THE BEHAVIOR AND ECOLOGY OF BIRDS IN TROPICAL FOREST RESTORATION
SITES.
By
Emily B. Morrison

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
DOCTOR OF PHILSOPHY
Zoology
Ecology, Evolutionary Biology and Behavior- Dual Major
2011

ABSTRACT
THE BEHAVIOR AND ECOLOGY OF BIRDS IN TROPICAL FOREST RESTORATION
SITES.
By
Emily B. Morrison
Forest restoration has become an important management tool that facilitates
forest regeneration. However there has been little study thus far that examines how well
restored plots, and particular planting designs, function over time in providing suitable
habitat for tropical animals.
I examined the effects of restoration patch size on the foraging behavior of four
resident tropical bird species. I also measure arthropod abundance and anti-predator
vigilance behavior to assess whether variation in food availability or predation risk could
explain patch size effects on foraging behavior. Prey attack rates were highest, and the
effort required to find prey lowest, in larger patches for three of the four bird species.
Arthropod density was approximately twice as great in larger patches compared to
smaller patches. Evidence for patch size differences in predation risk was more limited
but risk may be higher in smaller patches. The results indicate that food availability is
the primary mechanism driving patch size effects on foraging behavior.
I compared the habitat quality of the active restoration sites with passive
restoration sites, by measuring vegetation structure, arthropod biomass and the
foraging behavior of three resident bird species. Although vegetation measures such as
amount of understory cover and tree species richness and density differed between the
two restoration strategies, arthropod biomass and foraging behavioral measures were

2

similar. Results suggest that, while active and passive restoration strategies may lead to
different vegetation structure, they may support similar biomass of foliage-dwelling
arthropods and be similarly used by foraging insectivorous birds.
I also examined the effects of predator vocalizations (direct cues) on the
vigilance behavior of birds foraging in different locations: small patches and the center
and edges of large patches (indirect cues). Within feeder location, the time birds spent
alert increased significantly after a predator compared to a non-predator (control)
vocalization in all three locations. However, pairwise comparisons across feeder
location showed that the change in time spent alert in response to the predator call was
significantly greater in small patches and the edges of large patches compared to the
center of large patches. Results show that birds perceive small patches and potentially,
the edges of large patches as more risky habitats than the center of large patches. Birds
responded more strongly to the direct cue of predation risk when foraging in these
locations by increasing vigilance.
I then used an exclosure experiment to examine the top-down effects of birds
and bats on arthropods and herbivory. Arthropod biomass was highest on tree branches
where both birds and bats were excluded and lowest on branches without exclosures
where both birds and bats were present. However, birds and bats had equal impacts on
arthropod populations. The size of restoration patches had no significant effect on the
impact of birds and bats on arthropod biomass, but did affect levels of leaf damage.
These results suggest that both birds and bats play important roles as top-predators in
restoration systems by reducing herbivorous arthropods and their damage to planted
trees.

3

ACKNOWLEDGEMENTS
Many, many thanks go to my advisor, Dr. Catherine Lindell, who has been a
tremendous source of knowledge, patience, inspiration and guidance through this stage
of my graduate career. I would also like to thank the rest of my committee, Dr. Thomas
Getty, Dr. Kay Holekamp and Dr. Kelly Millenbah, for excellent guidance and support. I
would also like to thank my family, Mom, Dad, Tyler and Julia, who have always
encouraged and supported my interests. Finally, to my wonderful husband, Scott Lynn,
thank you so much for your love and support throughout this endeavor, I couldn't have
done it without you.
This work was supported by funds from the Michigan State University College of
Natural Science, Department of Zoology and Program in Ecology, Evolutionary Biology
and Behavior, American Ornithologists Union, Animal Behavior Society, Organization
for Tropical Studies and a National Science Foundation Doctoral Dissertation
Improvement Award (DEB-0909640). I thank K.D. Holl and R.A. Zahawi for use of the
restoration sites. Funding for the development and maintenance of the restoration sites
was provided by the National Science Foundation (DEB 0515577) and the Earthwatch
Institute.

iv

TABLE OF CONTENTS

LIST OF TABLES ............................................................................................................vi
LIST OF FIGURES .........................................................................................................vii
GENERAL INTRODUCTION ...........................................................................................1
Overview of Chapters............................................................................................3
CHAPTER 1
PATCH SIZE EFFECTS ON AVIAN FORAGING BEHAVIOR: IMPLICATIONS FOR
TROPICAL FOREST RESTORATION DESIGN………………………..............................8
Introduction............................................................................................................8
Methods...............................................................................................................10
Results.................................................................................................................23
Discussion............................................................................................................32
CHAPTER 2
ACTIVE OR PASSIVE FOREST RESTORATION? ASSESSING RESTORATION
ALTERNATIVES WITH AVIAN FORAGING BEHAVIOR...............................................36
Introduction..........................................................................................................36
Methods...............................................................................................................38
Results.................................................................................................................46
Discussion............................................................................................................51
CHAPTER 3
VIGILANCE BEHAVIOR OF A TROPICAL BIRD IN RESPONSE TO INDIRECT AND
DIRECT CUES OF PREDATION RISK..........................................................................55
Introduction..........................................................................................................55
Methods...............................................................................................................58
Results.................................................................................................................66
Discussion............................................................................................................73
CHAPTER 4
BIRD AND BAT EFFECTS ON ARTHROPD ABUNDANCES AND LEAF DAMAGE IN
TROPICAL RESTORATION SITES...............................................................................78
Introduction..........................................................................................................78
Methods...............................................................................................................81
Results.................................................................................................................89
Discussion............................................................................................................95
LITERATURE CITED....................................................................................................100

v

LIST OF TABLES
Table 1.1: The percentage of prey attacks and percentage of arthropod samples in each
tree species in 2007 and 2008........................................................................................20
Table 1.2: ANOVA results of the effect of patch size on attack rate and search effort for
all four bird species in 2007 and 2008............................................................................24
Table 3.1: Sample sizes by treatment for the dependent measures of vigilance behavior.
Total number of individuals tested was 36. For most birds, measurements for all five
vigilance behaviors were collected. For those individuals that left the feeder less than
five seconds after playback initiation, only time spent alert and latency to flee were
measured. 'All others' includes head turn frequency, average duration of head-up bouts
and frequency of head-up bouts.....................................................................................64

vi

LIST OF FIGURES
Figure 1.1: Planting arrangement at each restoration site. Each site had one 50 x 50 m
extra-large patch (right) and one 50 x 50 treatment with six smaller patches (left): two
large (12x12 m), two medium (8x8 m) and two small (4x4 m). Each side of the second
treatment had one small, one medium and one large patch, although the arrangement
of the patches varied among sites. The number and arrangement of the tree species
within the patches was constant across all sites. Shaded areas are planted with trees.
Tree species are indicated by letters: Erythrina (E), Vochysia (V), Inga (I), and
Terminalia (T). Non shaded areas are non-woody vegetation, primarily grasses.
Sm=small patch..............................................................................................................12
Figure 1.2: Effect of patch size on attack rates in (a) 2007 and (b) 2008 and search
effort in (c) 2007 and (d) 2008. E=extra-large patches, L=large patches, M=medium
patches and S=small patches. Standard error bars are presented. Means with different
letters are significantly different (Tukey, P<0.05). CHTA=Cherrie's Tanager,
PLWR=Plain Wren, CTFL=Common Tody-flycatcher, RCWA=Rufous-capped Warbler.
There were no observations in the small patches of plain wrens in 2007 or common
tody-flycatchers and rufous-capped warblers in 2008....................................................25
Figure 1.3: Effect of patch size on insect abundance [(g of arthropod biomass/ g of
clipped vegetation)*1000] in 2007 and 2008. E=extra-large patches, L=large patches,
M=medium patches and S=small patches. Standard error bars are presented. Means
with different letters are significantly different (Tukey, P<0.05)......................................28
Figure 1.4: Effect of patch size on predation risk (measured as vigilance) in (a) 2007
and (b) 2008. There were significant differences among patch sizes (indicated by
asterisk) for only rufous-capped warblers in 2007. E=extra-large patches, L=large
patches, M=medium patches and S=small patches. Standard error bars are presented.
CHTA=Cherrie's Tanager, PLWR=Plain Wren, CTFL=Common Tody-flycatcher,
RCWA=Rufous-capped Warbler. There were no observations in the small patches of
plain wrens in 2007 or common tody-flycatchers and rufous-capped warblers in 2008.
Vigilance was zero in the medium patches for plain wrens in 2008...............................30
Figure 2.1: Principle components 1 and 2 from the PCA of habitat variables in the two
restoration treatments (Active = 8 sampling plots, Passive = 6 sampling plots).
Treatments are represented by symbols........................................................................47
Figure 2.2: Foliage height profiles of the active and passive restoration treatments.
These values were averaged across the multiple sampling plots per site and used to
produce vertical foliage height profiles. Standard error bars are presented...................49
Figure 3.1: Mean change in (a) proportion of time spent alert, (b) frequency of head-turn
bouts (measured as head-turns per minute) and (c) average duration of head-up bouts

vii

(measured in seconds) for birds exposed to the Falcon and Wren treatments at each
feeder location (Small Patch, Large Patch Edge, Large Patch Center). Asterisks
represent significant differences (P<0.05) between the Falcon and Wren playback
treatments within feeder location. Letters represent pairwise comparisons among the
Falcon treatments across locations (different letters represent significant differences at
P<0.05). If no letters are present, there were no significant differences across
locations..........................................................................................................................69
Figure 3.2: Mean latency to flee (measured as the number of seconds the bird stayed at
the feeder after initiation of the playback; log transformed) for birds exposed to the
Falcon and Wren treatments at each feeder location (Small Patch, Large Patch Center,
Large Patch Edge)..........................................................................................................73
Figure 4.1: Planting arrangement at each restoration site. Each site had one 50 x 50 m
extra-large patch (right) and one 50 x 50 treatment with six smaller patches (left): two
large (12x12 m), two medium (8x8 m) and two small (4x4 m). Each side of the second
treatment had one small, one medium and one large patch, although the arrangement
of the patches varied among sites. The number and arrangement of the tree species
within the patches was constant across all sites. Shaded areas are planted with trees.
Tree species are indicated by letters: Erythrina (E), Vochysia (V), Inga (I), and
Terminalia (T). Non shaded areas are non-woody vegetation, primarily grasses.
Sm=small patch..............................................................................................................83
Figure 4.2: Data from islands and plantations showing average arthropod biomass (mg
of arthropods/leaf area m2) per census in each of the four exclosure treatments. None
present = exclosure 24-hrs, excluded both birds and bats, Birds+bats present = no
exclosure, neither birds nor bats excluded, Bats present = birds only excluded, bats
present, Birds present = bats only excluded, birds present. Different letters represent
significant pairwise (P<0.05) differences based on post-hoc tests from the repeated
measures mixed effect model. There were no significant pairwise differences in leaf
damage between islands and plantations within any of the four exclosure treatments.
The letters A-D are for the island treatment only, and the letters W-Z are for the
plantations treatment. Standard error bars are shown...................................................90
Figure 4.3. Variation in arthropod biomass over the five sampling periods in the islands
(A) and plantations (B). Pre-exclosure measurements were taken two days before the
experiment began. Each of the other samples were taken once weekly thereafter for
four weeks. None present = exclosure 24-hrs, excluded both birds and bats, Birds+bats
present = no exclosure, neither birds nor bats excluded, Bats present = birds only
excluded, bats present, Birds present = bats only excluded, birds present. Standard
error bars are shown.......................................................................................................92
Figure 4.4: Proportion damage due to herbivory for leaves in the four exclosure
treatments within the two planting treatments (Island and Plantation). None present =
exclosure 24-hrs, excluded both birds and bats, Birds+bats present = no exclosure,

viii

neither birds nor bats excluded, Bats present = birds only excluded, bats present, Birds
present = bats only excluded, birds present. Different letters represent significant
pairwise (P<0.05) differences based on post-hoc tests from the linear mixed effect
model. There were no significant pairwise differences in leaf damage between islands
and plantations within any of the four exclosure treatments. Standard error bars are
shown.............................................................................................................................95

ix

GENERAL INTRODUCTION
Conversion of forests to pastures for cattle grazing and other agriculture has
created large areas of deforested and degraded land in the humid tropics (FAO 2005).
Developing effective strategies to restore these lands is critical because of their high
biodiversity and their significant role in processes such as water and carbon cycles
(Brown 1993). Forest restoration has become an important management tool that
facilitates forest regeneration by buffering microclimates in replanted areas, shading out
fast-growing grasses and making key nutrients available (Kus 1998, Holl 1999).
However there has been little study thus far that examines how well restored plots, and
particular planting designs, function over time in providing suitable habitat for tropical
animals, whose presence and activities are often key to successful ecosystem recovery.
Animals are important contributors to forest regeneration processes, for example
as pollinators (Feldman 2006), seed dispersers (Wunderle 1997) and consumers of
herbivorous insects (Van Beal et al. 2003). However, most studies that consider animals
in evaluations of restoration success measure species presence/absence and calculate
diversity indices (Ruiz-Jaen and Aide 2005). Although valuable, such measures do not
provide information about the factors that influence why animals use some sites rather
than others. In contrast, behavior patterns demonstrate the importance of particular
resources for species (Persson and Stenberg 2006), are strongly linked to fitness
(Alcock 2005) and indicate why some habitats are preferred over others (Pomara et al.
2003). Such behavioral studies are important but underutilized tools in the evaluation of
restoration success (Lindell 2008).

1

Many studies have shown that a bird's use of a habitat is sensitive to landscape
features (Freemark et al. 1995), patch size (Lee et al. 2000; Zanette et al. 2000) and the
availability of specific tree species or foraging substrates (Holmes and Robinson 1981;
Parrish 1995; Whelan 2001). One proximate explanation for variation in avian habitat
use is that these levels of habitat structure can interact to mediate the abundance and
availability of arthropod prey (Wiens et al. 1987; Whelan 1989; Johnson and Sherry
2001). Forest structure, plant species composition (Raman et al. 1998) and prey
abundance are inter-related components of a forest habitat quality (Hutto 1990) that
affect foraging behavior. The availability of food in a habitat can influence foraging
opportunities and foraging behavior in insectivorous birds.
While prey availability can significantly influence avian foraging behavior, other
factors such as predation risk may limit an individual's ability to forage in an optimal
way. To maximize energy gain, individuals should forage in habitat that provides high
food density (Nonacs 2001) and minimizes predation risk (Olsson et al. 2002).
Predation risk may be influenced by habitat characteristics such as canopy cover, patch
size, and amount of edge (Murcia 1995; Whelan and Maina 2005). In poor quality
habitats where predation risk is high, the trade-off between time spent alert for
predators and time spent foraging may have significant negative fitness consequences.
Birds and other top predators, especially bats, in tropical systems have the
potential to impact other components of tropical restoration systems. Both of these
predators consume large quantities of arthropods, many of which are herbivorous and
potentially cause damage to vegetation (Marquis and Whelan 1994; Kalko and Kalka
2006). Insectivorous birds and bats could have significant direct effects on herbivorous

2

arthropod communities and resulting leaf damage to plants (Williams-Guillen et al.
2008; Kalko et al. 2008). Leaf damage due to herbivorous arthropods can negatively
affect the growth and survival of tropical plants (Norghauer et al. 2006; Eichorn et al.
2010). Tree seedling survival and growth are critical to the success of tropical forest
restoration efforts (Holl et al. 2000), so arthropod predators could play an important role
in controlling leaf damage due to herbivory.
My research was conducted in a replicated tropical forest restoration experiment
in southern Costa Rica. Sixteen restoration sites were planted between 2004-2006 with
four tree seedling species: I used two experimental treatments at each site. Treatment
plots were 50 x 50 m and either had (1) trees planted in rows to cover the entire plot
2

(extra-large patch, 2500 m ), or (2) trees planted in six patches of three different sizes:
2

2

two small (16 m ), two medium (64 m ) and two large patches (144 m2).
My research used this replicated restoration project to investigate (1) the effect of
restoration patch size on foraging behavioral measures such as prey attack rate and
prey search effort, and whether food abundance or predation risk were influencing the
variation in foraging behavior among patch sizes, (2) differences avian foraging
behavior between active restoration sites and naturally regenerating sites, (3) the
effects of direct and indirect cues of predation risk on vigilance in birds foraging in the
restoration sites, and (4) the top-down effects of birds and bats on arthropod biomass
and leaf damage due to arthropod herbivory in the restoration sites.

OVERVIEW OF CHAPTERS

3

In Chapter 1, I examine the effects of restoration patch size on the foraging
behavior of four resident tropical bird species. I also measure arthropod abundance and
anti-predator vigilance behavior to assess whether variation in food availability or
predation risk could explain patch size effects on foraging behavior. Prey attack rates
were highest, and the effort required to find prey lowest, in larger patches for three of
the four bird species. Arthropod density was approximately twice as great in larger
2

2

patches (> 3500 m ) compared to smaller patches (< 350 m ). Evidence for patch size
differences in predation risk was more limited but risk may be higher in smaller patches.
The results indicate that food availability is the primary mechanism driving patch size
effects on foraging behavior, with predation risk being an additional influence for some
species in some years. As demonstrated in this study, patches of tens to a few
hundreds of metres squared are likely to provide fewer food resources and potentially
less cover from predators for vertebrates that use woody habitat, compared to patches
of a few thousand metres squared. The more limited resources in smaller patches are
likely to have short-term and, potentially, long-term consequences for the fitness of
organisms. When considering restoration project design, the potential economic and
other benefits of planting in smaller patches must be weighed with the potentially
negative ecological effects on some taxonomic groups.
Results from Chapter 1 show that using behavioral measures can reveal what
habitat factors are important for birds in restoration systems. Behavioral measures can
also be used to compare different types of restoration approaches. Active and passive
restoration are two important strategies to aid the recovery of large areas of deforested
and degraded tropical lands. Active restoration is where management techniques such

4

as planting seeds or seedlings are implemented, and passive restoration is when no
action is taken except to cease environmental stressors such as agriculture or grazing
(Rey Benayas et al. 2008).
In Chapter 2, I compare the habitat quality of the active restoration sites
described above, with passive restoration sites, by measuring vegetation structure,
arthropod biomass and the foraging behavior of three resident bird species. Both
restoration sites had similar land-use histories and times since abandonment. Although
vegetation measures such as amount of understory cover and tree species richness
and density differed between the two restoration strategies, arthropod biomass and
foraging behavioral measures were similar. Results suggest that, while active and
passive restoration strategies may lead to different vegetation structure, they may
support similar biomass of foliage-dwelling arthropods and be similarly used by foraging
insectivorous birds. Passive restoration is generally less costly than active restoration
and, if local and landscape characteristics do not impede recovery, may be a viable
alternative from the perspective of birds using the sites.
Results from Chapter 1 suggested that predation risk varies among patch size for
birds. I explore this idea further in Chapter 3. I examine the effects of predator
vocalizations (direct cues) on the vigilance behavior of birds foraging in different
locations: small patches and the center and edges of large patches (indirect cues).
Within feeder location, the time birds spent alert increased significantly after a predator
compared to a non-predator (control) vocalization in all three locations. However,
pairwise comparisons across feeder location showed that the change in time spent alert
in response to the predator call was significantly greater in small patches and the edges

5

of large patches compared to the center of large patches. For birds foraging in small
patches, the frequency of lateral head turns and duration of head-up bouts increased
significantly after the predator compared to the non-predator vocalization. Additionally,
the change in head-turn frequency and head-up duration in response to the predator
call was significantly greater in small patches compared to the center of large patches.
Results show that birds perceive small patches and potentially, the edges of large
patches as more risky habitats than the center of large patches. Birds responded more
strongly to the direct cue of predation risk when foraging in these locations by
increasing a number of vigilance behaviors. These results highlight the importance of
investigating avian behavioral responses to the characteristics of restoration sites, many
of which consist of small patches of habitat.
My research in Chapters 1-3 demonstrates that the quality of restoration sites
can significantly affect the behaviors of birds using the sites. However, birds, and also
bats, are themselves likely to impact other components of the restoration system
including herbivorous arthropods and the extent of their damage to vegetation
(Williams-Guillen et al. 2008; Kalko et al. 2008). The influence of birds and bats on
arthropod populations and their indirect effects on leaf damage have not previously
been investigated in tropical restoration sites. In Chapter 4, I use an exclosure
experiment to examine the top-down effects of birds and bats on arthropods and
herbivory. Arthropod biomass was highest on tree branches where both birds and bats
were excluded and lowest on branches without exclosures where both birds and bats
were present. However, birds and bats had equal impacts on arthropod populations, in
contrast to other studies which have found that one or the other predator has stronger

6

effects on arthropods. The size of restoration patches had no significant effect on the
impact of birds and bats on arthropod biomass, but did affect levels of leaf damage. In
large patches, levels of leaf damage were higher on branches where birds and bats
were excluded than on branches where both had access, but in small patches, no
significant differences in herbivory between exclosure treatments were found. These
results suggest that both birds and bats play important roles as top-predators in
restoration systems by reducing herbivorous arthropods and their damage to planted
trees. Tropical restoration projects should include efforts to attract and provide suitable
habitat for birds and bats, given their demonstrated ecological importance.
Throughout the rest of this dissertation, I use the term "we" instead of "I", which
reflects the true collaborative nature of my work and the fact that all chapters were
prepared in manuscript format.

