Treefall Gaps and Plant Species Diversity in Forests 205resource partitioning across the landscape. Addi-
tionally, the negative impact of lianas on tree
recruitment and growth of shade-tolerant trees
may be far lower in aseasonal wet forests because
liana abundance is much lower in these forests
(Schnitzer2005).Overall,wepredictthatgapswill
have the strongest impact on the maintenance of
diversity and forest regeneration in aseasonal wet
forests because light in the understory is most lim-
iting, nutrient gradients are more dramatic, and
gaps are likely more frequent. Comparative studies
of gaps along precipitation gradients are required
to more thoroughly address this hypothesis.TREEFALL GAPS PROVIDE A POOR
MODEL TO TEST THE
INTERMEDIATE DISTURBANCE
HYPOTHESIS
The intermediate disturbance hypothesis (IDH)
states that diversity will be highest at intermediate
levels of disturbance size and frequency, and time
since the last disturbance (Connell 1978). Accord-
ing to the IDH, large and frequent disturbances
reduce diversity by physically removing individ-
uals and thus extirpating species. When distur-
bances are too infrequent, competitive exclusion
occurs, which also leads to lower species diversity.
Only at some intermediate level of disturbance will
diversity peak (see Connell 1978).
If treefall gaps represent an intermediate level
of disturbance in tropical forests, then the gap
hypothesis might be considered within the frame-
work of the IDH (e.g., Hubbell 1999, Molino and
Sabatier 2002). The fundamental difficulty with
this approach, however, is that it is doubtful that
treefall gaps are sufficiently large enough or range
in size enough to rigorously test the IDH. Even
if gaps spanned a large enough range of dis-
turbance so that diversity varied with ga psize,
testing the IDH with treefall gaps is still problem-
atic because the absence of a unimodal response
does not reject the IDH. For instance, a linear
increase in diversity with ga psize may indicate
that the ga pdisturbance is on the low distur-
bance side of the unimodal IDH curve (Sheil and
Burslem 2003). However, a positive relationship
between ga psize and diversity provides no infor-
mation on whether the true curve will eventually
become unimodal with decreasing disturbance.
For example, Hubbell (1999) refuted the IDH
with species-individual curves from BCI, showing
that species accumulation (diversity) increased
as ga psize decreased, with the non-ga pforest
having the highest diversity accumulation. The
IDH may still hold, however, if diversity decreases
when levels of disturbance become lower than
the background (non-gap) disturbance regime on
BCI (Sheil and Burslem 2003). Even the old-
growth forest on BCI has a history of disturbance,
possibly indicating that it is still undergoing suc-
cession, and thus diversity could still decrease
over time in the non-ga pforest, which would
be consistent with the IDH (Sheil and Burslem
2003). In addition, with the exception of pio-
neer trees, Hubbellet al. (1999) omitted growth
forms most likely to be disturbance dependent
in light-limited forests (e.g., lianas, shrubs, and
herbs; Schnitzer and Carson 2000). A rigorous
test of the IDH requires the consideration of key
relevant growth forms along a disturbance gradi-
ent ranging from minimal to catastrophic. These
considerations make rigorous tests of the IDH
extremely challenging.CONCLUSIONS
Treefall gaps provide both equilibrium and
non-equilibrium explanations for the mainte-
nance of species diversity in tropical forests.
Although the ga phy pothesis is one of the major
hypotheses proposed to explain the maintenance
of species diversity in tropical forests, consider-
able work remains to be done to test the full
range of predictions that stem from this hypoth-
esis. Currently, the degree to which gaps maintain
diversity likely depends on the growth form and
life-history characteristics of the species exam-
ined. Liana and pioneer tree diversity, and possibly
that of shrubs and herbaceous plants, appears
to be maintained by treefall gaps to a signifi-
cant degree. Little evidence suggests that gaps
maintain shade-tolerant tree diversity, apparently
because of dispersal and recruitment limita-
tion, which may be exacerbated by competition