Patterns of seed dispersal at a variety of scales in a tropical forest system

MISSING IMAGE

Material Information

Title:
Patterns of seed dispersal at a variety of scales in a tropical forest system do post-dispersal processes disrupt patterns established by frugivores?
Physical Description:
ix, 114 leaves : ill. ; 29 cm.
Language:
English
Creator:
Balcomb, Sophia Robb
Publication Date:

Subjects

Subjects / Keywords:
Seeds -- Dispersal -- Uganda -- Kibale Forest Reserve   ( lcsh )
Frugivores -- Behavior -- Uganda -- Kibale Forest Reserve   ( lcsh )
Forest ecology -- Tropics   ( lcsh )
Nutmeg tree -- Seedlings -- Uganda -- Kibale Forest Reserve   ( lcsh )
Zoology thesis, Ph.D   ( lcsh )
Dissertations, Academic -- Zoology -- UF   ( lcsh )
Genre:
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Thesis:
Thesis (Ph.D.)--University of Florida, 2001.
Bibliography:
Includes bibliographical references (leaves 101-111).
Statement of Responsibility:
by Sophia Robb Balcomb.
General Note:
Printout.
General Note:
Vita.

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 026974288
oclc - 47243261
System ID:
AA00020444:00001

Full Text










PATTERNS OF SEED DISPERSAL AT A VARIETY OF SCALES IN A TROPICAL
FOREST SYSTEM: DO POST-DISPERSAL PROCESSES DISRUPT PATTERNS
ESTABLISHED BY FRUGIVORES?















By

SOPHIA ROBB BALCOMB


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA


2001




















To my mother, Victoria Balcomb,
who instilled her love for travel and natural history in me as a child,

and to my husband, James Vonesh,
whose love for travel, outdoors, and science, helps keep me going.














ACKNOWLEDGMENTS

I would like to thank my advisor, Colin Chapman, for his encouragement, advice,

patience, support, and generosity throughout this process. He helped me realize a dream

I had always had of returning to East Africa to conduct scientific research in the field.

Over the years we have been through a lot and, in addition to his invaluable help in all

stages of this research project, from the initial proposal development to the final writing

stage, I thank him for his flexibility in dealing with my forthright nature. I also thank my

committee members Lauren Chapman, Doug Levey, Carmine Lanciani, and Sue Boinski,

for their help in this project, their insightful comments and suggestions, and their

encouragement. Many thanks also go to Ben Bolker whose support and advice regarding

statistical matters were invaluable, and Joanna Lambert, whose seed dispersal studies in

Kibale helped pave the way for mine.

This research would not have been possible without the invaluable field assistance

from Peter Adyeeri Irumba with whom I spent endless hours and days walking up and

down hills, counting seeds, and measuring plants. I also thank him for his support

through the long, rainy days sloughing through the mud, and for his humor. Katusabe

Swaibu and Chris Amooti also provided support in data collection when I would have

given anything to have been able to be in 10 places at once and instead, to my great relief,

realized I could rely on these great Ugandans. In addition, I would never have known

which tree species I was looking at without Tusiimi Atwooki's patient instructions on

how to tell one round green leaf from another.








Crucial financial support for my fieldwork came from a Dissertation Improvement

Grant from the National Science Foundation (SBR 9628055). Additional funding for my

research in the field came from Sigma Xi, Association for Women in Science Educational

Foundation, and a number of sources from the University of Florida including the

Department of Zoology, Center for African Studies, Office of Research, Technology and

Graduate Education, and the College of Liberal Arts and Sciences. Further support from

the University of Florida during the writing stage came from the Association for

Academic Women at the University of Florida and a Dissertation Fellowship through the

College of Liberal Arts and Sciences. Colin and Lauren Chapman contributed logistic

support in the field, including field equipment and, most importantly, the use of their

Hilux truck, without which I would not have been able to conduct my field research at the

Dura site with such reliability. I am most grateful for their generosity in allowing me to

add wear and tear to their vehicles in the name of science (and my sanity).

Permission to conduct this research was given by the Office of the President,

Uganda, the Uganda Wildlife Authority, and the Makerere University Biological Field

Station. I would also like to thank the staff at the field station for their logistical support

during my stay, Jimmy in the post-office for conveying all my faxes back and forth to

Colin, and the women in the market and Andrew for their cheerful greetings every time I

came to town.

Many graduate students in zoology and botany provided critical support, advice,

and good times throughout this process. I would like to thank Amy Zanne for making me

stick to plants at the beginning of this project, Susan Moegenberg for all her

encouragement and support in the final year, John Paul and Scot Duncan for feedback on








trees and other Kibale matters, Nat Seavy and Chrissy Apodaca for many enjoyable

drives down to the Dura site together, and Manjula Tiwari, April Randle, Ben Miner,

Becca Hale, Daphne Onderdonk, Suhel Quader, and countless others who have helped

me in innumerable ways. I thank them all for scientific discussions, editing proposals

and manuscripts, statistical advice, feedback on oral presentations, and conversations

regarding both serious and fun aspects of life as a graduate student, and hopes and plans

for our futures. Thanks to everyone who came out for regular pick-up soccer games,

which helped me keep my sanity during data analysis and writing up.

Finally, I would like to thank my immediate family for their continual support in

my pursuing this degree. In particular, I would like to thank my mother for her continued

interest in my research projects, her concern with my well-being in the field, expressed

through many care packages and cards, and her more than generous financial

contributions, particularly upon my return to Florida. I would to thank the Vonesh family

for their support; their generous contribution of a computer facilitated the writing of this

dissertation, for which I am grateful. And last, but certainly not least, I would like to

thank my husband and academic colleague, James Vonesh, who helped with innumerable

aspects of every stage of this dissertation, who provided constant friendship, support, and

encouragement during the stressful, and the fun, times, and whose belief in me has helped

carry me through this process.














TABLE OF CONTENTS

page

ACKN OW LEDGM EN TS ................................................................................................. in

ABSTRACT.....................................................................................................................viii

CHAPTERS

1. GENERAL INTRODUCTION ....................................................................................... 1

2. BRIDGING THE SEED DISPERSAL GAP: DOES SEED DISPERSAL
DETERMINE PATTERNS OF TREE SEEDLING RECRUITMENT IN A PRIMATE-
PLAN T INTERA CTION ? ...................................................................................................7

Introduction ..................................................................................... ........................ 7
Study Sites .................................................................................................................... 13
Study Species................................................................................................... .. . 14
M methods ......................................................................................................................... 18
Fruit Rem oval ........................................................................................................... 18
Fate of Dropped Fruits .......................................... ........................ ......................... 20
Post-Deposition Fate of Seeds Through to Seedling Survival .................................. 22
Plant Size-Class Densities...................................................................................... 25
Fate of Natural Seedlings and Saplings................................................................... 25
Frugivorous Prim ate Densities ............................................................................... 26
Statistical Analyses................................................................................................... 27
Overall Probability of Dispersal and Seedling Recruitment................................. 29
Results........................................................................................................................... 32
Prim ate Density ........................................................................................................32
Plant Fecundity ......................................................................................................... 33
Fruit Rem oval ........................................................................................................... 33
Im m ature fruits ...................................................................................................... 33
M ature fruits.......................................................................................................... 34
Frugivore Feeding Behavior..................................................................................... 35
Fate of Seeds from Dropped Fruits ........................................................................... 36
Fate of Spit Seeds...................................................................................................... 38
Fate of Seeds in Dung ............................................................................................... 39
Seedling Survival from Spit Seeds........................................................................... 40
Seedling Survival from Seeds in Dung ..................................................................... 41
Natural Seedling Density and Survival ..................................................................... 42








Overall Probability of Dispersal and Seedling Recruitment.................................... 44
D iscussion..................................................................................................................... 47
Effect of Changes in Frugivore Abundance on Seedling Abundance ...................... 50
Ecological Im plications ............................................................................................ 52
Evolutionary Im plications.......................................... .............................................. 55
Conclusion ................................................................................................................ 57

3. DO FRUGIVORES DETERMINE SPATIAL DISPERSION PATTERNS OF
TROPICAL FOREST TREES?: A COMPARISON BETWEEN TWO SITES.........76

Introduction............................................................................................................... .... 76
Study Sites .................................................................................................................... 80
Study Species ........................................................ ........................................................ 81
M methods ................................................................................................... 83
Results........................................................................................................................... 86
A ggregation Patterns................................................................................................. 86
Densities.................................................................................................................... 87
Discussion..................................................................................................................... 88

4. SU M M ARY AN D CON CLU SION S ...................................................... ..................... 97

LIST OF REFEREN CES.................................................................................................101

BIOGRAPHICAL SKETCH ....................................................... .................... ............... 112














Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

PATTERNS OF SEED DISPERSAL AT A VARIETY OF SCALES IN A TROPICAL
FOREST SYSTEM: DO POST-DISPERSAL PROCESSES DISRUPT PATTERNS
ESTABLISHED BY FRUGIVORES?

By

Sophia Robb Balcomb

May 2001
Chairman: Dr. Colin Chapman
Major Department: Zoology

The ubiquity of fleshy-fruited plants in the tropics suggests that animal-mediated

seed dispersal is important for plant recruitment. By creating the initial spatial template

of seeds, frugivores are thought to influence recruitment and spatial dispersion patterns in

all subsequent plant life history stages. However, plant establishment depends on a series

of steps influenced not only by seed dispersal but also by post-dispersal processes, such

as seed survival, seedling establishment, and recruitment into juvenile and adult stages.

Thus, it is possible that initial patterns established by frugivores are largely erased by

subsequent processes. Few studies have studied successive life-history stages of animal-

dispersed plants to determine if initial patterns established by frugivores persist to later

stages. I addressed this issue by quantifying patterns of dispersal and recruitment in a

tropical forest in Kibale National Park, Uganda. First, to better understand the link

between frugivore behavior and subsequent seedling recruitment, I examined seed

dispersal and seedling establishment of Monodora myristica, a tree species whose fruit








morphology suggests it is dispersed by only the largest arboreal frugivores. I found that

large-bodied primates played a critical role in initial stages of M myristica seed dispersal

by opening the hard-husked fruits and dispersing the large seeds away from parent plants.

However, spatial and temporal variation in post-dispersal processes, such as seed

predation, subsequently influenced recruitment patterns. At some spatial and temporal

scales, this variation in post-dispersal processes reduces the predictability of frugivore

actions on seedling recruitment. Second, to assess whether patterns of plant dispersion

established by frugivores persist to older stages, I examined whether changes in pattern of

dispersion from seedlings to trees differed between stands for six animal-dispersed

species. I found that, for some species, spatial patterns across size classes were similar

between stands, indicating that processes likely operate similarly between stands.

However, in other species, patterns differed between stands, indicating that initial spatial

patterns may be disrupted in later stages. Thus, similar to the first study's findings,

spatial and temporal variation in processes operating at later stages may result in little

concordance between frugivore seed deposition and plant recruitment and spatial

dispersion patterns.













CHAPTER 1
GENERAL INTRODUCTION

In tropical forests, 70-90% of tree and woody shrub species produce fleshy fruits

with morphologies that suggest they are adapted for avian and mammalian dispersal

(Frankie et al. 1974, Howe and Smallwood 1982, Jordano 1992). Frugivores that

consume the pulp and spit out, regurgitate, defecate, or drop seeds of these plants set the

stage for all subsequent demographic processes (Schupp and Fuentes 1995). Thus, seed

dispersers are thought to play an important role in plant population dynamics by creating

the initial spatial arrangement of seeds.

Although many studies conclude that frugivores are important for seed dispersal

of particular plant species, most have examined only one or two stages in the plant

recruitment process and have not connected early stages with recruitment into later life

history stages (Garber 1986, Stevenson 2000). Thus, few studies have examined both the

fate of dispersed seeds and the likelihood that those individual seeds will recruit into

seedlings, juveniles, and adults (Herrera et al. 1994, Schupp and Fuentes 1995).

Studies have demonstrated that foraging patterns of frugivores can influence the

distribution patterns of seeds (Kalina 1988, Lambert 1999). Increased survival of seeds

deposited in particular locations by dispersers indicates that frugivore deposition patterns

can influence local seedling composition (Tutin et al. 1991, Pacheco and Simonetti 1998,

Rogers et al. 1998, Wenny and Levey 1998, Wenny 2000). For example, gorillas

(Gorilla gorilla) in the Lope Reserve, Gabon, deposit dense clumps of seeds of many tree

species at their ground sleeping nests (Rogers et al. 1998). In a study of six of these








species, three were found to have increased seedling survival and growth from these nest

sites compared with dung depositions along trails and seeds spat out by other consumers

(Voysey et al. 1999). Studies such as these often extrapolate their findings, suggesting

that frugivore seed dispersal influences species composition and patterns of spatial

distribution of forest trees. Other studies come to similar conclusions, based on observed

correlations between current seed dispersion patterns and current adult dispersion patterns

(Garber 1986).

Although seed deposition by frugivores provides the initial dispersion pattern of

seeds, this spatial template interacts with subsequent processes, such as predation and

competition, to determine recruitment into the seedling stage, which in turn determines

recruitment into juvenile and then adult stages (Nathan and Muller-Landau 2000). Adult

plant recruitment and spatial dispersion patterns depend on a series of stage-sequential

processes that begin with the production of immature seeds, includes seed dispersal and

seedling establishment, and ends with recruitment into reproductive adults.

Patterns of seed rain initially established by frugivores may be disrupted by post-

dispersal processes. Thus, there may be little concordance between spatial patterns of

seed dispersal by frugivores and patterns of recruitment and aggregation at later life-

history stages. For example, although large-bodied frugivores tend to deposit seeds in

dense clumps (Mack 1993, Lambert 1999), rearrangement of these seeds by secondary

dispersal agents, such as dung beetles, ants, or rodents (Estrada and Coates-Estrada 1991,

Levey and Bymrne 1993, Hoshizaki et al. 1997, Wenny 1999), may scatter these seeds,

thus altering the primary seed-deposition pattern. Seedlings that establish from sites of

dense seed deposition, such as gorilla nest sites (Rogers et al. 1998) or howler monkey








sleeping sites (Julliot 1996), may suffer intense competition or increased predation,

further diluting initially clumped patterns. Moreover, spatial and temporal variation in

these post-dispersal processes (Willson and Whelan 1990, Forget et al. 1998) may further

remove the impact of frugivores on subsequent stages, resulting in little connection

between patterns at early stages to those at later stages. Therefore, to assess the degree to

which particular events at an early stage, such as dispersal of seeds by frugivores, impact

later stages, such as seedling density or tree dispersion patterns, we must connect stages

in seed dispersal and plant establishment processes.

Few studies have linked these processes in a single system to determine the

degree to which initial spatial patterns, set by frugivores at the seed stage, are connected

to spatial patterns of later stages, including seedling, juvenile, and, ultimately,

reproductive stages. Fragoso (1997) found that Maximiliana maripa seeds dispersed by

tapirs, and secondarily dispersed by rodents, gave rise to high densities of seedlings

located around tapir latrines. Moreover, density of fifth-year M maripa saplings was

higher at latrines than around parent trees. Thus, he attributed the patchy distribution of

M maripa palms in the Amazon to the influence of dispersal by tapirs. In contrast, Paul

(2001) found that increased deposition of hornbill-dispersed seeds below nesting trees in

Uganda did not result in altered juvenile composition 12-16 years later. The

contradictory nature of these and other studies' findings (Herrera et al. 1994, Chapman

and Onderdonk 1998, Pacheco and Simonetti 1998, Rey and Alcantara 2000, Wright et

al. 2000), emphasizes the importance of both connecting the stages and examining later

size classes to determine the impact of frugivores on plant recruitment and spatial

dispersion patterns.








I addressed these issues by quantifying whether patterns of seed deposition

generated by frugivores were disrupted at later life history stages (seedlings, saplings, and

poles) of tropical fleshy-fruited tree species in a moist evergreen forest, Kibale National

Park, Uganda, East Africa. Previous studies at this site provided background data crucial

to this study, including extensive phenological data on tree species, behavioral ecology

data on frugivores, and data on plant-frugivore interactions (Struhsaker 1975, 1997,

Wrangham et al. 1994, Chapman and Chapman 1995, 1996, Chapman et al. 1995, 1997,

Lambert 1997, 1999, Chapman and Onderdonk 1998, Lambert and Garber 1998,

Shepherd and Chapman 1998).

To better understand the link between frugivore activity and subsequent seedling

recruitment, I first focused on a single tree species, Monodora myristica (Annonaceae).

The potential to detect the effect of frugivore seed dispersal behavior on subsequent plant

recruitment should be great for this species because the hard-husked fruit and large-sized

seeds suggests the set of frugivores potentially capable of dispersing its seeds is limited.

I quantified the stage-sequential steps in the seed dispersal process, from seed removal to

seedling establishment, at two sites separated by 15 km, through two consecutive fruiting

seasons. At each study site I located and monitored fruiting M myristica trees to

determine immature and mature fruit fate and the identity of frugivores that dispersed the

seeds. I conducted a series of experiments to determine the fate of seeds handled by

frugivores in different fashions. The four treatments mimicked the following deposition

processes: (1) seeds embedded in fruit pulp, that had been dropped below the parent tree

canopy in partially eaten fruits; (2) seeds that had their fruit pulp removed and had been

spit out singly under parent canopies; (3) seeds that had their fruit pulp removed and had








been spit out singly away from parent canopies; and (4) seeds that were swallowed and

defecated in clumps surrounded with dung away from parent canopies. Fates of these

seeds were followed for 13 months to determine cause of death and rates of survivorship

to the established seedling stage. I used the seedling recruitment probabilities obtained

from these seed experiments to calculate expected seedling density at each site in each

year. I then compared expected seedling densities to natural seedling densities at each

site, and also quantified sapling, pole and tree densities to make inferences regarding the

impact of frugivores on older size classes of M myristica.

