Carsonet al. (Chapter 13, this volume; also
Leigh Chapter 8, this volume). Many studies
of individual species have found evidence for
the Janzen–Connell hypothesis, often with direct
information on the pests or pathogens causing
the effect (Hammond and Brown 1998, Carson
et al. Chapter 13, this volume). Nonetheless, a
species by species approach makes it difficult
to determine how widespread this effect is in
structuring tropical forest communities. Negative
density dependence is implicitly spatially depen-
dent, and therefore the data from the FDPs on
tree size and mapped locations makes them ide-
ally suited to investigate this explanation. It is
important to understand, however, that studies of
demography and community structure can only
reveal the existence of NDD, not the mechanism
causing NDD.
Many of the first tests of NDD in an FDP
were conducted using the BCI and Pasoh plots
by looking for spatial patterns in the abundance
of recruits or mortality, with respect to distance
between individuals, or density of conspecifics
within subplots of the FDP (Hubbellet al. 1990,
2001, Willset al. 1997, Wills and Condit 1999).
The impression from these studies was that only
the most common species, encompassing about
10% of the community, exhibited NDD (see also
Conditet al. 1992). However, using subplots of
an FDP to test density dependence can be crit-
icized because trees that are close to the edge
of a subplot present little information about the
tree composition of the real neighborhood around
each individual for analysis. Using an individual-
based analysis, where neighborhoods of varying
radii are constructed around each individual tree,
Peters (2003) detected density-dependent mortal-
ityinapatternconsistentwiththeJanzen–Connell
hypothesis, in more than 80% of species he inves-
tigated in the BCI and Pasoh plots. Peters’ (2003)
methods, however, did not account for spatial
autocorrelationinthedata(seeHubbellet al.2001
as a contrast), among other things, and the study
likely overstates the number of species exhibit-
ing NDD. Using a different statistical approach
to that of Peters (2003), for individual-based
neighborhood analyses, Uriarteet al. (2004a,
2005a) analyzed patterns of growth on the BCI
plot and found that 26 of 60 species exhib-
ited negative conspecific interactions, while 16 of
50 species exhibited negative conspecific interac-
tionsforsurvival.Inthehurricane-drivenLuquillo
FDP, Uriarteet al. (2004b) found negative conspe-
cific growth effects in all of the 11 species tested
and, for mortality, in 7 of 12 species tested. Taken
together, these results (Uriarteet al.2004a,b,
2005b) suggest that NDD may be common in
tropical forests. The individual-based approaches
are limited by the sample sizes needed to detect
an effect of density on plant performance, and
they may not reveal the full extent of den-
sity dependence in tropical forests (Uriarteet al.
2005a). Returning to the subplot approach, Wills
et al. (2006) asked how the diversity of subplots
changed as a result of the survival and recruit-
ment of trees in different dbh size classes in
repeated censuses, and consistently found that in
most of the forests tested diversity increased with
time, consistent with the predictions of NDD and
two other potential explanations (see below; Wills
et al. 2006). This study encompassed seven of the
CTFS plots and the results were robust; only one
comparison in one FDP failed to show the expected
pattern.
The key difference between neutral theory
(Hubbell 2001, Volkovet al. 2003) and NDD is
that they make fundamentally different predic-
tions about the status of rare species (Volkovet al.
2005). Neutral theory implies that rare species,
compared with species of more modest abun-
dance, are on the verge of “winking out” of the
community, while the rare species advantage (as a
result of NDD) specifies that they are on their way
to becoming more common (Volkovet al. 2005).