7

CHAPTER 1
PATCH SIZE EFFECTS ON AVIAN FORAGING BEHAVIOR: IMPLICATIONS FOR
TROPICAL FOREST RESTORATION DESIGN
INTRODUCTION
Tropical forest cover is disappearing at a rapid pace. More than seven million
hectares of tropical forest were lost between 2000 and 2005, primarily because of direct
conversion of forest to agricultural land covers (FAO 2005). Active restoration, by
planting native vegetation, has gained momentum globally and contributed to the
reduction in net loss of forest area in many countries (FAO 2005; Chazdon 2008).
Developing effective tropical forest restoration strategies will increase the value of
restoration efforts and increase the rate at which habitats regenerate (Lamb et al. 2005).
Planting woody vegetation over large areas is an effective restoration technique
but is also costly (Parrotta and Engel 2001; Rey Benayas et al. 2008). Planting
seedlings in small patches of tens to a few hundred metres squared has been
suggested as a potentially cost-effective alternative although this restoration strategy
has not been widely implemented nor rigorously tested in comparison with other
methods, until very recently (Zahawi and Augspurger 2006; Rey Benayas et al. 2008).
Planting small blocks of vegetation mimics the natural regeneration process of
'nucleation', where early successional vegetation establishes patchily and patches
coalesce over time (Yarranton and Morrison 1974). This strategy could reduce the cost
of restoration by reducing the amount of vegetation planted. However, patch size then
becomes an important consideration of restoration design because it may influence the
attractiveness of restoration sites to animals. For example, the rate and duration of bird

8

visits was greater in larger restoration patches in previous studies (Zahawi and
Augspurger 2006; Fink et al. 2009). However no previous work has examined why
some patch sizes are more attractive to animals than others.
Animals are important contributors to forest regeneration processes, for example
as pollinators (Feldman 2006), seed dispersers (Wunderle 1997) and consumers of
herbivorous insects (Van Beal et al. 2003). However, most studies that consider animals
in evaluations of restoration success measure species presence/absence and calculate
diversity indices (Ruiz-Jaen and Aide 2005). Although valuable, such measures do not
provide information about the factors that influence why animals use some sites rather
than others. In contrast, behaviour patterns demonstrate the importance of particular
resources for species (Persson and Stenberg 2006), are strongly linked to fitness
(Alcock 2005) and indicate why some habitats are preferred over others (Pomara et al.
2003). Such behavioral studies are important but underutilized tools in the evaluation of
restoration success (Lindell 2008).
Foraging behavior measures, such as prey attack rate (number of attacks on prey
per unit time) and search effort (locomotor movements per attack) are probably
influenced by patch size through food availability (Burke and Nol 1998) and predation
risk (Butler et al. 2005). Studies across a variety of spatial scales show that larger
patches have greater densities of arthropods (Bach 1988; Burke and Nol 1998; Zanette
et al. 2000). Predation risk may be higher in small habitat patches because they offer
less cover from predators (Suhonen 1993; Kotler et al. 2004). When predation risk is
high, animals increase the proportion of time spent vigilant, negatively affecting foraging
efficiency (Caraco et al. 1980; Elgar 1989).

9

In this study, we examined the effect of restoration patch size on the foraging
behavior of four tropical resident bird species in a replicated forest restoration
experiment in southern Costa Rica. We tested the hypothesis that patch size influences
the foraging behavior of birds and that this effect is due to differences in arthropod
abundance and predation risk. We predicted that prey attack rates would be highest and
prey search effort lowest in the larger patches because arthropod abundance would be
higher (Burke and Nol 1998) and predation risk lower in larger patches (Suhonen 1993).

METHODS
Study Site
The study was conducted from 5 May-15 August 2007 and 11 May-8 August
2008 in restoration sites located near the town of Agua Buena (8° 44' N, 82° 58' W;
elevation 1100-1300 m) in southern Costa Rica. Six restoration sites were planted in
2

2004 throughout a 30 km area. The closest two sites were located 0.9 km apart and all
other sites were at least 1.2 km apart. The next closest restoration site to each of the six
sites ranged from 0.9-4.1 km away (mean ± SD: 1.95 ± 1.24 km). In the past 50 years
most land surrounding Agua Buena was cleared for agricultural use, primarily for the
production of sun coffee with some land used for pasture and subsistence agriculture.
Currently, less than 27% of the original forest remains in the region, mostly in small
fragments along waterways or on the mountains at higher elevations (Daily et al. 2001).
The landscape surrounding the restoration sites is dominated by pasture, sparselyshaded coffee and other agriculture, small human settlements and some patches of
secondary and mature Pacific premontane humid forest (Daily et al. 2001; Sekercioglu

10

et al. 2007). All restoration sites had previously been used for ≥18 years for agriculture
(usually coffee first and then pasture).
The restoration sites were planted with four seedling species: Terminalia
amazonia (J.F. Gmel.) Exell [Combretaceae], Vochysia guatemalensis Donn. Sm.
[Vochysiaceae], Erythrina poeppigiana (Walp.) O.F. Cook [Fabaceae], and Inga edulis
Mart. [Fabaceae]. All species are native or naturalized to the region and are widely used
in tropical forest restoration (Nichols et al. 2001; Carpenter et al. 2004). We conducted
this study in all six sites in 2007, and in four of the six sites in 2008 due to logistical
constraints. We used two experimental treatments at each site. Treatment plots were
50 x 50 m and either had (1) trees planted in rows to cover the entire plot (extra-large
2

patch, 2500 m ), or (2) trees planted in six patches of three different sizes: two small (16
2

2

2

m ), two medium (64 m ) and two large patches (144 m ; Figure 1.1). Vegetation was
cleared around planted trees every 2-3 months for the first two years of planting to allow
seedlings to become established. The small number of seedlings that died during this
time were replanted. No further clearing or planting was conducted after May 2006.
Tree heights ranged from 2-5 m in 2007. Extra-large patches had an almost
continuous canopy over the 50 x 50 m treatment area. The large, medium and small
patches remained distinct patches within a matrix of mostly herbaceous vegetation.

11

Figure 1.1: Planting arrangement at each restoration site. Each site had one 50 x
50 m extra-large patch (bottom) and one 50 x 50 treatment with six smaller patches
(top): two large (12x12 m), two medium (8x8 m) and two small (4x4 m). Each side of the
second treatment had one small, one medium and one large patch, although the
arrangement of the patches varied among sites. The number and arrangement of the
tree species within the patches was constant across all sites. Shaded areas are planted
with trees. Tree species are indicated by letters: Erythrina (E), Vochysia (V), Inga (I),
and Terminalia (T). Non shaded areas are non-woody vegetation, primarily grasses.
Sm=small patch.

12

13

Patch Size Reclassification
Although the number of seedlings planted in each plot was initially the same,
seedling growth and mortality varied, resulting in variation in canopy cover among patch
sizes and tree species. Thus we classified patch sizes based on canopy area, which
sometimes differed from the original size classification. In both 2007 and 2008, we
measured canopy cover for every tree in the small, medium and large patches and onethird of the trees (selected at random) in each extra-large patch. To measure individual
tree canopy cover, we used the formula for the area of an ellipse (πd1d2/4); where
d1=the longest canopy diameter and d2=the diameter perpendicular to d1. This method
of canopy cover approximation correlates with above-ground biomass (Peek 1970) and
total leaf area (Kohyama and Hotta 1990) and has been used to describe forest
structure in tropical secondary forests (Montgomery and Chazdon 2001). To calculate
total canopy cover for each small, medium and large patch, we summed the canopy
cover of all trees within each patch. For extra-large patches, we multiplied the average
canopy cover of each tree species by the number of each species in that patch (Fink et
al. 2009). We used the classification based on canopy cover in all further analyses.
There were several patches where birds were never observed. In 2007, no birds
were observed in 15 of the 41 patches and in 2008, no birds were observed in 9 out of
the 28 patches. The patches where no birds were observed were significantly smaller
than those in which birds were observed (mean ± SE, 2007: patches with birds: 979.9 ±
2

2

310.9 m ; patches without birds: 42.5 ± 12.9 m ; t39=3.01, P=0.006; 2008: patches with
2

2

birds: 1251.1 ± 473.2 m ; patches without birds: 62.1 ± 25.8 m ; t26=3.12, P=0.005).
Patches without bird observations were not included in further analyses.

14

Patches were reclassified based on canopy cover (see above). Patches that
2

2

2

were initially planted as small (16 m ), medium (64 m ), and large (144 m ) were
2

2

reclassified to small (mean ± SE, 2007: 62.9 ± 8.2 m , N=7; 2008: 78.3 ± 5.5 m , N=4),
2

2

medium (2007: 143.4 ± 6.1 m , N=5; 2008: 134.0 ± 10.6 m , N=3), and large (2007:
2

2

300.6 ± 31.7 m , N=7; 2008: 342.0 ± 39.7 m , N=8) using total canopy cover. Since all
2

extra-large patches had much greater total canopy cover (2007: 3672.2 ± 449.2 m ,
2

N=6; 2008: 5080.1 ± 407.4 m , N=4) than small, medium and large patches, their
original classification of "extra-large" was retained. Reclassified patch sizes were used
in all analyses.

Foraging Observations
We conducted foraging observations of four species of insectivorous and
omnivorous bird species in the restoration sites: Cherrie's tanager Ramphocelus
costaricensis, rufous-capped warbler Basileuterus rufifrons, common tody-flycatcher
Todirostrum cinereum and plain wren Thryothorus modestus. Cherrie's tanager is a
frugivore-insectivore while the other species are primarily insectivores (Stiles and
Skutch 1989). Cherrie's tanagers, rufous-capped warblers and plain wrens forage
primarily by gleaning (picking food items from nearby substrates from a standing
position, no acrobatic movements; Remson and Robinson 1982; Stiles and Skutch
1989). Common tody-flycatchers use both gleaning and sally-strike (flight directly at a
prey item on substrate) manoeuvres, but rarely fly out to capture aerial insects (Stiles
and Skutch 1989). These four species were the most commonly observed foliage-

15

gleaning insectivores in the restoration sites. There was no fruit available in the sites
during the study, thus all observations were of birds foraging for arthropods.
In 2007, foraging observations were conducted from 5 May-15 August at five of
the six restoration sites. Due to logistical constraints, foraging observations were
conducted at the sixth site from 24 June-21 July 2007 only. In 2008, observations were
conducted from 11 May-8 August at four of the six sites used in 2007. In both years,
observations were conducted at each site approximately once every 10 days between
sunrise and 1100 hours. Each site was systematically searched to locate foraging birds
and observations were made on birds throughout each patch. Foraging observations
lasted as long as the bird was in sight. For each observation, we recorded the bird
species, number of prey attacks (attempts to capture insect prey), tree species where
attacks occurred, and movements between attacks: hops, short flights (<1 m), and long
flights (>1 m) (Pomara et al. 2003). Only those foraging bouts of at least 20 seconds
were included for analysis (Robinson and Holmes 1982). All observations were
continuously recorded on a digital voice recorder by E.B Morrison.
Time spent in non-foraging states (e.g. preening) was removed from the total
observation time before calculation of attack rate, search effort, and vigilance (Lyons
2005). We calculated attack rate (number of attacks per minute) by dividing the total
number of attacks by the total observation time (Pomara et al. 2003; Lyons 2005).
Search effort (number of movements per attack) was calculated by dividing the total
number of movements by the total number of attacks in each observation (Pomara et al.
2003). Vigilance was considered a measure of predation risk (Caraco et al. 1980; Glück

16

1987) and was recorded as the percentage of the observation a bird spent alert or
scanning with bill horizontal or above (Repasky 1996; Bednekoff and Lima 2002).
Some birds were colorbanded which allowed individual identification. In 2007, ten
Cherrie's tanagers, six rufous-capped warblers, two common tody-flycatchers, and two
plain wrens were banded. In 2008, nine Cherrie's tanagers, seven rufous-capped
warblers, three common tody-flycatchers and three plain wrens were banded. Rufouscapped warblers, plain wrens and common tody-flycatchers were paired and held
territories during the study period (Stiles and Skutch 1989). Our observations of banded
and paired individuals suggested that territory size was 1 ha or less for plain wrens,
common tody-flycatchers and rufous-capped warblers. Jedlicka et al. (2006) reported a
similar territory size for rufous-capped warblers. We identified two pairs of each of these
three species at some sites, while other sites had only one pair.
Cherrie's tanagers are sexually dimorphic and found in loose groups of 4-10
individuals year-round (Stiles and Skutch 1989). Based on observations of banded
individuals, the territories of Cherrie's tanager groups in our study included an entire
restoration site (0.5 ha), with 6-10 individuals per site. In the same region a density of
12 Cherrie's tanagers per ha was reported by Krueger et al. (2008). The size of
Cherrie's tanager group territories has not been reported in the literature. However
extensive mist-netting in the restoration sites between 2006-2008 resulted in no
recaptures of individuals between sites, although individuals are often recaptured within
a site (Lindell, unpubl. data).
We estimated the number of individuals observed from sightings of paired and
banded individuals and the locations of their territories within the restoration sites.

17

Additionally, we assumed that unbanded individuals at one site were different from
unbanded individuals at another site because the sites were separated by distances
much greater than the territory sizes of the bird species.
No more than one foraging observation was made per banded individual per day
and we waited at least one day between recording observations of unbanded individuals
of the same species (or sex for Cherrie's tanager) on the same site (Wagner 1981;
Wunderle and Latta 1998; Johnson 2000; Pomara et al. 2003). We included multiple
observations of the same individuals in our analysis because it ensured the inclusion of
variability in the behavior of individual birds among patch sizes. Such observations were
not truly independent, and their inclusion may have inflated sample sizes somewhat and
led to an increased Type I error rate in statistical tests (Hurlbert 1984). However, the
use of only a single observation of each individual would have weakened our ability to
separate the effects of patch size on behavior from differences among individuals
(Pomara et al. 2003). Thus we believe that the value of using multiple observations of
single individuals outweighed the potential problems (e.g. Holmes, Bonney and Pacala
1979; Johnson 2000; Pelletier 2005; Lyons 2005; Jedlicka et al. 2006).

Arthropod Sampling
General Methods - The four bird species we studied are primarily surface
gleaners, taking insects directly from vegetation substrates such as leaves or branches
(Stiles and Skutch 1989). Branch clipping was used to collect arthropod samples from
branches and leaves (Johnson 2000). A collapsible bag was positioned around the
outer 0.75-1.0 m of a selected branch, the mouth of the bag was closed, and the branch

18

clipped free with a tree pruner. Bags were visually inspected for arthropods and all
arthropods were immediately categorized into 1 mm size intervals and identified to at
least the Order level. The clipped vegetation was weighed to the nearest g using a
Pesola scale (wet mass). We standardized the amount of arthropod biomass per
vegetation biomass by dividing the wet mass of the arthropods in each sample by the
wet mass of the clipped vegetation. Each sample is expressed as [(g of arthropod
biomass/ g of clipped vegetation)*1000]. This method is commonly used to measure
food availability for birds (Johnson 2000; Johnson et al. 2005; Smith et al. 2006; Vitz
and Rodewald 2006).
We identified arthropod species to the following categories: Aranae, Orthoptera,
Dermaptera, Hemiptera, Homoptera, Coleoptera, Lepidoptera adults, Lepidoptera
larvae, Diptera, Formicidae and other Hymenoptera (bees and wasps), unknown adults,
and unknown larvae. Length-weight regressions from Johnson and Strong (2000) and
Schoener (1980) were used to calculate arthropod biomass for all Orders. These Orders
have been reported in the stomach samples of these species and analogous species in
similar habitats (C.H. Sekercioglu pers. comm.; Poulin et al. 1994; Poulin and Lefebvre
1996).
Weighted Insect Abundances - Foraging observations and branch clip sampling
were conducted concurrently. We selected the proportions of each of the four tree
species for arthropod sampling based on estimations of tree species use by the four
focal bird species, such that tree species with greater proportionate use formed a
greater proportion of the invertebrate samples. Branch clip samples were taken from
17-30 trees in each extra-large patch, 5-8 trees in each large patch, 3-7 trees in each

19

medium patch, and 3-6 trees in each small patch. Numbers of trees sampled varied
within patch sizes because of variation in tree mortality and growth rates. Trees used for
arthropod sampling were located throughout each patch. Branch clip samples were
taken from a range of heights that matched the range of prey attack heights for the four
bird species in each year.
After the study period concluded, we found that the percentage of prey attacks
was greater in Inga than in the other three tree species and the proportion of branch
clippings among the four tree species did not match the percentage of prey attacks in
each tree species. In both years we undersampled Inga and oversampled the other
three tree species (Table 1.1).
Table 1.1: The percentage of prey attacks and percentage of arthropod samples in each
tree species in 2007 and 2008.
2007
2008
Tree
% prey attacks % arthropod % prey attacks
% arthropod
Species
samples
samples
Terminalia
1.8%
13.8%
1.6%
17.5%
Vochysia
6.6%
12.7%
2.9%
16.9%
Erythrina
4.5%
18.4%
10.4%
16.3%
Inga
87.1%
55.1%
85.1%
49.3%
To correct for this, we weighted branch clip samples from each tree species to
match the proportions of prey attacks in each tree species. We used the following
formula (adapted from Johnson 2000) for each tree species (i):

Wi = nideal(i) / nactual(i) eqn 1
nideal(i) = Ui * ntotal
where

20

eqn 2

Wi = the weighting factor by which all branch clip samples within tree
species (i) were multiplied,

nideal(i) = the number of branch clip samples from tree species i if the match
between proportions of attacks and branch clip sample distributions
as precise,

nactual(i) = the actual number of branch clip samples taken from tree species i,

Ui = the per cent of attacks in tree species i,

ntotal = the total number of branch clip samples from all tree species.

