Chance and Determinism in Tropical Forest Succession 395animal-dispersed species, but not for two wind-
dispersed species (Howlett and Davidson 2003).
Even when secondary forests are close to mature
forests, seed dispersal can be a major limitation
(Gorchovet al.1983, Corlett 1992, Turneret al.
1997, Wunderle 1997, Duncan and Chapman
1999, Holl 1999, Ingle 2003). Martinez-Garza
and Gonzalez-Montagut (1999) found that dis-
persal limitation of forest interior species resulted
in pioneer dominance for 30–70 years in aban-
doned pastures of lowland tropical regions of
Mexico.
Under ideal conditions, the early arrival and
establishment of some shade-tolerant canopy tree
species (includin gpalms) can increase the rate of
succession, as these species often grow rapidly in
height and reach reproductive maturity within
20–30 years, when they begin to produce their
own seedlin gcohorts (Sezenet al.2005). Many
of these species are capable of recruitment into
canopy tree size classes (≥25 cm dbh) in the
absence of gaps (Chazdon unpublished data).
If seedlings of shade-tolerant and slow-growing
species colonize later, durin gthe stem exclusion
or understory reinitiation phase, their recruit-
ment to the canopy may require several decades
or longer (Finegan and Chazdon unpublished
data).
Few studies have statistically compared species
composition across a tropical forest chronose-
quence. Terborghet al.(1996) compared species
composition in early, middle, and late succes-
sional floodplain vegetation with mature flood-
plain forests of the Manu River in Peru. In this
study, cluster analysis showed that the five mature
floodplain forests were most similar to each other
in species composition and that they differed con-
siderably from successional forests. A detrended
correspondence analysis suggested a clear direc-
tionality to species compositional changes during
floodplain succession in this region. A similar
approach was used by Sheil (1999) to com-
pare canopy tree species composition for the 10
sites in Eggeling’s (1947) study of forests in
Budongo, Uganda. This analysis indicated a con-
sistent compositional progression across the plot
series, with the rankin gof plots conformin gpre-
cisely to Eggeling’s original successional sequence.
Within this set of plots, there was stron gevidence
for compositional convergence towards a species-
poor forest dominated byCynometra alexandri
(ironwood; Sheil 1999). An alternative interpre-
tation, suggested by Sheil (1999), is that Eggeling
originally selected the plots to fit his preconceived
model of an ordered developmental successional
series in Budongo Forest.Life-forms, functional groups, and
life-history traitsDurin gsuccession, life-form composition shifts
dramatically, particularly durin gthe stand ini-
tiation phase. Durin gthe first 5 years of post
slash-and-burn succession, Ewel and Bigelow
(1996) documented decreases in herbaceous
vines, increases in shrubs and trees, and a dra-
matic increase in epiphytes between 30 and
36 months. Grass and forb dominance peaked
after 3 years in an abandoned pasture in Puerto
Rico (Myster 2003). Vines, ferns, and persistent
grasses can impede establishment and growth
of woody shrubs and trees in abandoned pas-
tures (Hollet al.2000, Hooperet al.2004). More
often, however, early dominance of large-leaved
herbaceous species facilitates establishment of
shade-tolerant woody species (Denslow 1978,
Ewel 1983). Across a sequence of stands from
20 to over 100 years old in Barro Colorado
Island and surroundin gareas, liana abundance
decreased as a function of stand age (Dewalt
et al.2000). Liana size increased durin gsuc-
cession, however, resultin gin a lack of corre-
spondence between stand age and liana basal
area. Liana diversity (as measured by Fisher’s
α) was higher in young stands than in older
stands, up to 70 years in age (Dewaltet al.
2000).
Considerin gonly woody life-forms in wet tropi-
cal forests of northeastern Costa Rica, Guariguata
et al.(1997) found that shrub abundance was
significantly higher whereas understory palm
abundance was significantly lower in young sec-
ondary stands (15–20 years old) compared with
old-growth stands. Mature canopy palms (stems
≥10 cm dbh) were also significantly more abun-
dant in old-growth stands. Woody seedlings in
second-growth permanent plots in this region