In a second study, to explore the suggestion commonly proposed in frugivore

foraging studies that patterns of seed dispersion established by frugivores persist to older

stages, I examined patterns of dispersion among 4 size classes for six animal-dispersed

tree species. The six species were Monodora myristica (Annonaceae), Mimusops

bagshawei (Sapotaceae), Pseudospondias microcarpa (Anacardiaceae),

Tabernaemontana holstii (Apocynaceae), Uvariopsis congensis (Annonaceae), and Celtis

durandii (Ulmaceae). For each species, I assessed whether overall dispersion patterns,

from seedlings to trees, changed between two stands, separated by 15 km, within the

same continuous forest. At each of two sites I quantified the number of individuals of

each species in each size class (seedlings, saplings, poles, and trees) found in 50 plots

located at random points along a 4-km loop transect. To evaluate whether patterns

established in early stages are disrupted by processes acting in later stages, I determined

whether degree of dispersion differed between sites for each size class, and then assessed

whether patterns in between-site differences were consistent with increasing size class. If

initial patterns persist over time, then changes in aggregation pattern from one size class








to the next should be similar between stands. Alternatively, if initial patterns do not

persist, then changes in aggregation across size classes should differ between stands.

The main objective of these two studies was to link patterns of dispersal with

patterns of recruitment and to assess if initial patterns established by frugivores were

disrupted by processes acting on later stages. Many researchers advocate that

maintenance of frugivore populations facilitating seed dispersal is critical for

regeneration of tropical forest trees (Howe 1984, Wrangham et al. 1994, Chapman and

Chapman 1995). Studies have found changes in seedling density and distribution

patterns in areas that have reduced frugivore abundances due to anthropogenic

disturbances, such as hunting or habitat fragmentation (Dirzo and Miranda 1991,

Chapman and Onderdonk 1998, Pacheco and Simonetti 1998, Wright et al. 2000).

However, conflicting outcomes from these studies render it difficult to make predictions

regarding consequences of changing frugivore abundance on tropical forest seedlings

and, ultimately, trees. Assessing the role frugivores play in determining tree dispersal

and recruitment patterns in multiple sites and years, and evaluating whether these patterns

are consistent at these scales, should yield insights into the role of seed dispersers in

determining plant recruitment and spatial dispersion patterns. A critical component of

this type of research is linking steps in the system such that the findings can contribute to

a better understanding of the role of dispersers and post-deposition processes in

determining tree recruitment patterns. Although these data are time consuming and

difficult to collect, they yield insight into ecological and evolutionary processes of

tropical forest tree recruitment, and provide information valuable for conservation

decisions regarding these forest ecosystems.













CHAPTER 2
BRIDGING THE SEED DISPERSAL GAP: DOES SEED DISPERSAL DETERMINE
PATTERNS OF TREE SEEDLING RECRUITMENT IN A PRIMATE-PLANT
INTERACTION?


Introduction

Frugivores are thought to strongly impact the composition of tropical forests by

acting as seed dispersers (Howe and Smallwood 1982). Their impact is thought to be

highly significant for forest regeneration due to the ubiquity of fleshy-fruited plant

species; as many as 75% of tropical tree species produce fruits presumably adapted for

animal dispersal (Frankie et al. 1974, Howe and Smallwood 1982). Moreover, tropical

ecologists advocate that reduction or disappearance of frugivore species will profoundly

affect tropical forest composition and could eventually lead to extinction of local tree

populations (Howe 1984, Chapman et al. 1992, Chapman and Chapman 1995, Wright et

al. 2000).

Frugivores interact with fleshy-fruited plants by removing fruits and depositing

seeds away from the parent tree. Evidence supporting the importance of this interaction

for plant recruitment has been based primarily on studies that examine the quantity of

fruit removed by vertebrates and the percentage of viable seed species that they deposit

(McKey 1975, Howe 1982, Estrada and Coates-Estrada 1984, Wrangham et al. 1994).

Additional support for the importance of removing seeds away from the parent tree has

come from numerous studies testing the Janzen-Connell model, which proposes that seed

and seedling survival increases with distance from the parent tree (Schupp 1988,








Augspurger and Kitajima 1992, Condit et al. 1992, Webb and Peart 1999, Connell and

Green 2000, see Hammond and Brown 1998 for review). Seeds deposited far from

parents and/or in low densities should have greater probabilities of surviving and

establishing as seedlings due to escape from distance- and/or density-dependent predation

associated with proximity to parent trees.

It is often implicitly assumed that high fruit removal and seed dissemination by

frugivores lead to high adult recruitment. For example, Herrera (1998) stated that

variation in disperser abundance probably translated into annual fluctuations in the

number and composition of seeds dispersed and eventual number of seedlings recruited.

However, this need not be so given that fruit removal and seed dissemination by

dispersers are just two events in a series of steps necessary for tree establishment.

Factors that influence later life-history stages, such as seed germination, seedling

establishment, and juvenile survivorship, may also play critical roles in determining

recruitment of adult trees. Despite the fact that plant mortality is highest at seed and

seedling stages (Harper 1977), studies that focus on particular frugivores often claim that

seed deposition patterns generated by these frugivores create observed adult plant

distributions (e.g., Stevenson 2000). These claims are made despite the fact that stages

between seed dissemination and adult recruitment are not simultaneously examined.

Recruitment of reproductive adults into the plant population should be viewed as

the cumulative probability of making a series of transitions between successive life

history stages including seed maturation, seed deposition, seed germination, seedling

establishment, juvenile survival, and adult tree recruitment. Plant recruitment may be

limited by processes that occur at any one of these stages, or through the interaction of








these stages (Jordano and Herrera 1995). For any given stage, multiple selective

pressures determine the probability of survivorship and thus the probability of making the

transition into the next stage. As studies that have focused on one or two stages have

shown, considerable spatial and temporal variation exits in the selective pressures that

influence each stage (Willson 1988, Willson and Whelan 1990, Forget et al. 1998,

Kollmann et al. 1998). In addition, the suitability of any site may vary from stage to

stage, such that a good site for seed germination may not be a good site for seedling

establishment (Schupp 1995, Kollmann and Schill 1996, Rey and Alcantara 2000).

Furthermore, selective pressures on sequential stages may be conflicting (Herrera 1985,

Herrera et al. 1994). For instance, at the seed removal stage, smaller seed size may

increase the likelihood of being dispersed away from the parent, yet at the seedling

establishment stage, larger seeds are more likely to produce larger, more competitive

seedlings due to larger seed reserves (Schupp 1995). Most studies focus on one or a few

of these stages without considering the influence of previous stages or examining the

consequences of later stages (Herrera et al. 1994, Schupp and Fuentes 1995). Few

studies have linked these successive stages in a single system and bridged this

demographic gap (Schupp and Fuentes 1995).

Herrera et al's (1994) exceptional study was the first to examine successive stages

from seed rain to seed and seedling survival during a single reproductive event to assess

whether patterns of seed rain by birds produced a predictable seedling shadow for a

Mediterranean forest tree, Phillyrea latifolia. They found that seed rain by birds directly

impacted seedling distribution in scrubland, but in forest habitat, seed deposition patterns

by birds were overshadowed by forces after deposition; rodent predation or variation in








germination success resulted in spatial discordance between seed rain and seedling

distribution. Jordano and Herrera (1995) further emphasized the complexity of the multi-

staged nature of recruitment in vertebrate-dispersed plants. Using the same data set as

Herrera et al. (1994), Jordano and Herrera (1995) found that differences in seed rain

intensity in different microsites were unrelated to second-year seedling recruitment

despite concordance between other stages (e.g., seedling emergence was positively

correlated with second-year seedling survival). Thus, estimations of probabilities of

seedling recruitment based on one or two consecutive stages may be erroneous due to the

"uncoupling" of stages in the recruitment process. Furthermore, the degree to which

stages in the seed dispersal process are coupled will influence the degree to which

predictable regeneration patterns can be assessed for different plant species in different

habitats or communities.

Herrera et al. (1994) showed de-coupling between current avian seed dispersers

and subsequent plant life history stages. This may reflect the fact that these frugivore-

plant interactions are of relatively recent origin (Herrera 1984). The woody genus

Phillyrea is present in the pre-Pliocene (> 5 mya) fossil record (Herrera 1992), whereas

most extant avian species present in the Mediterranean region arose during the

Quaternary (2 mya to present; Herrera 1984). The slow rate of evolutionary change of

long-lived plants (about 27 my for woody shrubs) relative to their animal dispersers (0.5

my for birds) causes the adaptations of plants, such as P. latifolia, to their current

dispersers to be diffuse in nature (Herrera 1985). As such, fruit and seed traits of present

day P. latifolia may result from selective pressures of the predecessors of their current

bird dispersers, and thus one might expect a certain degree of discordance between these








frugivores' actions and P. latifolia demography. In addition, in a system where many

dispersers interact with a plant species, it may be difficult to detect the influence of any

given disperser on subsequent plant recruitment. In contrast, in a simpler system where a

plant species interacts with a small set of dispersers, and has potentially evolved in

response to selective pressures by these frugivores, the ability to detect the influence of

disperser activity on subsequent plant recruitment should be great.

I chose to examine the seed dispersal and seedling recruitment processes of an

Afrotropical tree species, Monodora myristica (Annonaceae), whose fruit morphology

strongly suggests the access to its seeds is restricted to a subset of the frugivore

community, namely, large-bodied vertebrates. The genus Monodora has a specialized

syncarpous fruit that is thought to be the result of selection for probable dispersal by large

vertebrates, and it is one of the few genera of Annonaceae with this advanced trait

(Schatz and Le Thomas 1993). Large diaspores (1,000 mm3) first started to appear in

numbers during the late Paleocene to early Eocene (60 mya; Tiffihey 1984) which

corresponds to the appearance and radiation of modem primate lineages (Sussman 1991).

Monodora myristica produces large (16 cm diameter), green fruit that contain numerous

large seeds (1.9 cm ave. length). The thick (1.8 cm), woody husk is difficult to open,

initially rendering seeds and pulp inaccessible to smaller frugivores, and the large seed

size limits ingestion and endozoochorous dispersal to larger-bodied frugivores. The

specialized nature of M myristica's fruit morphology and its potential long-term co-

existence with large-bodied frugivores make this an ideal system in which to examine the

potential importance of a limited set of frugivores to the recruitment of a plant species.








The objective of this study was to assess the relative importance of large-bodied

primates in determining the seed dispersal and seedling recruitment patterns of M

myristica, and to make inferences regarding the recruitment of older size classes from

juveniles to adults. The general approach was to link patterns of fruit and seed removal

by frugivores with seed germination and seedling establishment, and to quantify the post-

deposition fate of seeds through to the established seedling stage. Based on these data,

cumulative probabilities of seedling recruitment were used to generate estimates of

expected seedling density, which were compared with natural seedling, sapling, and pole

densities.

Previous studies examining whether spatial patterns of seed rain are coupled with

seedling recruitment have often focused on two spatial scales, comparing different habitat

types, such as forest vs. scrubland, or different microhabitats, such as under canopies of

different tree species (Herrera et al. 1994, Kollmann and Schill 1996, Rey and Alcantara

2000). I assessed whether patterns that primates generate with regard to M myristica

seedling recruitment were consistent within a single continuous forest ecosystem at two

sites separated by 15 km. In addition, I assessed whether these patterns were consistent

across a temporal scale of two years by repeating the study at each site for two

consecutive yearly fruiting seasons. Thus, I address the following questions for M

myristica: (1) What role do large-bodied frugivorous primates play in initial stages of

seed dispersal (fruit removal and seed deposition)? (2) Do patterns established by these

primates in initial stages consistently link with patterns in subsequent stages (seed

germination, seedling emergence, seedling recruitment, and seedling density)? (3)

Across the spatial scale of 15 kin, do differences in large-bodied primate density








correspond to differences in seedling density? and (4) How does interannual variation in

post-deposition processes affect patterns seen in the seed dispersal process?




Study Sites

This study was conducted in Kibale National Park (KNP), western Uganda (0 13'

- 0 41' N and 30 19' 30 32' E), a moist-evergreen forest transitional between lowland

rain forest and montane forest, situated approximately 25 km east of the Ruwenzori

Mountains (Wing and Buss 1970, Struhsaker 1975, Howard 1991). The park covers an

area of 766 km2 of undulating terrain and is drained by the Mpanga and Dura rivers

(Howard 1991, Chapman and Lambert 2000). Two rainfall peaks occur each year, from

March to May and September to November. Mean annual rainfall in the region

(measured at Kanyawara) is 1807 mm (1990-1997), the mean daily minimum

temperature is 15.5 C, and the mean daily maximum temperature is 23.7 C (1990-1996;

Chapman and Chapman unpublished data). Within the park, an elevational gradient from

1,590 m a.s.l. in the north to 920 m in the south corresponds to a north to south increase

in temperature and decrease in rainfall (Howard 1991, Struhsaker 1997, Seavy et al.

2001).

Two sites, Kanyawara (Forestry Compartment K-30) and Dura River (hereafter

referred to as Dura), were selected for the study. These two sites are located 15 km apart

on a north-south gradient within KNP. The Makerere University Biological Field Station

is located at the Kanyawara site, where more than 20 years of continuous research have

been conducted, including extensive meteorological, phenological, and primate

behavioral data (Struhsaker 1975, 1997, Chapman et al. 1995). This site is situated at an








elevation of 1500 m (Chapman et al. 1997) and consists of a series of moderately

undulating valleys with an average slope of 15.8 (Balcomb unpublished data). Although

the Dura site has been the location of a few short-term studies and surveys (Clutton-

Brock 1975, Struhsaker 1975, Ghiglieri 1984), no long-term monitoring was conducted

there until 1995 (Chapman et al. 1997). This site is situated at 1250 m elevation and is

drained by the Dura River (Chapman et al. 1997). The topography is similar to that of

Kanyawara, with an average slope of 5.90 (Balcomb unpublished data). Although the

sites occur within the same forest, the dominant tree species differ between the two sites.

Foresters classify the Kanyawara forest as dominated by Parinari excelsa, Aningeria

altissima, Olea welwitschii, Newtonia buchanaii, and Chrysophyllum gorungosanum,

while these tree species are less common at the Dura forest, which is dominated by

Pterygota mildbraidii, Cola gigantea, Pipadeniastrum africanum, and Chrysophyllum

albidum (Chapman and Lambert 2000).




Study Species

Monodora myristica is a canopy level tree reaching 30 m in height and is

restricted to evergreen forest (Verdcourt 1971). The flowers, as is typical of the

Annonaceae, are pollinated by beetles (Judd et al. 1999). Monodora myristica produces

large, green, spherical fruit (average diameter 16 cm) with a thick (1.8 cm), woody

pericarp and numerous (175-750) large seeds (dry mass: 0.92 g [Zanne unpublished

data]; length: 18.9 2.3 mm; width: 12.0 1.1 mm, n = 20 [Balcomb unpublished data];

(Kalina 1988, Hamilton 1991, Chapman and Chapman 1996). Fruiting pedicels may be

25-60 cm long and 1-3.5 cm thick (Verdcourt 1971).








In 1996 and 1997, fruit production started in June, and fruits were present on trees

for 7 months. Although M myristica fruits gradually increased in size during these

months, the husk remained green, rendering it difficult to determine maturation based on

color or size alone. However, the maturation state could be assessed once the seeds and

pulp were exposed by vertebrate frugivores (i.e., from partially eaten fruits in the tree or

on the ground). Thus, I judged fruits to be immature when the seeds had soft, white outer

coats and the pulp was hard; fruits were usually less than 10 cm diameter. I judged fruits

to be mature when the seeds had hard, brown outer coats and the pulp was soft; fruits

could be as large as 25 cm diameter. Fruits ripened during the second rainfall peak of the

year. Mature fruits were first eaten in November in 1996 and in September in 1997 and

by mid-January no fruits remained on trees in either year.

Initial access to mature M myristica fruits is most likely restricted to large-bodied

arboreal frugivores. The most common large-bodied arboreal frugivores in KNP are

primates; Kibale has one of the highest recorded primate biomass estimates in the world

(Chapman and Lambert 2000). The three largest-bodied primates, chimpanzees (Pan

troglodytes, 25-40 kg), baboons (Papio anubis, 11-50 kg), and grey-cheeked mangabeys

(Lophocebus albigena, 4-11 kg; Kingdon 1997), all have large gape widths and possess

large, strong jaws capable of biting through hard husks, such as those found on M

myristica. Chimpanzees and baboons often swallow large seeds and tend to defecate

them intact in clumps surrounded by substantial amounts of dung (Wrangham et al. 1994,

Lambert 1999). Mangabeys typically store large seeds in their cheek pouches and later

clean off the pulp and spit the seeds, although they may also act as seed predators when

M myristica seeds are immature (Lambert 1997, Lambert pers. comn., Olupot pers. com.).








The three smaller-bodied frugivorous primates, redtail monkeys (Cercopithecus ascanius,

2-6 kg), blue monkeys (C. mitis, 4-12 kg), and l'hoesti monkeys (C. l'hoesti, 3-10 kg;

Kingdon 1997), cannot break open hard-husked fruits and typically do not swallow large

seeds, but instead remove the pulp and spit out the seeds (Lambert 1999). Both of the

large-bodied avian frugivores in KNP, the Black-and-White Casqued Hornbill

(Ceratogymna subcylindricus) and the Great Blue Turaco (Corythaeola cristata), can

swallow large seeds the size of M myristica (Kalina 1988, Sun et al. 1997). While

hombills have been observed bringing M myristica seeds back to their nests (Kalina

1988), there are no published observations of turacos feeding on M myristica.

Other frugivores may disperse M myristica once the seeds and fruits are dropped

to the ground. Elephants (Loxodonta africana) may be an important, but less reliable,

disperser of M myristica seeds; M. myristica seeds were recovered in 5.0% of elephant

dung in KNP (Cochrane unpublished data), and M myristica seedlings were observed in

older dung in 8.7% of dung censuses (Cochrane unpublished data). African civets

(Civetticus civetta), palm civets (Nandinia binotata), and genets (Genetta sp.) are all

frugivorous and deposit viable seeds in latrines (Engel 2000). African civets can swallow

large seeds; however, of 123 African civet defecations collected at Kanyawara and Dura

between July 1996 and May 1997, only one contained seeds of M myristica (Balcomb

unpublished data). Moreover, seedling recruitment from African civet latrines is low due

to repeated use of the same location and the fact that these latrines are often located on

hard, exposed soil (Pendje 1994). Little is known about the feeding ecology of genets

and palm civets in KNP; however, genets are reported to share latrine cites with African

civets (Engel 2000), and palm civets are reported to eat small- and medium-sized seeds








(Charles-Dominique 1978, Engel 2000). Small ruminants, such as the red and blue

duikers (Cephalophus harveyi and C. moniticola), can swallow and regurgitate large

seeds; however, only seeds with a hard, stoney exocarp are likely to survive gut

rumination (Gautier-Hion et al. 1985, Feer 1995). Thus, duikers are likely predators of

M myristica seeds.

The major vertebrate seed predators of M myristica are likely rodents. Fruits and

seeds constitute 48 65% of the diet of the three most common rodents in KNP

(Praomys stella, P. jacksoni, and Hybomys univattatus) and 60% of the diet of Thomas's

rope squirrel (Funisciurus anerythrus), the one species of squirrel that was trapped on the

ground (Isabirye-Basuta 1979). Rodents were the most important cause of seed mortality

of two large-seeded canopy tree species (Mimusops bagshawei and Stombosia scheffleri;

Lwanga 1994); however, in captive feeding trials, P. stella and P. jacksoni tended to

avoid or eat very little of M myristica seeds (Kasenene 1980).

Rodents browse on the roots and shoots of young seedlings and thus are also

significant seedling predators (Kasenene 1980, 1984, Lwanga 1994). In captive feeding

experiments, P. stella and P. jacksoni ate 70-80% of the seedlings presented to them

(Kasenene 1984), and 40% of M bagshawei and S. scheffleri seedlings planted in the

forest died due to rodent predation (Lwanga 1994). Bushbuck (Tragelaphus scriptus)

prefer young leaves and tender shoots (Dubost 1984), and duikers are also browsers,

although their role as tree seedling predators is thought to be minor compared to rodents

(Struhsaker 1997).








Methods

This study was conducted from May 1996 June 1998, encompassing two fruiting

seasons for M myristica. Unless otherwise noted, all methods used at Kanyawara and

Dura were identical and repeated for the two fruiting seasons. Since the fruiting season

started in June of one calendar year, and seed germination and seedling establishment

occurred in the next calendar year, "Year 1" and "Year 2" will refer to the 12-month

period starting June 1996 and June 1997, respectively.



Fruit Removal

To quantify fruit removal by frugivores, I determined the fate of all fruits of a set

of focal M myristica trees by counting all fruits present in the tree and on the ground

underneath the canopy. In Year 1,1 was primarily interested in dispersal events and thus

monitored the quantity of mature fruit removed by frugivores. At the beginning of Year

2,1 observed heavy predation of immature fruit by some of these same frugivores, and

thus I monitored the quantity of immature fruit preyed upon. Although in Year 21

obtained an estimate of fruit crop size after all fruits had matured, due to logistical

contraints I did not quantify removal of mature fruits from focal trees; I assumed fate of

mature fruit was similar between years.

In June 1996,1 surveyed 41 M myristica trees at Kanyawara and 40 trees at Dura

for presence of fruits. Fifteen trees at Kanyawara and 18 trees at Dura contained at least

one fruit and thus were selected as focal trees. Once a month, from June 1996 to January

1997,1 counted the number of fruits in each focal fruiting tree, and mapped each fruit's

position in the tree for future reference. Due to the large fruit size, I was able to count all

fruits. To determine the fate of fruits missing from the tree since the previous count, I








searched the ground underneath each tree to locate husks or partially eaten fruit that had

been dropped. When a fruit was found on the ground, its location was tagged and its

condition was recorded including size of fruit, amount of fruit eaten (estimated in 25%

increments), tooth marks on husk, and number and size of exposed seeds. Fruits that

disappeared entirely from the tree with no husk remnants on the ground, were assumed to

have been removed by large-bodied primates (see Results: Frugivore feeding behavior).

In June 1997, the same 15 and 18 focal trees at Kanyawara and Dura,

respectively, produced fruit. Once a month, from June to October 1997, I counted the

number of immature fruit on these focal trees and searched the ground for evidence of

fruit fate; after October, no fruits in the immature stage were observed in the trees. The

condition of each immature fruit found on the ground was recorded including whether it

had been torn from the tree by a vertebrate pedicell attached to fruit and ripped from tree)

or had been aborted (no pedicel attached to fruit, clean abscission mark, no teeth marks

on fruit). In addition, size of fruit, size and color of exposed seeds, amount of fruit eaten

(estimated in 25% increments), and teeth marks on the husk were recorded.

To substantiate results of mature fruit fate obtained from fruit counts, I, with the

aid of three field assistants, quantified diurnal frugivore visitation and feeding behavior

during all day watches on a subset of the focal trees. Three trees were each monitored by

a single observer from approximately 0800 h 1700 h at each site on alternating days.

Focal trees were selected based on high fruit abundance and presence of opened fruits in

the tree. These data were collected during the peak of mature fruit removal from 2

December 1996 14 January 1997 (Year 1). Due to the low frugivore visitation rate in

Year 1, these data were also collected in Year 2, from 27 October 11 December 1997.








Observations were collected for 809 hrs (Dura = 350 hrs, Kanyawara = 459 hrs) over 115

days, both years combined.

To determine frugivore visitation rates, the observer scanned the tree every 15

minutes for the presence of frugivores. In addition, presence of frugivores in the vicinity

of the tree was noted when they were seen in neighboring trees or heard calling. To

determine frugivore fruit- and seed-handling behavior, when a frugivore was observed

feeding on a fruit, opportunistic feeding samples were obtained (average duration of 40

35 seconds, range 4-252 seconds). The following data were collected during these

samples: type of fruit fed from (whole or already opened), number of bites taken per

fruit, and whether the seeds were swallowed immediately, spit out under the canopy of

the parent tree, spit out in a neighboring tree, or consumed (visibly chewed upon). For

primates, seeds may be stored in cheek pouches and processed later. Thus, if animals left

the area before the fate of removed seeds was observed, I assumed these seeds were spit

out away from conspecific canopies. Since M myristica fruit may contain several

hundred seeds, these data were collected to obtain estimates of the proportion of seeds

that were deposited in these different conditions (see below: Overall probability of

dispersal and seedling recruitment).



Fate of Dropped Fruits

Large-bodied primates often drop opened M myristica fruit to the ground

underneath the parent tree. These seeds may subsequently be dispersed by terrestrial

frugivores, be preyed upon, or germinate under the parent tree. To determine the fate of

these seeds, I collected partially eaten fruits from the ground, cut them into sections, and








placed them at tracking stations. Each section consisted of 20 pulp-covered seeds pressed

firmly onto a piece of the outer husk of the fruit (approx. 5 x 10 cm). A tracking station

(1 x 1 m plot cleared of vegetation and smoothed over with dirt), with one section placed

at its center, was placed at a randomly location under the canopy of each of 13 fruiting M

myristica trees at each site.

At each site, 13 stations were established in Year 1, from January 6 8, 1997 (at

the end of the fruiting season), and 13 in Year 2, from October 27 November 10 1997

(in the peak of the fruiting season). Due to logistical constraints I was unable to establish

these stations in the same month in both years. The stations were checked once a day for

7 days, then once a week for a month, and then once a month until all seeds had been

removed or had died. If any seeds produced seedlings, the fate of these seedlings was

followed until 13 months after the start of the experiment.

The identity of the seed remover was determined by (1) teeth marks in the pulp,

on the husk, or on the seeds, (2) pieces of the outer seed coat left behind (small rodents),

and (3) footprints in the dirt. Teeth marks of large primates, small primates, squirrels,

and small rodents could be distinguished. I assumed primates were dispersers and

rodents were predators ofM. myristica seeds. Arthropod infestation of seeds was

determined by presence of entry or exit holes. Seeds were considered dead once they

were soft. They were subsequently opened and examined for evidence of current or past

arthropod infestation. These data were used to obtain estimates of the proportion of seeds

in dropped fruits that subsequently remained under the canopy or were dispersed away

from the canopy by different frugivores.








Post-Deposition Fate of Seeds Through to Seedling Survival

To quantify the fate of seeds once they were processed by primates, I established

experiments that simulated the conditions under which primates deposit seeds. Small-

bodied primates typically spit out single seeds, cleaned of pulp, under or within ten

meters of the canopy of parent trees, while large-bodied primates typically swallow seeds

and deposit them intact in large clumps in dung (Lambert 1999). These experiments

were established at both sites in January 1997 (Year 1) and November 1997 (Year 2), at

the same time as the experiment to determine the fate of dropped fruit.

To mimic seed handling by a seed-spitting frugivore, I obtained seeds from ripe

fruit and washed them to remove pulp. To quantify the fate of seeds spit under parent

canopies, I placed 10 single seeds at random locations under each of 10 fruiting M.

myristica trees at each site (n = 100 seeds per site in each year). To quantify the fate of

seeds spit away from parent canopies, I placed 120 single seeds per site, in each year, at

10 m intervals, 1 m off of trails (7 transects at Kanyawara, 5 at Dura), in the area where

adult M myristica were located; in some cases I increased the interval between stations to

avoid placing seeds under conspecific canopies. For both experiments, at each station a

small section (approx. 5 x 5 cm) of leaf litter was removed, a seed was placed on the soil

surface, and the location was marked with flagging tied one meter above ground to

minimize visual attraction of rodents to the station. In Year 2, thread was glued to seeds

and one end was tied to vegetation to minimize seed loss by factors other than predation.

I assumed the presence of glue had no effects on predation or germination.

A typical chimpanzee defecation weighs 80 g (range 1-350 g) and contains an

average of 10.7 intact M myristica seeds (Wrangham et al. 1994). To mimic deposition

by large-bodied primates, I collected M myristica seeds from baboon dung (readily








available at the Dura site), and at each station I placed 10 seeds in 60 g of fresh baboon

dung from which all large seeds had been removed. Stations were established on the

same transects as the spit-seed experiment. A stratified-random design was used such

that 5 seeds-in-dung stations were present on each transect and were separated by 10 to

40 m. Within a transect, each seeds-in-dung station was placed 1 m from a randomly

selected spit-seed station. A small section (approx. 10 x 10 cm) of leaf litter was

removed and seeds in dung were placed on the soil surface. Twenty-four and 26

replicates were established per site in Year 1 and Year 2, respectively.

For all three experiments (spit under, spit away, in dung), seeds were monitored

once a week for the first month and then once a month for 13 months. Seeds were

monitored for fungal attack, beetle infestation (indicated by entry holes), and rodent

predation (indicated by rodent tooth marks on seeds or pieces of outer seed coat left

behind). Seeds that disappeared were assumed to have been taken and eaten by rodents

(Chapman 1989). Although survival of seeds cached by rodents may be a source of plant

recruitment, few studies have tracked the fate of seeds removed by small rodents in the

tropics (Brewer and Rejmanek 1999), and little is known about caching behavior of small

African rodents. The proportion of removed seeds that may be cached versus eaten, may

change radically across small spatial scales (several hundred meters), among habitats, and

as a function of food availability (Forget et al. 1998). Thus, I assumed that all removed

seeds were killed.

Dung beetles may have removed some seeds from the seeds-in-dung experiment.

However, in Kibale, dung beetle movement of larger seeds, such as M myristica, is

primarily accomplished by a species of burrowing beetle, which tends to tunnel under the








dung pile (Shepherd and Chapman 1998). Seeds are often unintentionally brought down

with large chunks of dung (Shepherd and Chapman 1998) or fall into the tunnels created

by these beetles (as seen in this study). A large species of roller beetle typically removes

small pieces of dung to construct its dung balls, and tends to avoid seeds as large as M

myristica (Shepherd and Chapman 1998). Thus, when seeds were removed from the

seeds-in-dung station, I searched the soil under the station to a depth of approximately 10

cm to determine the number of seeds buried by dung beetles. These seeds were re-buried

and checked on each subsequent monitoring, along with the other seeds in the

experiment. I assumed that seeds not located had been removed and killed by rodents.

For all three experiments, seeds that germinated and became seedlings continued

to be monitored monthly until 13 months after the start of the experiment. Germination

was defined as the production of a radicle. Seedlings were categorized as "emerged

seedlings" for as long as they were still physically attached to the seed. Once the

seedling was no longer attached to the seed and its root was firmly in the soil it was

deemed to be an "established seedling". Upon production of a shoot, seedling height

(measured as aboveground stem length), number of leaves, and cause of death

desiccationn, herbivory, crushed from fallen branches, or unknown) was also recorded.

In a separate experiment, I investigated the effect of seed burial on seedling

establishment. Although dung beetles tend to avoid large seeds and leave them on the

surface, Shepherd and Chapman (1998) found that of M myristica seeds that were buried,

the majority were found at a depth of 1 cm. In general, this is also the ideal depth at

which seeds avoid predation and are able to germinate (Shepherd and Chapman 1998).

Thus, 50 single M. myristica seeds were each buried under 1 cm of soil at 5 meter








intervals along a 250-m transect at each site in May June 1997. After 9 months, the

proportion of established seedlings at each site was recorded.



Plant Size-Class Densities

The densities of M myristica seedlings (small: < 0.2 m tall; large: 0.2 0.5 m

tall), saplings (0.5 2.0 m tall), poles (> 2.0 m tall and DBH < 20 cm), and adult trees

(DBH > 20 cm) were quantified at both sites from July 1996 October 1997. Plots (50 x

60 m) were placed at 50 randomly-selected points along a 4-km loop route (see below

Frugivorous primate densities). The trail was used as the central axis for each plot.

Adult tree density was measured within the entire plot. Due to the expected greater

density of stems with decreasing size class, sub-plots, oriented along the central axis of

the plot, were used to measure densities of small seedlings (10 1 x 1 m sub-plots every 4

m), large seedlings and saplings (one 4 x 20 min sub-plot), and poles (one 10 x 20 m sub-

plot). No individuals were included within 0.5 m of either side of the trail. The 10 small-

seedling sub-plots were combined to obtain one value per plot. For all individuals,

except adults, distance to the edge of the canopy of the nearest adult conspecific, if

present within the plot, was estimated.



Fate of Natural Seedlings and Saplings

The growth rate and survival of naturally-emerged seedlings and saplings were

quantified at both sites by monitoring the first five individuals of each size class

encountered in the first 30 plots used to determine plant density. For some size classes,

sample size was increased by including individuals found outside the plots. A total of

355 individuals (Kanyawara: 36 small seedlings, 58 large seedlings, 47 saplings; Dura:








99 small seedlings, 100 large seedlings, 15 saplings ) were tagged by placing flagging on

a nearby stem, and, for seedlings, a loop of green string was loosely tied around the base

of the focal stem. Surveys were conducted every six months for 18 months (mean

interval = 148 54 days, range 25 235 days) to assess survival, stem height, and

number of leaves. Cause of death included desiccation, herbivory (by insects and

mammals), crushing by fallen branches, and unknown (seedling missing). Relative

Height Growth Rate (RHGR) was determined as (In H2 In HI)/(H2 Hi), where Hi is

the initial height and H2 is the final height; this accounts for differences in initial plant

sizes (Hunt 1982, Hutchings 1997).



Frugivorous Primate Densities

Since changes in seedling density have been attributed to changes in disperser

abundance (Chapman and Onderdonk 1998, Pacheco and Simonetti 1998, Wright et al.

2000) the density of diurnal frugivorous primates was assessed at each site by line-

transect (Deflor and Pintor 1985, Chapman et al. 1988, Whitesides et al. 1988). This is

thought to be the most appropriate method for estimating densities of large-bodied,

diurnal primates (Council 1981), other than the very time-consuming effort of mapping

home ranges of known groups. At both sites, a 4-km primate census route was

established, and censuses were conducted biweekly from June 1996 to July 1997

(Kanyawara, n = 26; Dura, n = 23). Censuses were conducted between 0700 h and 1400

h at a speed of approximately one km/h. Data collected included primate species

observed, time of observation, straight line distance between the animal and observer








(visually estimated), and mode of detection. For details regarding group density

estimates for the five cercopithecine primates, see Chapman et al. (2000).

Chimpanzees are particularly difficult to census using the standard line-transect

method due to their fission-fusion societies and large home ranges. However,

chimpanzees construct individual sleeping nests each night leaving evidence of their

presence and numbers. Thus, counts of nests were conducted from the line transects

(Plumptre and Reynolds 1997). For each census, new nests were flagged perpendicular

to their location along the transects, and the following were recorded: perpendicular

distance from transect to nest, height of nest, age-class of nest, and tree species in which

the nest was located. This methodology is equivalent to the marked nest counts in which

only new nests are counted (Plumptre and Reynolds 1996). For information regarding

density estimates for chimpanzees, see Balcomb et al. (2000).



Statistical Analyses

Due to the high prevalence of zeros in seed experiments (e.g., seeds not

germinated, see below), data did not meet assumptions of normality and equal variance.

Since traditional arc sin transformation of proportional data (Sokal and Rohlf 1995) did

not improve these violations, data were analyzed with PROC GENMOD (SAS Institute

1998), a procedure of categorical data modeling analogous to ANOVA. In cases where

the response variable was binary [e.g., fruit fate (eaten or not), seed fate (alive or dead)],

a logistic regression model with binomial error distribution was used (Agresti 1996).

Most of the analyses included two main effects, site and year, and the site-by-year

interaction. For spit seeds, a third main effect, location (under vs. away), as well as all

possible two-way interactions, were included.








In all experiments, the fate of each seed was given a score of 1 or 0 for each

dependent variable (germinated or not, removed by rodents or not, etc.) and compiled as

number of "successes" over number of seeds in each station (1, 10, or 20). For example,

germination was assumed to be a dichotomous factor with a probability of success, p, and

a binomial distribution. The dependent variable was logit transformed as part of the

logistic regression analysis; the logit transformation is the natural logarithm of the ratio

of the proportion of successes to failures (Trexler 1985).

For each analysis of a dependent variable, the goodness-of-fit test was used for

the maximal model to determine if residuals (error terms) were consistent with the

binomial distribution. In cases where there was a lack of fit, data were assumed to be

overdispersed, most likely due to clustering, and deviances were rescaled to make tests

for each term more conservative (Crawley 1993). The significance of each main effect

and interaction term was assessed through the likelihood ratio tests with a chi-square

distribution, where likelihood of the maximal model is compared with likelihood of a

simpler model. If, when a factor or interaction term is removed from the maximal model,

the change in deviance is small, then the more complex model does not fit the data

significantly better than the simpler model, and the term is deleted from the model

(Crawley 1993). Higher-order interactions, such as trails nested within site (for seeds in

dung) or trees nested within site (for spit seeds), were not included as predictor factors

due to the difficulty in interpreting strength of main effects in models with multiple

interactions. However, variation explained by trails or trees is incorporated in the

overdispersion factor (error term) and thus is accounted for in the final model.








For count-type response variables (e.g., fruit crop size, seedling density), a log-

linear model with poisson error distribution was used (Agresti 1996). PROC GENMOD

(SAS Institute 1998) was used to analyze effect of site, year, and site-by-year on fruit

crop size, and effect of site on density of seedlings, saplings, poles, and trees.

All other data were analyzed using the SPSS statistical package (SPSS 1999). I

used t-tests or, when assumptions of normality and unequal variance were unable to be

met, Mann Whitney U tests, to examine differences between sites in seedling and sapling

relative-height growth rates. For between-site differences in proportion of trees fruiting

and proportion of seedlings and saplings surviving I used chi-square and G-tests, when

row totals were fixed, of independence (Sokal and Rohlf 1995). Yates' correction for

continuity was applied due to low sample size resulting in some expected cell frequencies

of< 5 (Sokal and Rohlf 1995).

For all statistical tests, degrees of freedom equals one and thus are not noted in the

text, with the exception of the t-test. In addition, all values are presented as the average

one standard deviation, unless otherwise noted.



Overall Probability of Dispersal and Seedling Recruitment

To determine the cumulative probability of an immature seed becoming an

established seedling, one needs to determine the probability of making the transition from

each stage in the seed dispersal and seedling establishment process. A fate diagram of

stages and processes influencing M myristica recruitment was constructed (Fig. 1) and

the stage-specific and cumulative probabilities of seeds passing from one stage to the

next were determined (Figs. 2-3).








The probability of seed fate from the immature seed stage to the seed deposition

stage (Fig. 2) was based on data from fruit crop counts, frugivore feeding behavior, and

the experiment determining fate of fruits dropped under the canopy. The probability that

seeds became mature was determined as the product of the probabilities that fruits

escaped abortion and pre-dispersal predation by vertebrates (all immature seeds within a

single fruit were destroyed in both cases).

The probabilities that seeds were removed by primates and other frugivores and

deposited in different conditions and locations, relative to conspecific canopies, were

extrapolated from data collected on frugivore feeding behavior (during focal tree

watches) and fruit crop counts. Since it is infeasible to follow the fate of every seed in

every tree, and since individual frugivores were not followed once they left the vicinity of

the tree, I made a number of assumptions when calculating the probability of seed

deposition in different conditions. I determined the probability that a seed was deposited

away from the parent canopy without dung as the sum of the probability that each

frugivore deposited seeds in this manner. For each frugivore this was calculated as

illustrated by the following example. The probability that a seed was deposited by redtail

monkeys away from the parent canopy as a single seed without dung was the product of

the average number of partially eaten fruits in the tree, the proportion of the fruit that was

eaten [set at 0.5, since the majority of partially eaten fruits (26/37) were estimated to be

half eaten], the proportion of all feeding records from open fruits where redtails were

seen feeding, and the proportion of all feeding records for redtails where they were seen

spitting single seeds away from the parent tree. If a redtail moved out of the area before

seed fate was observed, I assumed the seeds were spit away from conspecifics.








The probability that seeds were deposited in clumps surrounded by dung was

estimated in a manner similar to that for singly spit seeds, using data on fruit crop

removal and chimpanzee feeding samples (baboons were assumed to behave similarly;

see Results: Frugivore feeding behavior). I assumed that seeds from fruits entirely

removed from trees were swallowed and, since chimpanzees were never observed

defecating M myristica seeds while feeding in conspecific trees, defecated away from

parent trees in all cases. Since it is possible that chimpanzees only consume part of a

fruit they carry away, this may overestimate the probability that seeds were deposited

clumped in dung.

The probability that seeds from opened fruits dropped under the canopy remained

non-dispersed under the canopy or were dispersed by frugivores, incorporated additional

data from the experiment examining fate of seeds from fruits dropped below

conspecifics. These probabilities were determined as the product of the average number

of partially eaten fruits on the ground, proportion of fruit remaining (set at 0.5), and

average number of seeds from the fruit experiment that remained under the canopy, or

were removed by each frugivore, respectively.

Stage-specific probabilities of seed germination, seedling emergence, and

seedling establishment (Fig. 3) were determined from the experiments quantifying fate of

seeds under conspecific canopies (spit singly and in dropped, partially-eaten fruits) and

away from conspecific canopies (spit singly and defecated in clumps with dung). For

each seed deposition type (dropped in fruit, spit singly under, spit singly away, or

defecated in clumps) germination probability was the average number of seeds that

germinated from each experiment. For each seed deposition type, emerged-seedling








probability was determined as the average number of seedlings that survived the emerged

seedling stage out of the number of seeds that germinated (e.g., emerged seedling

survivorship). Similarly, established-seedling probability was determined as the average

number of seedlings that survived until 13 months post-experiment initiation out of the

number of emerged seedlings (e.g., established seedling survivorship). Cumulative

probability of seedling recruitment for each deposition type was then determined for each

site in each year as the product of the three stage-specific probabilities (Fig. 3).

The absolute number of seedlings per tree expected to recruit from each

deposition type was estimated as the product of the average seed crop size and the

probabilities that seeds were deposited in, and seedlings recruited to, each type/location.

To estimate expected seedling density for the population as a whole (i.e., Kanyawara vs.

Dura) in each year, at each site, I multiplied the sum of the number of seedlings per tree

for each type/location by adult tree density and proportion of trees fruiting in that year.




Results

Primate Density

Frugivorous primate group density was higher at the Dura site for all three large-

bodied primates (P. troglodytes, P. anubis, and L. albigena) and two of the small-bodied

primates (C. ascanius and C. I'hoesti, Table 1). In contrast, while C mitis was present at

the Kanyawara site, it was never seen at the Dura site during this study (although

sightings of a single individual at this site have occurred on two occasions; Chapman

pers. comrn.).








Plant Fecundity

Adult M myristica density was higher at Dura (3.6 4.0 trees/ha) than at

Kanyawara (1.0 1.9 trees/ha; logistic regression with poisson error and log link for

effect of site: y2 = 19.53, P < 0.001). The proportion of trees that fruited (carried fruit to

maturity) did not differ between sites for either Year 1 (Kanyawara = 39.0%, N = 41

trees; Dura = 45.0%, N = 40 trees; chi-square with Yates' correction term = 0.19, P >

0.05) or Year 2 (Kanyawara = 34.1%, N = 41 trees; Dura = 27.5%, N = 40 trees; X2 =

0.15, P > 0.05). Number of mature fruits produced per tree differed between years

(logistic regression with binary error and logit link: X2 = 5.37, P = 0.02), but not between

sites (x2 = 0.43, P = 0.51); trees on average produced larger fruit crops in Year 2,

regardless of site (Table 2). The increase in fruit crop size from Year I to 2 was

consistent between sites (i.e., the site-by-year interaction was non-significant; X2 = 0.00,

P = 0.98). When fruit crop was defined as number of immature and mature fruits

combined (data available only for Year 2) there was still no significant difference

between sites.(x2 = 0.63, P = 0.43; Table 2).



Fruit Removal

Immature fruits

Based on fruit crop counts in Year 2, 50.2 28.7% (N = 15 trees) and 66.3

33.8% (N = 18 trees) of M myristica fruit crops at Kanyawara and Dura, respectively,

were lost prior to maturation (effect of site: X2 = 1.68, P = 0.19). Predation of immature

fruits by primates did not differ between sites (X2 = 1.53, P = 0.22), although primate fruit

predation was the primary cause of pre-dispersal fruit loss at both sites; evidence from








immature fruit husks indicated that mangabeys consumed 47.9 29.6% and 55.8

35.1% of fruit crops at Kanyawara and Dura, respectively. Overall, 97.0% of trees

(32/33) experienced some level of immature fruit predation by mangabeys; the one

exception was a tree that produced only 1 fruit that was aborted early in the season. In

contrast, only 33.3% and 50.0% of trees at Kanyawara and Dura, respectively, lost

immature fruit due to abortion; of those trees that aborted fruits, 5 trees at Kanyawara

aborted 7.1 4.8% of their fruit crop, while at Dura 8 trees aborted 11.1 10.9% of their

fruit crop, and 1 additional tree aborted the single fruit it produced. The overall

proportion of immature fruits aborted at both sites was relatively low and did not differ

between sites (Kanyawara = 2.4 4.3%, Dura = 10.5 24.1%; x2 = 0.00, P = 0.97).



Mature fruits

In Year 1, based on fruit crop counts and husk remnants, all mature fruit were

handled (bitten into or removed) by large primates. Chimpanzees and baboons were

primarily responsible for removing 76.3 32.5% (N = 15 trees) and 79.6 29.1% (N =

17 trees) of M myristica mature fruit crops at Kanyawara and Dura, respectively (site

effect: X2 = 2.31, P = 0.13). The remaining mature fruits (Kanyawara = 23.7 32.5%,

Dura = 20.4 29.1%) were partially (approximately 50%) consumed, and either left

attached to the tree or dropped below the parent canopy. Although number of partially

eaten fruits dropped to the ground did not differ between sites (Kanyawara = 3.9 7.0%;

Dura = 9.0 17.7%; X2 = 0.00, P = 0.99), more fruits were left attached to the tree at

Kanyawara (19.8 29.7%) than at Dura (11.4 26.1%; X2 = 4.23, P = 0.04). From








observations on non-focal trees, the few fruits not handled by primates remained attached

to the tree and eventually turned black and rotted.



Frugivore Feeding Behavior

Frugivores were present in the vicinity of focal fruiting trees on all observation

days, although visitation rates to focal trees were low (12.6% of 3,288 15-min scans,

Table 3). All species of frugivorous primates were observed feeding from ripe fruits in

focal trees, except the primarily terrestrial l'hoesti monkey which is a secretive species

(Kaplin and Moermond 1998). Primates treated fruits differently depending primarily on

their body size. Chimpanzees and mangabeys were the only frugivores observed biting

through the outer husk of unopened mature fruits (Table 3), although teeth marks on

fallen fruit indicated that baboons are also capable of opening fruits. Chimpanzees

assessed state of fruit ripeness by smelling the fruit prior to biting into or removing it; in

general they did not bite into unripe fruit nor did they drop unopened fruit to the ground.

They were observed on occasion carrying a whole fruit away from the area. Because

chimpanzees and baboons are not habituated at the Dura site, only a few observations of

feeding on mature M myristica fruits were collected for these species at this site.

However, teeth marks present on 80.8% (N = 52) and 75.7% (N = 70) of M myristica

husks at Kanyawara and Dura, respectively, indicated that chimpanzees and baboons

were responsible for removing the majority of mature fruit at both sites. None of the

small-bodied primates were ever observed opening fruits; 100% of their feeding samples

were from fruits previously opened and left attached by their pedicels to the tree.

Chimpanzees primarily swallowed pulp and seeds and only occasionally spat

seeds out under the parent tree (Table 4). Thus, the majority of seeds they handled








passed through their gut and were deposited intact in clumps with dung on the forest floor

away from the parent tree (see also Lambert 1999). Based on dung collections, it was

evident that this was also the fate of seeds consumed by baboons. In contrast, the

cercopithecines never swallowed M myristica seeds whole. Mangabeys, the largest of

the cercopithecines, were primarily responsible for depositing single seeds away from the

parent tree; mangabeys stored mature seeds in their cheek pouches, then processed the

pulp and spat seeds out once they left the parent tree (Table 4). However, one on

occasion they were observed cracking open mature seeds and feeding on the endosperm.

Redtail monkeys, in contrast, processed pulp and spat the majority of mature seeds

directly under the parent M myristica tree (Table 4; see also Lambert 1999). Blue

monkeys at Kanyawara processed seeds similarly, although they deposited seeds both

under and away from the parent tree with roughly equal frequency (Table 4). The only

other frugivore observed feeding from an open fruit was a large-bodied bird, the Great

Blue Turaco. On this one occasion, the turaco removed only pulp from the fruit, and no

seeds were removed or dropped in the process.



Fate of Seeds from Dropped Fruits

In general, removal of experimental seeds and pulp from under the parent tree

occurred primarily within the first week. Dispersal or predation of seeds that remained

thereafter was relatively rare. The logistic regression indicated that both site and year

strongly affected the average number of seeds removed (Table 5); overall more seeds

were removed from fruit experiments at Kanyawara than at Dura and in Year 1 than 2.

There was no significant year-by-site effect (Table 5), indicating that the decrease in

removal from Year 1 to 2 was consistent between sites. Because experiments were








established at different times during the fruiting season in the two years, seasonality may

also contribute to differences observed between years.

The primary removal agent differed between sites (Table 6). The logistic

regression indicated that site strongly affected the average number of seeds removed by

primates and by rodents but year did not (Table 5). Regardless of year, primates removed

more seeds at Dura than at Kanyawara, and rodents removed more seeds at Kanyawara

than at Dura (Tables 5 and 6). Greater seed and pulp removal by primates at Dura is in

accordance with higher terrestrial frugivorous primate densities at that site (Table 1).

Chimpanzees were the only primates who removed seeds from fruit experiments at

Kanyawara. In contrast, at Dura, of seeds removed by primates, chimpanzees and

baboons removed the majority (82%) in Year 1, with l'hoestis removing the remaining;

this pattern was reversed in Year 2 with l'hoestis removing 60% of seeds, and

chimpanzees and baboons removing the remaining 40%. The African civet and the red

duiker removed only a small percentage of seeds from Kanyawara in Year 1 and from

Dura in Year 2, respectively (Table 6).

In Year 1, less than 10% of seeds from fruit experiments remained under the

parent tree at both sites, and all were infested by curculionid beetles which led to seed

death (Table 6). In Year 2, although beetle infestation of seeds that remained under the

parent tree was lower, and germination was higher (Table 6), only 1.2% of seeds placed

out in seed experiments (3/260) at Dura, and none at Kanyawara, survived under the

parent tree to become established seedlings.








Fate of Spit Seeds

The two main causes of mortality to experimental spit seeds were curculionid

beetles and rodents (Table 7). Entrance holes where adult beetles had bored into the

seeds to lay eggs were seen within 1 week of the experiment's start. The logistic

regression indicated there was a non-significant trend in the effect of location (under vs.

away from the "parent" tree canopy) on seed mortality due to beetles (Table 8). There

were significant site and year effects on seed mortality due to beetle infestation (Table 8);

beetle infestation was greater at Dura than at Kanyawara and in Year 1 than in Year 2

(Table 7). These trends were consistent among sites, years, and locations as indicated by

non-significant interaction terms (Table 8).

Direct evidence of rodents consuming seeds, such as outer seed coat remnants,

was relatively rare. However, when consumed and removed seeds were combined,

logistic regression indicated there was a significant effect of location on seed removal by

rodents; the proportion of seeds removed by rodents was greater under the canopy than

away (Table 8). Site and year also affected rodent seed removal (Table 8); removal was

greater at Kanyawara than at Dura and in Year 1 than in Year 2 (Table 7). These trends

were consistent among sites, years, and locations (n.s. interactive effects; Table 8).

Spit seeds began germinating 10 weeks after initiation of the experiment in Year

1, and 4 weeks after initiation in Year 2, regardless of site or location. Location

significantly influenced seed germination (Table 8); seed germination was significantly

lower under the canopy than away (Table 7). The increased probability of germination

away from the canopy was consistent between sites (i.e., n.s. location x site interaction;

Table 8). Site and year also significantly influenced seed germination (Table 8). Seed

germination was significantly higher at Kanyawara than at Dura, and in Year 2 than in








Year 1 (Table 7). However, proportional changes in germination between years were not

consistent, likely due to low germination of seeds in Year 1 at Dura away from the

canopy [i.e., the increase in germination from Dura to Kanyawara was greater in Year 1

than in Year 2 (sig. site x year) and the increase in germination from under to away was

greater in Year 1 than in Year 2 (sig. location x year; Table 8)].



Fate of Seeds in Dung

Primary causes of mortality to seeds placed in baboon dung differed between sites

and varied between years within sites (Table 9). As with spit seeds, the two main causes

of mortality were curculionid beetles and rodents. Beetle entrance holes were seen within

1 week of the experiment's initiation (the exception was Kanyawara in Year 1, where

they were not seen until the third week). Proportion of seeds infested and subsequently

killed by beetles was greater at Dura than at Kanyawara and in Year 1 than in Year 2

(Tables 9 and 10). The decrease in beetle infestation from Year 1 to 2 was consistent

between sites (n.s. site x year; Table 10). Beetle infestation was the main cause of seed

mortality at Dura in both years, whereas at Kanyawara, beetles infested and killed

relatively fewer seeds in Year 1 and none in Year 2 (Table 9).

In contrast, rodent consumption of seeds was greatest at Kanyawara (Table 9). As

with the spit seeds, direct evidence of rodent seed consumption was rare. However, when

consumed and removed seeds were combined, logistic regression indicated that seed

removal by rodents was greater at Kanyawara than at Dura and in Year 1 than in Year 2

(Table 10). There was no year-by-site interaction effect (Table 10). Thus, rodent

predation was the main cause of seed mortality at Kanyawara in both years, whereas they

removed relatively fewer seeds at Dura (Table 9).








Dung beetle activity differed between sites and years (Table 9). In the seeds-in-

dung experiment, seeds buried as a result of dung beetle activity (N = 73 of 1000 seeds

overall) were typically found between 0.1 and 2.0 cm below the soil surface directly

under the dung pile; in one case, a seed was found buried 3.0 cm deep. Although number

of seeds buried by dung beetles was relatively low at both sites in both years (Table 9), a

greater proportion of seeds were buried at Kanyawara than at Dura and in Year 2 than in

Year 1 (Table 10). There was no year-by-site interaction effect (Table 10).

In Year 1, seeds began germinating 6 and 10 weeks after the seeds-in-dung

experiment was initiated at Kanyawara and Dura, respectively. In Year 2, seeds began

germinating 4 weeks after the experiment was initiated at both sites. In all cases, by the

time seeds germinated, dung had either been completely removed by dung beetles or had

turned to soil. Both site and year strongly affected the average number of seeds that

germinated from 10 seeds placed in baboon dung (Table 10). Germination was greater at

Kanyawara than at Dura and in Year 2 than in Year 1 (Table 9). The increase in

germination from Year 1 to 2 was consistent between sites (n.s. site x year; Table 10).



Seedling Survival from Spit Seeds

In Year 1 at Dura, only 1 seed under the canopy (N = 100) germinated; it was

infested by curculionid beetles and did not survive to become a seedling (Table 11).

Therefore, seedling survival for experimentally spit seeds at Dura in Year I was not

analyzed further. At Kanyawara, emerged seedling survivorship was greater away from

the canopy than under, and in Year 2 than in Year 1 (Table 12). In Year 2, survivorship

was greater away from the canopy than under and at Kanyawara than at Dura (Table 12).

The main identifiable causes of mortality to emerged seedlings were beetle or rodent








damage to the seed upon which the emergent seedling was still dependent for its reserves

(Table 11). Differences in prevalence of these factors among location, site, and year

were not tested due to small sample sizes. However, there appears to be a trend of

greater beetle damage under the canopy than away at each site within each year, and in

Year 1 than in Year 2 (Table 11). Rodent predation on seeds was more prevalent at

Kanyawara and virtually non-existent at Dura, although there was no consistent pattern

with location or year (Table 11).

Established seedling survivorship for experimentally spit seeds at Kanyawara was

greater away from the canopy than under but did not differ between years (Table 12). In

Year 2, established seedling survivorship was greater away from the canopy than under

and at Kanyawara than at Dura (Table 12). The two main identifiable causes of mortality

to established seedlings were desiccation and herbivory, either by insects eating the root

or shoot or by mammals eating leaves and stems. Differences in their prevalences among

location, site, and year were not tested due to small sample size, and no consistent trends

were apparent (Table 11).



Seedling Survival from Seeds in Dung

In Year 1 at Dura, only 5 seeds, all from a single station, germinated and none

survived to become seedlings (Table 13). Therefore, as with spit seeds, seedling survival

for seeds-in-dung at Dura in Year 1 was not analyzed further. At Kanyawara, emerged

seedling survivorship was greater in Year 2 than in Year 1 (Table 12). In Year 2,

survivorship was greater at Kanyawara than at Dura (Table 12). There was no trail effect

at this stage. Because of the small number of seedlings emerging per station, data

regarding causes of mortality were pooled across stations. As with spit seeds, the main








identifiable cause of mortality at both sites in both years was beetle infestation of seeds

upon which seedlings were still dependent (Table 13). Rodent predation of seeds was

greatest at Kanyawara in Year 1 and non-existent at Dura. Desiccation of emerged

seedlings was prevalent at Kanyawara in Year 2 and at Dura in Year 1.

At Kanyawara, as with spit seeds, there was no difference in established seedling

survivorship from seeds-in-dung between years, and survivorship was greater at

Kanyawara than at Dura in Year 2 (Table 12). Sources of mortality were pooled across

stations in the same manner as with emerged seedlings. Dessication was the major cause

of mortality to established seedlings in Year 1 (data only available for Kanyawara), while

herbivory was the major cause of mortality in Year 2 at both sites (Table 13).

The pattern of seedling establishment from seeds that were buried under 1 cm of

soil was similar to that of the spit-seed and seeds in dung experiments. More seedlings

became established at Kanyawara (28.0%) relative to Dura (10.0%; n = 50 seeds per site;

X2 = 5.26, P = 0.02).



Natural Seedling Density and Survival

Data on M myristica seedling growth from the spit-seed and seeds-in-dung

experiments were used to determine average height of seedlings one year after seeds land

on the forest floor. Final seedling height 13 months after seeds were placed on the forest

floor for all surviving seedlings was 12.9 4.2 cm (N = 159) at Kanyawara and 12.5 +

2.3 cm (N = 22) at Dura (t = 0.45, df= 179, P = 0.65); seedling height ranged from 2.0 -

23.0 cm, both sites combined. Since age of naturally occurring seedlings was unknown, I

used these data to determine the smallest size class that would best represent a single








year's cohort of seedlings. Thus, seedlings < 0.2 m tall ("small" seedlings) were those

considered to have established within the previous year.

Monodora myristica density differed between the two sites for all size classes

(Table 14). While seedling and tree densities were lower at Kanyawara compared with

Dura, sapling and pole densities were higher. At Kanyawara it was rare to see seedlings

under conspecific canopies (only one small seedling was found under an adult

conspecific in 50 plots), while at Dura there were often carpets of young seedlings

beneath adults (as many as 265 small seedlings were found in a 10 m2 area under an

adult). This resulted in a 30-fold difference in small seedling density between

Kanyawara and Dura.

Small seedlings under adults had lower survivorship than did those located at least

0.5 m away from the edge of an adult canopy (data only available for Dura; under =

40.0%, N = 45; away = 63.0%, N = 54; X2 = 5.19, P = 0.02). In contrast, large seedlings

(0.2 0.5 m tall) did not differ in survivorship between locations (under = 83.3%, N =

36; away = 85.9%, N = 64; x2 =0.12, P =0.73). Saplings (0.5 2 m tall) were rarely

under adults (2 saplings at Kanyawara and 0 at Dura), and poles (> 2 m tall) were never

under adults at either site.

When only those individuals located away from adults were included in the

analyses, densities of all size classes were still different between sites (Table 14).

Although density of both small and large seedlings was lower at Kanyawara than at Dura

(Table 14), small-seedling survivorship and relative-height growth rate (RHGR) over an

18-month period were greater at Kanyawara, with a similar trend for large seedlings

(Table 15). In contrast, while sapling and pole densities were higher at Kanyawara than








at Dura (Table 14), sapling survivorship and RHGR did not differ between sites (Table

15).

Desiccation, although low, was the primary identifiable cause of mortality for all

size classes at both sites (Table 15). Cause of mortality for remaining individuals was

unable to be determined since they disappeared during the six months between

monitoring.



Overall Probability of Dispersal and Seedling Recruitment

Although pre-dispersal predation by primates caused substantial reductions in the

initial fruit crop, once the seeds were mature the estimated probability that primates

removed seeds from mature fruit was high (0.95) at both sites (Fig. 2). Furthermore, the

probability of dispersal away from the parent tree canopy was greater than 0.90 at both

sites. I estimated that seeds had a much higher chance of being deposited by large-bodied

primates in clumps surrounded by dung (> 0.85) than being spat out singly (< 0.06) at

both sites. The probability that seeds from fruits dropped under the canopy were

dispersed by primates was low overall; however, I estimated that a seed had a higher

chance of being dispersed from these dropped fruits at Dura than at Kanyawara.

The probability that deposited seeds germinated, became emerged seedlings, and

eventually established seedlings, varied greatly depending on deposition type, location,

year, and site (Fig. 3). If the lowest probability indicates the most critical stage in the

seedling recruitment process (Clark et al. 1999, Rey and Alcantara 2000), then at both

sites the most critical stage, in Year 1, was seed germination and, in Year 2, established-

seedling survival. This trend held regardless of deposition type or location with two

exceptions: at Dura in Year 1, seeds spit singly under the canopy and seeds deposited in








clumps surrounded by dung both had the lowest survival probability of emerged

seedlings. However, when within-site, inter-annual differences in stage-specific

probabilities were compared, the magnitude of change from Year 1 to Year 2 in the

probability of making it to each stage was greatest at the seed germination stage and

decreased with the progression through the stages. Despite the large inter-annual

differences in germination probabilities within a site, no significant difference in

established seedling probabilities existed between the two years. Therefore, seed

germination sets the stage for the absolute number of seedlings that emerge, and the

probability of seed germination is more critical than the probability of established

seedling survivorship. This evaluation thus indicates that germination is the limiting

stage regardless of site or year.

Cumulative probability of seedling recruitment under the parent canopy was very

low, although in three year/site combinations, experimental single-spit seeds (with pulp

removed) had a higher estimated recruitment probability than did those that remained in

pulp in partially eaten fruits (Fig. 3). Cumulative probability of seedling recruitment

from the seed experiments was greatest for single seeds away from the parent at both

sites in both years, although at Kanyawara in Year 2, cumulative probability of seedling

recruitment was equally high for seeds in clumps with dung.

When the absolute number of seedlings recruiting into the population was

calculated based on average fruit crop size per tree, total expected seedling recruitment

was 33 times greater at Kanyawara than at Dura in Year 1, and 2 times greater in Year 2.

Despite the high cumulative probability of seedling recruitment from seeds spit singly

away from the parent, 86 94% of seedlings contributing to that year's seedling








population were estimated to come from seeds deposited in clumps with dung. This was

the case at Kanyawara in both years and at Dura in Year 2; however, at Dura in Year 1,

the only seedlings projected to recruit into the population were from single seeds spit

away from the canopy.

Another potential source of seedling recruitment is secondary dispersal of seeds

by rodents (Forget et al. 1998, Brewer and Rejmanek 1999, Wenny 1999). Since M

myristica seed removal by rodents varied between sites and years, the potential

contribution to seedling recruitment through seed caching may also vary. Although little

is known about caching behavior of small rodents in the tropics, in Belize, spiny pocket

mice (Heteromys desmarestianus) scatter-hoarded on average 9.5% of seeds removed

(Brewer and Rejmanek 1999). If I assume that 10% ofM. myristica seeds removed by

rodents were buried, and use data on seedling survival from M myristica seeds buried 1

cm below the soil surface at each site (28 and 10% survival at Kanyawara and Dura

respectively), I can extrapolate potential seedling recruitment from seeds buried by

rodents. Although these may be overestimates, using these data, rodents could

potentially contribute an estimated 76% and 98% of seedling recruitment in Year 1 at

Kanyawara and Dura respectively. This would reduce the relative contribution by both

large- and small-bodied primates to seedling recruitment in Year 1. However, in Year 2,

potential caching of seeds by rodents would not change estimated seedling recruitment at

either site, since few seeds were removed. Thus, large-bodied primates were still

estimated to be the major contributors to seedling recruitment in Year 2, contributing

94% of the seedlings at both sites.








Discussion

For frugivores to reliably contribute to plant population recruitment, the steps

leading from frugivore seed removal to seedling recruitment must be predictable in terms

of their demographic consequences (Herrera et al. 1994). Thus, the more reliably a seed

disperser visits fruiting trees, disperses seeds, and deposits them intact in sites that are

good for establishment, the more likely it will contribute to seedling recruitment and,

ultimately, the number of new adults added to the population (Howe 1989, Jordano 1992,

Schupp 1993). This study clearly demonstrates that M myristica, with its large, hard-

husked fruit, initially restricts access to its seeds to all but the largest-bodied arboreal

primates. Mature fruit do not drop from the tree unaided; when the few fruits not handled

by large-bodied primates eventually fell, the seeds were dried and rotten inside.

Moreover, large-bodied primates visited all fruiting M myristica trees monitored,

removed over 75% of the mature fruits, and could potentially have dispersed over 85% of

mature seeds. Thus, not only do these frugivores reliably visit and disperse mature seeds

of M myristica, but they are critical for initiating the seed dispersal process; only after

mature fruits have been handled by large-bodied primates are these seeds accessible to

other frugivores in the community.

The frugivorous primate community was estimated to have moved 90 97% of

mature M myristica seeds away from the parent tree. Removal of seeds away from

parent trees is particularly important since seeds located under conspecific canopies had

lower germination and seedling establishment probabilities. Furthermore, the lack of M

myristica poles under conspecifics indicates that recruitment into the adult population

may be difficult without seed dispersal (Chapman and Chapman 1995).








For dispersers to consistently contribute positively to future seedling recruitment,

they must reliably deposit seeds in sites suitable for seedling establishment (Schupp

1993). In three of four year-by-site combinations, 86-94% of seedlings estimated to

recruit into the population were from seeds deposited in dung by large-bodied primates

(based on potential seed crop production and seed deposition and seedling recruitment

probabilities). Despite lower cumulative probability of seedling recruitment from

experimental clumps of seeds in dung compared with single seeds, the vast majority of

seeds were likely ingested and defecated by large-bodied primates. Thus, not only are

large-bodied primates vital for initiating seed dispersal by opening mature fruits, but they

also play a major role disseminating seeds away from parent trees and can contribute

substantially to recruitment into the seedling population.

Frugivores may also contribute negatively to seedling recruitment. If individuals

move from one fruiting M myristica tree to the next, they may carry seeds away from the

parent tree yet still deposit them under conspecific canopies. In Kibale, Lambert (1999)

found that 25% of chimpanzee dung samples were found under the crowns of tree species

represented by seeds in the dung sample. However, of the 81 dung samples collected,

only two contained M myristica seeds and these were not located under M myristica

trees (Lambert pers. com.). Another negative effect of frugivores on plant recruitment

can occur through pre-dispersal seed loss via feeding on immature fruits. For example,

mangabeys were responsible for approximately 50% of M myristica fruit loss. Although

seed loss prior to maturation could have substantial effects on plant recruitment, in

general these consequences are not well documented (Louda 1995). Interestingly,

mangabeys also play a potentially important positive role in M myristica seed dispersal








since they are one of the few species capable of opening mature fruits and thus initiating

dispersal of mature seeds.

Despite the initial reliable seed dispersal interaction ofM. myristica with a limited

set of frugivores once the seeds are mature, spatial and temporal variation in processes

influencing mortality and growth of later stages disrupts the continuity of earlier

processes. This variation thus reduces the predictability of demographic consequences of

frugivore actions (Herrera et al. 1994). Although seed-deposition patterns were similar in

experiments between sites and years, the magnitude of effects of location, site, and year

on all post-deposition stages, from seed survivorship to seedling establishment, were

highly variable. For example, at Dura, the probability of germination from the seeds-in-

dung experiment varied from 2.1% in Year 1 to 69.2% in Year 2. Secondly, seeds had

the greatest probability of germinating at Kanyawara if they were experimentally

deposited away from the canopy as single seeds in Year 1 (35.8%), and as clumps of

seeds in dung in Year 2 (86.5%). Other studies have similarly shown that post-dispersal

seed survival varies among habitats and among years (Schupp 1988, 1990, Willson 1988)

and that spatial and temporal patterns of seed survival are generally unpredictable

(Willson and Whelan 1990).

By linking stage-sequential steps in the seed dispersal and seedling establishment

processes for M myristica, this study illustrates that although large-bodied primates are

crucial to initial stages of seed dispersal, once seeds are deposited, sources of mortality

and probability of survival vary greatly across a spatial scale of 15 km and a temporal

scale of 2 consecutive years. Based on these findings, I first evaluate several studies that

previously examined the effect of changes in frugivore abundances on seedling








abundances; second, discuss the ecological implications of this study; and third, discuss

the evolutionary implications of this study.



Effect of Changes in Frugivore Abundance on Seedling Abundance

Given M myristica's potential long-term co-existence with large-bodied

frugivores and its dependency on them to open mature fruits and initiate seed dispersal,

this system, if any, should exhibit a detectable link between disperser and plant densities.

When primate and standing-seedling densities were compared between sites, this may at

first appear to be the case; frugivorous primate density was three times greater, and M

myristica seedling density (< 0.2 m tall, away from conspecifics) was four times greater

at Dura compared with Kanyawara. However, when expected seedling densities were

calculated based on the average fruit-crop size and cumulative seedling-recruitment

probabilities, Dura had lower expected seedling density in both years than did

Kanyawara.

Previous studies have shown that changes in disperser abundances are associated

with changes in seedling densities (Dirzo and Miranda 1991, Chapman and Onderdonk

1998, Pacheco and Simonetti 1998, Wright et al. 2000) as well as spatial patterns of

seedling recruitment (Pacheco and Simonetti 1998). However, these studies had

conflicting outcomes. In Uganda and Bolivia, reduced numbers of large-bodied primates

were correlated with lower seedling densities of large-seeded forest trees species

(Chapman and Onderdonk 1998, Pacheco and Simonetti 1998) and higher seedling

aggregations around parent trees (Pacheco and Simonetti 1998). In contrast, in Mexico

and Panama, seedling densities were higher in areas with depleted mammalian

communities (Dirzo and Miranda 1991, Wright et al. 2000).








The different outcomes of these studies may be influenced by the fact that

seedling recruitment is a consequence not only of frugivore abundance, and thereby

dispersal of seeds, but also of a multitude of processes that operate after seed deposition.

Since these studies focused on changes in mammalian communities due to habitat

fragmentation or hunting, the effect of changes in frugivore populations on seedling

density was confounded by other possible effects of the anthropogenic disturbance. For

instance, poaching also reduces the abundance of seedling predators which may result in

increased seedling survival in areas with reduced frugivore density (Dirzo and Miranda

1991, Wright et al. 2000).

My study of M myristica demonstrated that, within an intact forest system,

variance in post-deposition processes, such as seed predation and seedling growth,

resulted in spatial and temporal variability in seedling recruitment across sites. For

instance, although no obvious changes in intra-site frugivorous primate densities were

discerned between the two years (i.e., no hunting occurs at either site), based on

experimental data, overall projected M myristica seedling recruitment from the

frugivorous primate community increased from Year 1 to Year 2 by 383% at Dura and

23% at Kanyawara. Moreover, fluctuations in recruitment between sites were variable as

well. Dura, the site with the higher large-bodied primate density, had either a 33 (Year 1)

or 2 (Year 2) times lower expected seedling density than did Kanyawara. Thus, even in a

system where the frugivores are reliable seed dispersers, such as with M myristica, it is

difficult to predict the number of seedlings recruiting into a given year's cohort at a given

site based on frugivore density alone.








Positive or negative associations between frugivore abundances and seedling

densities, as has been found in the studies discussed above, can be caused by processes

that operate independently of differences in number of seeds dispersed by frugivores.

Studies that compare standing-seedling densities among sites may also need to account

for differential growth and survivorship in seedlings among sites. Once seeds germinate,

rates of seedling growth and survivorship determine number of individuals recruiting into

the seedling bank, and thus seedling densities. In this study, M myristica seedlings in the

< 0.2 m size-class at Kanyawara survived better and grew faster then did those at Dura.

Thus, densities of seedlings in this size-class at the two sites may not represent equivalent

recruitment from a single fruiting season. Moreover, after an initial burst in height,

seedlings may remain within a given size-class for many years. For example, Connell

and Green (2000) found that after the first 2 years of growth, Chrysophyllum sp.

seedlings only doubled in height over a 27-year period (from 17.5 to 34.9 cm). Thus,

there may be an accumulation of many years of seedling recruitment into a single size-

class. Consequently, the lower M myristica seedling density at Kanyawara, compared

with Dura, may not result from lower densities of primates and subsequently lower seed

dispersal rates.



Ecological Implications

Several studies claim that seed deposition patterns presently generated by

frugivores create currently observed adult plant distributions (Garber 1986, Fragoso

1997, Stevenson 2000). Yet environmental variation can produce wide fluctuations in

seed production, dispersal, predation, germination, and seedling growth and survival, all

of which result in variation in recruitment patterns (Clark et al. 1999). This study found








temporal variation in fruit-crop production and both temporal and spatial variation in seed

predation, seed germination, and seedling emergence. Moreover, differences in

magnitude of change in these processes between sites, years, and deposition types varied

to a large degree. For example, beetle infestation and rodent predation of M myristica

seeds interacted to influence seedling recruitment probabilities, yet the magnitude with

which they affected a given year's seedling crop varied greatly. In Year 1 at Dura,

beetles were primarily responsible for lack of seedling recruitment through mortality both

at seed (85.0%) and emerged-seedling (60%) stages, while rodents played a more minor

role (12.9% and 0% respectively). In the same year at Kanyawara, beetles played a

minor role in mortality at the seed stage (19.2%) compared with rodents (52.9%).

However, at the emerged-seedling stage, they killed proportionally more seeds (31.4%)

than did rodents (11.8%).

Fluctuations in vectors of mortality and survival in consecutive steps, such as

illustrated above, may result in spatial and temporal variation in seedling recruitment

patterns. Thus, the relative importance of seed deposition in different conditions by

different frugivores may vary between years and sites. For example, although the

majority of M myristica seeds were projected to be deposited by large-bodied primates in

clumps with dung, the contribution of these seeds to the seedling cohort at Dura was

estimated to be 0% in Year 1 and 94% in Year 2. In Year 1 at Dura, 100% of seedling

recruitment was estimated to come from single seeds deposited by smaller-bodied

primates. Other studies have similarly found that initial seed-dispersion patterns can be

obscured by variation in post-deposition seed survival, germination, and seedling

recruitment (Herrera et al. 1994, Kollmann and Schill 1996, Houle 1998, Rey and








Alcantara 2000, Wenny 2000). This study has demonstrated that this is the case even for

a fruit-frugivore system where one would expect the link between disperser activity and

later stages in plant recruitment to be apparent.

Examining densities and size-class structure of older seedlings, saplings, and

poles may shed light on the long-term influence of frugivorous dispersers on plant

recruitment. In KNP, although substantial dispersal of seeds by Black-and-White

Hombills to sites below their nests altered seedling composition, advanced regeneration

(individuals > 0.5 m tall) at these sites 12-16 years later did not differ from control sites

(Paul 2001). Thus, Paul (2001) suggested that hornbill seed dispersal did not strongly

influence local species composition at later stages for tree species dispersed by hornbills.

Although current per capital M myristica seedling recruitment is similar at

Kanyawara and Dura (500 and 555 seedlings per adult, respectively), per capital sapling

and pole recruitment is higher at Kanyawara (170 saplings and 50 poles per adult) than at

Dura (22 saplings and 0.5 poles per adult). At Dura, low sapling-to-adult and pole-to-

adult ratios suggests poor recruitment into larger size classes (De Steven 1994) and thus

perhaps a decline in the population. At Kanyawara (K-30 compartment), in contrast,

despite the fact that frugivorous primate densities were three times lower at Kanyawara

compared with Dura, and have remained relatively unchanged at this site for at least the

last 28 years (Chapman et al. 2000), high sapling- and pole-to-adult ratios suggests

successful recruitment into later stages and thus an increase in the population.

Furthermore, temporal fluctuations in seed and seedling survivorship may result in

episodic recruitment into the seedling bank, which in turn may influence the structure of

subsequent size classes. Thus, even in a species that relies on a limited set of frugivores,








spatiotemporal variation in post-deposition processes contributes to a lack of concordance

between actions by frugivores and subsequent plant recruitment.



Evolutionary Implications

Frugivores may exert selective pressures on seed characteristics through

differential seed treatment and seed survival in deposition sites. Despite the fact that

large-bodied primates are reliable seed dispersers for M myristica, selective pressures on

seeds, once they are deposited, varied both on the temporal scale of 2 years and on the

spatial scale of 15 km. For instance, at Dura in Year 2, an estimated 96% of seedlings

recruiting into the population were from clumps of seeds in dung, whereas in Year 1,

100% of seedling recruitment was from single seeds.

Differences in the primary source of post-deposition seed mortality may result in

different selective pressures. For example, rodent predation at Kanyawara may select for

different seed characteristics than do beetle infestation at Dura. Since caching increases

with seed weight (Forget et al. 1998), rodent predation pressure at Kanyawara may select

for larger seeds. Larger seeds at Dura, however, may increase the infestation of seeds by

curculionid beetles; multiple adults can infest a single seed, and instances of a seed with

one or two beetle entry holes surviving to produce a seedling were observed during this

study. Although these two M myristica populations are separated by 15 km, they are

within the same continuous forest, and thus some gene flow between them is very likely;

elephants range throughout the park, and dispersal of individuals between neighboring

chimpanzee communities is probable. Thus, conflicting selective pressures (such as seed

mortality due to rodents vs. beetles), extensive gene flow, and spatial and temporal








variation in germination and recruitment sites all weaken potential selective pressures

exerted by primates on M myristica seed traits (Herrera 1985).

Differential fruit choice by primates, however, may exert considerable selective

pressure on traits that affect M myristica fitness. For example, through their actions as

pre-dispersal seed predators, primates may influence M. myristica husk thickness.

Transitions from unhusked to husked fruit may be due to selective pressures exerted by

mammals capable of feeding on, and dispersing the seeds of, husked forms (Janson

1992). However, present-day large-bodied primates uniformly depleted ripe M myristica

fruit crops at both sites, resulting in little variance in percent fruit removal among

individual trees.

On the other hand, pre-dispersal predation by some of the same species of large-

bodied primates, can significantly impact tree fitness. For example, mangabeys ate 27/28

immature fruits on one tree, and several trees with smaller fruit crops lost their entire crop

at the immature stage to mangabeys. Moreover, mangabey pre-dispersal predation of

young fruits and seeds was prevalent, affecting 97% of trees, and varied widely in

intensity, ranging from 4 100% of an individual tree's fruit crop. If there is differential

seed predation among plants and there is genetic variation for traits affecting rate of seed

predation (e.g., husk thickness), then the potential exists for evolution of seed-defending

mechanisms (Crawley 1992). Thus, primates may be selecting for harder husked M

myristica fruits through their actions as pre-dispersal predators.

An interesting conflict for M myristica arises once fruits are mature. The same

hard husk that may have served to protect soft, immature seeds may subsequently prevent

mature seed dispersal if frugivores cannot access the seeds and pulp. Thus, it is perhaps








not surprising that the primate species primarily responsible for pre-dispersal seed loss

was also among the few frugivore species capable of opening hard husks of mature fruit,

thus initiating seed dispersal of M myristica.



Conclusion

Even in a situation where vertebrate frugivores are critical to seed dispersal of a

fleshy-fruited plant, these frugivores' effect on that plant's recruitment varies due to

spatial and temporal variation in post-deposition processes. Thus, it is difficult to predict

densities of different life-history stages based only on frugivore densities or frugivore

behavior. Nevertheless, frugivores do have an important influence on plant populations.

For example, large-bodied primates are critical for initiating the seed dispersal process of

M myristica since they open fruits, potentially move the majority of seeds away from the

parent tree crown, and deposit seeds where they have a relatively high chance of

germinating and becoming established seedlings. Removal of seeds away from the

parent crown is important for M myristica since germination of seeds is lower under

conspecifics, and few saplings and poles were found under conspecific canopies.

Although dispersal by large-bodied primates plays an important role, post-

dispersal seed predators and seedling recruitment also influence density of M myristica

seedlings. Thus, it is difficult to predict consequences of changes in primate abundance

for M myristica recruitment due to inconsistent patterns of seedling recruitment at

varying spatial and temporal scales. Despite the fact that studies have justified examining

effects of frugivores on plant fitness by examining the seedling stage, there is little

evidence that patterns of seedling recruitment laid down by frugivores translate into

larger size classes and thus adult tree demography.











Pre-dispersal predation
Abortion


Figure 1. Fate diagram of M myristica recruitment indicating processes (ovals) influencing the immature seed to adult stages (boxes).












5Escape from pre-dispersal-_
predation by primates


I Immature seed
K D ^^ '*^ K D ^-------<
10.6160.461 Mature seed ^
K D



I,
I Handled by pri mates


Seed still in fruit
K 1)


Single spit seed Clumped seeds in dung
K D K !)
1 0.05610.026 1 0.854 0.941
2 0.056 0.034 2 0.857 0.927
Figure 2. Diagram of stage-specific probabilities for M myristica seed maturation, removal, and deposition in different locations and
conditions by primates at Kanyawara (K) and Dura (D), KNP, Uganda, in each year (1 and 2). Probabilities are extrapolated from data
on fruit crop counts and frugivore feeding samples (see text for details).









1Deposition location


[ Deposition tyeJ Dropped

Site K D
Germination 0
Year probability 2 0131 0.125


Site K D
Emerged I -
Year seedling
probability 2 0.189 0.90

Site K D
[Established -- T
Year seedling I 1 ..
probability 2 0 0.075


Spit singly
K* D
1 0.040 0.010
2* 0.590 0.560


K* D
1 0.500 0
2* 0.356 0.143


K* D
1
1 0.500 2,
2 0.238 0.125


Site
Cumulative K D K D K D K D

recruitment 2 0 0.002 2 0.0500.010 2 0.4080.075 2 0.412 0.054

Site
K D K D K D K D
Expected IIII
Year edl 1 0 0 1 0.5 0 7 0.4 1 44 0
Year 22eedllng ------ --- ---
desity/tree 2 0 0.1 2 4 0.7 2 69 11 2 1239 204

+ I I II


K D
1 0.002 0.00006
2 0.045 0.023


Figure 3. Diagram of stage-specific recruitment probabilities for M myristica seed
germination, seedling emergence, and seedling establishment for each seed deposition
type and location at Kanyawara (K) and Dura (D) in Years 1 and 2. Logistic regressions
were performed separately for each deposition site/type within each stage. If the test was
significant (P < 0.05), a indicates which site or year was greater. Shaded boxes were
not analyzed due to small sample size. The absolute number of seedlings per tree was
calculated based on the average fruit crop size per site per year using the probability of
seed deposition (see Fig. 2) and cumulative seedling recruitment. Total seedling density
per m2 further incorporated tree fruiting density/site/year.


( Away from canopy'


E | Cumped in
K* D K* D
1 0.358 0.067 1 0.213 0.021
2* 0.8170.692 2* 0.865 0.692


K* D K* D
1 0 442 0.250 1 0.366 0
2* 0.755 0.422 2* 0.736 0.315


K* D K* D

1 0421 0.500 1 0.322 -
2 0.662 0.257 2 0.647 0.247








Table 1. Group densities, or number of sightings, of diurnal frugivorous primates at
Kanyawara and Dura, KNP, Uganda, derived from line transect surveys (monkeys;
groups/km2) and nest counts (chimpanzees; nests/km2/day). Body weights are from
Kingdon (1997).

Species Density Body Mode of
weight (kg) travel

Kanyawara Dura
Cercopithecus ascanius 4.83 12.19 2-6 arboreal
C. mitis 1.00 0 4-12 arboreal
C. l'hoesti 3 sightings 5 sightings 3-10 terrestrial
Lophocebus albigena 1.13 3.29 4-11 arboreal
Papio anubis 1 sighting 5 sightings 11-50 terrestrial
Pan troglodytes 1.78 4.81 25-40 terrestrial








Table 2. Fruit crop size for adult M myristica trees at Kanyawara and Dura, KNP,
Uganda, for Year 1 and Year 2. Values are given as average number of fruit per tree 1
SD (number of fruiting trees in parentheses).


Kanyawara


Dura


Mature fruits only
Year 1


Year 2


Mature + immature fruits


Year 2


7.93 11.99
(15)
13.42 12.65
(14)


26.47 34.06


6.41 10.11
(17)
16.36 25.98
(12)


20.67 14.25


(15) (18)


(15)


(18)








Table 3. Interactions between different frugivores and fruiting M myristica trees as determined by focal tree observations conducted
at Kanyawara and Dura, KNP, Uganda. Presence in the area was indicated by sightings and calls of frugivores in the vicinity of the
focal tree. For frugivores that were never observed in the focal tree, no feeding data were recorded "---".

Site/Interaction Chimpanzee Baboon Mangabey Redtail Blue L'hoestis Great Blue B&W Squirrel
monkey monkey monkey Turaco Homrnbill


Kanyawara
Days present in area (%)
(n=62 days)
Scans present in tree (%)
(n=1877 scans)
Feeding rate (bites/min)*
(n)
No. times observed
opening fruits


Dura
Days present in are,
(n=53 days)
Scans present in tre
(n=1411 sca
Feeding rate (bites/i
(n)
No. times observed
opening fruits
* Mean SD


a (%)

e(%)
mns)
min)*


45.2

1.7

9.1 4.1
(52)


49.1

0


53.2

5.0


--- 11.4 5.1
(25)


43.4

0.3

12
(1)

0


75.5

0.9

8.6 7.6
(17)

2


53.2

3.1

7.0 3.0
(37)

0

69.8

3.0

6.7 5.3
(23)


48.0

2.0

6.9 4.7
(20)

0

0

0


8.1

0.1

0


0

17.0

0.1


77.4

0.2


50.9

0


85.5

0


66.0

0


14.5

0.1

0


0

7.6








Table 4. Fate of M myristica seeds processed by primate frugivores observed during focal tree watches at Kanyawara and Dura,
KNP, Uganda. N = number of feeding samples in which seed fate was recorded. Seed fate percentage is expressed as a function of
total number of feeding samples for each species where fate was observed.

Seed Fate (%)

Frugivore Site N Swallowed Spit under Spit away from Preyed upon

parent canopy parent canopy (killed)

Chimpanzee Kanyawara 51 92.2 7.8 0 0

Mangabey Dura 15 0 13.3 86.7 0

Kanyawara 22 0 0 95.5 4.5

Redtail Dura 15 0 93.3 6.7 0

Kanyawara 27 0 74.1 25.9 0

Blue Monkey Kanyawara 11 0 45.5 54.5 0








Table 5. Effect of site and year on fate of M myristica seeds from fruit experiments
where 20 seeds with pulp were placed in tracking stations under fruiting adult trees in
KNP, Uganda. Logistic regressions with binomial error distribution and logit link were
used; df = 1 in all cases. Directional differences are indicated when P < 0.05 (K =
Kanyawara, D = Dura, 1 = Year 1, 2 = Year 2). For averages see Tables 6.

Seed fate X P-value Direction
Factors and interactions
Overall removal
Site 21.74 < 0.001 K>D
Year 12.79 < 0.001 1>2
Site x Year 0.17 0.68
Removal by primates
Site 24.44 < 0.001 K Year 1.50 0.22
Site x Year 1.73 0.19
Removal by rodents
Site 20.38 < 0.001 K>D
Year 1.82 0.18
Site x Year 0.03 0.87








Table 6. Fate of M myristica seeds from experiments designed to mimic fruit dropped by primates under fruiting trees at Kanyawara
and Dura, KNP, Uganda for Year 1 and Year 2. Percentages are expressed as the mean proportion of seeds per 20 seeds 1 SD. N =
number of replicates. Seeds that remained under the tree were monitored until 13 months after the experiment was started.

Seeds removed Seeds remaining

Year Site N Primates Other Rodents Missing Infested by beetles Germinated Germinated and

(%) frugivore (%) (%) (%) (dead) (%) and died (%) survived (%)

1 Kanyawara 13 7.7 27.7 14.6 35.7* 68.8 47.8 8.1 27.7 0.8 2.8 0.0 0.0

Dura 13 78.8 28.4 0.0 12.3 25.2 0.4 1.4 8.5 11.3 0.0 0.0

2 Kanyawara 13 12.3 30.1 0.0 49.6 48.4 21.2 31.3 5.0 12.4 11.9 18.3 0.0

Dura 13 50.8 49.2 3.9 13.9t 4.6 15.2 5.8 13.7 27.7 34.4 6.5 9.7 1.2 3.0

* African civet
t Red duiker








Table 7. Fate of M. myristica seeds from experiments where single seeds were placed under and away from conspecific adult
canopies to mimic spitting by small-bodied primates at Kanyawara and Dura, KNP, Uganda, for Year 1 and Year 2. Percentages are
expressed out of the total number of seeds placed out per site per year (under: N = 100 seeds/site/year; Away: N = 120
seeds/site/year).

Kanyawara Dura
Year Year 2 Year 1 Year 2
Fate Under Away Under Away Under Away Under Away
Infested by beetles 37.0 30.0 6.0 1.7 83.0 80.8 35.0 20.0
Consumed by rodents* 13.0 2.5 1.0 0.0 6.0 3.3 2.0 2.5
Removed 46.0 30.8 33.0 15.0 10.0 9.2 7.0 6.7
Germinated 4.0 35.8 59.0 81.7 1.0 6.7 56.0 69.2
* remains of eaten seeds at the station as evidence
t seeds removed from the stations (presumably by rodents)








Table 8. Effect of location, site, and year on fate of M myristica seeds from experiments
where single seeds were placed under and away from adult trees, at KNP, Uganda.
Logistic regressions with binomial error distribution and logit link were used; df= 1 in all
cases. Directional differences are indicated when P < 0.05 for main effects (U = under
adult canopy, A = away from adult canopy, K = Kanyawara, D = Dura, 1 = Year 1,2 =
Year 2). For averages see Table 7.

Seed fate; X P-value Direction
Factors and interactions


Beetle infestation
Location


Site
Year
Location x Site
Location x Year
Site x Year
Rodent removal
Location
Site
Year
Location x Site
Location x Year
Site x Year
Germination
Location
Site
Year
Location x Site
Location x Year


Site x Year


3.16
69.65
91.68


0.15
1.13
0.05


8.44


29.84
10.34


2.56
0.03
0.76


32.08
19.14
183.15


1.12
6.78
8.25


0.07


< 0.001
< 0.001


K 1>2


0.70
0.29
0.83


<0.01
< 0.001
0.001


U>A
K>D
1>2


0.11
0.86
0.38


< 0.001
< 0.001
< 0.001
0.29
<0.01
<0.01


U K>D
1<2








Table 9. Fate of M myristica seeds from experiments where 10 seeds were placed in 60 grams of baboon dung at Kanyawara and
Dura, KNP, Uganda for Year 1 and Year 2. Percentages are expressed as mean proportion of seeds per 10 seeds 1 S.D. N = 24
replicates per site in Year 1, and 26 replicates per site in Year 2.


Kanyawara Dura
Fate Yearl Year2 Year Year 2
Infested by beetles 19.2 25.4 0.0 85.0 25.7 23.1 29.5
Consumed by rodents* 23.3 36.3 0.0 0.4 2.0 0.4 2.0
Removed 29.6 39.2 12.7 23.1 12.5 24.9 5.0 8.6
Buried by dung beetles 7.5 19.4 15.0 22.3 1.7 4.8 4.6 11.4
Germinated 21.3 23.7 86.5 22.8 2.1 10.2 69.2 29.7
* remains of eaten seeds at the station as evidence
t seeds removed from the stations (presumably by rodents)
Buried directly under the station by tunneler dung beetles








Table 10. Effect of site and year on fate of M myristica seeds from experiments where
10 seeds were placed in 60 grams of baboon dung, at KNP, Uganda. Logistic regressions
with binomial error distribution and logit link were used; df= 1 in all cases. Directional
differences are indicated when P < 0.05 (K = Kanyawara, D = Dura, 1 = Year 1, 2 = Year
2). For averages see Table 9.

Seed fate P-value Direction
Factors and interactions
Beetle infestation


Site


Year


Site x Year


88.66
77.93


1.65


< 0.001
< 0.001


K 1>2


0.20


Rodent removal
Site
Year
Site x Year
Dung beetle removal
Site
Year
Site x Year


Germination
Site


Year


Site x Year 2.84


21.05
22.31


1.53


< 0.001
< 0.001


K>D
1>2


0.22


< 0.001
0.04


K>D
1<2


10.99
4.41
0.07


18.05
148.94


0.79


< 0.001
< 0.001
0.09


K>D
1<2


Site x Year


2.84








Table 11. Causes of M myristica seedling mortality for emerged and established seedlings from seed experiments of single seeds
placed under and away from conspecific canopies at Kanyawara and Dura, KNP, Uganda, for Year 1 and Year 2. Percentages are
expressed with respect to total number of seedlings in that stage per site and year (in parentheses).
Kanyawara Dura
Year Year2 Year Year 2
Fate Under Away Under Away Under Away Under Away
Emerged seedlings (4) (43) (59) (98) (1) (8) (56) (83)
Survived 50.0 44.2 35.6 75.5 0 25.0 14.3 42.2
Died
Beetles* 50.0 34.9 1.7 0 100.0 75.0 5.4 0
Rodents* 0 11.6 13.6 6.1 0 0 1.8 0
Herbivoryt 0 2.3 0 0 0 0 0 0
Dessication 0 4.7 0 1.0 0 0 0 0
Unknown 0 2.3 49.2 17.4 0 0 78.6 57.8
Established seedlings (2) (19) (21) (74) (0) (2) (8) (35)
Survived 50.0 42.1 23.8 66.2 -- 50.0 12.5 25.7
Died
Herbivoryt 0 15.8 9.5 12.2 0 12.5 8.6
Dessication 50.0 21.1 42.9 9.5 0 25.0 25.7
Unknown 0 21.1 23.8 12.2 50.0 50.0 40.0
* seed destroyed while emerged seedling was still dependent on its reserves
t stems, leaves or roots eaten or damaged by either mammalian and insect herbivores








Table 12. Effect of location, site, and year on seedling survival of M myristica seeds
from spit seed and seeds in dung experiments in KNP, Uganda (see text for details).
Logistic regressions with binomial error distribution and logit link were used; df= 1 in all
cases. Directional differences are indicated when P < 0.05 (U = under adult canopy, A =
away from adult canopy, K = Kanyawara, D = Dura, 1 = Year 1,2 = Year 2). For
percentage values see Tables 11 and 13.

Factor P-value Direction
Spit seed experiment
Emerged seedling survival
Kanyawara Location 21.91 < 0.001 U Year 10.17 <0.01 1<2
Year 2 Location 37.57 < 0.001 U Site 28.13 < 0.001 K>D
Established seedling survival
Kanyawara Location 10.57 <0.01 K> D
Year 2.52 0.11
Year 2 Location 12.42 < 0.001 U Site 16.09 < 0.001 K>D
Seeds in dung experiment
Emerged seedling survival
Kanyawara Year 4.52 0.04 1 < 2
Year 2 Site 23.21 < 0.001 K>D
Established seedling survival
Kanyawara Year 0.10 0.76
Year 2 Site 6.81 <0.01 K>D








Table 13. Causes of M myristica seedling mortality for emerged and established
seedlings from seed experiments where 10 seeds were placed in 60 g of baboon dung at
Kanyawara and Dura, KNP, Uganda, for Year 1 and Year 2. Percentages are expressed
with respect to total number of seedlings in that stage per site and year (in parentheses).

Kanyawara Dura
Fate Year Year 2 Year 1 Year 2
Emerged seedlings (51) (225) (5) (180)
Survived 47.1 74.7 0 35.0
Died
Beetles* 31.4 2.7 60.0 28.9
Rodents* 11.8 0.4 0 0
Herbivoryt 0 0.9 0 0
Dessication 0 3.1 40.0 0.6
Unknown 9.8 18.2 0 35.6


Established seedlings (24) (168) (0) (63)
Survived 29.2 61.9 -- 25.4
Died
Herbivoryt 20.8 13.1 46.0
Dessication 33.3 8.3 11.1
Crushed: 0 8.9 0
Unknown 16.7 7.7 19.1
* seed destroyed while emerged seedling was still dependent on its reserves
t stems, leaves or roots eaten or damaged by either mammalian and insect herbivores
: crushed by fallen branches








Table 14. Density of seedlings, saplings, poles, and trees ofM. myristica from vegetation plots randomly located along the 4-km
primate census route at Kanyawara and Dura, KNP, Uganda. Density estimates are given for all individuals regardless of location
relative to adult conspecific canopies and for only those individuals located > 0.5 m away from the edge of an adult conspecific
canopy. Density estimates are mean individuals/m2 for small seedlings poles and individuals/ha for trees 1 S.D. Logistic
regressions with poisson errors and log links were used; df= 1.

Size Class All locations Away only
Kanyawara Dura P Kanyawara Dura X! P

Small seedlings 0.05 0.16 1.47 4.86 27.56 < 0.001 0.05 0.16 0.20 0.33 15.29 < 0.001
(< 0.2 m tall)
Large seedlings 0.05 0.08 0.36 0.85 24.75 < 0.001 0.05 0.08 0.12 0.15 10.55 < 0.001
(0.2 0.5 m tall)
Saplings 0.0180.018 0.008 0.015 11.27 < 0.001 0.0170.017 0.008 0.015 10.59 < 0.001
(0.5 2.0 m tall)
Poles (> 2.0 m tall 0.005 0.007 0.0002 0.001 54.42 < 0.001 0.005 0.007 0.0002 0.001
and < 20 cm DBH)
Trees (> 20 cm DBH) 1.0 1.9 3.6 4.0 19.53 < 0.001








Table 15. Relative-height growth rate, survivorship, and causes of mortality of small seedlings, large seedlings, and saplings at
Kanyawara and Dura, KNP, Uganda, located at least 0.5 m away from adult conspecific canopies. Individuals were selected at the
time densities were quantified (see methods) and monitored for the following 18 months. Percentages are expressed with respect to
total number of individuals monitored per size class.

Size Class (height) N RHGR P* Survived Test Died
(cm/cm/yr) (%) (%)
Desiccation Herbivory Crushed Unknown
Seedlings (< 0.2 m)
Kanyawara 35 0.23 0.18 88.6 G = 6.27 5.7 0 0 5.7
Dura 54 0.04 0.46 0.04 63.0 P = 0.01 5.6 3.7 1.9 25.9
Seedlings (0.2 0.5 m)
Kanyawara 57 0.08 0.13 96.5 G = 3.06 0 0 0 3.5
Dura 64 0.01 0.25 0.09 85.9 P = 0.08 4.7 0 0 9.3
Saplings (0.5 2.0 m)
Kanyawara 46 0.05 0.08 91.3 G = 0.08 2.2 0 0 6.5
Dura 15 -0.06 0.34 0.25 93.3 P = 0.77 0 0 0 6.7
* Values based on Mann Whitney tests













CHAPTER 3
DO FRUGIVORES DETERMINE SPATIAL DISPERSION PATTERNS OF
TROPICAL FOREST TREES?: A COMPARISON BETWEEN TWO SITES

Introduction

Seed dispersal is an important factor in the generation of plant spatial patterns. In

tropical forests, where up to 75% of tree species produce fleshy fruits, frugivores are

thought to exert a significant influence on the spatial dispersion and population dynamics

of tropical forest communities (Howe and Smallwood 1982, Herrera 1985, Howe 1986,

Kinnaird 1998). Although studies often claim that deposition patterns by frugivores

cause aggregated plant distributions (Garber 1986, Fragoso 1997, Stevenson 2000), the

spatial dispersion pattern of mature plants reflects recruitment from previous stages with

changes due to mortality (Hutchings 1997). Thus, spatial dispersion patterns arise from

the dispersal of seeds followed by post-dispersal attrition (Lieberman and Lierberman

1994).

By dispersing seeds, frugivores largely determine the initial dispersion patterns of

seeds. This provides the spatial template from which processes arise, such as predation

and competition, that influence spatial distribution of subsequent life-history stages

(Schupp and Fuentes 1995, Nathan and Muller-Landau 2000). Increased survival of

seeds deposited in particular locations by dispersers indicates that frugivore deposition

patterns can influence recruitment and establishment patterns of seedlings (Herrera et al.

1994, Schupp and Fuentes 1995, Julliot 1997, Rogers et al. 1998, Voysey et al. 1999).

For example, howler monkeys (Alouatta seniculus) in French Guiana are thought to








significantly impact seedling distributions of species whose seeds they disperse; seedlings

of four such species were found in clumped distributions under howler monkey sleeping

sites (Julliot 1997). Few studies however, have demonstrated that patterns established by

frugivores persist into later stages, such as saplings and trees (Paul 2001). Studies

inferring the importance of frugivores to plant aggregation patterns are usually based on

correlations between seed deposition and seedling or adult dispersion patterns (Garber

1986, Stevenson 2000). For example, Stevenson (2000) found that seed deposition by

woolly monkeys (Lagothrix lagothricha) was not spatially random, and that preferred

resting places received the majority of dispersed seeds. Thus, he postulated that

deposition patterns by woolly monkeys were the cause of aggregated plant distribution

patterns found in tropical forests.

However, although patterns of seed rain provide the initial spatial template of

individuals, subsequent processes may disrupt or erase the original patterns. For

example, initial patterns of seed dispersion may be modified by many factors including

secondary dispersal of seeds by rodents (Hoshizaki et al. 1997, Wenny 1999), ants

(Levey and Bymrne 1993), or dung beetles (Estrada and Coates-Estrada 1991, Shepherd

and Chapman 1998); density-dependent mortality of seeds and seedlings caused by

herbivores, fungal pathogens, or competition (Augsburger 1983); or abiotic

heterogeneity, such as variable microhabitats and unpredictable canopy disturbances (De

Steven 1994). These processes may cause changes in the spatial patterns of conspecific

individuals over time. Stochastic events, such as canopy disturbances, may result in

unpredictable changes in spatial distribution between successive life-history stages.

Other processes, such as density-dependent mortality, may be species-specific, and thus








may induce relatively more predictable changes in spatial distribution of subsequent life

history stages. For example, tree species whose seeds are normally dispersed in clumps

by frugivores, may have evolved resistance to density-dependent sources of mortality,

resulting in clumped seedling distributions (Howe 1989).

Despite the fact that numerous post-dispersal processes can alter spatial patterns

established by dispersers, there is some evidence that initial patterns may persist. For

example, on Barro Colorado Island, Panama, the degree of clumping of forest trees has

been documented to be related to mode of seed dispersal, such that heavier-seeded,

mammal-dispersed species have higher juvenile and adult densities closer to conspecifics,

than do bird- and bat-dispersed species (Hubbell 1979). In a detailed study of

Maximiliana maripa palms in the Amazon, Fragoso (1997) found that seeds dispersed by

tapirs, and secondarily dispersed by rodents, gave rise to seedlings located around tapir

latrines, and densities of seedlings to fifth year saplings were higher at latrines than

around parent trees. Thus, he attributed the patchy distribution of M maripa to the

influence of dispersal by tapirs.

Given the long life span of most woody species, and the high levels of post-

deposition seed mortality (Schupp 1988, 1990, Willson and Whelan 1990, Balcomb

Chapter 2), it is often difficult, or impossible, to follow adequate samples of seedlings

arising from frugivore depositions over a sufficient time period to determine how

dispersers influence adult dispersion patterns. Thus, to assess the degree to which

frugivores influence spatial aggregation patterns of plants in a general sense, other

approaches must be taken. Since spatial dispersion patterns of adult plants arise from

seed dispersal patterns with subsequent changes due to mortality at each life history








stage, changes in patterns of aggregation over time can be inferred from a static

examination of different size classes (Lieberman and Lierberman 1994). Comparing

changes in patterns of plant aggregation with increasing size class across multiple stands

(Lieberman and Lierberman 1994, Clark et al. 1999), may elucidate the potential

influence that frugivores have on spatial dispersion patterns at different life-history

stages.

To explore the idea commonly found in frugivore foraging studies, that patterns

of seed dispersion established by frugivores persist to older stages, for six animal-

dispersed tree species, I compared patterns of spatial dispersion of four size classes

(seedlings, saplings, poles, and trees) between two stands, separated by 15 km, with

similar frugivore communities in Kibale National Park, Uganda. If, within a species, the

seedling-to-tree aggregation pattern differs between distinct stands, this suggests that

processes differentially influence these size-classes at these stands. Thus, initial patterns

are altered by processes occurring at later stages, and potential patterns initially

established by frugivores are not likely to persist to older stages. Alternatively, if

changes in aggregation patterns from seedlings to trees are similar between distinct stands

this suggests that processes operate similarly between stands and thus initial patterns may

persist over time. Without data on seed rain, however, this finding does not necessarily

indicate that frugivores determine adult spatial patterns. Nevertheless, if most tree

species follow this pattern, this would indicate that, in cases where frugivores do

determine initial seedling aggregation, these frugivores may also play an important role

in determining spatial patterns of older size-classes.








Changes in aggregation patterns among size classes however, can be influenced

by both initial spatial patterns of seed dispersion and post-deposition processes. Ideally

these processes would be measured directly at each site. However, quantifying variation

in spatial deposition patterns of seeds by following individuals of numerous disperser

species, and quantifying post-dispersal fate of several life-history stages for multiple tree

species at multiple sites, requires logistical resources not normally available.

Alternatively, determining whether densities of consecutive size-classes within a species

are correlated, and then comparing patterns of correlation between sites may shed light on

whether processes operating at later stages differ between sites. Thus, for example, if

density dependent mortality operates similarly between sites for a species, then a similar

pattern of correlation between seedling and sapling densities should be evident at both

sites.




Study Sites
Kibale National Park (KNP; 766 km2) is located at the foothills of the Rwenzori

Mountains in Western Uganda (0 13' 0 41' N and 30 19' 30 32' E; Chapman and

Lambert 2000). Kibale is a moist, mid-elevation, evergreen forest, transitional between

lowland rain forest and montane forest (Struhsaker 1975). An elevational gradient from

1,590 m a.s.l. in the north to 920 m in the south corresponds to a north to south increase

in temperature and decrease in rainfall (Howard 1991, Struhsaker 1997, Seavy et al.

2001).

I selected two sites within the primary forest, Kanyawara (Forestry Compartment

K-30) and Dura River. The Kanyawara site is situated at 1500 m (Chapman et al. 1997),








receives 1976 mm of rainfall (data for 1997; Chapman and Chapman unpublished data),
0
and consists of a series of moderately undulating valleys with an average slope of 15.8

(Balcomb unpublished data). Located 15 km to the south of Kanyawara, the Dura River

site (hereafter referred to as Dura) is situated at 1250 m (Chapman et al. 1997), receives

1500 mm of rainfall (data for 1997; Chapman and Chapman unpublished data), and has a

similar topography to that of Kanyawara, with an average slope of 5.90 (Balcomb

unpublished data). Although the sites occur within the same forest and have many tree

species in common (Chapman et al. 1997), the dominant species differ. Foresters classify

the Kanyawara forest as dominated by Parinari excelsa, Aningeria altissima, Olea

welwitschii, Newtonia buchananii, and Chrysophyllum gorungosanum (Chapman and

Lambert 2000). These tree species are less common at Dura, which is dominated by

Pterygota mildbraedii, Cola gigantea, Piptadeniastrum africanum, and Chrysophyllum

albidum (Chapman and Lambert 2000).




Study Species

The six tree species chosen for this study were Mimusops bagshawei

(Sapotaceae), Monodora myristica (Annonaceae), Pseudospondias microcarpa

(Anacardiaceae), Tabernaemontana holstii (Apocynaceae), Uvariopsis congensis

(Annonaceae), and Celtis durandii (Ulmaceae). The first three species are canopy trees

(30 40 m height), while the latter three are common understory trees, 10, 15, and 25 m

tall, respectively (Hamilton 1991). All six species produce fleshy fruits (Table 16)

attractive to a range of frugivores.








Due to the large seed size of most of these species (Table 16), their most common

seed dispersers in KNP are primates; Kibale has one of the highest primate biomass

estimates in the world (Chapman and Lambert 2000). Of the six species, M myristica

and T. holstii potentially have the most restricted suite of disperser species, due to their

protective fruit husks. Initial access to the hard-husked fruits of M myristica is restricted

to large-bodied primates, namely chimpanzees (Pan troglodytes), baboons (Papio

anubis), and grey-cheeked mangabeys (Lophocebus albigena; Balcomb Chapter 2).

Although the majority of M myristica seeds are dispersed by these primates, once these

species open fruits, smaller-bodied primates, such as redtail monkeys (Cercopithecus

ascanius), and blue monkeys (Cercopithecus mitis) also disperse the seeds (Lambert

1999, Balcomb Chapter 2). Likewise, the thick husk and sticky latex of T. holstii may

restrict access to its seeds. Mangabeys have been observed feeding on T. holstii fruits,

and seeds ofT. holstii have been found in chimpanzee dung (Waser 1975, Wrangham et

al. 1994). There are no published observations of other frugivores feeding on T. holstii

fruits in KNP. The soft, fleshy fruits of the other species are accessible to a wider range

of frugivores, and their seeds are known to be dispersed by both large- and small-bodied

diurnal primates, including chimpanzees, mangabeys, redtail monkeys, and blue monkeys

(Waser 1975, Rudran 1978, Wrangham et al. 1994, Chapman and Chapman 1996, Oluput

1998, Lambert 1999).

Other frugivores may also disperse seeds of these species. In KNP, seeds of all of

the focal species, except T. holstii, have been observed being dispersed by Black-and-

White Casqued Hornbills (Ceratogymna subcylindricus; Kalina 1988), in the dung of

elephants (Loxodonta africana; Cochrane unpublished data), and in African civet








(Civetticus civetta) latrines (Balcomb Chapter 2). Although little is known about the

feeding ecology of palm civets (Nandinia binotata) and genets (Genetta sp.) in KNP,

these species also deposit viable seeds in latrines (Engel 2000). However, seedling

recruitment from these latrines is low due to their regular use and the fact that they are

often located on hard, exposed soil (Pendje 1994). Great Blue Turacos (Corythaeola

cristata) in Rwanda can swallow and disperse seeds of comparable sizes to those of all

the focal species (Sun et al. 1997). Small ruminants, such as the red and blue duikers

(Cephalophus harveyi and C moniticola, respectively), can swallow and regurgitate

seeds of these species; however, only seeds with a hard, stoney exocarp are likely to

survive gut rumination (Gautier-Hion et al. 1985, Feer 1995). Thus, duikers are likely

predators of seeds of all these species with the possible exception of P. microcarpa.

Of all the species, C. durandii has the most general suite of dispersers. Due to its

small, unprotected fruit and small seed size, many species of small frugivorous birds, in

addition to the larger-bodied frugivores mentioned above, disperse its seeds (Balcomb

unpublished data).




Methods
At each site, data were collected from plots (50 x 60 m) placed at 50 random

(non-overlapping) locations along a 4-km loop route. The trail was used as the central

axis for each plot. Individuals were placed into the following size classes: seedlings (<

0.3 m tall), saplings (0.3 2.0 m tall), poles (> 2.0 m tall and < 10 cm diameter at breast

height), and trees (> 10 cm dbh). In each plot, data were quantified for all six species.

The entire 50 x 60 m area was used to measure trees. Due to the expected greater density








of stems with decreasing size class, concentric sub-plots, oriented along the central axis

of the plot, were used to measure poles (one 10 x 20 m sub-plot), saplings (one 4 x 20 m

sub-plot), and seedlings (10 1 x 1 m sub-plots every 4 m). The 10 seedling sub-plots

were combined to obtain one value per plot. No individuals within 0.5 m of the trail were

included. Since I was interested in patterns resulting from dispersal away from parent

trees, only individuals that were located > 0.5 m away from the edge of conspecific adult

tree canopies were included in the analyses. Due to its abundance, U congensis was

counted only in the first 30 plots at each site. At the Dura site, four plots intersected a

river, and I was not able to quantify trees in their full area.

Morisita's index of dispersion was used to calculate relative dispersion patterns of

each size class (Krebs 1989). Morisita's index uses counts of individuals in quadrats to

provide a useful, simple descriptive index of dispersion (Hurlbert 1990). This index is a

function of the variance to mean ratio, and measures how many times more likely it is

that two randomly selected individuals will be from the same quadrat than it would be if

the X individuals in the population were distributed at random (Hurlbert 1990). An index

value of one indicates a random dispersion pattern. The index is greater than one for a

contagious distribution, and less than one for a regular distribution. The departure from

random for each size class within each species was determined using a chi-square

distribution (Krebs 1989).

Morisita's index is independent of the sample mean, total numbers in the sample,

and number of sampling units (Elliot 1977, Hurlbert 1990, Tonhasca et al. 1994).

However, the degree of aggregation is a function of the spatial scale measurements are

made on (i.e., it is plot size dependent). In a study of tree dispersion patterns on Barro








Colorado Island, Hubbell (1979) found that for the 30 most common species, juveniles

were more clumped than adults at all plot sizes, ranging from 2 196 m on a side,

although degree of clumping decreased with increasing plot size. Determining a single

plot size that would appropriately assess individuals ranging from small seedlings to

reproductive adults would be difficult since it would require using multiple-sized plots

for each size class of each species. Thus, due to increasing density with decreasing size

class I quantified smaller individuals in 10 1 x 1 m2 sub-plots, and larger individuals at

increasing increments of plot size (see above). For seedlings, since the 10 1 x I m

subplots within each larger plot were combined, if one or more subplots had seedlings

located under adult conspecifics, the entire plot was not included in the analysis so that

the quadrat area was always the same.

I used a bootstrap method to examine intraspecific within-size class differences in

aggregation between sites (Dixon 1993). For each size class, I pooled data from the two

sites and randomly resampled to obtain a data set to represent each site; each resampled

data set had the same number of plots as the observed original. I then took the difference

between the Morisita index for the two data sets. This was repeated 1,000 times, and

95% confidence intervals were generated from the subsequent distribution of values. If

the two sites are from the same population, then the observed difference between their

Morisita indices should lie within 95% confidence intervals generated by the

bootstrapping procedure.

To determine if densities of individuals in a given size class were correlated with

densities of conspecifics in other size classes, Spearman rank correlations were used








(SPSS 1999). When plots were reduced in size (e.g., some 1 x 1 m sub-plots located

under conspecfic adults were excluded), densities were adjusted for sampling areas.




Results

Aggregation Patterns

In five of the six species, seedlings and saplings were significantly clumped at

both sites (Table 17). In the one exception, P. microcarpa, seedling aggregation could

not be calculated since 1 and 0 seedlings were found at Dura and Kanyawara,

respectively; saplings were clumped at Dura and randomly distributed at Kanyawara.

In four species, poles were clumped at both sites (Table 17); the two exceptions

were M. bagshawei and M myristica, in which poles were randomly distributed at the

Dura site. Trees of the three understory species, C. durandii, T. holstii, and U congensis

were clumped at both study sites. Trees of the three canopy species were clumped at one

site and randomly distributed at the other; M bagshawei and P. microcarpa were

clumped at Kanyawara, while M myristica was clumped at Dura.

The bootstrap analysis indicated that there were no inter-site differences in the

degree of aggregation within each of the size classes for C. durandii, M bagshawei, and

P. microcarpa (Table 17 and Fig.4). For T. holstii, seedling and sapling aggregation

patterns did not differ between sites; however, poles and trees were more clumped at

Kanyawara than Dura. Similarly, U congensis aggregation patterns did not differ

between sites for seedlings, while saplings, poles, and trees were more clumped at

Kanyawara than Dura. In contrast, M myristica seedlings were more clumped at








Kanyawara than Dura, while inter-site aggregation patterns did not differ for saplings,

poles, and trees.



Densities

Seedling density varied among the six species; at both sites the lowest seedling

density was in P. microcarpa and the highest was in U congensis (Table 18). Intra-

specific seedling densities varied between the sites from a five-fold difference in M

myristica to roughly equivalent densities in U congensis. In addition, intra-specific

seedling-to-tree ratios differed between the sites for all species except U congensis and

C. durandii.

When correlations in consecutive size-class densities were compared between

sites, for three species, patterns of correlation did not differ between sites (Fig. 5). Thus,

for C durandii at Kanyawara, seedling density was positively correlated with sapling

density, sapling density was not correlated with pole density, and pole density was not

correlated with tree density. This was also the case for this species at Dura. Similarly, at

both sites, M bagshawei seedling-sapling and sapling-pole densities were positively

correlated and pole-tree densities were not correlated, and P. microcarpa showed no

correlation between any size class densities at both sites. In contrast, for both T. holstii

and U congensis, seedling-sapling density correlation was similar between sites (no

correlation and positive correlation, respectively), but, for sapling-pole and pole-tree

densities, correlation patterns differed between sites showing positive correlation at one

site and no correlation at the other site. Monodora myristica exhibited the opposite

pattern; while seedling-sapling and sapling-pole density correlation patterns differed

between sites, pole-tree density correlation pattern was the same at both sites.











Discussion

The overall influence of frugivores on fleshy-fruited plant density and spatial

distribution depends on the effectiveness of seed dispersal in influencing patterns of plant

population recruitment (Howe 1989, Jordano 1992). Although arrival of seeds does not

guarantee recruitment (Nathan and Muller-Landau 2000), Howe (1989) suggested that

the initial pattern of seed deposition by frugivores should be reflected in the distribution

patterns of seedlings. Thus, for example, tree species that have fruits dispersed by

frugivores that deposit seeds in an aggregated fashion should correspondingly exhibit

aggregated seedling distributions. These patterns might be further evident in older size

classes, such as saplings or poles. Moreover, other studies have suggested that adult

aggregation patterns result from frugivore seed deposition patterns (Garber 1986, Fragoso

1997, Stevenson 2000).

I examined whether seedling-to-tree aggregation patterns changed between stands

separated by 15 km to explore the idea commonly found in frugivore foraging studies,

that patterns of seed dispersion established by frugivores persist to older stages. For three

of the six species, using the bootstrap analysis, I found changes in the spatial dispersion

pattern between sites with increasing size class, indicating that initial patterns are not

maintained in a similar fashion at the two sites. Two different directions of change were

evident. First, spatial dispersion patterns were similar between sites at the initial,

seedling, stage yet differed at later stages (poles and adults). This was the case for T.

holstii and U. congensis (Fig. 4). Second, spatial dispersion patterns differed between

sites at the initial seedling stage yet were similar at later stages. This was the case for M








myristica (Fig. 4). Thus, in both situations, regardless of whether initial seedling patterns

differed between the sites, processes occurring at later stages disrupted earlier patterns

resulting in different patterns between the sites.

Comparing differences between sites in patterns of correlation in densities of

consecutive size-classes for these three species, T. holstii, U. congensis, and M myristica,

provides additional support for the importance of between-site variation in post-

deposition processes. Thus, for T. holstii and U congensis, the two species whose spatial

dispersion patterns were similar initially yet differed at later stages, seedling-sapling

densities had similar between-site correlation patterns, while sapling-pole and pole-tree

densities had different between-site correlation patterns. This suggests that, for these

species, while processes influencing recruitment from seedlings to saplings are similar

between sites, processes operating at the sapling and pole stages differ between sites.

Moreover, since these later-acting processes play an important role in determining the

spatial dispersion pattern of subsequent stages, they serve to disrupt patterns observed at

earlier stages.

Disruption of patterns established at early stages may not always occur. Changes

in the spatial dispersion pattern between sites did not differ for increasing size-classes for

the remaining three tree species, C durandii, M bagshawei, and P. microcarpa. For

these species, processes affecting mortality and recruitment are most likely operating in a

similar fashion between the two sites. Similar between-site patterns in correlation

between densities of consecutive size-classes for all three of these species further

supports this idea. Thus, it is possible that spatial dispersion patterns established by

frugivore seed deposition persisted into later stages. However, since data on seed rain








were not available for these species, it is difficult to separate the influence of frugivore

seed deposition from other processes, such as patchy distributions of critical resources,

that may affect seedling recruitment and thus patterns seen at seedling and older stages.

In general, the findings from this study indicate that biotic and abiotic processes

can influence post-deposition stages in plant density and spatial distributions patterns.

Any number of a multitude of factors operating post-dispersal may change the initial

dispersion pattern of seeds established by frugivores and thus disrupt patterns observed at

earlier stages. Variation in these processes likely occur both spatially and temporally.

Houle (1998) found that post-dispersal seed density of yellow birch was positively

correlated with subsequent first-year seedling density in one cohort, but not another.

Moreover, spatial patterns in recruitment distances of cohorts relative to the parent tree

may change over time due to the interaction of biotic processes, such as density

dependence, and abiotic processes, such as chance openings of light-gaps in the forest

(Augsburger 1983). In addition, the relative importance of these processes may differ

between sites. Thus, the location of spatial foci of juveniles may shift in the community

over time (De Steven 1994), and shifts in spatial patterns may differ over larger spatial

scales, such as between sites.

In summary, this study showed that changes in aggregation patterns across

different life history stages of six fleshy-fruited tree species differed between two sites

separated by 15 km. Although initial spatial patterns persisted into later stages in some

species, in others, initial patterns were disrupted, resulting either in convergence or

divergence of patterns in older stages between sites. Interestingly, the two species that

potentially have the most restrictive suite of dispersers, T holstii and M. myristica, were








two of the three species whose initial seedling patterns changed between sites. Thus,

despite the fact that frugivores often play a very important role in the initial steps of the

seed dispersal process (Balcomb Chapter 2), biotic and abiotic post-dispersal processes,

such as predation or gap formation, can influence later stages in plant population

dynamics. Spatial or temporal variation in any one of these processes may affect patterns

of recruitment, thereby influencing spatial aggregation patterns. Thus, although seed

dispersers may influence the initial spatial patterns, they do not necessarily drive the

patterns of adult tree spatial distributions in this system.