The weighted sample totals were used in calculations of arthropod abundance (Johnson
2000).
Insect Sampling Period - In 2007, five of the sites were sampled twice during the
study; once between 21 May-10 June and once between 10-22 July. Branch clip
samples at the sixth site were taken only once between 5-8 July because foraging
observations were only conducted from 24 June-21 July. In 2008, every site was
sampled twice; once between 5-8 June and once between 7-11 July. Arthropod
abundance was calculated for each sampling period for each patch by summing the
arthropod biomass from all branch samples taken in each patch and then dividing by the

21

sum of the branch weights for that particular patch. Arthropod abundance values for
patches from the two sampling periods were strongly correlated (2007: Pearson's
r=0.53, P=0.01; 2008: r=0.70, P=0.001). Therefore, within each year, arthropod samples
from the two sampling periods were averaged, so that each patch had one arthropod
abundance value.

Data Analysis
Some patches changed size classes between years, i.e. they grew from medium
in 2007 to large in 2008, so we analysed data from 2007 and 2008 separately in order
to use data from the same patches in both years. All dependent variables (attack rate,
search effort, arthropod abundance, vigilance) were checked for normality and
homoscedasticity. Attack rate and search effort were log transformed to meet these
assumptions. General linear models (GLM) were used to examine the effects of site,
sex (Cherrie's tanager only) and patch size (small, medium, large and extra-large) on
attack rates and search efforts for each of the four species. All factors except patch size
were not significant (P>0.35 for all factors) and removed from the models.
Vigilance data could not be normalized using transformation. Thus we used nonparametric Kruskal-Wallis ANOVAs to analyze the effects of patch size on per cent time
spent vigilant. A GLM was used to examine the effect of site and patch size on
arthropod abundance. The effect of site was not significant and was removed from the
model. Tukey's post hoc analyses were used to examine pair-wise differences among
the patch sizes for arthropod and foraging behavior data. All data were analysed using

22

SAS 9.1 (SAS Institute 2004). Throughout, we present results as mean ± SE.
Differences between groups were considered significant when P<0.05.

RESULTS
Foraging Observations
Total observation time across all sites was 237 hours in 2007 and 116 hours in
2008. Mean duration of foraging observations (mean ± SE averaged across 2007 and
2008) was 82.8 ± 3.6 sec for Cherrie's tanagers, 85.8 ± 4.8 sec for rufous-capped
warblers, 65.4 ± 5.4 sec for common tody-flycatchers and 80.4 ± 7.2 sec for plain
wrens.
In 2007 (at six sites), we conducted 75 observations of at least 43 Cherrie's
tanager individuals, 55 observations of at least 18 rufous-capped warblers, 22
observations of at least 14 common tody-flycatchers and 30 observations of at least 16
plain wrens. In 2008 (at four sites), we conducted 54 observations of at least 22
Cherrie's tanagers, 38 observations of at least 14 rufous-capped warblers, 28
observations of at least 12 common tody-flycatchers and 23 observations of at least 12
plain wrens.

Foraging Behavior
In both 2007 and 2008, attack rate was significantly higher and search effort was
significantly lower in larger patches than in smaller patches for Cherrie's tanagers, plain
wrens and rufous-capped warblers (Figure 1.2; Table 1.2). Common tody-flycatchers

23

showed no differences among patch sizes for both search effort and attack rate in both
years (Figure 1.2; Table 1.2).
Table 1.2: ANOVA results of the effect of patch size on attack rate and search effort for
all four bird species in 2007 and 2008.

Species
Cherrie's tanager
attack rate

F

2007
d.f.

P

F

2008
d.f.

P

5.6

3,71

<0.01

6.4

3,50

<0.01

search effort

7.0

3,71

<0.01

9.2

3,50

<0.01

4.0

2,27

0.03

5.5

3,19

<0.01

search effort

3.6

2,27

0.04

5.1

3,19

0.01

Common todyflycatcher
attack rate

2.1

3,18

0.14

0.3

2,25

0.74

0.9

3,18

0.44

0.8

2,25

0.50

12.1

3,51

<0.01

7.1

2,35

<0.01

7.6

3,51

<0.01

4.4

2,35

0.02

Plain wren
attack rate

search effort
Rufous-capped warbler
attack rate
search effort

24

Figure 1.2: Effect of patch size on attack rates in (a) 2007 and (b) 2008 and search
effort in (c) 2007 and (d) 2008. E=extra-large patches, L=large patches, M=medium
patches and S=small patches. Standard error bars are presented. Means with different
letters are significantly different (Tukey, P<0.05). CHTA=Cherrie's Tanager,
PLWR=Plain Wren, CTFL=Common Tody-flycatcher, RCWA=Rufous-capped Warbler.
There were no observations in the small patches of plain wrens in 2007 or common
tody-flycatchers and rufous-capped warblers in 2008.

25

Attack rate (attacks/minute)

3.5

(a) 2007

3.0
2.5
2.0

a

a

a
ab ab
b

b b

EL MS
CHTA

E L M
PLWR

1.5

b
b b

1.0
0.5
0.0

Attack rate (attacks/minute)

5.0

E L MS
CTFL

E L MS
RCWA

(b) 2008

4.5
4.0
3.5

a
a

a
ab

3.0
2.0
1.5

b

ab

2.5

ab

b

b

b

b

1.0
0.5
0.0

E LM S
CHTA

E L MS
PLWR

E L M
CTFL

26

E L M
RCWA

Search effort (movements/attack)

Search effort (movements/attack)

Figure 1.2 (cont'd)

50 (c) 2007
45

b
b

40
35

b

ab b

30

ab

25
20

ab

ab

a

a

a

15
10
5
0

E L MS
CHTA

E LM
PLWR

E L MS
CTFL

E L MS
RCWA

40 (d) 2008
35

b

30

b
ab

25

b b

20
15

a

ab

abab
a
a

10
5
0

E L MS
CHTA

E L MS
PLWR

E L M
CTFL

27

E L M
RCWA

Arthropod Abundance
Patch size significantly influenced arthropod abundance in both years (Figure
1.3; 2007: F3,22=14.5, P<0.0001; 2008: F3,15=11.3, P<0.0001). In 2007, extra-large and
large patches had higher arthropod abundance than either medium or small patches
(Tukey, P<0.05). In 2008, arthropod abundances in extra-large patches were greater
than medium and small patches and arthropod abundances in large patches were

Arthropod abundance
(g arthropods/g vegetation * 1000)

greater than small patches (Tukey, P<0.05).

0.18
a

0.16
0.14

a

0.12

ab
a
b

0.10
0.08

b

0.06

c

b

0.04
0.02
0.00
E

L

M

S

2007

E

L

M

S

2008

Figure 1.3: Effect of patch size on insect abundance [(g of arthropod biomass/ g
of clipped vegetation)*1000] in 2007 and 2008. E=extra-large patches, L=large
patches, M=medium patches and S=small patches. Standard error bars are presented.
Means with different letters are significantly different (Tukey, P<0.05).

Predation Risk

28

In 2007, vigilance was significantly higher in smaller patches for rufous-capped
warblers (Figure 1.4a; Kruskal-Wallis ANOVA: H=18.8, d.f.=3, P=0.003). Cherrie's
tanagers had marginally higher vigilance in smaller patches (Figure 1.4a; H=6.6, d.f.=3,
P=0.08). Patch size did not affect vigilance for either plain wrens (Figure 1.4a; H=5.5,
d.f.=2, P=0.11) or common tody-flycatchers (Figure 1.4a; H=5.1, d.f.=3, P=0.16). In
2008, there were no effects of patch size on vigilance for any of the four species (Figure
1.4b; Cherrie's tanagers: H=2.9, d.f.=3, P=0.40; plain wrens: H=2.5, d.f.=3, P=0.48;
common tody-flycatchers: H=1.1, d.f.=2, P=0.57; rufous-capped warblers: H=0.3, d.f.=2,
P=0.85).

29

Figure 1.4: Effect of patch size on predation risk (measured as vigilance) in (a) 2007
and (b) 2008. There were significant differences among patch sizes (indicated by
asterisk) for only rufous-capped warblers in 2007. E=extra-large patches, L=large
patches, M=medium patches and S=small patches. Standard error bars are presented.
CHTA=Cherrie's Tanager, PLWR=Plain Wren, CTFL=Common Tody-flycatcher,
RCWA=Rufous-capped Warbler. There were no observations in the small patches of
plain wrens in 2007 or common tody-flycatchers and rufous-capped warblers in 2008.
Vigilance was zero in the medium patches for plain wrens in 2008.

30

Percent time spent vigilant

35 (a) 2007
30
25

*

20
15
10
5
0

E

L

M S

Percent time spent vigilant

CHTA

22 (b) 2008
20
18
16
14
12
10
8
6
4
2
0
E

L

M S

CHTA

E

L

M

E

L

PLWR

E

L

M S

CTFL

M

S

PLWR

E

L

M

CTFL

31

E

L

M S

RCWA

E

L

M

RCWA

DISCUSSION
Planting small patches of vegetation, of tens to hundreds of metres squared, is a
potential alternative restoration strategy to planting over large areas (Rey Benayas et al.
2008). However this restoration strategy has not been widely implemented nor
rigorously tested in comparison with other restoration methods. Using Web of Science,
we systematically searched articles published in Restoration Ecology, Ecological
Applications and the Journal of Applied Ecology from 1999-2009 to find studies that
specifically examined patch size effects of planted woody vegetation on at least one
restoration process. We also encountered papers from other journals on an ad-hoc
basis. We found only four studies that considered patch size in analyses (Zahawi and
Augspurger 2006; Fink et al. 2009; Munro et al. 2009; Selwood et al. 2009). Numerous
studies, however, used replicated patches of planted seedlings equivalent in size to our
small, medium and large patches to evaluate or compare restoration techniques such
as irrigation and different soil preparation methods and their effects on phenomena such
as seedling survival and growth (e.g. Burkett et al. 2005; Eränen and Kozlov 2006). The
implicit assumption of these studies is that their results would be similar for planted
patches of larger sizes. Given the results of this study and others, it is clear that patch
size is an important variable to consider and that results from one patch size may not
necessarily scale up or down to other patch sizes.
The effects of restoration patch size on the behavior of animals have received
almost no attention (Fink et al. 2009) although animal activities such as foraging
contribute to ecosystem recovery processes such as pollination, seed dispersal, and the
regulation of herbivorous insect populations (Sekercioglu 2006). Our results showed

32

greater foraging success, i.e. higher prey attack rates and lower search effort, in larger
patches (a few thousand metres squared) than in smaller patches (a few hundred
metres squared or less) for three of the four species we studied. Patch size effects on
foraging success are likely to have consequences for rates of energy acquisition and
fitness (Lima and Dill 1990; Olsson et al. 2002). Thus, from the perspective of
insectivorous birds, larger patches of planted trees provide more food for a given effort
than smaller patches. We expect patch size effects for other insectivorous taxa, for
example bats, would be similar. While planted patches of hundreds of metres squared
may facilitate ecosystem recovery in some ways (Rey Benayas et al. 2008; Cole et al.
2010), our results indicate they will provide lower quality habitat than patches of several
thousand metres squared for some animals whose presence and activities contribute to
successful restoration. These differences may diminish over time as the patches spread
and additional woody vegetation establishes naturally. Our results also demonstrate the
value of using a behavioral ecological approach in the assessment of restoration patch
quality. Behavioral observations, unlike species presence/absence data, provide
information on the mechanisms likely to be driving patterns of patch use (Lindell 2008).
Previous studies on birds conducted in non-restoration contexts have used
foraging behavior as a proxy measure of food availability to document a wide range of
patterns, including variation in habitat characteristics (e.g. Pomara et al.; Lyons 2005).
However, it is important to concurrently measure arthropod abundance to verify that
foraging patterns reflect food availability. We found that arthropod abundance was
higher in larger patches, indicating that our findings of higher prey attack rates and
lower search effort in larger patches reflect variation in food availability among patch

33

sizes. Thus, attack rates and search effort are likely to be good indicators of food
availability and habitat characteristics and a useful index of the value of a patch from the
perspective of foraging animals (Johnson 2000; Pomara et al. 2003; Lewis et al. 2008).
We assumed that structural characteristics of smaller patches, including low
canopy cover and high proportions of edge, would expose foraging individuals to
greater predation risk from aerial predators than when they foraged in larger patches
(Suhonen 1993; Kotler et al. 2004). Thus, we predicted that patch size would be a
significant factor affecting vigilance, our measure of predation risk, but only rufouscapped warbler showed greater vigilance in smaller patches. Our data indicate that food
availability is the primary driver of foraging differences across patch sizes while
predation risk may be important only for some species in some years. However,
methodological consideration may also have influenced the results on predation risk.
When conducting observations, we assumed that foraging and vigilance were mutually
exclusive and only measured overt scanning (head up with bill horizontal or above) as
evidence of anti-predator vigilance (Fernández-Juricic et al. 2005). Recent studies show
that some foraging birds use peripheral vigilance (while head is down foraging) in
addition to overt scanning vigilance (Lima and Bednekoff 1999). The birds we studied
may employ peripheral vigilance while foraging, which we were not able to measure. In
addition, vigilance strategies may differ depending on the type of predators present. We
assumed that predatory birds were the most important predators because we regularly
observed them in the sites and the study species responded to them. Head-up
vigilance is likely to be a good reflection of predation risk from predatory birds but may

34

be less representative of risk from mammals and reptiles, which we observed
occasionally.
The objective of planting in small patches, i.e. tens to a few hundred metres
squared, is to provide colonization foci in degraded lands where little native vegetation
exists, using a fraction of the financial and labor resources required for planting
seedlings to cover entire areas (Zahawi and Augspurger 2006). However, in choosing a
restoration design it is important to balance economic benefits with ecological
consequences and to prioritize the processes and species to be restored. We found
2

that, in 3-4 year old restoration sites, larger patches (> 3500 m of canopy cover)
2

provide approximately twice the density of arthropods as smaller patches (< 350 m ).
This difference is reflected in the greater foraging success of insectivorous birds in
larger patches. Tropical birds are key players in ecosystem functions (e.g. Van Bael et
al. 2008) and forest birds are often a focus of conservation efforts because of tropical
deforestation (Sodhi et al. 2008). Thus, forest restoration efforts using a patch-based
planting scheme should consider patch size as an important factor that is likely to affect
the quality of the patches as habitat for birds that use woody habitat. We recommend
that small patches be avoided and that patches of at least a few thousand metres
squared be planted when resources are available.

35

CHAPTER 2
ACTIVE OR PASSIVE FOREST RESTORATION? ASSESSING RESTORATION
ALTERNATIVES WITH AVIAN FORAGING BEHAVIOR
INTRODUCTION
Conversion of forests to pastures for cattle grazing and other agriculture has
created large areas of deforested and degraded land in the humid tropics (FAO 2005).
Developing effective strategies to restore these lands is critical because of their high
biodiversity and their significant role in processes such as water and carbon cycles
(Brown 1993). Ecological restoration strategies can be divided into two broad
categories. Passive restoration is where environmental stressors such as cattle grazing
and agriculture are removed and colonization by shrubs and trees and secondary
succession takes place naturally. Active restoration is where the land is managed by
planting vegetation, weeding, burning and/or thinning to achieve a desired structure
(Rey Benayas et al. 2008).
A limited number of studies have directly compared active and passive
restoration and found that the regeneration pathways and resulting vegetation structure
are likely to be significantly different (Parrotta 1992; Carnevale and Montagnini 2002;
Florentine and Westbrooke 2004). Tropical restoration sites planted with tree seedlings
had higher biomass (Parrotta 1992), understory tree density and diversity (Carnevale
and Montagnini 2002) and greater seedling recruitment (Parrotta 1992; Florentine and
Westbrooke 2004) than control sites that were not planted. However, Powers et al.
(1997) found that seven-year old abandoned pastures were intermediate in understory
vegetation species richness compared to monocultures of several different tree species.

36

Resulting differences in vegetation structure may influence several components
of the habitat quality of passively versus actively restored sites for animals, including
food availability. Canopy cover (Perfecto et al. 1997; Johnson 2000), structural
heterogeneity (Basset et al. 2001; Richards and Windsor 2007) and plant species
diversity (Holmes and Schultz 1988; Johnson 2000) are correlated with the biomass and
distribution of arthropods. In turn, arthropod biomass influences the foraging patterns of
animals that prey on arthropods, such as insectivorous birds (Maurer and Whitmore
1981; Robinson and Holmes 1982 Johnson 2000). Foraging attack rates (number of
attacks on prey per unit time) have been shown to be positively associated with food
availability for birds in both tropical (Johnson 2000; Morrison et al. 2010) and temperate
habitats (Kilgo 2005; Lyons 2005). Attack rates, along with search effort (locomotor
movements per attack), affect rates of energy acquisition and expenditure that can
significantly affect fitness (Pomara et al. 2003; Butler et al. 2005; Lewis et al. 2008).
Measuring these components of the habitat quality of restoration sites for birds is
important because birds are significant contributors to ecological functions as pollinators
(Feldman 2006), seed dispersers (Wunderle 1997) and consumers of herbivorous
insects (Van Beal et al. 2003).
We evaluated vegetation structure and food availability (measured using both
foraging behavior and arthropod biomass) between active and passive restoration
treatments for three resident tropical bird species. Active and passive restoration sites in
this study were part of a replicated forest restoration experiment in southern Costa Rica.
Active restoration treatments were former pastures planted in 2004 with four species of
tree seedlings in a uniform planting design. Passive restoration treatments were former

37

pastures burned once in 2003 and again in 2004, then abandoned to regenerate
naturally over time.
We expected that the active treatment sites would have greater vegetation
density in the highest canopy strata and greater heterogeneity in the vertical structure of
the vegetation than the passive restoration treatments (Parrotta 1992; Carnevale and
Montagnini 2002; Florentine and Westbrooke 2004). These vegetation characteristics
have been shown to affect the biomass and distribution of arthropod prey, which in turn
affect the foraging patterns of insectivorous birds (Holmes and Robinson 1981; Holmes
and Schultz 1988, Johnson 2000). Thus we predicted that arthropod biomass and prey
attack rates would be highest, and search efforts lowest, in the active restoration sites.

METHODS
Study Sites
The study was conducted from 15 May - 15 August 2008 in active and passive
restoration sites in southern Costa Rica. The study sites were located between the
Organization for Tropical Studies' Las Cruces Biological Reserve (hereafter Las Cruces;
8° 47' N, 82° 57' W) and the town of Agua Buena (8° 44' N, 82° 58' W; elevation 11001300 m). In the past 50 years, most land surrounding Agua Buena was cleared for
agricultural use, primarily for the production of sun coffee with some land used for
pasture and subsistence agriculture. Less than 27% of the original forest remains in the
region, mostly in small fragments along waterways or in the mountains at higher
elevations (Daily et al. 2001).

38

The active restoration sites (n = 4) had previously been used for greater than 18
years for agriculture (typically coffee first and then pasture). Active sites were planted in
2004 with four species of tree seedlings: Terminalia amazonia (J.F. Gmel.) Exell
[Combretaceae], Vochysia guatemalensis Donn. Sm. [Vochysiaceae], Erythrina
poeppigiana (Walp.) O. F. Cook [Fabaceae], and Inga edulis Mart. [Fabaceae]. All
species are native or naturalized to the region and are widely used in tropical forest
restoration (Nichols et al. 2001; Carpenter et al. 2004). At each site a 50 x 50 m area
was planted with trees in rows to cover the entire plot. Vegetation was cleared around
planted trees every 2-3 months for the first two years after planting to allow seedlings to
become established. The small number of seedlings that died during this time were
replanted. No further clearing or planting was conducted after May 2006. Tree heights
ranged from 2-10 m in 2008. Trees had formed an almost continuous canopy over each
50 x 50 m area.
The passive restoration sites (n = 2) were located in the Melissa's Meadow (MM)
section of Las Cruces. MM is a 31 ha tract of land that had been planted with
Pennisetum grass species and used for cattle grazing until 1998, when Las Cruces
acquired it. MM is located adjacent to the primary and secondary forest (250 ha) of Las
Cruces. The passive sites were two 1 ha square plots (100 m apart) that were burned to
clear the grass cover in February of 2003 and again in February of 2004 and
subsequently abandoned. No vegetation was cleared prior to the burn treatments
except for a 12 m fire break surrounding each 1 ha plot. After the burning, there was no
vegetation standing, except for 2-3 remnant trees per plot that did not burn. Currently
the sites are dominated by the native, woody species Heliocarpus appendiculatus

39

Turcz. [Tiliaceae], Miconia trinervia (Sw.) D. Don ex Loudon [Melastomataceae],
Cecropia spp. (mostly C. obtusifolia and C. peltata) and several Piper spp.
One of the active restoration sites is also located within the MM section of Las
Cruces, within 50 m of the mature forest of Las Cruces and between 100-300 m from
the two passive sites. The other three active restoration sites are located 3-10 km away
from Las Cruces. One of these three sites is located adjacent to a 43 ha mature forest.
The other two sites were located in landscapes with small amounts of forest (10-20%
secondary forest within a 500 m radius surrounding the site, Cole et al. in press). All
four active restoration sites were at least 3 km apart from each other.

Foraging Observations
We conducted foraging observations of three species of insectivorous and
omnivorous birds in the restoration sites: Cherrie's Tanager (Ramphocelus
costaricensis), Rufous-capped Warbler (Basileuterus rufifrons) and Common Todyflycatcher (Todirostrum cinereum). Cherrie's Tanager is a frugivore-insectivore while the
other species are primarily insectivores (Stiles and Skutch 1989). These three species
were the most commonly observed foliage-gleaning insectivores in the restoration sites.
All observations were of birds foraging for arthropods rather than fruit.
Observations were conducted from 11 May - 8 August 2008 and were conducted
at each site approximately once every 7-10 days between sunrise and 1100 hours.
Each site was systematically searched to locate foraging birds and observations were
made on birds throughout each site. Foraging observations lasted as long as the bird
was in sight. For each observation, we recorded the bird species, number of prey

40

attacks (attempts to capture insect prey), and movements between attacks: hops, short
flights (<1 m), and long flights (>1 m) (Pomara et al. 2003). Only foraging bouts of at
least 20 seconds were included in analyses (Robinson and Holmes 1982). All
observations were continuously recorded on a digital voice recorder by E.B Morrison.
We defined 'total observation time' as the number of minutes during which a focal
bird was watched, minus the number of minutes in which it was engaged in activities
other than foraging (i.e. preening; Lyons 2005). We calculated attack rate (number of
attacks per minute) by dividing the total number of attacks by the total observation time
(Pomara et al. 2003; Lyons 2005). Search effort (number of movements per attack) was
calculated by dividing the total number of movements by the total number of attacks in
each observation (Pomara et al. 2003).
Some birds were color-banded which allowed individual identification. A total of
ten Cherrie's Tanagers, nine Rufous-capped Warblers and five Common Todyflycatchers were colorbanded across all sites used in this study. Based on sightings of
color-banded individuals and birds in pairs, we estimated that two pairs of Rufouscapped Warblers and Common Tody-flycatchers were using each of the active
restoration sites. Each passive restoration sites had at least two pairs of each of these
species. Cherrie's Tanagers are sexually dimorphic and found in loose groups yearround (Stiles and Skutch 1989). Based on observations of banded individuals, the
Cherrie's Tanager groups in our study were 6-10 individuals in size, which matched
reported group sizes in the same region (Krueger et al. 2008). The one active and two
passive restoration sites located within the MM section of Las Cruces may have been
close enough that the same Cherrie's Tanagers used all three sites. The other three

41

active sites were located far enough apart from each other and from MM (3-10 km) that
the individuals at these sites were different.
No more than one foraging observation was made per banded individual per day
and we waited at least one day between recording observations of unbanded individuals
of the same species (or sex for Cherrie's tanager; Wunderle and Latta 1998; Johnson
2000; Pomara et al. 2003). We included multiple observations of the same individuals in
our analysis because it ensured the inclusion of variability in the behavior of individual
birds (e.g. Johnson 2000; Lyons 2005; Jedlicka et al. 2006). Such observations were
not truly independent, and their inclusion may have inflated sample sizes somewhat and
led to an increased Type I error rate in statistical tests (Hurlbert 1984).

Arthropod Sampling
General Methods - The focal bird species are primarily surface gleaners, taking
insects directly from vegetation substrates such as leaves or branches (Stiles and
Skutch 1989). A method called branch clipping was used to collect arthropod samples
from branches and leaves (Johnson 2000). A collapsible bag was positioned around the
outer 0.75-1.0 m of a selected branch, the mouth of the bag was closed and the branch
clipped free with a tree pruner. Bags were visually inspected for arthropods and the
clipped vegetation weighed to the nearest g using a Pesola scale. All arthropods were
immediately categorized into 1 mm size intervals and identified at least to the level of
order. To standardize for variable amounts of clipped vegetation, each sample is
expressed as [(g of arthropod biomass/ g of clipped vegetation)*1000].

42

Every site was sampled for arthropods twice; once between 4-8 June and once
between 7-11 July 2008. We identified species to the following categories: Aranae,
Orthoptera, Dermaptera, suborder Heteroptera (order Hemiptera), suborder Homoptera
(order Hemiptera), Coleoptera, Lepidoptera adults, Lepidoptera larvae, Diptera,
Formicidae (ants) and other Hymenoptera (bees and wasps), unknown adults, and
unknown larvae. Length-weight regressions from Johnson and Strong (2000) and
Schoener (1980) were used to calculate arthropod biomass for all insect categories.
These insect groups have been reported in the stomach samples of the focal bird
species and analogous species in similar habitats (C.H. Sekercioglu 2009, personal
communication; Sekercioglu et al. 2002; Poulin et al. 1994; Poulin and Lefebvre 1996).
Foraging observations and branch clip sampling were conducted concurrently.
As a result, we selected the proportions of tree species for arthropod sampling based on
approximations of tree species use by the three focal bird species (measured as
proportion of attacks that occurred in each tree species). Branch clip samples were
taken from 25-30 trees in each active site and 60 trees in each passive restoration site.
In the passive restoration sites, trees used for sampling included mostly H.
appendiculatus, M. trinervia, Cecropia spp. and Piper spp. These tree species were the
most common species in the sites and the most commonly used by foraging birds.
Trees used for arthropod sampling were located throughout each site. Branch clip
samples were taken from heights that matched the range of prey attack heights for the
three bird species
Weighted Branch Samples - After the field work concluded, we found that in the
actively restored sites, the percentage of prey attacks was greater in Inga than in the

43

other three tree species for each of the bird species. The proportion of branch clippings
among the four tree species did not match the percent of prey attacks in each tree
species (see Morrison et al. 2010). We undersampled Inga and oversampled the other
three tree species. Measures of food availability should take into account the bird's
foraging activity, use of certain tree species or locations within the habitat (e.g., Hutto
1990; Poulin and Lefebvre 1997). Thus, we weighted branch clip samples from each
tree species to match the proportions of prey attacks in each tree species (see Morrison
et al. 2010 for details of the weighting procedure).
The weighted sample totals were used in calculations of arthropod biomass for
active sites (Johnson 2000). The proportions of branch clip samples in trees in the
passive sites approximately matched the proportions of prey attacks in those tree
species. Thus we did not correct the arthropod samples from the passive restoration
sites.

Vegetation Sampling
We characterized the vegetation structure of each treatment by measuring tree
size (diameter at breast height, dbh), tree species diversity, tree density and vertical
foliage profiles to measure the percent vegetation cover at different subcanopy strata at
each site. Circular sampling plots of 16 m diameter (0.02 ha) were established in each
of the four active sites (two plots in each site) and in the two passive sites (three plots in
each). The location of the vegetation sampling plots were chosen at random. In each
plot, vegetation was sampled at 21 points located at 1.6 m intervals along the north,
south, east and west radii of the plot (Wunderle and Waide 1993). A 3 m tall pole

44

marked at 0.5 m intervals was placed vertically at each sample point. We recorded the
presence or absence of foliage touching the pole within each height class. For height
intervals above 3 m, we sighted along the pole (the observer stood at the base of the
pole) and recorded the presence/absence of foliage in each of the following estimated
height intervals: 3-4, 4-6, 6-8, 8-10, 10-12 and12-15 m. For each height interval, percent
vegetation cover was calculated by dividing the number of points in which foliage was
present in that height interval by the total number of sample points (n = 21 per plot) and
multiplying by 100.
We also measured the dbh of each tree greater than or equal to 3 cm in diameter
within the circular plot and calculated an average dbh for each sampling plot. We
counted the number of trees (≥ 3 cm dbh only) in each circular plot and used those data
to calculate average tree density per hectare. We estimated tree species richness by
counting the number of different tree species in each circular plot and averaging over
the multiple plots per site.

Data Analysis
We used Principle Components Analysis (PCA) to determine which vegetation
characteristics explained most of the variance among restoration treatments. We
condensed the 12 height intervals into four categories: 0-1.5 m (understory), 1.5-3 m
(midstory), 3-8 m (subcanopy) and >8 m (canopy) (Pejchar et al. 2005). These four
height intervals were included in the PCA along with average dbh and tree density per
site. Before performing the PCA analyses, we standardized the variables to account for
differences in variable units (Gotelli and Ellison 2004).

45

All dependent variables were checked for normality and homoscedasticity. Attack
rate and search effort were log transformed to meet these assumptions. We used t-tests
to examine the effect of restoration treatment (active or passive) on average tree dbh
and density in the vegetation sampling plots. We used t-tests to examine the effect of
restoration treatment on attack rate and search effort for the three bird species, using
the values from individual observations as the dependent variable. We also ran t-tests
using the mean foraging behavioral values per plot (four active and two passive
treatment plots) as the dependent variable, to compare average foraging behavior
between the active and passive treatments. This latter test was done to assess the
potential influence of having multiple observations from some individual birds in some of
the plots. We used general linear models (GLM) to examine the effect of restoration
treatment, the time of day at which arthropod samples were collected and sampling
period (June and July) on arthropod biomass (measured as g of arthropod biomass/ g of
clipped vegetation)*1000) . All data were analyzed using SYSTAT (SAS Institute 2004).
Throughout, we present results as mean ± SE and differences between groups were
considered significant when p < 0.05.

RESULTS
Vegetation
The two restoration treatments showed substantial differences in vegetation
structure (Figure 2.1). The first two principle components combined explained 75.2% of
the variance. The first principle component explained 46.9% of the variance and
separated the vegetation sampling plots in the two restoration treatments based on

46

average tree dbh and percent vegetation cover in the understory vs. overstory. Passive
treatments had smaller tree dbh and greater percent vegetation cover in the understory
and midstory, while the active treatments had greater tree dbh and a much sparser
understory and midstory. The second principle component explained 28.3% of the
variance and separated the vegetation sampling plots in the treatments based on tree
density. The active treatments consistently had low density while the passive sites
showed substantial variability in tree density.

3
P
assive

Principle component 2

2

Active

1
0
-1
-2
-3
-4
-5
-4

-3

-2

-1

0

1

2

3

4

5

6

Principle component 1
Figure 2.1: Principle components 1 and 2 from the PCA of habitat variables in the
two restoration treatments (Active = 8 sampling plots, Passive = 6 sampling
plots). Treatments are represented by symbols.

47

The active treatments had only four species of trees, while the number of tree
species in the passive treatments ranged from 7-12. The vertical foliage height profiles
confirmed results from the PCA and showed that the active treatments had much
sparser understory and midstory while the passive treatments had greater understory
and midstory cover (Figure 2.2). While both treatments had almost complete canopy
cover, active treatments had a fairly uniform canopy height of 6-8 m tall while passive
treatments had less uniform canopy height (Figure 2.2). Average dbh of trees in the
active restoration treatments was significantly larger than in the passive treatments (t = 5.15, df = 12, p < 0.001). Tree density was significantly higher in the passive treatments
(t = 3.21, df = 12, p = 0.007).

48

> 12

Active

10 - 12
8 - 10
6-8
4-6
3-4
2.5 - 3
2 - 2.5

Height intervals (m)

1.5 - 2
1 - 1.5
0.5 - 1
0 - 0.5

0

10

20

30

40

50

60

70

> 12

80

90

100

Passive

10 - 12
8 - 10
6-8
4-6
3-4
2.5 - 3
2 - 2.5
1.5 - 2
1 - 1.5
0.5 - 1
0 - 0.5

0

10

20

30

40

50

60

70

80

90

100

Percent of height intervals with foliage
Figure 2.2: Foliage height profiles of the active and passive restoration
treatments. These values were averaged across the multiple sampling plots per
site and used to produce vertical foliage height profiles. Standard error bars are
presented.
49

Arthropod Biomass
There was no difference in arthropod biomass between active and passive
restoration sties (F1,453 = 1.98, p = 0.23). There was no effect of time of day (F1,453 =
0.037, p = 0.20) or sampling period (June or July) on arthropod biomass (F1,453 = 2.32,
p = 0.18).

Foraging Behavior
Total observation time across all sites was 169 hours. Mean duration of foraging
observations (mean ± SE) was 0.98 ± 0.05 min for Cherrie's Tanagers (n = 43), 1.07 ±
0.06 min for Rufous-capped Warblers (n = 45) and 0.89 ± 0.07 min for Common Todyflycatchers (n = 22).
Cherrie's Tanager showed no difference in either attack rate (t = -1.79, df = 41, p
= 0.08) or search effort (t = 1.80, df = 41, p = 0.08) between the active and passive
restoration treatments. Results were similar for Rufous-capped Warbler (attack rate: t =
-0.20, df = 43, p = 0.84; search effort: t = -1.84, df = 43, p = 0.07) and Common Todyflycatcher (attack rate: t = 0.45, df = 20, p = 0.66; search effort: t = -0.47, df = 20, p =
0.65).
Mean foraging behavioral values showed no significant differences for all three
species for either attack rate (Cherrie's Tanager: t = 0.85, df = 4, p = 0.45; Rufouscapped Warbler: t = 0.40, df = 4, p = 0.71; Common Tody-flycatcher: t = 0.01, df = 4, p
= 0.94) or search effort (Cherrie's Tanager: t = 1.35, df = 4, p = 0.25; Rufous-capped
Warbler: t = 0.64, df = 4, p = 0.56; Common Tody-flycatcher: t = 0.12, df = 4, p = 0.91).

50

These results suggest that multiple observations from individual birds within treatments
did not skew foraging behavior patterns.

DISCUSSION
Food availability is an important component of habitat quality that is influenced by
vegetation structure (Holmes and Robinson 1981; Johnson 2000; Johnson 2007). In
our study area, passive restoration treatments had greater tree species diversity, denser
and smaller trees, and a greater percent vegetation cover in the understory than did the
active treatments. Despite these differences, arthropod biomass was not different
between the active and passive treatments and avian foraging behavioral measures
reflected these patterns of food availability. All three bird species showed no significant
differences in prey attack rate and search effort between the two restoration treatments,
indicating rates of energy acquisition are likely similar between the treatments.
Numerous studies have found that attack rates are greater and search effort
lower when arthropod availability is higher. For example, Hooded Warblers (Wilsonia
citrina) had lower attack rates near hardwood forest gaps than in the interiors, which
corresponded with lower arthropod abundance (Kilgo 2005). Johnson (2000) found that
attack rates of several Neotropical migrant species closely matched the patterns of
arthropod biomass among shade coffee farms, citrus plantations and dry forests in
Jamaica. American Redstart (Setophaga ruticilla) attack rates and arthropod densities
were highest on temperate lake shorelines when compared to inland areas (Smith et al.
2007). Our previous work in the study sites showed that attack rates were higher and
search efforts lower in restoration patches with the highest arthropod biomass (Morrison

51

et al. 2010). Thus, arthropod biomass and bird attack rates are consistently positively
related.
The availability of food in a habitat affects an organism's energy intake and
expenditure, which ultimately affects fitness (Pomara et al. 2003; Butler et al. 2005;
Lewis et al. 2008). Our data on foraging behavior and arthropod biomass suggest that
active and passive sites are approximately equal in food availability for insectivorous
birds. However, other important components of habitat quality for insectivorous birds,
including nest site availability, predation risk and reproductive success (Johnson 2007)
were not measured in this study and could have differed between active and passive
restoration sites.
The active and passive restoration sites had similar land-use histories and time
since abandonment. Therefore, these potential influences on vegetation trajectories
were controlled. The restoration sites located within the MM section of Las Cruces (one
active and the two passive sites) showed strikingly different structure, suggesting that
restoration strategy (passive or active) was the primary influence on vegetation
structure. This is in accord with previous work that shows that tree species diversity,
density and vegetation structure in planted sites are significantly different when
compared to passive sites (Cusak and Montagnini 2004; Florentine and Westbrooke
2004). Passive sites tend to be influenced most by the characteristics of the local
vegetation surrounding the sites (Günter et al. 2007), while the vegetation structure and
tree species found on active restoration sites depends strongly on the characteristics
(pioneer vs. mature-phase) and diversity of the tree species planted (Carnevale and

52

Montagnini 2004; Firn et al. 2007), the planting density (Vesk et al. 2008) and the type
of maintenance done on the sites after planting (Piotto et al. 2003).
One potential influence on vegetation structure and, potentially, on arthropod
biomass and foraging behavior, that we were not able to address was the effect of
landscape. The active restoration sites were located in landscapes that varied in forest
cover within a surrounding 500 m radius. The seedlings planted in all the active sites
grew well and each of the sites developed a closed canopy. In addition, the patterns of
arthropod biomass and foraging behavior of birds in the active sites were consistent
over a range of landscapes (high to low forest), suggesting that landscape was not a
significant influence on vegetation, arthropods, or birds in the active sites. In contrast,
both passive sites were located adjacent to mature forest. Previous studies have shown
that distance to forest is an important influence on forest regeneration in degraded
lands, in part because of the greater availability of seeds in sites close to forest (Holl
1999). Thus, it is possible that passive sites in our study region located in less forested
landscapes would have had less tree diversity and smaller trees, which could influence
arthropod biomass and avian foraging behavior. Future work should compare the
effectiveness of active versus passive restoration strategies in a paired design in
landscapes with a range of forest cover.
The decision to use active or passive strategies in restoration projects should be
based on the restoration project goals, budget, and site and landscape characteristics.
Given that the recovery of passive restoration sites will depend strongly on land use
history , time since abandonment and, potentially, landscape factors (Chazdon 2003),
lands that have been used intensively, are severely degraded, and that are far from

53

seed sources may not recover naturally and could benefit from a more active restoration
strategy ( Wijdeven and Kuzee 2000; Günter et al. 2007). Our results suggest that
passive restoration in high-forest cover landscapes in our study region is equally as
effective as active restoration, from the perspective of insectivorous birds.
The costs of the two restoration strategies are substantially different (Chazdon
2008). Passive restoration requires no management costs except to eliminate the
factors that caused the original degradation (i.e. grazing, agriculture). The passive sites
in this study were burned twice to remove grass cover and were then left to regenerate.
Each burning event took place over one day, required labor only to clear a fire break
around each plot, and set and monitor the burn. No further labor or monitoring were
needed after the final burn (R. Quiros 2009, personal communication). Active
restoration is generally more costly and labor intensive than passive restoration (Rey
Benayas et al. 2008). For the active restoration treatments in this study, costs included
the purchase of tree seedlings, labor costs to clear sites and plant seedlings, the
purchase and application of fertilizers and post-planting site maintenance (including
clearing around seedlings). Including planting and labor, costs were $120-180 per 100
tree seedlings planted, or approximately $900-1400 per hectare (for two years of site
maintenance; Zahawi and Holl 2009). Given the differences in costs, and the similar
effectiveness of the two strategies, passive restoration would be preferred in high-forest
cover landscapes in the study region.

54

CHAPTER 3
VIGILANCE BEHAVIOR OF A TROPICAL BIRD IN RESPONSE TO INDIRECT AND
DIRECT CUES OF PREDATION RISK
INTRODUCTION
Animals must balance the costs and benefits of foraging in particular habitat
patches (Brown 1988; Kilpatrick 2003). Costs such as predation risk can have
significant effects on an individual's foraging decisions (Lima and Dill 1990; Nystrand
2007). Animals use a number of behavioral strategies to enhance the probability of
predator detection, for instance, by increasing vigilance (Barnard 1980; Lima 1992;
Lima and Bednekoff 1999; Whittingham et al. 2004), foraging in less risky habitats, or
reducing travel distances to food sources (Cresswell 1999).
Vigilance is a common anti-predator defense mechanism to reduce predation risk
in the short-term. However, vigilance often reduces food intake and foraging efficiency
(Elgar 1989) and over the long-term, increased vigilance could potentially have a
negative impact on fitness through a higher risk of starvation (Watson et al. 2007) or
lack of adequate energetic resources to allocate to reproduction (Wolff and Van Horn
2003; but see Lind and Cresswell 2005).
Animals likely use multiple environmental cues to assess predation risk (Lima
and Dill 1990). Indirect cues of predation risk are not provided by the predator itself, but
are correlated with the likelihood of detecting, encountering and succumbing to a
predator (Thorson et al. 1998). For example, habitat structure, such as canopy cover or
habitat patch size, might be an indirect cue of an individual's risk of detection by a
predator or the probability of escape from a predator (Thorson et al. 1998; Orrock et al.
2004).

55

Indirect cues such as the physical characteristics of a habitat may provide
relatively consistent information about predation risk from a variety of potential predation
threats (Blumstein et al. 2000; Orrock et al. 2004). However, indirect cues only allow
animals to assess their potential, but not actual, predation risk at any given moment and
they tend to overestimate risk when predators are absent and underestimate risk when
predators are present (Sih 1992; Orrock et al. 2004).
In contrast, direct cues, particularly auditory and visual cues of a predator, are
highly reliable. However, by the time potential prey detect a predator by a direct cue,
they may not be able to respond successfully. Thus animals likely integrate both indirect
and direct cues to evaluate their risk of predation. Studies in rodents show that animals
respond more strongly to live or model predators when foraging in risky environments
(open or illuminated habitats) than in safe environments (covered or dark habitats)
(Kotler et al.1991; Schmidt 2006). Other rodent studies have shown that foragers
perceive a greater predation risk in response to indirect cues, such as degree of habitat
cover or illumination, than to direct cues, such as predator scent, when both are present
(Thorson et al. 1998; Orrock et al. 2004).
Birds are also likely to use multiple cues to assess predation risk. Much evidence
shows that birds increase vigilance in response to both indirect cues, such as visual
obstructions (Elgar 1989; Whittingham et al. 2004) and amount or type of vegetation
cover (Eckman 1987; Carrascal and Alonso 2006), and direct cues of predator
presence, including vocalizations (Hauser and Caffrey 1994) and model predators
(Sullivan 1994). However no study has combined both types of cues to empirically

56

examine whether birds rely more on direct or indirect cues, or a combination of both
types.
I examined the effects of indirect and direct cues of predation risk on the
vigilance behavior of a tropical omnivorous bird, Cherrie's Tanager (Ramphocelus
costaricensis), in a controlled field experiment. Habitat patches of different sizes were
planted with four species of trees in 2005 as part of a large-scale experimental tropical
forest restoration study. Restoring woody habitat by planting small patches of
vegetation, of tens to hundreds of square meters, is an increasingly common type of
forest restoration strategy (Rey Benayas et al. 2008). It is important that forest
restoration efforts using a patch-based planting scheme consider patch size as an
important factor that is likely to affect the quality of the patches as habitat for birds
(Morrison et al. 2010). The habitat structure of these types of restoration are likely to
influence behaviors, such as vigilance, that are could have potential negative fitness
effects (Alcock 2005).
I used feeders placed at three different locations, the center and edge of a large
patch and at the center of a small patch, as indirect cues of predation risk from an aerial
predator. We assumed that small patches, and the edges of large patches, represented
riskier habitat for foraging birds than the center of large patches because the structural
characteristics of these locations, including low amount of canopy cover and high
proportions of edge, would expose foraging individuals to greater predation risk then if
they foraged in the center of large patches (e.g. Suhonen 1993; Kotler et al. 2004;
Carrascal and Alonso 2006). Vocalization playbacks of a common falcon predator and
non-predator bird (control) were used as direct cues of predation risk. I then recorded

57

the vigilance response of birds to the two playback treatments in each of the three
feeder locations.
I tested the hypothesis that indirect cues of predation risk would influence
several different measures of vigilance behavior, including time spent alert, head-turn
frequency, the frequency and duration of head-up bouts, and latency to flee the feeder,
in response to a direct cue of predation risk. If birds are using both types of cues to
assess predation risk, then the change in vigilance response from pre- to post-playback
of the predator vocalization will be highest in small patches and at the edges of large
patches and lowest in large patches.

METHODS
Research Sites
The study was conducted from June to July 2008 in restoration sites located near
the town of Agua Buena (8° 44' N, 82° 58' W; elevation 1100-1300 m) in southern
2

Costa Rica. Four restoration sites were planted in 2005 throughout a 30 km area. The
closest two sites were located 3 km apart and all other sites were at least 5 km apart. In
the past 50 years most land surrounding Agua Buena was cleared for agricultural use,
primarily for the production of sun coffee with some land used for pasture and
subsistence agriculture. Currently, less than 27% of the original forest remains in the
region, mostly in small fragments along waterways or on the mountains at higher
elevations (Daily et al. 2001). The landscape surrounding the restoration sites is
dominated by pasture, sparsely-shaded coffee and other agriculture, small human
settlements and some patches of secondary and mature Pacific premontane humid

58

forest (Daily et al. 2001; Sekercioglu et al. 2007). All restoration sites had previously
been used for more than18 years for agriculture.
The restoration sites were planted with four seedling species: Terminalia
amazonia (J.F. Gmel.) Exell [Combretaceae], Vochysia guatemalensis Donn. Sm.
[Vochysiaceae], Erythrina poeppigiana (Walp.) O.F. Cook [Fabaceae], and Inga edulis
Mart. [Fabaceae]. All species are native or naturalized to the region and are widely used
in tropical forest restoration (Nichols et al. 2001; Carpenter et al 2004).
I used two experimental treatments at each site. Treatment plots were 50 x 50 m
2

and either had (1) trees planted in rows to cover the entire plot (plantation, 2500 m ), or
2

(2) trees planted in six patches of three different sizes: two small (16 m ), two medium
2

2

(64 m ) and two large patches (144 m ). Vegetation was cleared around planted trees
every 2-3 months for the first two years of planting to allow seedlings to become
established. The small number of seedlings that died during this time were replanted.
No further clearing or planting was conducted after January 2007.
Tree heights ranged from 3-9 m in 2008. Plantation plots had an almost
continuous canopy over the 50 x 50 m treatment area. The large, medium and small
patches within the island plots remained distinct patches within a matrix of herbaceous
vegetation.

Study Species
Cherrie's Tanagers are sexually dimorphic and found in loose groups of 4-10
individuals year-round (Stiles and Skutch 1989). Based on observations of banded
individuals, the territories of Cherrie's Tanager groups in our study included an entire

59

restoration site (0.5 ha), with approximately 6-10 individuals per site. In the same region
a density of 12 Cherrie's Tanagers per ha was reported by Krueger et al. (2008). The
size of Cherrie's Tanager group territories has not been reported in the literature.
However extensive mist-netting in the restoration sites between 2006-2008 resulted in
no recaptures of individuals between sites, although individuals are often recaptured
within a site (Lindell, unpubl. data). Individuals in this study were colorbanded for
identification.

Experimental Design
Feeders baited with bananas were placed at three locations within each of the
four restoration sites: center of the plantation (hereafter Large Center), within 4m of the
edge of the plantation (hereafter Large Edge) and within 4m of the edge of one large
island (hereafter Small; selected at random from the two large islands in the island
treatment). Thus N=4 for each of the three feeder location treatments. Large Edge
feeders were located along an edge of the plantation that bordered fields consisting of
mostly non-woody vegetation with few to no trees present. Small feeders were placed
4m from the edge of the island. The feeders were small wooden platforms
(approximately 10x25 cm) placed on a stand 1.5m tall with three nails hammered into
the surface of the platform to secure bananas. Feeders were set up in all sites in midMay and left in place for the duration of the experiment. Feeders were baited with
bananas for several days immediately after set-up to attract birds.
There were two experimental periods of four days each: 15-18 June and 5-8 July
2008. On Day 1 of each experimental period, the experiment was conducted at Site 1,

60

on Day 2, Site 2 was used, etc. Feeders were baited with bananas one day before
each experimental trial. On the day of each trial, fresh bananas were placed on the
feeders at 5:00am.
I simulated high predation risk using calls of a Barred Forest-falcon (Micrastur
rufficolis), a common raptor in the landscape that preys on birds (Thorstrom 2000;
Panasci and Whitacre 2002). The low predation risk treatment used calls of a Plain
Wren (Thryothorus modestus), a common insectivore found in the restoration sites that
has a loud song and is unlikely to be a food or territorial competitor with Cherrie's
Tanager (Stiles and Skutch 1989; E.B Morrison pers. obs). One ten second recording
was made for the Barred Forest-falcon and one for the Plain Wren. Each recording
comprised two different songs for Plain Wren and two different call types for the Barred
Forest-falcon. Songs for both species were taken from Ross (1998) and had been
recorded in Puntarenas, Costa Rica. Hereafter I refer to the high predation risk playback
as the Falcon treatment and the low predation risk (or control) treatment as the Wren
treatment.
On each day, feeder locations (Small, Large Center, Large Edge) were randomly
chosen for each of three two-hour time blocks. Between 5:30-7:30am, playbacks were
conducted at one feeder location, the second feeder location between 7:30-9:30am and
the third feeder location between 9:30-11:30am. Feeder locations were rotated among
time blocks for both experimental periods so that each location was used in each time
block at least twice (across all sites).
At each feeder location, I randomly chose a playback order (Falcon then Wren,
or vice versa). When a bird landed and began feeding at the feeder, I started recording

61

using a digital video camera. After 25 seconds, either the Falcon or Wren call was
played (10 s duration) and I continued recording until the bird left the feeder. I then
waited at least 20 minutes, and repeated the playback procedure on a different
individual using the other playback type. No more than two individuals were tested per
feeder location per time block. Thus on one day, three Falcon calls and three Wren calls
were played at each site. The experiment was only conducted on birds that were alone
on the feeder platform. No individual was used more than once in this experiment.
The observer and video camera (JVC Everio Digital Camcorder GZ-MG360)
were set up at least 20m from the feeder and hidden behind vegetation. The playbacks
were broadcast at a standard sound level using a small speaker (LG MSP-100 Portable
Stereo Speaker) placed 5 m from the feeder and connected via a cable to an MP3
player (Sansa Clip 1GB MP3 Player) controlled by the observer.

Behavioral Observations
Windows Movie Maker 2.1 was used to analyze videos frame by frame at 0.1
second intervals. I measured behavioral variables for a 50-s observation period (Martin
and Bateson 1986; Adams et al. 2006) that was divided into two 25-s periods; a 25-s
pre-playback period, immediately followed by a 25-s observation post-playback period
that included the 10-s playback of the vocalizations.
I measured five types of vigilance behavior for both the pre-playback and postplayback periods. First, the proportion of time a bird spends alert, with head-up and bill
horizontal, is positively correlated with a response to increased predation risk in birds
and other animals (Elgar 1989). I also measured the frequency of head-up bouts

62

(number of times bird raised its head from below horizontal to horizontal and above) and
average duration of each head-up bout, which should increase with increased predation
risk to facilitate quick detection of predators (Whittingham et al. 2004). The frequency of
'head-turns', scored as a lateral turn of at least 90 degrees of the focal bird's head, while
vigilant in an upright posture with bill at least horizontal, can increase the quantity and
the quality of information gathered by allowing individuals to scan a wider portion of
space (Jones et al. 2007). I also measured the latency to flee the feeder as the time
from the playback onset until the bird left the feeding platform. Latency to flee may
reflect the level of fear birds have of the predation threat (Ramakrishnan and Coss
2000; Adams et al. 2006).
Some birds left the feeder before the 25-s post-playback period had elapsed.
Thus I measured both pre- and post-playback behavioral variables as a proportion of
the pre- and post-playback observation times. Fourteen of 36 individuals left the feeder
before the 25-s post-playback period ended (mean ± SE of post-playback duration for
birds that left early, 12.06 ± 2.50 s). Birds from six of the 14 observations (all six from
the Falcon call treatment) left within 5-s (2.38 ± 0.63 s) of the initiation of the playback. I
felt that this was not sufficient time to record meaningful data for most of the behavioral
measures. For these six observations I recorded the proportion of time spent alert as
1.0 (100%) and did not record data for the other variables (i.e. frequency of head turns,
head-up bouts, etc.). Sample sizes for each of the dependent measures of vigilance
behavior are reported in Table 3.1.

63

Table 3.1: Sample sizes by treatment for the dependent measures of vigilance behavior.
Total number of individuals tested was 36. For most birds, measurements for all five
vigilance behaviors were collected. For those individuals that left the feeder less than
five seconds after playback initiation, only time spent alert and latency to flee were
measured. 'All others' includes head turn frequency, average duration of head-up bouts
and frequency of head-up bouts.
Treatments
Small
Falcon
Wren
Large Center
Falcon
Wren
Large Edge
Falcon
Wren

Dependent measures of vigilance behavior
Time spent alert and latency to flee All others
7
5

4
5

5
7

5
7

6
6

3
6

Data Analysis
I used repeated measures general linear models (GLM) to examine the effects of
the indirect cue of feeder location (Large Center, Large Edge and Small) and the direct
cue of playback treatment (Falcon, Wren) on the change in vigilance behavioral
measures from pre-playback to post-playback. The vigilance behaviors (dependent
variables) examined were (1) proportion of time spent alert, (2) head turn frequency
(measured as the number of lateral head turns per minute), (3) frequency of head-up
bouts (measured as the number of head-ups per minute) and (4) the average duration
(in seconds) of each head-up bout. The change in behavior from pre- to post-playback
was the within-subject factor. I included experimental period (June or July), site (one of
four sites), time block (5:30-7:30, 7:30-9:30 or 9:30-11:30 am), sex (M or F) and the

64

interaction between feeder location and playback treatment as between-subject factors
in the models.
I used ANOVA to examine the effects of feeder location and playback treatment
on latency to flee (number of seconds the bird stayed at the feeder after initiation of the
playback). I included experimental period, site, time block and sex in the models as
independent variables. I used interaction contrasts to do pairwise comparisons of the
effects of playback treatment on the vigilance behavior variables both within and across
the three feeder locations.
All data were analysed using SPSS 18 for Windows (SPSS, Inc. 2009).
Throughout, I present results as mean ± SE. Differences between groups were
considered significant when P<0.05. All variables were checked for normality. Latency
to flee was log transformed. All repeated measures GLM models met the sphericity
assumption.

65

RESULTS
Proportion of time spent alert
Both feeder location (F1,23=3.70, P=0.04) and playback treatment (F1,23=31.4,
P<0.001) significantly affected the change in proportion of time spent alert from pre- to
post-playback. All other factors in the model were not significant (all F≤0.98, all P≥0.42).
When comparing within feeder location, the proportion of time spent alert
increased significantly after the Falcon playback compared to after the Wren playback in
all three feeder locations (Small: P<0.01; Large Edge: P<0.01; Large Center: P=0.03;
Figure 3.2a). For birds exposed to the Falcon treatment, pairwise comparisons across
feeder locations showed that the change in the proportion of time spent alert was
significantly greater in both the Small and the Large Edge locations when compared to
Large Center locations (P≤0.01 for both; Figure 3.2a). However, difference between the
Small and Large Edge locations were not significant (P=0.67; Figure 3.2a). For birds
exposed to the Wren playback, the change in proportion of time spent alert did not vary
significantly among feeder locations (all pairwise comparisons P≥0.48; Figure 3.2a).

Frequency of head turns
Playback treatment significantly affected the change in head turn frequency from
pre- to post-playback (F1,17=4.80, P=0.04). However, feeder location did not affect the
change in the frequency of head turns (F2,17=2.49, P=0.11). All other factors were not
significant (all F ≤2.04, all P ≥0.16).
When comparing within feeder location, head turn frequency increased
significantly after the Falcon playback compared to after the Wren playback for birds at

66

feeders in the Small locations (P=0.02), but not in the Large Edge (P=0.65) or Large
Center locations (P=0.23; Figure 2b). For birds exposed to the Falcon treatment,
pairwise comparisons across feeder locations showed that the change in the frequency
of head turns was greater in Small locations than in Large Center locations, but this
difference was not significant (P=0.06, Figure 3.2b). There were no significant
differences between the Small and Large Edge locations (P=0.90) or between the Large
Edge and Large Center locations (P=0.10; Figure 3.2b). Head-turn frequency did not
vary significantly among the three feeder locations for birds exposed to the Wren
playback (all pairwise comparisons P≥0.43; Figure 3.2b).

Average duration of head-up bouts
Playback treatment significantly affected the change in average duration of headup bouts from pre- to post-playback (F1,17=8.40, P=0.01). Feeder location did not affect
the average duration of head-up bouts (F2,17=0.97, P=0.40). All other factors were not
significant (all F ≤2.58, all P ≥0.13).
There was a significant increase in the average duration of head-up bouts after
the Falcon playback than after the Wren playback within the Small feeder locations
(P=0.01; Figure 3.1c). Birds in the Large Center and Large Edge locations also showed
a greater increase in average duration of head-up bouts after the Falcon playback then
after the Wren playback, but these differences were not significant (Large Center,
P=0.06; Large Edge, P=0.07; Figure 3.1c). For birds exposed to the Falcon treatment,
pairwise comparisons across feeder locations showed that the change in the frequency
of head turns was greater in Small locations compared to Large Center locations, but

67

this difference was not significant (P=0.07, Figure 3.1c). There were no significant
differences between the Small and Large Edge locations (P=0.15) or the Large Edge
and Large Center locations (P=0.86; Figure 3.1c). Head-up bouts did not differ across
the three feeder locations for birds exposed to the Wren playback (all pairwise
comparisons P≥0.88; Figure 3.1c).

68

Figure 3.1: Mean change in (a) proportion of time spent alert, (b) frequency of head-turn
bouts (measured as head-turns per minute) and (c) average duration of head-up bouts
(measured in seconds) from pre- to post-playback for birds exposed to the Falcon and
Wren treatments at each feeder location (Small Patch, Large Patch Edge, Large Patch
Center). Asterisks represent significant differences (P<0.05) between the Falcon and
Wren playback treatments within feeder location. Letters represent pairwise
comparisons among the Falcon treatments across locations (different letters represent
significant differences at P<0.05). If no letters are present, there were no significant
differences across locations.

69

Change in proportion of time spent alert

0.5

a.
0.4

A

*

*

0.3
0.2

B

*

0.1
0.0
-0.1

Falcon Wren
Small Patch

Change in frequency of head-turns

A

Falcon Wren
Large Patch Edge

Falcon Wren
Large Patch Center

3.0

b.
2.5

*

2.0
1.5
1.0
0.5
0.0
-0.5

Falcon Wren

Falcon Wren

Falcon Wren

Small Patch

Large Patch Edge

Large Patch Center

70

Change in average duration of headup bouts

Figure 3.1 (cont'd)
20

c.
15

*

10
5
0
-5

-10

Falcon Wren
Small Patch

Falcon Wren
Large Patch Edge

71

Falcon Wren
Large Patch Center

Frequency of head-up bouts
Neither playback treatment (F1,16=0.83, P=0.37) nor feeder location (F2,16=0.76,
P=0.48) had a significant effect on the change in frequency of head-ups (measured as
the number of head-ups per minute) from pre- to post-playback) and no other factors
had a significant effect on the model (all F≤1.66 and P≥0.22).

Latency to flee
Birds left the feeder significantly faster after the Falcon treatment then after the
Wren treatment (mean ± SE across all feeder locations: Falcon = 27.5 ± 8.8 s; Wren =
50.9 ± 16.6 s; F1,24=4.53, P=0.04). However, there was no effect of feeder location on
latency to flee (F2,24=2.25, P=0.13). All other model terms were not significant (all
F≤0.74, all P≥0.49; Figure 3.2).

72

120

Latency to flee (seconds)

110
100
90
80
70
60
50
40
30
20
10
0

Wren Falcon
N=5
N=7
Small
Patch

Wren Falcon
N=7
N=5
Large Patch
Center

Wren Falcon
N=6
N=6
Large Patch
Edge

Figure 3.2: Mean latency to flee (measured as the number of seconds the bird stayed at
the feeder after initiation of the playback; log transformed) for birds exposed to the
Falcon and Wren treatments at each feeder location (Small Patch, Large Patch Center,
Large Patch Edge).
DISCUSSION
Both direct and indirect cues of predation risk influenced the vigilance behavior of
foraging Cherrie's Tanagers. Birds in all three feeder locations increased time spent
alert in response to the predator call compared to the non-predator call, similar to other
studies that show birds increase vigilance in response to direct cues of predation risk
including vocalizations (Hauser and Caffrey 1994) and model predators (Sullivan 1994).
However, the indirect cues of patch size and location within the patch mediated the alert
response to the direct cue of predator vocalizations. The change in time spent alert in
73

response to the predator call was significantly greater in the presumably riskier small
patches and the edges of large patches compared to the relatively safe center of large
patches. These results match studies in rodents showing that animals respond more
strongly to live predators or predator vocalizations when foraging in risky environments,
such as open or illuminated habitats, than safe environments, including covered or dark
habitats (Kotler et al. 1991; Schmidt 2006). The structural characteristics of smaller
patches, including low amount of canopy cover and high proportions of edge, would
expose foraging individuals to greater predation risk then when they foraged in the
center of large patches(e.g. Suhonen 1993; Kotler et al. 2004; Carrascal and Alonso
2006). Head-up alertness is an important component of the vigilance response to
threats of predation risk, allowing animals to better detect and respond to predators.
Results for some, but not all, of the other measures of vigilance behavior also
support the hypothesis that birds use both indirect and direct cues to evaluate predation
risk. Birds responded more strongly to direct cues of predation risk while foraging in
small patches, than in the center of large patches. Head-turn frequency increased
significantly in response to the predator compared to the non-predator call for birds
foraging in small patches, but not for birds in the other two locations. Additionally, the
change in head-turn frequency in response to the predator call was greater in small
patches than in the center of large patches. Several studies have shown that head-turn
frequency is highest for birds foraging alone, when they are at greatest risk of predation,
and decreases in larger flocks (Lazarus 1979; Franklin and Lima 2001). Head-turning
has been suggested as an important component of the vigilance response to predation
risk and could have several anti-predator functions. First, head turns may be a predator

74

deterrent by informing the predator it has been detected and that the prey bird is alert
(Woodland et al. 1980; Hasson 1991; Caro 1995). In birds, lateral movements of the
head are also important for object and pattern recognition (Dawkins and Woodington
2000; Dittrich and Lea 2001) and might therefore play an important role in identification
of an animal as a potential predator and threat. Head turns may also increase the
effectiveness of head-up vigilance bouts by increasing the area of the environment
scanned and allowing birds to better judge distances (Jones et al. 2007).
Birds foraging in the riskier small patches also increased the duration of head-up
bouts in response to the predator compared to the non-predator call. Additionally, the
change in head-up duration in response to the predator call was greater in small
patches compared to the center of large patches. However, neither feeder location nor
playback treatment influenced the frequency of head-up bouts. Thus the increase in
time spent alert after the predator playback was likely due to an increase in the duration,
not the rate, of head-up bouts. The potential cost of increasing head-up duration is a
decrease in feeding rate (Fritz et al. 2002). For birds in this study, head-up frequency
was likely correlated with food intake, because when the birds' heads were down, they
were taking bites of food on the feeders. Thus if birds were trying to maintain their food
intake rate post-playback, then head-up frequency would have increased (Cresswell et
al. 2003). Instead, birds increased head-up duration at the cost of food intake, which
could potentially have a negative effect on individual fitness. An increase in head-up
bout duration may allow for a more effective assessment of the environment and could
therefore be useful in situations when predators are known to be present (Glück 1987),
such as after a direct cue of predator presence. Other studies have found that birds

75

increased their head-up duration at the expense of food consumption when indirect
cues (foraging in visually obstructing habitats) indicate greater risk of predation (Elgar
1989; Whittingham et al. 2004).
Rather than responding to the threat of predation by immediately fleeing, birds in
the more risky locations increased their vigilance. This response is aligned with optimal
escape theory, which predicts that prey tolerate higher predation risk when food is
present due to the potential loss of feeding opportunities if they flee upon immediately
detecting a threat (Ydenberg and Dill 1986; Cooper 2003). Research on lizards shows
that, when approached by a predator, individuals wait longer to flee if food is present
than when it is absent (Cooper 2003) and when the quantity of food is high than when it
is low (Cooper et al. 2006). Thus birds in this study may have responded to a cue of
predator presence by increasing vigilance, rather then expending energy to flee and
potentially losing the opportunity to continue foraging at a known, plentiful food source.
Small patches of forest habitat are common in tropical landscapes due to severe
degradation and conversion to agriculture (Gaston et al. 2003) Additionally, forest
restoration efforts are increasingly planting seedlings in small patches of tens to a few
hundred metres squared as a cost-effective alternative to planting over large areas
(Zahawi and Augspurger 2006; Rey Benayas et al. 2008). However, in choosing a
restoration design, it is important to balance economic benefits with ecological
consequences. Results from this study suggest that patch size can have a significant
affect on vigilance behavior that could result in potential fitness costs to birds (Lima and
Dill 1990; Cimprich et al. 2005; Carrascal and Alonso 2006). Thus forest restoration
efforts using a patch-based planting scheme should consider that patch size is likely to

76

affect the quality of the patches as habitat for birds.

77

CHAPTER 4
BIRD AND BAT EFFECTS ON ARTHROPD ABUNDANCES AND LEAF DAMAGE IN
TROPICAL RESTORATION SITES
INTRODUCTION
Planting native vegetation is an increasingly common method to restore degraded
lands (Jansen 2005). In tropical forest restoration projects, the growth and survival of
planted trees is critical to restoration success (Holl et al. 2000). One potential obstacle
to the success of restoration projects is the negative effects of herbivorous insects on
planted vegetation. Herbivorous arthropods cause significant leaf damage in the tropics,
with rates ranging from 10-40% (Coley and Barone 1996). Resulting leaf damage can
reduce the photosynthetic abilities of leaves (Zangrel et al. 2002), negatively affecting
growth, and can increase the likelihood of mortality in saplings and seedlings (Clark and
Clark 1985; Norghauer et al. 2006; Eichhorn et al. 2010).
The top-down limitation of herbivorous insects by predators is an important
ecological function that could facilitate growth of vegetation planted for restoration.
Insectivorous birds can significantly reduce numbers of their insect prey (e.g., Holmes et
al. 1979; Marquis and Whelan 1994). Birds have been shown to reduce herbivorous
arthropod densities and their damage to plants in temperate forests (Marquis and
Whelan 1994; Murakami and Nakano 2000; Strong et al. 2000), neotropical forests (Van
Bael et al. 2003; Van Bael and Brawn 2005), and neotropical agroecosystems
(Greenberg et al. 2000; Mols and Visser 2002; Van Bael et al. 2007). However, the
relative effects of predators on arthropods and leaf damage in restoration systems has

78

not been investigated, despite the importance of vegetation growth and survival for the
success of restoration projects.
Past research on the role of insectivorous birds in ecosystems has used
exclosure experiments, where plants or branches are covered in netting, to show that
avian insectivores influence the abundance of arthropod herbivores and levels of
damage to plants. These studies have attributed all insectivory to birds, although
exclosures were in place 24 hours a day (e.g. Greenberg et al. 2000; Van Bael et al.
2003; Boege and Marquis 2006; Koh 2008). In these cases, insectivory that was
attributed to birds could have also included effects of bat insectivory. Bats are likely to
play a significant role as predators of herbivorous arthropods, because many
herbivorous arthropods in the tropics are active mainly at night and hide during the day
(Kalka and Kalko 2006). Both foliage-gleaning bats (Subfamily Phyllostominae) and
insectivorous birds consume arthropods in similar size-ranges (~4-30mm) and from
similar taxa (LaVal and LaVal 1980). Bats also can consume large quantities of
arthropods: gleaning bats consume over half of their body mass in arthropods nightly
(Kalka and Kalko 2006).
Two recent studies have distinguished bird and bat predation. Both bird and bat
predation reduced arthropod abundances, but not leaf damage, in a Mexican shade
coffee system (Williams-Guillén et al. 2008). In the tropical forest understory of Barro
Colorado Island, Panama, birds and bats reduced arthropods and leaf damage levels
(Kalka et al. 2008). The relative role of birds and bats were different in these two
systems: bat predation exceeded that of birds in the forest understory, while in the

79

coffee plantation, bird predation exceeded that of bats during the dry season, but the
pattern was reversed during the wet season.
We used an exclosure experiment to examine the top-down effects of birds and
bats on arthropods and leaf damage due to herbivory in restoration sites. Our study site
is a large-scale restoration project where trees were planted in two different
experimental treatments: islands, where several small patches of trees were planted
and plantations, where trees were planted in one large patch to create continuous
canopy cover. Thus we also investigated whether there were differences in the effects
of bird and bat predation between the smaller patches found in the island treatment and
the larger patch size of the plantation treatment.
We predicted that arthropod densities and leaf damage levels would be lowest on
tree branches exposed to both bird and bat predation and highest on branches where
both birds and bats were prevented from foraging. Although many species of foliagegleaning birds are commonly detected in the restoration sites (Lindell et al. unpubl.
data), we have little data on the use of the restoration sites by foliage-gleaning bats.
Additionally. previous work on the relative effects of birds and bats has been variable
(Williams-Guillén et al. 2008; Kalka et al. 2008). Thus we did not have any predictions
for which predator might have stronger effects on arthropods and leaf damage. Birds,
and potentially bats, are less likely to visit small than large patches (Zahawi and
Augspuger 2006; Fink et al. 2008: Schulze et al. 2000; Bernard and Fenton 2003). Thus
we predicted that differences in arthropods and leaf damage levels between branches
that excluded both birds and bats and branches with no exclosures would be greater in

80

larger patches, where birds and bats are more likely to be found, than branches in
smaller patches.

METHODS
Sites and Location: The study was conducted from June to July 2009 in four
restoration sites located near the town of Agua Buena (8° 44' N, 82° 58' W; elevation
1100-1300 m) in southern Costa Rica. One site was planted in 2004 and the other three
in 2005. All sites were at least 3 km apart. All restoration sites had previously been used
for ≥18 years for agriculture (Cole et al. 2010). In the past 50 years most land
surrounding Agua Buena was cleared for agricultural use, primarily for the production of
sun coffee with some land used for pasture and subsistence agriculture. Currently, less
than 27% of the original forest remains in the region, mostly in small riparian fragments
or on mountains at higher elevations. The landscape surrounding the restoration sites is
dominated by pasture, sparsely-shaded coffee and other agriculture, small human
settlements and some patches of secondary and mature Pacific premontane humid
forest (Daily et al. 2001; Sekercioglu et al. 2007).
The restoration sites were planted with four seedling species: Terminalia
amazonia (J.F. Gmel.) Exell [Combretaceae], Vochysia guatemalensis Donn. Sm.
[Vochysiaceae], Erythrina poeppigiana (Walp.) O.F. Cook [Fabaceae], and Inga edulis
Mart. [Fabaceae]. All species are native or naturalized to the region and are widely used
in tropical forest restoration (Nichols et al. 2001; Carpenter et al 2004).
We used two experimental treatments at each site. Treatment plots were 50 x 50
2

m and either had (1) trees planted in rows to cover the entire plot (plantation, 2500 m ),

81

or (2) trees planted in six patches of three different sizes (Figure 4.1). Vegetation was
cleared around planted trees every 2-3 months for the first two and a half years after
planting to allow seedlings to become established. The small number of seedlings that
died during this time were replanted. No further clearing or planting was conducted after
January 2007. Tree heights ranged from 3-9 m in 2008. Plantation plots had an almost
continuous canopy over the 50 x 50 m treatment area. The large, medium and small
patches within the island plots remained distinct patches within a matrix of herbaceous
vegetation.

82

Figure 4.1: Planting arrangement at each restoration site. Each site had one 50 x 50 m
extra-large patch (bottom) and one 50 x 50 treatment with six smaller patches (top): two
large (12x12 m), two medium (8x8 m) and two small (4x4 m). Each side of the second
treatment had one small, one medium and one large patch, although the arrangement
of the patches varied among sites. The number and arrangement of the tree species
within the patches was constant across all sites. Shaded areas are planted with trees.
Tree species are indicated by letters: Erythrina (E), Vochysia (V), Inga (I), and
Terminalia (T). Non shaded areas are non-woody vegetation, primarily grasses.
Sm=small patch.

83

84

Although the number of seedlings planted in each plot was initially the same,
seedling growth and mortality varied, resulting in variation in canopy cover among patch
sizes and tree species. Thus we measured island and plantation size by measuring
canopy cover for every tree in the medium and large islands and one-third of the trees
(selected at random) in each plantation. To measure individual tree canopy cover, we
used the formula for the area of an ellipse (πd1d2/4); where d1=the longest canopy
diameter and d2=the diameter perpendicular to d1. To calculate total canopy cover for
each island, we summed the canopy cover of all trees within each island. For
plantations, we multiplied the average canopy cover of each tree species by the number
of each species in that patch (Fink et al. 2009).
Exclosures: The experiment was conducted during the Costa Rican wet season,
from 31 May - 2 July 2009 (33 days) at two of the four sites and from 6 July - 7 August
2009 (33 days) on the other two sites. We conducted the exclosure experiments using
Inga edulis trees only. Inga trees had the greatest canopy cover in the restoration sites
and were the most commonly used trees by foraging birds in 2007 and 2008 (Fink et al.
2009; Morrison et al. unpubl data).
Exclosures were placed in a total of 30 Inga trees in the plantation (between 7-8
trees per plantation). Due to logistical constraints, we only used nine patches in the
island treatment, with a total of 16 Inga trees used for exclosures (between 1-4 Inga
trees per island). The total canopy cover of the four plantations treatments ranged from
2

1659.3 - 3139.4 m , and the total canopy cover of the nine different islands of the island
2

treatment ranged from 101.0 - 352.0 m .

85

The Inga from the plantation treatment were randomly selected from those
individuals that were within 5 m of the edge of the plantation to control for distance to
edge (Ingas in the islands were within 5 m of the edge of the island). On each tree we
selected four branches: one control branch (no exclosure, both birds and bats present),
one bat-only exclosure branch (exclosure at night only, birds present), one bird-only
exclosure branch (exclosure during the day only, bats present) and one bird+bat
exclosure branch (exclosure was in place during both day and night, neither birds nor
bats present). Branches were located between 1-2. 5m off the ground and were located
on the outer edge of the tree. Exclosures were constructed around an approximately
0.75 m-length of branch, using agricultural netting (mesh opening 2 x 2 cm; Van Bael et
al. 2003; Kalka et al. 2008; Williams-Guillén et al. 2008). Exclosures restricted access
by birds and bats but allowed access by arthropods, including spiders and large
Lepidoptera larvae (Morrison pers. obs.)
Each day during the experimental period, exclosures were placed on bat
excluded branches at dusk and removed at dawn, while exclosures were placed on bird
excluded branches at dawn and removed at dusk. Control branches had no exclosure
for the duration of the experiment. Exclosures on bird+bat branches were left in place
for the duration of the experiment.
Arthropod and Leaf damage Census: Arthropods were visually censused on all
experimental branches two days before the exclosures were placed on the branches
and then once weekly thereafter for four weeks (Kalka et al. 2008; Williams- Guillén et
al. 2008). All arthropods were categorized into 1 mm size intervals and identified to at
least the order level: Aranae, Orthoptera, Dermaptera, Hemiptera, Homoptera,

86

Coleoptera, Lepidoptera adults, Lepidoptera larvae, Diptera, Formicidae and other
Hymenoptera (bees and wasps), unknown adults, and unknown larvae. Length-weight
regressions from Johnson and Strong (2000) and Schoener (1980) were used to
calculate arthropod biomass for all Orders.
Only some arthropod orders are herbivorous (chewing insects). These include
Coleoptera, Orthoptera, Dermaptera and larvae from Lepidoptera or Coleoptera. Other
orders are sucking insects that feed on plant phloem, prey on other arthropods, or are in
feeding guilds other than herbivores. Thus we only used the chewing insects orders to
calculate arthropod biomass.
In order to standardize arthropod data for leaf area, we created a regression
equation by measuring the lengths, widths and areas for 200 leaves (50 leaves taken
from each of the four sites). We then measured the length and width of all leaves on
each exclosure branch and used the regression equation to estimate the total leaf area
of all leaves for each exclosure. By using the regression equation we did not have to
destructively sample leaves in the exclosures. Measures of arthropod abundance were
then expressed as insect mg/leaf m2 for each exclosure. Leaf length and width
measurements were taken immediately before the exclosures were set up.
Leaf damage due to herbivory was assessed at the end of the experiment and
was measured only on new leaves that flushed after the experiment began (Van Bael et
al 2003; Van Bael and Brawn 2005). We took digital photographs of these leaves and
used the software ImageJ v1.43 (National Institutes of Health, USA) to measure the
proportion of the leaf that was missing. Thus our measure of leaf damage included only
missing leaf tissue that was assumed to be from herbivorous insects, not other damage

87

(such as texture changes due to fungi or viruses). Estimates of leaf damage from this
study are likely conservative, because we did not take into account the effects of leaf
mining (brown spots due to sucking insects). Evidence of fungi or viruses on leaves was
rare.
Data Analysis: We measured the effects of plot type (island or plantation) and
treatment (both birds and bats present, birds only present, bats only present, neither
birds or bats present) on arthropod abundance (arthropod mass/leaf area) using
repeated measures mixed effects models. The four insect sampling periods were the
repeated within-subject factor, and treatment, plot and the treatment by plot interaction
were between-subject factors. Site was included as a random factor. Sampling period
(June or July) was initially included as a fixed factor, but was not significant, likely
because it was correlated with the effects of site. We initially included the patch size
(measured as m2 of canopy cover, see above) of each plantation and island patch in the
model, but this variable was not significant and was removed from subsequent analysis.
We used interaction contrasts to do pairwise comparisons of the effects of exclosure
treatment on arthropod abundance in the island and plantation locations.
We used linear mixed effects models to examine the fixed effects of plot type
(island or plantation), treatment, and the interaction between plot and treatment on
proportion of leaf damage. Site was included as a random factor. As above, sampling
period and patch size were not significant factors in the models and were removed from
subsequent analyses. We used interaction contrasts to do pairwise comparisons of the
effects of exclosure treatment on proportion leaf damage in the island and plantation
locations.

88

All analyses were conducted using SAS v9.1 (SAS Institute, Inc.). All variables
were examined for normality and homoscedasticity and proportion leaf damage and
arthropod biomass (arthropod mass /leaf area) were square root transformed.
Differences between groups were considered significant when P < 0.05.

RESULTS
Pre-exclosure arthropod biomass
Arthropod biomasses measured before the start of the exclosure experiment did
not vary significantly different among the branches used for experimental treatments
(F3,168 =0.76, P=0.52) nor did they differ between islands and plantations (F1,169 = 3.09,
P = 0.08). The plot by treatment interaction was not significant (F3,168 = 0.44 , P = 0.72).
Arthropod biomass
Arthropod biomass was significantly affected by exclosure treatment (F3,716
=23.2, P <0.001), but not different between islands and plantations (F1,717 =0.11, P
=0.74; Figure 4.2). The interaction between exclosure and planting treatments was not
significant (F3,716 = 1.33, P = 0.26).
In both the island and plantation treatments, branches where both birds and bats
were excluded (no predators present) had significantly higher arthropod biomass than
branches where both predators were present (no exclosure), branches where only birds
were present (bat-excluded) and branches where only bats were present (birdexcluded) (all P≤0.02; Figure 4.2). Arthropod biomasses on bat-only and bird-only
branches did not differ significantly from one another in either islands (P=0.89) or
plantations (P=0.13; Figure 4.2). In the islands, both bat-only and bird-only branches
89

had significantly higher arthropod biomass than braches where both were present (all
P≤0.01; Figure 4.2). However in the plantations, only bird-only branches had
significantly greater biomass than the branches where both were present (P<0.01;

Arthropod biomass (mg) per leaf area (m2)

Figure 4.2).

0.030

0.025

A
W

0.020

C
YZ

Y

C

0.015

XZ
0.010

B
0.005

0.000

I

P

No predators

I

P

I

P

Birds+bats present Bats present

I

P

Birds present

Exclosure Treatment
Figure 4.2: Data from islands and plantations showing average arthropod biomass (mg
of arthropods/leaf area m2) per census in each of the four exclosure treatments. None
present = exclosure 24-hrs, excluded both birds and bats, Birds+bats present = no
exclosure, neither birds nor bats excluded, Bats present = birds only excluded, bats
present, Birds present = bats only excluded, birds present. Different letters represent
significant pairwise (P<0.05) differences based on post-hoc tests from the repeated
measures mixed effect model. There were no significant pairwise differences in leaf
damage between islands and plantations within any of the four exclosure treatments.
The letters A-D are for the island treatment only, and the letters W-Z are for the
plantations treatment. Standard error bars are shown.

90

Arthropod biomass on branches where both birds and bats were present
remained relatively consistent across all sampling points for both islands and
plantations, while arthropod biomass increased on branches where both birds and bats
were excluded after the first few sampling periods (Figures 4.3a-b). Arthropod biomass
for bird-only and bat-only branches also increased, but at an intermediate level
compared to the other treatments (Figures 4.3a-b).

91

Figure 4.3. Variation in arthropod biomass over the five sampling periods in the islands
(A) and plantations (B). Pre-exclosure measurements were taken two days before the
experiment began. Each of the other samples were taken once weekly thereafter for
four weeks. None present = exclosure 24-hrs, excluded both birds and bats, Birds+bats
present = no exclosure, neither birds nor bats excluded, Bats present = birds only
excluded, bats present, Birds present = bats only excluded, birds present. Standard
error bars are shown.

92

Arthropod mass (mg) per leaf area (m2)

0.035

None present
Birds+bats present
Bats present

0.030

(A)

Birds present

0.025
0.020
0.015
0.010
0.005
0.000

Pre-exclosure

Week 1

Week 2

Week 3

Week 4

Sampling period
0.035
(B)

Arthropod mass (mg) per leaf area (m2)

None present

0.030

Birds+bats present
Bats present

0.025

Birds present

0.020
0.015
0.010
0.005
0.000
Pre-exclosure

Week 1

Week 2
Sampling period

93

Week 3

Week 4

Leaf damage due to herbivory
Both exclosure treatment (F3,173 =3.28, P=0.02) and planting treatment
(F1,174=7.03, P=0.001) had significant effects on leaf damage. The interaction between
exclosure and planting treatment was not significant (F3,173=0.46, P=0.71).
Proportion of leaf damage across all exclosure treatments was higher in the
plantations (0.158 ± 0.009) than in the islands (0.125 ± 0.013; P=0.01). However, within
each exclosure treatment, there were no significant pairwise differences in leaf damage
between islands and plantations (all P≥0.31; Figure 4.4). In the islands, leaf damage did
not differ between the four treatments (all P ≥ 0.09; Figure 4.4). In the plantations,
however, leaf damage due to herbivory was significantly higher on the branches where
both birds and bats were excluded compared to each of the other three treatments (P ≤
0.02; Figure 4.4).

94

Proportion leaf damage due to herbivory

0.25

A

0.20

B

B

B
0.15

0.10

0.05

0.00

I

P

None Present

I

P

Birds+bats present

I

P

Bats Present

I

P

Birds Present

Exclosure Treatment

Figure 4.4: Proportion damage due to herbivory for leaves in the four exclosure
treatments within the two planting treatments (Island and Plantation). None present =
exclosure 24-hrs, excluded both birds and bats, Birds+bats present = no exclosure,
neither birds nor bats excluded, Bats present = birds only excluded, bats present, Birds
present = bats only excluded, birds present. Different letters represent significant
pairwise (P<0.05) differences based on post-hoc tests from the linear mixed effect
model. There were no significant pairwise differences in leaf damage between islands
and plantations within any of the four exclosure treatments. Standard error bars are
shown.
DISCUSSION
Foraging by both birds and bats had a significant impact on arthropod
populations and, in plantations these effects translated to lower levels of leaf damage.
In island and plantation patches combined, arthropod biomass on branches where both
predators were present was 70% lower than that observed on branches where both
predators were excluded. Leaf damage due to herbivory was 32% lower on branches

95

where birds and bats were present compared to branches where they were both
excluded.
When either birds or bats were present, arthropod biomass was lower than when
both were excluded and higher than when both were allowed access to branches.
However, our data show that the effects of birds and bats on the populations of
herbivorous insects are approximately equivalent in this system. This result differs from
the few other studies that have examined the relative differences between top-down
effects of bird and bat in tropical systems. Both bird and bat predation reduced
arthropod abundances in a Mexican shade coffee system, but the effect of bird
predation exceeded that of bats in the dry season, while the effect of bats exceeded that
of birds in the wet season (Williams-Guillén et al. 2008). In the tropical forest understory
of Barro Colorado Island, Panama, birds and bats reduced arthropods, however the
effect of bat predation exceeded that of birds (Kalka et al. 2008).
Restoration planting treatment, island (small patches) or plantation (large
patches), had no significant effect on the impact of birds and bats on arthropod
biomass, but did affect levels of leaf damage due to herbivory. In islands, no significant
differences in leaf damage between exclosure treatments were detected, but in
plantations, leaf damage levels were higher on branches where birds and bats were
excluded than on branches where both had access. However, neither predator by itself
was able to reduce leaf damage compared to control branches where both were
excluded. Birds and bats are clearly removing some of the population of herbivorous
arthropods, but the indirect effect of their consumption did not translate into strong
reduction of herbivore damage to leaves. Williams-Guillén et al. (2008) also found no

96

effects of birds or bats on leaf damage levels in coffee plantations, despite clear effects
of each predator on arthropod populations.
The few studies that have examined the top-down effects of bird and bat
insectivory show that the relative effects of birds and bats vary across habitat type and
season. There are several potential mechanisms that could explain these differences.
First, food chain theory suggests that trophic cascades will be strongest when and
where plant productivity is highest (Fretwell 1977; Oksanen et al. 1981; Van Bael and
Brawn 2005). Results from a tropical forest in Panama show that the effects of bird
predation on arthropods and leaf damage were strongest at locations where, and during
the times of year when, the highest level of leaf production occurred: during the wet
season, compared to the dry season, and in forest canopies, compared to the
understories (Van Bael and Brawn 2008).
Kalka et al. (2008) found that bats, and to a lesser extent birds, had strong
effects on reducing arthropods and leaf damage during the wet season, when plant
productivity is predicted to be highest. In contrast to trophic theory, however, these
strong trophic cascades were found on understory plants, a location predicted to have
low plant productivity. Birds and bats were also found to reduce arthropods, but not
herbivory, on understory coffee plants shaded by a taller overstory canopy (WilliamsGuillén et al. 2008). However, birds and bats had differing effects in the coffee
agroecosystem that were dependent on season (Williams-Guillen et al. 2008). The
present study was also conducted during the wet season, when arthropod herbivore
populations peak in response to new leaf production (Barone 1998; Wolda 1978), and
theory suggests that trophic cascades should be strongest. Although the relative impact

97

of bird and bat insectivory in the restoration sites resulted in detectable differences in
arthropod biomass among exclosure treatments, this did not translate into strong effects
on leaf damage. Plant productivity was not measured in any of these three studies, so it
is unknown how actual leaf production rates affected these results.
Other seasonally dependent events could also affect the strength of these trophic
cascades. In the neotropics, the wet season also corresponds to the nesting season for
many insectivorous birds, which have an increased demand for arthropods to feed their
young (Martin 1987). On the other hand, during the dry season, the neotropics
experience an influx of millions of overwintering migratory birds that could also affect
arthropod populations (Greenberg 1995; Jedlicka et al. 2006).
These recent studies on the relative effects of birds and bats on controlling
arthropod populations and leaf damage suggest that the relationships among trophic
levels in these ecosystems is complex and dependent on a number of different factors
including density and types of bird, bat and plant species present, the structure of the
habitat and variation in abiotic factors due to season. More studies are needed to
examine the exact nature of these trophic interactions. Additionally, information on the
habitat use, densities, foraging behavior and diet of tropical insectivorous birds is
understudied, and this information is extremely limited for bats(Kalka and Kalko 2006).
We likely underestimated the effects of birds' and bats' insectivory on leaf
damage. Leaf damage due to arthropods potentially extends beyond the phytophagous
guilds we studied here, to insect that consume plant phloem. Phloem-feeding insects
can also damage plants, leading to reduced photosynthetic ability, growth and survival
(Goggin 2007). Birds and bats may also have reduced the impact of phloem-feeders in

98

this study, however damage due to these insects is much more difficult to measure.
Also, recent work has shown that measuring holes in leaves, as in this study, may
significantly underestimate the actual damage caused by chewing arthropods (Zangerl
et al 2002). The leaf areas directly adjacent to holes left by herbivorous arthropods
suffers reduced photosynthetic ability and can cover an area larger than the area
directly damaged by leaf consumption. Thus it is likely that birds and bats are having
greater direct and indirect effects on arthropods and plants than what we could measure
in this study.
Previous research has established that some groups of birds and bats have
important ecological roles as seed-dispersers in restoration systems (Bianconi et al.
2007; Cole et al. 2010), but their roles as insect predators was previously unknown. Our
data show that these two top predators in tropical restoration systems can impact
arthropod populations. Our data also suggest that resulting indirect effects of bird and
bat insectivory on leaf damage, an important influence on plant growth and survival,
were strongly dependent on the presence of both predators. Studies that have
examined trophic cascades have suggested that the presence of predators can have
important effects on forest productivity, and this effect could also be important to
restoration success (Marquis and Whelan 1994; Terborgh et al. 2001). Thus restoration
efforts should emphasize the importance of attracting top predators communities, rather
than one or a few species, in order to establish crucial trophic processes important for
ecosystem function (Terborgh et al. 2001).

99

LITERATURE CITED

100

LITERATURE CITED
Alcock J. 2005. Animal Behavior: An Evolutionary Approach, 8th edition. Sunderland,
MA: Sinauer Associates, Inc.
Adams JL, Camelio KW, Orique MJ, Blumstein DT. 2006. Does information of
predators influence general wariness? Behavioral Ecology and Sociobiology 60:
742–747.
Bach CE. 1988. Effects of host plant patch size on herbivore density: Patterns. Ecology
69: 1090-1102.
Barnard CJ. 1980. Flock feeding and time budgets in the house sparrow (Passer
domesticus L.), Animal Behaviour 28: 295–309.
Barone JA. 1998. Host-specificity of folivorous insects in a moist tropical forest. Journal
of Animal Ecology: 67: 400–409.
Basset Y, Aberlenc HP, Barrios H, Curletti G, Bérenger JM, Vesco JP, Causse P, Haug
A, Hennion AS, Lesobre L, Marquès F, O'Meara R. 2001. Stratification and diel
activity of arthropods in lowland rainforest in Gabon. Biological Journal of the
Linnean Society 72: 585-607.
Bednekoff PA, Lima SL. 2002. Why are scanning patterns so variable? An overlooked
question in the study of anti-predator vigilance. Journal of Avian Ecology 33:143149.
Bernard E and Fenton MB. 2003. Bat mobility in a fragmented landscape in Central
Amazonia, Brazil. Biotropica 35:262-277.
Bianconi GV, Mikich SB, Teixeira SD, Maia BHLNS. 2007. Attraction of Fruit-Eating
Bats with Essential Oils of Fruits: A Potential Tool for Forest Restoration.
Biotropica 39: 136–140.
Blumstein DT, Daniel JC, Griffin AS, Evans CS. 2000. Insular tammar wallabies
(Macropus eugenii) respond to visual but not acoustic cues from predators.
Behavioral Ecology 115:28–535.
Boege K and Marquis RJ. 2006. Plant quality and predation risk mediated by plant
ontogeny: consequences for herbivores and plants. Oikos 115:559-572.
Brown JS. 1988. Patch use as an indicator of habitat preference, predation risk, and
competition. Behavioral Ecology and Sociobiology 22:37-47.
Brown S. 1993. Tropical forests and the global carbon cycle: the need for sustainable

101

land-use patterns. Agriculture, Ecosystems and Environment 46:31–44.
Burke DM, Nol E. 1998. Influence of food abundance, nest-site habitat, and forest
fragmentation of breeding ovenbirds. Auk 115: 96-104.
Burkett VR, Draugelis-Dale RO, Williams HM, Schoenholtz SH. 2005. Effects of flooding
regime and seedling treatment on early survival and growth of Nuttall Oak.
Restoration Ecology 13: 471-479.
Butler SJ, Whittingham MJ, Quinn JL, Cresswell W. 2005. Quantifying the
interaction between food density and habitat structure in determining patch
selection. Animal Behaviour 69:337–343.
Caraco T, Martindale S, Pulliam HR. 1980. Avian flocking in the presence of a predator.
Nature 285: 400-401.
Carnevale NJ, Montagnini F. 2002. Facilitating regeneration of secondary forests
with the use of mixed and pure plantations of indigenous tree species. Forest
Ecology and Management 163:217-227.
Caro TM. 1995. Pursuit-deterrence revisited. Trends in Ecology and Evolution 10: 500503.
Carpenter FL, Nichols D, Sandi E. 2004. Early growth of native and exotic trees planted
on degraded tropical pasture. Forest Ecology and Management 196:367-378.
Carrascal LM, Alonso CL. 2006. Habitat use under latent predation risk. A case study
with wintering forest birds. Oikos 112: 51-62
Chazdon RL. 2003. Tropical forest recovery: legacies of human impact and natural
disturbances. Perspectives in Plant Ecology, Evolution and Systematics 6: 51–
71.
Chazdon RL. 2008. Beyond deforestation: Restoring forests and ecosystem services
on degraded lands. Science 320:1458-1460.
Cimprich DA, Woodrey MS, Moore FR. 2005. Passerine migrants respond to variation in
predation risk during stopover. Animal Behaviour 69:1173–1179.
Clark DB. and Clark DA. 1985. Seedling dynamics of a tropical tree: impacts of
herbivory and meristem damage. Ecology 66:1884-1892.
Cole RJ, Holl KD, Zahawi RA 2010. Seed rain under tree islands planted to restore
degraded lands in a tropical agricultural landscape. Ecological Applications 20:
1255-1269.

102

Coley PD and Barrone JA.1996. Herbivory and plant defenses in tropical forests.
Annual Review of Ecology and Systematics 27:305–335
Cooper, Jr. WE, Perez-Mellado V, Hawlena D. 2006. Magnitude of food reward affects
escape behavior and acceptable risk in Balearic lizards, Podarcis lilfordi.
Behavioral Ecology 17: 554-559.
Cooper, Jr WE. 2000. Tradeoffs between predation risk and feeding in a lizard, the
broad-headed skink (Eumeces laticeps). Behaviour 137:1175–1189.
Cresswell W, Quinn JL, Whittingham MJ, Butler S. 2003. Good foragers can also be
good at detecting predators. Proceedings of the Royal Society of London, Series
B 270:1069-1076
Cusak D, Montagnini F. 2004. The role of native species plantations in recovery of
understory woody diversity in degraded pasturelands of Costa Rica. Forest
Ecology and Management 188:1–15.
Daily GC, Ehrlich PR, Sánchez-Azofeifa GA. 2001. Countryside biogeography: use of
human-dominated habitats by the avifauna of southern Costa Rica. Ecological
Applications 11:1-13.
Dawkins MS, Woodington A. 2000. Pattern recognition and active vision in chickens.
Nature. 403, 652-655.
Dittrich WH, Lea SEG. 2001. Motion discrimination and recognition. In: Avian Visual
Cognition (Ed. by R. G. Cook), (online). Available:
www.pigeon.psy.tufts.edu/avc/dittrich.
Eichhorn MP, Nilus R, Compton SG, Hartley SE, Burslem DF. 2010. Herbivory of
tropical rain forest tree seedlings correlates with future mortality. Ecology
91:1092-101.
Ekman J. 1986.Tree use and predator vulnerability of wintering passerines. Ornis
Scandinavia 17: 261-267.
Elgar MA. 1989. Predator vigilance and group size in mammals and birds: a critical
review of the empirical evidence. Biological Reviews. 64: 13–33.
Eränen JK, Kozlov MV. 2006. Physical sheltering and liming improve survival and
performance of Mountain Birch seedlings: a 5-year study in a heavily polluted
industrial barren. Restoration Ecology 14: 77-86.
Feldman TS. 2006. Pollinator aggregative and functional responses to flower density:
does pollinator response to patches of plants accelerate at low densities? Oikos
115:128–140.

103

Fernández-Juricic E, Smith R, Kacelnik A. 2005. Increasing the costs of conspecific
scanning in socially foraging starlings affects vigilance and foraging behaviour.
Animal Behaviour 69: 73–81.
Fink RD, Lindell CA, Morrison EB, Zahawi RA, Holl KD. 2009. Patch size and
canopy cover of planted tree patches influence birds’ duration and rate of visit.
Restoration Ecoloy 17: 479-486.
Firn J, Erskine PD, Lamb D. 2007. Woody species diversity influences productivity and
soil nutrient availability in tropical plantations. Oecologia 154:521-533.
Florentine SK, Westbrooke ME. 2004. Restoration on abandoned tropical
pasturelands—do we know enough? Journal for Nature Conservation 12:85-94.
FAO, Global Forest Resources Assessment 2005: Progress Towards Sustainable
Forest Management (FAO, Rome, 2005).
Franklin WE, Lima SL. 2001. Laterality in avian vigilance: do sparrows have a favourite
eye? Animal Behaviour 62: 879-885.
Fretwell SD .1977. The regulation of plant communities by the food chains exploiting
them. Perspec Biol Med 20:169–185.
Fritz H, Guilleain M, Durant D. 2002. The cost of vigilance for intake rate in the mallard
(Anas platyrhynchos): an approach through foraging experiments. Ethology
Ecology and Evolution 14: 91-97.
Gaston KJ, Blackburn TM, Kleiny Goldewijk K. 2003. Habitat conversion and global
avian biodiversity loss. Proceedings of the Royal Society Series B 270: 1293–
1300.
Glück E. 1987. An experimental study of feeding, vigilance and predator avoidance in a
single bird. Oecologia. 71:268-272.
Goggin FD. 2007. Plant-aphid interactions: molecular and ecological perspectives.
Current Opinions in Plant Biology 10: 399–408.
Gotelli NJ, Ellison AM. 2004. A primer of ecological statistics. Sinauer Associates, Inc.
Sunderland, MA.
Greenberg, R. 1995. Insectivorous migratory birds in tropical ecosystems: the breeding
currency hypothesis. Journal Avian Biology 26:260–264.
Greenberg R, Bichier P, Agnon AC, MacVean C, Perez P, Cano E. 2000. The impact of
avian insectivory on arthropods and leaf damage in some Guatemalan coffee

104

plantations. Ecology 81:1750-1755.
Günter S, Weber M, Erreis R, Aguirre N. 2007. Influence of distance to forest edges on
natural regeneration of abandoned pastures: a case study in the tropical
mountain rain forest of Southern Ecuador. European Journal of Forest Research
126:67–75.
Hasson O. 1991. Pursuit-deterrent signals: communication between prey and predator.
Trends in Ecology and Evolution 6: 325-329.
Hauser MD, Caffrey C.1994. Anti-predator response to raptor calls in wild crows,
Corvus brachyrhynchos hesperis. Animal Behaviour 48:1469–1471.
Holl KD. 1999. Factors liming tropical rain forest regeneration in abandoned pastures:
seed rain, seed germination, microclimate and soil. Biotropica 31:229-242.
Holl KD, Loik ME, Lin EHV, Samuels IA. 2000. Tropical montane forest restoration in
Costa Rica: Overcoming barriers to dispersal and establishment. Restoration
Ecology 8:339-349.
Holmes RT, Robinson SK. 1981. Tree species preferences of foraging insectivorous
birds in a Northern hardwoods forest. Oecologia 48:31-35.
Holmes RT, Bonney, Jr., RE, Pacala SW. 1979. Guild structure of the Hubbard Brook
bird community: A multivariate approach. Ecology. 60: 512-520.
Holmes RT, Schultz JC. 1988. Food availability for forest birds - effects of prey
distribution and abundance on bird foraging. Canadian Journal of Zoology
66:720-728.
Hurlbert SH. 1984. Pseudoreplication and the design of ecological field experiments.
Ecological Monographs 54:87–211.
Hutto RL. 1990. Measuring the availability of food resources. Studies in Avian Biology.
13: 20–28.
Jansen A. 2005. Avian use of restoration plantings along a creek linking rainforest
patches on the Atherton Tablelands, North Queensland. Restoration Ecology 13:
275-283.
Jedlicka JA, Greenberg R, Perfecto I, Philpott SM, Dietsch TV. 2006. Seasonal shift in
the foraging niche of a tropical avian resident: resource competition at work?
Journal of Tropical Ecology 22:385-395.
Johnson MD. 2000. Evaluation of an arthropod sampling technique for measuring food
availability for forest insectivorous birds. Journal of Field Ornithology 71:88-109.

105

Johnson MD, Sherry TW, Strong AM, Medori A. 2005. Migrants in Neotropical bird
communities: an assessment of the breeding currency hypothesis. Journal of
Animal Ecology 74: 333–341.
Johnson MD, Strong AM. 2000. Length-weight relationships of Jamaican arthropods.
Entomological News 111:270-281.
Johnson MD. 2007. Measuring habitat quality: a review. The Condor 109: 489-504.
Jones KA, Krebs JR, Whittingham MJ. 2007. Vigilance in the third dimension: head
movement not scan duration varies in response to different predator models
Animal Behaviour 74:1181–1187.
Kalka MB, Smith AR, Kalko EKV. 2008. Bats limit arthropods and herbivory in a tropical
forest. Science 320:71.
Kalka M, Kalko EKV. 2006. Gleaning bats as underestimated predators of herbivorous
insects: diet of Micronycteris microtis (Phyllostomidae) in Panama. Journal of
Tropical Ecology 22:1–10.
Kilgo JC. 2005. Harvest-related edge effects on prey availability and foraging of Hooded
Warblers in a bottomland hardwood forest. Condor 107:627-636.
Kilpatrick AM. 2003. The impact of thermoregulatory costs on foraging behaviour: a test
with American Crows (Corvus brachyrhynchos) and eastern grey squirrels
(Sciurus carolinensis). Evolutionary Ecology Research 5: 781–786.
Koh LP. 2008. Birds defend oil palms from herbivorous insects. Ecological Applications:
Vol. 18:821-825.
Kohyama T, Hotta M. 1990. Significance of allometry in tropical saplings. Functional
Ecology 4: 515-521.
Kotler BP, Brown JS, Bouskila A. 2004. Apprehension and time allocation in gerbils:
The effects of predatory risk and energetic state. Ecology: 85: 917-922.
Kotler BP, Brown JS, Hasson O. 1991. Factors affecting gerbil foraging behavior and
rates of owl predation. Ecology 72: 2249-2260.
Krueger TR, Williams DA, Searcy WA. 2008 The genetic mating system of a tropical
tanager. The Condor 110:559-562.
Lamb D, Erskine PD, Parrotta JA. 2005. Restoration of degraded tropical forest
landscapes. Science 310: 1628-1632.

106

LaVal RK. and LaVal ML. 1980. Ecological studies and management of Missouri bats,
with emphasis on cave-dwelling species. Terrestrial Series, Missouri Dept.
Conserv., Jefferson City 8:1-53.
Lazarus J. 1979. The early warning function of flocking in birds: an experimental study
with captive quelea. Animal Behaviour 27:855-865.
Lewis TL, Esler D, Boyd WS. 2008. Foraging behavior of Surf Scoters (Melanitta
perspicillata) and White-winged Scoters (M. fusca) in relation to clam density:
inferring food availability and habitat quality. Auk 125:149–157.
Lima SL. 1990. Protective cover and the use of space: different strategies in finches.
Oikos. 58:151–158.
Lima SL, Bednekoff BA. 1999. Temporal variation in danger drives antipredator
behavior: the predation risk allocation hypothesis. Am Nat. 153:649–659.
Lima SL, Dill LM. 1990. Behavioral decisions made under the risk of predation: a review
and prospectus. Can J Zoology. 68:619–640.
Lindell CA. 2008. The value of animal behavior in evaluations of restoration success.
Restoration Ecology 16: 197-203.
Lyons JE. 2005. Habitat specific foraging of Prothonotary warblers: Deducing habitat
quality. Condor 107:41-49.
Martin TE. 1987. Food as a limit on breeding birds: a life history perspective. Annual
Review of Ecology and Systematics 18:453-87.
Martin P, Bateson P. 1986. Measuring behaviour: an introductory guide. Cambridge,
UK: Cambridge University Press.
Marquis RJ and Whelan CJ.. 1994. Insectivorous birds increase growth of white oak
through consumption of leaf-chewing insects. Ecology 75:2007-2014.
Maurer BA, Whitmore RC. 1981. Foraging of five bird species in two forests with
different vegetation structure. Wilson Bulletin 93:478–490.
Mols C. and Visser M. 2002. Great tits can reduce caterpillar damage in apple orchards.
Journal of Applied Ecology 39: 888-899.
Montgomery RA, Chazdon RL. 2001. Forest structure, canopy architecture, and light
transmittance in tropical wet forests. Ecology 82: 2797-2718.
Munro NT, Fischer J, Wood J, Lindenmayer DB. 2009. Revegetation in agricultural
areas: the development of structural complexity and floristic diversity. Ecological

107

Applications 19: 1197- 1210.
Murakami M.and Nakano S. 2000. Species-specific bird functions in a forest-canopy
food web. Proceedings of the Royal Society of London, Series B 267:1597-1601.
Nichols JD, Rosemeyer ME, Carpenter FL and Kettler J. 2001. Intercropping legume
trees with native timber trees rapidly restores cover to eroded tropical pasture
without fertilization. Forest Ecology and Management 152:195-209.
Norghauer JM., Malcolm JR, and Zimmerman BL. 2006. Juvenile mortality and attacks
by a specialist herbivore increase with conspecific adult basal area of Amazonian
Swietenia macrophylla. Journal of Tropical Ecology 22:451-460.
Nystrand M. 2007. Associating with kin affects the trade-off between energy intake and
exposure to predators in a social bird species. Animal Behaviour 74:497–506.
Oksanen L, Fretwell S, Arruda J, Niemela P. 1981. Exploitation ecosystems in gradients
of primary productivity. American Naturalist 118:240–261.
Olsson O, Brown JS, Smith HG. 2002. Long- and short-term state-dependent foraging
under predation risk: an indication of habitat quality. Animal Behaviour 63: 981989.
Orrock JL, Danielson BJ, Brinkerhoff RJ. 2004. Rodent foraging is affected by indirect,
but not by direct, cues of predation risk. Behavioral Ecology 15: 433–437.
Panasci TA, Whitacre DF. 2002. Roadside Hawk breeding ecology in forest and farming
landscapes Wilson Bulletin 114: 114-121.
Parrotta JA. 1992. The role of plantation forests in rehabilitating degraded tropical
ecosystems. Agriculture, Ecosystems and Environment 41:115-133.
Peek JM. 1970. Relation of canopy area and volume to production of three woody
species. Ecology 51: 1098-1101.
Pejchar L, Holl KD, Lockwood JL. 2005. Hawaiian honeycreeper home range
size varies with habitat: implications for native Acacia koa forestry. Ecological
Applications 15:1053-1061.
Pelletier F. 2005 Foraging time of rutting bighorn rams varies with individual behavior,
not mating tactic. Behavioral Ecology, 16, 280-285.
Perfecto I, Vandermeer J, Hanson P, Cartnin V. 1997. Arthropod biodiversity loss
and the transformation of a tropical agro-ecosystem. Biodiversity and
Conservation 6:935-945.

108

Persson A, and Stenberg M. 2006. Linking patch-use behavior, resource density, and
growth expectations in fish. Ecology 87: 1953–1959.
Piotto D, Montagnini F, Ugalde L, Kanninen M. 2003. Growth and effects of
thinning of mixed and pure plantations with native trees in humid tropical Costa
Rica. Forest Ecology and Management 177:427-439.
Pomara LY, Cooper RJ, Petit LJ. 2003. Mixed-species flocking and foraging
behavior of four Neotropical warblers in Panamanian shade coffee fields
and forests. Auk 120:1000-1012.
Powers JS, Haggar PJ, Fisher RF. 1997. The effect of overstory composition
on understory woody regeneration and species richness in 7-year old plantations
in Costa Rica. Forest Ecology and Management 99:43-54.
Poulin B, Lefebvre G. 1996. Dietary relationships of migrant and resident birds
from a humid forest in central Panama. Auk 113: 277-287.
Poulin B, Lefebvre G. 1997. Estimation of arthropods available to birds: effect of
trapping technique, prey distribution, and bird diet. Journal of Field Ornithology
68: 426–442.
Poulin B, Lefebvre G, McNeil R. 1994. Diets of land birds from Northeastern
Venezuela. The Condor 96:354-367.
Ramakrishnan U, Coss RG. 2000. Early environment shapes the development of gaze
aversion by wild bonnet macaques (Macaca radiata). Ethology. 43: 217-222.
Remsen JV, Jr. Robinson SK. 1990. A classifications scheme for foraging behavior of
birds in terrestrial habitats. Studies in Avian Biology 13: 144-160.
Repasky RR 1996. Using vigilance behavior to test whether predation promotes habitat
partitioning. Ecology 77: 1880-1887.
Rey Benayas JM, Bullock JM, Newton AC. 2008. Creating woodland islets to
reconcile ecological restoration, conservation, and agricultural land use. Frontiers
in Ecology and the Environment 6:329-336.
Richards LA, Windsor DM. 2007. Seasonal variation of arthropod abundance in
gaps and the understorey of a low moist forest in Panama. Journal of Tropical
Ecology 23: 169-176.
Robinson GR, Handel SN. 2000 Directing spatial patterns of recruitment during an
experimental urban woodland reclamation. Ecological Application 10: 174-188.
Robinson SK, Holmes RT. 1982. Foraging behavior of forest birds: the

109

relationship among search tactics, diet, and habitat structure. Ecology 63:19181931.
Ross, Jr. DL. 1998. Costa Rican Bird Song Sampler. Ithaca, NY: Cornell Lab of
Ornithology.
Ruiz-Jaen MC, Aide TM. 2005. How is restoration success being measured?
Restoration Ecology 13: 569-577.
SAS Institute, Inc. SAS v9.1. Cary, NC.
Schmidt KA. 2006. Non-additivity among multiple cues of predation risk: a behaviorallydriven trophic cascade between owls and songbirds. Oikos. 113: 82-90.
Schoener TW. 1980. Length-weight regressions in tropical and temperate forestunderstory insects. Annals of the Entomological Society of America 73:106-109.
Schulze MD, Seavy NE, and Whitacre DF. 2000. A comparison of the phyllostomid bat
assemblages in undisturbed Neotropical forest and in forest fragments of a slashand-burn farming mosaic in Petén, Guatemala. Biotropica. 32:174–184.
Sekercioglu CH., Ehrlich PR, Daily GC, Aygen D, Goehring D, Sandí RF. 2002.
Disappearance of insectivorous birds from tropical forest fragments. Proceedings
of the National Academy of Sciences 99:263–267.
Sekercioglu CH. 2006. Increasing awareness of avian ecological function. Trends in
Ecology and Evolution 21: 464-471.
Sekercioglu CH, Loarie SR, Oviedo-Brenes F, Daily GC, Ehrlich PR. 2007. