ECOSYSTEM ECOLOGY - ORIGINAL PAPER

Improved recruitment of a lemur-dispersed tree in Malagasydry forests after the demise of vertebrates in forest fragments

K. H. Dausmann Æ J. Glos ÆK. E. Linsenmair Æ J. U. Ganzhorn

Received: 3 September 2007 / Accepted: 12 May 2008 / Published online: 4 June 2008

� Springer-Verlag 2008

Abstract The objective of this study was to examine how

the processes of seed dispersal and seed predation were

altered in forest fragments of the dry forest of Madagascar,

where the usual seed dispersers and vertebrate seed pre-

dators were absent, using a lemur-dispersed tree species

(Strychnos madagascariensis; Loganiaceae) as an example.

We then assessed how the changes in vertebrate commu-

nity composition alter the regeneration pattern and

establishment of this tree species and thus, ultimately, the

species composition of the forest fragments. By using size-

selective exclosures, data from forest fragments were

compared with results from continuous forest where ver-

tebrate dispersers and predators were abundant. Visits to

the exclosures by mammalian seed predators were moni-

tored with hair traps. In the continuous forest up to 100% of

the seeds were removed within the 7 days of the experi-

ments. A substantial proportion of them was lost to seed

predation by native rodents. In contrast, practically no

predation took place in the forest fragments and almost all

seeds removed were dispersed into the safety of ant nests

by Aphaenogaster swammerdami, which improves chances

of seedling establishment. In congruence with these find-

ings, the abundance of S. madagascariensis in the forest

fragments exceeded that of the continuous forest. Thus, the

lack of vertebrate seed dispersers in these forest fragments

did not lead to a decline in regeneration of this animal-

dispersed tree species as would have been expected, but

rather was counterbalanced by the concomitant demise of

vertebrate seed predators and an increased activity of ants

taking over the role of seed dispersers, and possibly even

out-doing the original candidates. This study provides an

example of a native vertebrate-dispersed species apparently

profiting from fragmentation due to flexible animal-plant

interactions in different facets, possibly resulting in an

impoverished tree species community.

Keywords Seed dispersal and predation �Fragmentation � Regeneration � Exclosure experiments �Madagascar

Introduction

Fragmentation is considered to be one of the most impor-

tant threats to biodiversity in tropical forest ecosystems

(Smith and Hellmann 2002). While early studies concen-

trated on the effect of habitat loss based on the hypotheses

of island biogeography (reviewed by Laurance and Bier-

regaard 1997), recent approaches emphasize synergistic

effects in fragmented landscapes that have negative effects

on native populations beyond the effect of area loss (Fahrig

2003). Plant and animal species that inhabit tropical forests

have developed complex ecological interactions with one

another over millions of years, and have thereby shaped

these forests. There is a sudden interruption of these

Communicated by Jacqui Shykoff.

Electronic supplementary material The online version of thisarticle (doi:10.1007/s00442-008-1070-6) contains supplementarymaterial, which is available to authorized users.

K. H. Dausmann � K. E. Linsenmair

Department of Animal Ecology and Tropical Biology,

Biozentrum, Am Hubland, Julius-Maximilians-University

Wurzburg, 97074 Wurzburg, Germany

K. H. Dausmann (&) � J. Glos � J. U. Ganzhorn

Department of Animal Ecology and Conservation,

Biozentrum Grindel, University of Hamburg,

Martin-Luther-King-Platz 3, 20146 Hamburg, Germany

e-mail: kathrin.dausmann@uni-hamburg.de

123

Oecologia (2008) 157:307–316

DOI 10.1007/s00442-008-1070-6

interactions when some species go locally extinct or are

reduced to very low abundances, and the question arises

how these communities persist with possibly important

links missing. Changes in community composition fol-

lowing fragmentation can have pronounced effects on

ecosystems ranging from changes in community structure

and function (Didham et al. 1996), reduced plant seed

production due to the lack of pollinators (e.g. Steffan-

Dewenter and Tscharntke 1999), to changes in the genetic

variation of single fruit crops and their germination success

(Cascante et al. 2002), and to reduced tree regeneration due

to the lack of seed dispersers (e.g. Redford 1992; Ganzhorn

et al. 1999; Cordeiro and Howe 2001; Wright et al. 2007).

The consequences of fragmentation on forest regenera-

tion are complex. Plant communities are altered whenever

animal communities are altered (Emmons 1989). On the

one hand potential seed dispersers and seed predators can

be decimated in forest fragments (Klein 1989; Burkey

1994; Laurance and Bierregaard 1997). On the other hand

they can increase in abundance once their natural enemies

have gone extinct locally (Emmons 1989; Peres and Pala-

cios 2007). If predator species can no longer sustain

themselves in the forest fragments, prey species may

become hyperabundant and cause overproportional damage

to the next trophic level, such as when seedlings cannot

establish themselves because of high numbers of herbi-

vores (Terborgh et al. 2001, 2006). All these changes affect

the diversity of tree species in isolated forest fragments and

can cause impoverishment of the community when some

species can no longer sustain themselves while others

thrive and begin to dominate the forest (Terborgh 1992;

Leigh et al. 1993; Chapman and Chapman 1995; Asquith

et al. 1997; Cascante et al. 2002; Wright 2003; Stoner et al.

2007). The limited knowledge of these processes is an

impediment to the understanding of forest regeneration and

the maintenance of diverse tree species communities.

Conservation of the remaining dry deciduous forests of

western Madagascar, highly fragmented and reduced to

about 3% of their initial surface area since the arrival of

humans some 2000 years ago (Smith 1997), has become a

matter of concern (Ganzhorn et al. 1997, 2001; Hannah et al.

1998; Sorg et al. 2003). Even though dry deciduous forests

rank among the most endangered major ecosystems of the

world (Janzen 1988; Lerdau et al. 1991) very little is known

about the ecological processes in these forest ecosystems.

Yet, the Malagasy dry forest represents an excellent system

for the study of the relationships between plants, seed dis-

persers, and seed predators for several reasons. Especially

among vertebrates the number of possible species involved

in primary and secondary seed dispersal of trees with large

seeds is very limited compared to other tropical ecosystems

(Langrand 1990; Goodman and Ganzhorn 1997). The island

has a very impoverished fauna of fruit-eating birds and, apart

from flying foxes (Bollen and van Elsacker 2002), lemurs are

the chief primary seed dispersers of the island (reviewed by

Birkinshaw 2003). This results in relatively simple interac-

tions between zoochorous trees, primary, and secondary

dispersers (Randrianasolo et al. 1996; Bohning-Gaese et al.

1999; Ganzhorn et al. 1999; Bleher and Bohning-Gaese

2001; Voigt et al. 2002; Bollen et al. 2005 and references

therein). On the community level at least one tenth of the tree

species of Madagascar’s dry forest seem to depend exclu-

sively on the brown lemur (Eulemur fulvus rufus) for primary

dispersal and successful regeneration. In forest fragments,

where these animals have disappeared, recruitment of these

tree species has been reduced significantly (Ganzhorn et al.

1999). While these studies stressed the importance of seed

dispersers, very little is known about effects of seed preda-

tors in Madagascar’s ecosystems (Du Puy 1996; Dolch and

Tscharntke 2000; Spehn and Ganzhorn 2000). The lemur

populations as well as those of potentially seed-predatory

rodents decline with increasing fragmentation leading to

predictable complete loss of some species in smaller frag-

ments (reviewed by Ganzhorn et al. 2003).Therefore the

objective of our study was to examine the role of seed pre-

dation and dispersal using a lemur-dispersed tree species

(Strychnos madagascariensis) as an example. By using size-

selective exclosures, we investigated how the processes of

seed dispersal and seed predation were altered in forest

fragments, where the usual seed dispersers and vertebrate

seed predators were absent. We then assessed how the

changes in vertebrate community composition alter the

regeneration pattern of this tree species and thus, ultimately,

the species composition of the forest fragments.

Materials and methods

Study area

Our study sites were located in the Kirindy Forest/Centre

de Formation Professionelle Forestiere de Morondava

(CFPF) (448400E, 20�040S; altitude, 18–40 m a.s.l.), a dry

deciduous forest of western Madagascar, 60 km north of

Morondava, and in two nearby forest fragments. The forest

fragments were surrounded by tree-shrub savannah, 5,700

and 6,100 m2 in size and approximately 200 m away from

the forest edge. They have been separated from the con-

tinuous forest for at least 20 years. A more detailed

description of the area is given by Ganzhorn and Sorg

(1996).

Study species

The tree species investigated in our experiments was

Strychnos madagascariensis (Loganiaceae). This species is

308 Oecologia (2008) 157:307–316

123

relatively common in the Kirindy Forest/CFPF with about

20 mature trees/ha [diameter at breast height (DBH) C

20 cm] (Dausmann 1997). It grows up to 20 m in height

with a trunk DBH of up to 60 cm (Leeuwenberg 1984).

The round fruits of S. madagascariensis ripen and fall from

December to March. They are enclosed in a firm shell and

divided into several segments, each comprising one seed

with pulp firmly attached to it (Fig. S1a, Electronic Sup-

plementary Material). Fruits contain between one and 19

seeds (3.9 mean ± 2.9 SD; n = 650). The diameters of

fruits vary correspondingly between 2 and 5 cm

(3.3 cm ± 0.6; n = 198) and fresh fruits weigh between 6

and 60 g (22.8 g ± 11.9; n = 206). Single seeds weigh

2.9 g ± 0.8 (n = 144) and are 1.90 9 1.49 9 0.97 cm in

size (SD = 0.20; 0.17; 0.15; n = 102). Fruit production

was similar in the two types of habitats with regard to

timing as well as amount of fruits both per tree and per area

(Dausmann 1997).

Seed dispersers and vertebrate seed predators

Three reptilian species, six avian species and 16 mamma-

lian species have been listed as potential vertebrate seed

dispersers and seed predators of the Kirindy Forest

(Ganzhorn et al. 1999). The fruit shells of S. madaga-

scariensis can probably only be opened by the two largest

species of lemurs, which either disperse (i.e. swallow

whole and defecate; Eulemur fulvus rufus; Fig. S1b,

Electronic Supplementary Material) or destroy (i.e. chew

and swallow; Propithecus verreauxi verreauxi) the seeds

(Ralisoamalala 1996). The fruits are mostly opened by

these lemurs in the fruiting trees and not all seeds included

in one fruit are eaten, but some seeds generally drop to the

ground in the process (either single, or still attached to a

half-opened fruit-shell). When fruits fall to the ground

unopened, they usually do not break open, but are soon

opened by rodents. If this is not the case, the fruits

decompose, spoiling the seeds inside. The presumably

introduced bushpig (Potamochoerus larvatus) also con-

sumes fallen fruits. Its impact on the seeds is unknown.

Seeds of S. madagascariensis are likely to attract different

kinds of seed dispersers and seed predators, vertebrates as

well as invertebrates, due to their comparatively large size

and the strong-smelling, orange, edible pulp that is firmly

attached to them (Fig. S1a–c, Electronic Supplementary

Material; Terborgh et al. 1993). The most likely vertebrate

seed predators are rodents [Hypogeomys antimena (body

mass 1,200 g) Eliurus myoxinus (80 g) and Macrotarsomys

bastardi (25–40 g); Goodman et al. 2003]. Based on sev-

eral thousand trap nights and supplementary trapping and

hair trap data collected in our study, none of the native

vertebrate seed dispersers or seed predators occur in small

forest fragments of the size considered in this study apart

from occasional visits by bushpigs (Dausmann 1997;

Ganzhorn et al. 1999; Ganzhorn 2003).

Experimental setup

The experiments were carried out during the rainy season

(January—March). This is within the natural period of seed

fall. Seeds were exposed to dispersers and predators by

placing them in three different types of size selective ex-

closure plots (Fig. 1): invertebrate accessible (IA), small

vertebrate accessible (SVA), and open control plots (CO).

The exclosures were constructed of fine wire 1 9 1-mm

mesh with four wooden stakes at the corners, 1 9 1 m in

size and 0.5 m in height. All experimental plots had a wire

mesh roof to keep out climbing animals and to exclude

naturally falling seeds. To prevent microclimatic changes,

the roofs were cleaned daily of twigs, leafs, and other litter.

The IA and the SVA exclosures had wire mesh at the four

sides. In the IA exclosures, the wire mesh was buried into

the ground and sewn to the roof, thus keeping out all

vertebrates. Invertebrates were given access to the IA ex-

closures by cutting holes of approximately 1 cm2 into the

wire mesh (100 per plot, distributed evenly across the wire

mesh). In the SVA exclosures the wire mesh ended about

2–3 cm above the ground and below the roof, making the

seeds available to rodents and other small animals, but not

to larger vertebrates. The base of the mesh was secured by

Fig. 1 Schematic diagram

showing the exclosures. IAInvertebrate accessible, SVAsmall vertebrate accessible, COopen control plots

Oecologia (2008) 157:307–316 309

123

additional stakes. The CO plots had a roof and were open

on all sides, thus allowing access to all possible seed dis-

persers and seed predators.

One of each type of experimental plot (IA, SVA, CO)

was installed within the crown perimeter below randomly

chosen mature fruiting trees of S. madagascariensis with

no conspecific within 10 m. The three types of plots were

assigned to cardinal directions (60�, 180�, and 300�) at

random under the different trees, but with the restriction

that each type was placed in each direction the same

number of times.

Exclosures were set up below six trees in the two types

of forest (continuous forest, forest fragments). Thus, there

were 18 experimental plots in the continuous forest (6 9 3)

and 18 in the forest fragments (2 9 3 9 3) altogether. All

experimental plots in the fragments were more than 5 m

away from the fragments’ edges.

Experimental procedure

Ripe fruits were collected outside the study area from

several parent trees and pooled to assure genetic variety.

The fruits were opened and the seeds checked individually.

Intact, apparently viable seeds were dried with the pulp

attached, all others were rejected. Twenty-five seeds were

placed in each of the exclosures. Seeds were spread out

evenly in five rows of five on the leaf litter, to allow for

individual monitoring of the seeds’ fate. The number of

seeds in each array mimicked the densities that are nor-

mally found under fruiting trees after visits by lemur

groups. The exclosures were censused daily for 7 days. The

length of the experiments had been determined in a pilot

study, which showed that there were no seeds left in most

of the CO plots after 4 days.

The fate of each seed was monitored individually, and

losses due to removal, damage by rodents, infestation by

insects and fungal growth were determined. In detail, first,

it was noted whether a seed was completely removed from

the exclosure or, in the IA exclosures, had been dragged

towards the wire mesh by ants (variable seed removal).

Second, for the remaining seeds it was noted whether the

pulp or the seed coat of the remaining seeds were partly or

completely removed or had tooth marks (number of

gnawed seeds). Third, it was noted whether a seed was

hollow, had interior parasites and/or entry/exit drilling

holes of insects (post-dispersal insect infestation). This was

done by dissecting the seeds that were not removed after

termination of each experiment. Fourth, it was noted

whether a seed showed signs of mould after termination of

the experiment. Furthermore, which animals were in con-

tact with the seeds at the time of the censuses was also

monitored. First, the number of seeds with ants in each plot

at the beginning of each experimental trial just after seeds

were set out (after 3 and 10 min, respectively) was recor-

ded. Second, the total number of ants in each plot at each

census, i.e. each day (number of ants/plot) was recorded.

Finally, the number of digging traces by H. antimena was

recorded each day at the SVA plots.

To correct for possible effects of weather conditions the

experiments were repeated 3 times, i.e., each experimental

plot was stocked 3 times with seeds (replicates in the fol-

lowing). Replicates were separated by at least 9 days to

avoid habituation of the animals to the exclosures as

feeding places. Data from the three replicates were aver-

aged, i.e. a second-order mean was calculated for each

experimental plot. Thus, for each experimental plot only

one data point was included in the analysis.

Identification of seed dispersers and seed predators

Hair traps were installed on the experimental plots in one

of the replicates in each habitat type. Hair traps consisted

of sticky tape stapled to the bottom of the wire mesh in the

SVA exclosures, so that the hair of passing animals stuck to

it (20 hair traps per plot). For the CO plots, the sticky tape

was fixed to a piece of wire 2–3 cm above the ground. The

collected hair samples were analysed under an electron

microscope (Zeiss DSM 962) and compared with our own

collection of reference hair samples (Fig. 2).

Additionally the following animals were tested for

whether they consumed seeds of S. madagascariensis in

the field camp: ants (Aphaenogaster swammerdami),

rodents (Macrotarsomys bastardi, E. myoxinus, Mus mus-

culus and Rattus rattus), lemurs (Cheirogaleus medius) and

tenrecs (Echinops telfairi, Setifer setosus and Tenrec

ecaudatus). Furthermore, the large lemurs E. fulvus rufus

and P. verreauxi verreauxi) were observed in the wild to

see whether or not they consumed the seeds.

Regeneration and establishment

To investigate whether differences in secondary seed dis-

persal and predation are reflected in the pattern of

regeneration and establishment, all individuals of S. mad-

agascariensis were located in 6 9 120-m (720 m2)

transects set up randomly, two in the continuous forest and

one in each of the two forest fragments. The plants were

assigned to three age classes: seedlings (circumference

B1 cm), young trees (circumference [1 cm to DBH

\20 cm), and mature trees (DBH C 20 cm).

Statistical analysis

It is difficult to differentiate experimentally between the

proportion of seeds that were dispersed and those that were

destroyed by seed predators (Barik et al. 1996). For the

310 Oecologia (2008) 157:307–316

123

analyses the seeds were thus classified into two categories:

‘‘remaining seeds’’ were those that were still found in the

exclosures and were, at least when inspected from the

outside, intact and thus were assumed to still be able to

germinate. The second category consisted of those seeds

that had been removed from the exclosures, i.e. those that

were dispersed or eaten, and those that still remained in the

exclosures, but had tooth marks, were infested by insects or

hollow and therefore presumably dead. Therefore, the term

‘‘seed loss’’ used in this paper includes predation as well as

transport and dispersal by animals, i.e. it includes the

variables seed removal, number of gnawed seeds, and

insect infestation.

Nonparametric tests were used for the analyses, because

the data did not meet the basic assumptions necessary for

parametric statistics. The comparison between data of one

tree (exclosure treatment, visiting rates of ants and rodents

on different days) were done by tests for dependent data

(Friedman-test with Wilcoxon–Wilcox post hoc compari-

sons; Wilcoxon-test). For other comparisons (continuous

forest vs. forest fragments) tests for independent data were

used (Mann–Whitney U-test). Nevertheless, to specify the

interaction of exclosure treatment and forest type, a two-

way ANOVA was performed with these two variables,

keeping in mind the statistical caveats. Frequencies were

compared with v2 goodness-of-fit tests. Wherever multiple

comparisons of data were necessary, Bonferroni’s correc-

tion of a was applied. Unless otherwise noted, analyses are

based on the number of seeds lost on the final day of

sampling.

Results

Seed loss in the continuous forest

Exclosure treatment had a highly significant effect on seed

loss rates in the continuous forest (Friedman: v2 = 12.0,

P \ 0.005; n = 6). There were significant differences

between all types of exclosures with seed loss rates

decreasing in the following order: CO plots [ SVA ex-

closures [ IA exclosures (Wilcoxon–Wilcox: P \ 0.01;

n = 6). In the CO plots up to 100% of the seeds were lost

within 7 days due to dispersal and seed predation (median:

90%; Fig. 3). Due to the small access holes in the IA ex-

closures, seeds could not be removed from these plots by

animals. Seed loss was therefore restricted to only very few

seeds, probably buried by beetles and other soil arthropods

(\1%). The fate of these could not be followed any further.

In the SVA exclosures seed loss was intermediate (63%).

Seed loss in the forest fragments

A different pattern of seed loss was revealed in the forest

fragments (Fig. 3; Friedman: v2 = 9.3, P \ 0.01; n = 6).

There was no difference in seed loss between the SVA

exclosures (65%) and the CO plots (73%) (Wilcoxon–

Wilcox: P [ 0.05; n = 6). Both differed significantly from

the IA exclosures where the seed loss was much lower

(6%) (P \ 0.01; n = 6).

Fig. 2 Hair structure of

Hypogeomys antimena (left) and

Eliurus myoxinus (right) under

the scanning electron

microscope

Fig. 3 Percentage of seeds of Strychnos madagascariensis that were

lost after 7 days in the continuous forest and in the forest fragments.

Values are medians, quartiles, minima and maxima (n = 6).

*P \ 0.05, **P \ 0.01. For abbreviations, see Fig. 1

Oecologia (2008) 157:307–316 311

123

When comparing seed loss between the fragments and

the continuous forest significant differences were found in

the IA exclosures and the CO plots. Seed loss was higher in

the forest fragments in the IA exclosures (Mann–Whitney

U: z = 2.6, P \ 0.01; n = 6), but lower in the CO plots

(z = 2.4, P \ 0.05; n = 6). There was no difference in the

SVA exclosures (z = 0.4, n.s.; n = 6) (Fig. 3).

A two-way ANOVA revealed a significant interaction

between exclosure treatment (IA, SVA, CO) and habitat

type (continuous forest vs. forest fragments) (df = 2;

F = 3.61; P = 0.039).

Dynamics of seed loss

The dynamics of seed loss differed between the continuous

forest and the forest fragments (Fig. 4). In both habitats the

largest numbers of seeds were lost on the first day of

sampling. In the continuous forest 43% of the total seed

loss had already taken place after 1 day. After that the seed

loss continued throughout the week, but to a much lesser

degree (on average 9.5% per day), matching the pattern of

increase in gnawed seeds (see below). In the forest frag-

ments, in contrast, there was hardly any more seed loss

after the initial period. Here, 97% of the total seed loss had

occurred during the first day. The pattern of seed loss in the

forest fragments thus resembled that part of seed loss in the

continuous forest that was attributed to the visits by ants

(see below).

Identification of different animal species

The results of the exclosure experiments suggest that

mainly rodents (H. antimena, E. myoxinus) were

responsible for predation of S. madagascariensis. This

assumption is supported by the feeding experiments and

the data obtained with the hair traps. Of the captive rodents

fed in camp E. myoxinus and M. bastardi ate the seeds.

They did, however, not hoard them, but ate them at once.

The other rodents took no notice of the seeds or only ate

the pulp. The lemur C. medius and the tenrecs ignored the

seeds completely. The lemurs E. fulvus rufus and P. ver-

reauxi verreauxi ate fruits and seeds in great amounts. It is

not known if the bushpig (P. larvatus) is involved in the

dispersal of large seeds in the Kirindy forest, but pigs’

digging traces were found twice in the CO plots. Moreover,

as seed loss in the CO plots was considerably higher than in

the SVA exclosures, it can be concluded that large verte-

brates also play an important role in the secondary

processes of seed dispersal and predation.

Thirty-one hair samples were collected in the contin-

uous forest and analysed. Of these, 21 could be assigned

to H. antimena, five to E. myoxinus, two to M. bastardi,

one to E. telfairi and one to S. setosus (Fig. 2). One hair

sample could not be identified. In the forest fragments

only three hair samples were found, two of which could

be assigned to E. telfairi and one that could not be

identified.

The only invertebrates that are able to move the seeds

of S. madagascariensis are the omnivorous ants A.

swammerdami (ca. 10 mm in size). These recruit strongly

in response to the seeds of S. madagascariensis: about 20

workers per seed transport them quickly into their

mounds (Fig. S1c, Electronic Supplementary Material).

This need not necessarily be the nearest ant nest and can

be up to 10 m away. The first ants arrived within less

than 1 min after the seeds had been placed, and tried to

carry the seeds away. After 3 min ants were found on up

to 18 out of 25 seeds (mean = 3.3, SD = 3.1, n = 36;

data from all experimental plots), and sometimes all 25

seeds were visited by ants after 10 min (mean = 9.1,

SD = 6.8, n = 36; data from all experimental plots). At

this time, usually several seeds were already transported

towards the ants’ nests, and the first seeds had often

already been deposited within the nests. Sometimes more

than 20 seeds were dragged to the wire mesh by ants in

the IA exclosures (mean = 5.0, SD = 0.78, n = 12, data

from all IA exclosures). An additional census in the

forest fragments 1 h after the beginning of the experiment

showed that between 72 and 100% of the total 7-day seed

loss had occurred in the SVA enclosures, and between 73

and 96% in the CO plots. As rodents were not active

during this time of day, the seed loss can be exclusively

attributed to A. swammerdami. About 4–6 days after the

deposition in the ant nests the seeds were carried to the

refuse pile with the pulp completely removed but other-

wise intact.

Fig. 4 Temporal patterns of seed loss over the course of the 7-day

experiments in the CO plots in the forest fragments (open squares)

and in the continuous forest (black squares). The cumulative seed loss

for each day is shown as a percentage of the total seed loss after

7 days for each plot (mean ± SD; n = 6)

312 Oecologia (2008) 157:307–316

123

The number of A. swammerdami that was found in

contact with the seeds during the daily censuses changed

considerably over the course of the 7 days of the experi-

ments (Fig. 5). The visiting rates were high during the first

2 days and declined significantly afterwards (Friedman:

v2 = 139.0, P \ 0.0001; n = 18, pooled data from all

plots in the continuous forest). In contrast, the increase in

gnawed seeds, and thus the visiting rates of rodents, was

fairly constant throughout the whole week [v2 = 4.7; not

significant (n.s.); n = 12, data from the SVA and CO plots

in the continuous forest], as was the occurrence of digging

traces of H. antimena (v2 = 2.6; n. s.; n = 6, data from the

SVA plots in the continuous forest).

Insect infestation and fungal growth

The rate of insect infestation (seeds not removed that had

interior parasites or entry/exit holes) was very low and

amounted to only 3.8% in the continuous forest (n = 732

seeds) and 2.7% in the forest fragments (n = 407). No

significant difference between the two habitat types was

found (v2 goodness-of-fit: v2 = 0.9; n.s.; df = 1).

Seeds from which the pulp had been partly removed by

ants and other invertebrates showed significantly less ten-

dency to mould than those where the pulp had been left

intact (v2 goodness-of-fit: v2 = 37.4; P \ 0.001; n = 450;

df = 1). In no case did a seed from which the pulp had

been completely removed start to mould.

Regeneration

Seedlings and young trees were more abundant in the

fragments than in the continuous forest, regardless of

whether the numbers are presented in absolute terms or as

individuals per mature tree (Fig. 6).

Discussion

In the tropics the majority of tree species are adapted for

seed dispersal by vertebrates (Willson et al. 1989). This has

led to substantial efforts to elucidate in particular the role

of vertebrate seed dispersal for forest regeneration and for

the maintenance of tropical forest diversity (Howe and

Smallwood 1982; Dirzo and Dominguez 1986; Chapman

1989; Terborgh et al. 1993; Chapman and Chapman 1995).

In view of the accelerating process of forest fragmentation

special emphasis has been placed on studying conse-

quences of fragmentation for forest regeneration in the

absence of primary seed dispersers. Undispersed seeds

inevitably fall beneath parent trees, where survival is typ-

ically lower (Howe et al. 1985) and most of these studies

are consistent with the expectation that the loss of verte-

brate seed dispersers results in reduced recruitment of

animal-dispersed trees in forest fragments (e.g. contribu-

tion to Laurance and Bierregaard 1997; Cordeiro and Howe

2001 and references therein). Especially in the dry decid-

uous forest of western Madagascar the very close

associations between trees and lemurs as primary dispers-

ers suggest substantial disadvantages for the successful

recruitment of trees in forests without lemurs.

However, these considerations fail to include the pos-

sibility that not only seed dispersers but also seed predators

are affected by fragmentation. Rodents have been identified

as important seed predators (Du Puy 1996). It is not known

to what extent the rodents of the dry forest also hoard intact

seeds and thus contribute to seed dispersal and recruitment

as can be observed in many other parts of the world,

including Madagascar (Goodman and Sterling 1996).

However, native as well as introduced rodent species are

thought to be mainly seed predators and not dispersers

(Bohning-Gaese et al. 1999; Forget et al. 1999). This

Fig. 5 Comparison of the presence of ants vs. rodents in the

experimental plots. Number of seeds lost to predation by rodents

per day (black squares ± SE; n = 12) and visiting rates of ants

(white squares ± SE; n = 12)Fig. 6 Abundance of S. madagascariensis in two 6 9 120-m

(720 m2) transects in continuous forest and two forest fragments

Oecologia (2008) 157:307–316 313

123

applies especially to the seeds of S. madagascariensis

which are surrounded by juicy pulp. Seeds like these are

not suitable for hoarding, as are dry seeds (Strasburger

et al. 1991). The substantial degree of seed removal by

rodents of S. madagascariensis in the continuous forest was

thus presumably equivalent to seed predation. This

assumption is supported by the high proportion of seeds

with tooth marks in the experimental plots, the gradual loss

of seeds each day, and the feeding experiments. In many

studies the rodent community increases following frag-

mentation due to predator release. However, in our study

rodents were also missing in the forest fragments, and seed

predation by rodents was therefore non existent. With

respect to seed survival, fragmentation with the associated

species reduction does therefore have positive, rather than

negative consequences for S. madagascariensis. Especially

as yet another species plays a role: the ant Aphaenogaster

swammerdami.

Ants are known to be important secondary dispersers in

the Malagasy dry forest (Randrianasolo et al. 1996; Boh-

ning-Gaese et al. 1999). A study comparing the role of ants

in the secondary dispersal of seeds from Commiphora spp.

in South Africa and Madagascar demonstrated a much

more important role of ants in the latter (Voigt et al. 2002).

About 48% of the seeds of Commiphora were dispersed

secondarily by ants in Madagascar while none were dis-

persed in South Africa. In the Kirindy Forest/CFPF this can

only be achieved by A. swammerdami, as they are the only

ant species that are able to move the seeds of S. madaga-

scariensis. The ants find and transport the seeds very

quickly. They carry them into their mounds and deposit

them there before they are found by rodents, so that at least

a proportion of the seeds are protected from seed predation.

As rodents orientate themselves mainly by odours, the

seeds are harder to detect and less attractive to rodents after

deposition on the ants’ refuse pile, with the pulp removed

but otherwise intact (Ruhren and Dudash 1996; Spehn and

Ganzhorn 2000). Therefore, seeds dispersed by A. swam-

merdami are not only moved from directly under the parent

tree, reducing clumping and parent tree–seedling conflicts,

but should also be exposed to a substantially reduced

predation risk. As S. madagascariensis does not seem to

have seed dormancy and the seeds usually germinate after

only 2–3 weeks, 1 week in the safety of an ants’ nest

represents a considerable portion of the time until

germination.

Lemur-ingested and defecated seeds are less attractive to

rodents and suffer lower mortality due to mould and insect

damage once the pulp has been removed than seeds that

have not passed through the digestive tract (Spehn and

Ganzhorn 2000). However, this additional advantage of

lemur dispersal can also be counterbalanced by the dis-

persal by A. swammerdami, because the removal of the

pulp in the ants’ mound also makes the seeds unsusceptible

to mould. Gut passage is not essential for successful ger-

mination in S. madagascariensis, as is typical for species

normally dispersed by nonvolant mammals (Traveset and

Verdu 2002).

In general, larger-seeded species, such as S. madaga-

scariensis, suffer greater reduction in dispersal once larger

vertebrates are decimated (Stoner et al. 2007). In the case

of S. madagascariensis, however, this disadvantage is

counteracted by reduced seed predation by rodents and

dispersal by ants. Even though most seeds were only dis-

persed a couple of metres, some were transported over

considerable distances. Greater dispersal distances by

vertebrates compared to ants is presumably not a pro-

nounced advantage for S. madagascariensis, due to its

relative abundance. On average, any dispersal over more

than 10 m (which is also achieved by the dispersal by A.

swammerdami) will only get the seed closer to the next

individual of S. madagascariensis. For the parent tree in a

stable habitat, dispersal to as many different sites as pos-

sible is often advantageous, because this maximizes the

chance that the seeds reach randomly dispersed favourable

sites for germination (Howe 1984; Whelan et al. 1990).

Mean dispersal distances are then of minor relevance and

even a small number of seeds that are transported over

longer distances could be disproportionally important for

recruitment (Andersen 1988; Burkey 1994).

Moreover, in a more intense competitive environment

where overall seed and seedling survival at the community

level is enhanced, large-seeded species are at an advantage.

In the continuous forest seeds of S. madagascariensis are

subjected to intense predation, contrary to the fragments,

where they are very successful, as shown by the increase in

seedling abundance in them. The change in seed survival

therefore seems to be carried over into changed seedling

abundance. This finding also shows, that, contrary to other

studies (e.g. Dolch and Tscharntke 2000; Harms et al.

2000), microbial and invertebrate natural enemies, or other

factors, did not cause negative density dependence in S.

madagascariensis, at least not at the stage of seed–seedling

transition.

The proportion of S. madagascariensis seeds transported

by ants can only be estimated, but it seems to be substantial

considering the behaviour of the ants during the censuses,

especially directly after we placed the seeds out, and the

large number of seeds that were dragged to the wire mesh

in the IA exclosures. The importance of ants is supported

by the observation that most of the high seed loss during

the first days was due to ants and not to rodents. While

removal of seeds by ants was concentrated in the first 2

days (while the pulp is still fresh), seed consumption by

rodents was much more evenly spread throughout the

week. Besides this, in the forest fragments which contain

314 Oecologia (2008) 157:307–316

123

many ants but lack a rodent community, the seed loss in the

first 2 days was at least as high as in the continuous forest.

In other parts of the world dispersal of tree seeds by ants

has generally been reported only rarely. In Kirindy/CFPF,

however, this phenomenon has already been observed in

different families in all vertebrate-dispersed tree species

investigated so far (Bohning-Gaese et al. 1999, Bursera-

ceae; Spehn and Ganzhorn 2000, Malvaceae; our study,

Loganiaceae). It remains to be clarified whether the

important role of ants in the dispersal of tree species

classified conventionally as ‘‘vertebrate-dispersed trees’’ is

a peculiarity of the dry forest ecosystem or whether ants

assume the role of seed dispersers in a faunal community

that is depauperate in terms of frugivorous vertebrates.

Conclusion

Practically no seed predation in forest fragments due to the

lack of granivorous rodents combined with the increased

activity of ants counterbalances the disadvantage of the

lack of lemurs as primary dispersers and leads to an

increase in S. madagascariensis abundance, possibly at the

expense of other trees. This in turn might eventually result

in an impoverished tree species community. While, in

general, fragmentation is considered to have negative

effects on the regeneration of native vertebrate-dispersed

tree species, the present study provides an example of the

reverse due to flexible animal-plant interactions in different

facets.

Acknowledgements We are grateful to the Commission Tripartite

of the Malagasy Government, the Laboratoire de Primatologie et des

Vertebres de l’Universite d’Antananarivo and the Ministere pour la

Production Animale and the Department des Eaux et Forets for per-

mits to work in Madagascar. We also thank the Centre de Formation

Professionelle Forestiere de Morondava for their hospitality and

permission to work on their concession. The late B. Rakotosamima-

nana, R. Rasoloarison and L. Razafimanantsoa supported the field

project in numerous ways. Aid from the Deutsche Forschungsgeme-

inschaft (Ga 342/3-1, 3-2) and the German Primate Centre is

gratefully acknowledged. All experiments comply with the laws of

Madagascar.

References

Andersen AN (1988) Dispersal distance as a benefit of myrmecoch-

ory. Oecologia 75:507–511

Asquith NM, Wright S, Clauss MJ (1997) Does mammal community

composition control recruitment in neotropical forests? Evidence

from Panama. Ecology 78:941–946

Barik SK, Tripathi RS, Pandey HN, Rao P (1996) Tree regeneration

in a subtropical humid forest: effect of cultural disturbance on

seed production, dispersal and germination. J Appl Ecol

33:1551–1560

Birkinshaw C (2003) Lemur dispersed trees. In: Goodman SM,

Benstead JP (eds) The natural history of Madagascar. University

of Chicago Press, Chicago

Bleher B, Bohning-Gaese K (2001) Consequences of frugivore

diversity for seed dispersal, seedling establishment and the

spatial pattern of seedlings and trees. Oecologia 129:385–394

Bohning-Gaese K, Gaese BH, Rabemanantsoa SB (1999) Importance

of primary and secondary seed dispersal in the Malagasy tree

Commiphora guillaumini. Ecology 80:821–832

Bollen A, van Elsacker L (2002) Feeding ecology of Pteropus rufus(Pteropodidae) in the littoral forest of Sainte Luce, SE Mada-

gascar. Acta Chiropt 4:33–47

Bollen A, Donati D, Fietz J, Schwab D, Ramanamanjato J-B,

Randrihasipara L, van Elsaker L, Ganzhorn JU (2005) Fruit

characteristics in a dry deciduous and a humid littoral forest of

Madagascar. In: Dewand L, Boubli JP (eds) Fruits and Frugi-

vores. Kluwer, New York, pp 92–118

Burkey TV (1994) Tropical tree species diversity: a test of the

Janzen–Connell model. Oecologia 97:533–540

Cascante A, Quesada M, Lobo JJ, Fuchs EA (2002) Effects of dry

tropical forest fragmentation on the reproductive success and

genetic structure of the tree Samanea saman. Conserv Biol

16:137–147

Chapman CA (1989) Primate seed dispersal: the fate of dispersed

seeds. Biotropica 21:148–154

Chapman CA, Chapman LJ (1995) Survival without dispersers:

seedling recruitment under parents. Conserv Biol 9:675–678

Cordeiro NJ, Howe HF (2001) Low recruitment of trees dispersed by

animals in African forest fragments. Conserv Biol 15:1733–1741

Dausmann KH (1997) Sekundare Samenausbreitung und Samenpra-

dation im westmadagassischen Trockenwald—Das Schicksal

Lemuren-ausgebreiteter Samen. Diploma thesis, Bayerische

Julius-Maximilians-Universitat, Wurzburg

Didham RK, Ghazoul J, Stork NE, Davis AJ (1996) Insects in

fragmented forests: a functional approach. Trends Ecol Evol

11:255–260

Dirzo R, Dominguez CA (1986) Seed shadows, seed predation and

the advantages of dispersal. In: Estrada A, Fleming TH (eds)

Frugivores and seed dispersal. Junk, Dordrecht, pp 237–249

Dolch R, Tscharntke T (2000) Insect communities of Malagasy legumes

(Fabaceae): patterns of species richness, resource fragmentation,

and tropical-temperate differences. In: Lourenco WR, Goodman

SM (eds) Diversity and endemism in Madagascar. Memoires de la

Societe de Biogeographie, Paris, pp 285–293

Du Puy B (1996) Faunal interactions with the genus Adansonia in the

Kirindy Forest. In Ganzhorn JU, Sorg JP (eds) Ecology and

economy of a tropical dry forest in Madagascar. Primate Report,

46-1, Gottingen, pp 329–334

Emmons LH (1989) Tropical rain forests: why they have so many

species and how we may lose this biodiversity without cutting a

single tree. Orion 8:8–14

Fahrig L (2003) Effects of habitat fragmentation on biodiversity.

Annu Rev Ecol Evol Syst 34:487–515

Forget P-M, Kitajima K, Foster RB (1999) Pre- and post-dispersal

seed predation in Tachigali versicolor (Caesalpiniaceae): effects

of timing of fruiting and variation among trees. J Trop Ecol

15:61–81

Ganzhorn JU (2003) Effects of introduced Rattus rattus on endemic

small mammals in dry deciduous forest fragments of western

Madagascar. Anim Conserv 6:147–157

Ganzhorn JU, Sorg J-P (1996) Ecology and economy of a tropical dry

forest, vol 46-1. Primate report, special volume. Gottingen

Ganzhorn JU, Rakotosamimanana B, Hannah L, Hough J, Iyer L,Olivieri S, Rajaobelina S, Rodstrom C, Tilkin G (1997) Priorities

for biodiversity conservation in Madagascar. Primate Report 48-

1:1–81

Ganzhorn JU, Fietz J, Rakotovao E, Schwab D, Zinner D (1999)

Lemurs and the regeneration of dry deciduous forest in

Madagascar. Conserv Biol 13:794–804

Oecologia (2008) 157:307–316 315

123

Ganzhorn JU, Lowry PI, Schatz G, Sommer S (2001) The biodiversity

of Madagascar: one of the world’s hottest hotspots on its way

out. Oryx 35:346–348

Ganzhorn JU, Goodman SM, Deghan A (2003) Effects of fragmen-

tation on small mammals and lemurs. In: Goodman SM,

Benstead JP (eds) The natural history of Madagascar. University

of Chicago Press, Chicago, pp 1228–1234

Goodman SM, Ganzhorn JU (1997) Rarity of figs (Ficus) on

Madagascar and its relationship to a depauperate frugivore

community. Rev Ecol 52:321–329

Goodman SM, Sterling EJ (1996) The utilization of Canarium(Burseraceae) seeds by vertebrates in the RNI d’Andringitra,

Madagascar. In: Goodman SM (ed) A floral and faunal inventory

of the eastern side of the Reserve Naturelle Integrale d’Andr-

ingitra, Madagascar: with reference to elevational variation.

Fieldiana: zoology edition, volume 85. Field Museum Natural

History, Chicago, pp 83–89

Goodman SM, Ganzhorn JU, Rakotondravony D (2003) Mammals. In:

Goodman SM, Benstead JP (eds) The natural history of Mada-

gascar. University of Chicago Press, Chicago, pp 1159–1186

Hannah L, Rakotosamimanana B, Ganzhorn JU, Mittermeier RA,

Olivieri S, Iyer L, Rajaobelina S, Hough J, Andriamialisoa F,

Bowles I, Tilkin G (1998) Participatory planning, scientific

priorities, and landscape conservation in Madagascar. Environ

Conserv 25:30–36

Harms KE, Wright SJ, Calderon O, Hernandez A, Herre EA (2000)

Pervasive density-dependent recruitment enhances seedling

diversity in a tropical forest. Nature 404:493–495

Howe HF (1984) Implications of seed dispersal by animals for

tropical reserve management. Biol Conserv 30:261–281

Howe HF, Smallwood J (1982) Ecology of seed dispersal. Annu Rev

Ecol Syst 13:201–228

Howe HF, Schupp EW, Westley LC (1985) Early consequences of

seed dispersal for a neotropical tree (Virola surinamensis).

Ecology 66:781–791

Janzen DH (1988) Tropical dry forests: the most endangered major

tropical ecosystem. In: Wilson EO (ed) Biodiversity. National

Academy Press, Washington, pp 130–137

Klein BC (1989) Effects of forest fragmentation on dung and carrion

beetle communities in Central Amazonia. Ecology 70:1715–

1725

Langrand O (1990) Guide to the birds of Madagascar. Yale University

Press, New Haven

Laurance WF, Bierregaard RO (1997) Tropical forest remnants.

University of Chicago Press, Chicago

Leeuwenberg AJM (1984) Famille 167 - Loganiacees. In: Humbert H

(ed) Flore de Madagaskar et des Comores. Museum National

d’Histoire Naturelle, Paris

Leigh EG Jr, Wright SJ, Herre EA, Putz FE (1993) The decline of tree

diversity on newly isolated tropical islands: a test of a null

hypothesis and some implications. Evol Ecol 7:76–102

Lerdau M, Whitbeck J, Holbrook NM (1991) Tropical deciduous

forest: death of a biome. Trends Ecol Evol 6:201–202

Peres CA, Palacios E (2007) Basin-wide effects of game harvest on

vertebrate population densities in Amazonian forests: implica-

tions for animal-mediated seed dispersal. Biotropica 39:304–315

Ralisoamalala RC (1996) Role de Eulemur fulvus rufus (Audeberg,

1799) et de Propithecus verreauxi verreauxi (A. Grandidier

1867) dans la dissemination des graines. In: Ganzhorn JU, Sorg

J-P (eds) Ecology and economy of a tropical dry forest in

Madagascar. Primate Report, 46-1, Gottingen, pp 285–293

Randrianasolo J, Rakotovao P, Deleporte P, Rarivoson C, Sorg J-P,

Rohner U (1996) Local tree species in the tree nursery. In:

Ganzhorn JU, Sorg J-P (eds) Ecology and economy of a tropical

dry forest in Madagascar. Primate Report 46-1, Gottingen, pp

117–132

Redford KH (1992) The empty forest. Bioscience 42:412–422

Ruhren S, Dudash MR (1996) Consequences of the timing of seed

release of Erythronium americanum (Liliaceae), a deciduous

forest myrmecochore. Am J Bot 83:633–640

Smith AP (1997) Deforestation, fragmentation, and reserve design in

western Madagascar. In: Laurance WF, Bierregaard RO (eds)

Tropical forest remnants: ecology management and conservation

of fragmented communities. University of Chicago Press,

Chicago, pp 415–441

Smith JNM, Hellmann JJ (2002) Population persistence in fragmented

landscapes. Trends Ecol Evol 17:397–399

Sorg JP, Ganzhorn JU, Kappeler PM (2003) Forestry and research in

the Kirindy Forest/CFPF. In: Goodman SM, Benstead JP (eds)

The natural history of Madagascar. University of Chicago Press,

Chicago, pp 1512–1519

Spehn S, Ganzhorn JU (2000) Influence of seed dispersal by Brown

Lemurs on removal rates of three Grewia species (Tiliaceae) in

the dry deciduous forest of Madagascar. Ecotropica 6:13–21

Steffan-Dewenter I, Tscharntke T (1999) Effects of habitat isolation

on pollinator communities and seed set. Oecologia 121:432–440

Stoner KE, Vulineck Wright SJ, Peres CA (2007) Hunting and plant

community dynamics in tropical forests: A synthesis and future

directions. Biotropica 39:385–392

Strasburger EB, Sitte P, Ziegler H, Ehrendorfer F, Bresinsky AB

(1991) Strasburger Lehrbuch der Botanik. Fischer, Stuttgart

Terborgh J (1992) Maintenance of diversity in tropical forests.

Biotropica 24:283–292

Terborgh J, Losos E, Riley MP, Bolanos Riley M (1993) Predation by

vertebrates and invertebrates on the seeds of five canopy species

of an Amazonian forest. In: Fleming TH, Estrada A (eds)

Frugivory and seed dispersal: ecological and evolutionary

aspects. Kluwer, Dordrecht, pp 375–386

Terborgh J, Lopez L, Nunez P, Rao M, Shahabuddin G, Orihuela G,

Riveros M, Ascanio R, Adler GH, Lambert TD, Balbas L (2001)

Ecological meltdown in predator-free forest fragments. Science

294:1923–1926

Terborgh J, Feeley K, Silman M, Nunez P, Balukjan B (2006)

Vegetation dynamics of predator-free land-bridge islands. J Ecol

94:253–263

Traveset A, Verdu M (2002) A meta-analysis of the effect of gut

treatment on seed germination. In: Levey DJ, Silva WR, Galetti

M (eds) Seed dispersal and frugivory: ecology evolution and

conservation. CABI International, Wallingford, pp 339–350

Voigt FA, Burkhardt JF, Verhaagh M, Bohning-Gaese K (2002)

Regional differences in ant community structure and conse-

quences for secondary dispersal of Commiphora seeds.

Ecotropica 8:59–66

Whelan CJ, Willson MF, Tuma CA, Souza-Pinto I (1990) Spatial and

temporal patterns of postdispersal seed predation. Can J Bot

69:428–436

Willson MF, Irvine AK, Walsh NG (1989) Vertebrate dispersal

syndromes in some Australian and New Zealand plant commu-

nities with geographic comparisons. Biotropica 21:133–147

Wright SJ (2003) The myriad consequences of hunting for vertebrates

and plants in tropical forests. Perspect Plant Ecol Evol Syst

6:73–86

Wright SJ, Stoner KE, Beckman N, Corlett RT, Dirzo R, Muller-

Landau HC, Nunez-Iturri G, Peres CA, Wang BC (2007) The

plight of large animals in tropical forests and the consequences

for plant regeneration. Biotropica 39:289–291

316 Oecologia (2008) 157:307–316

123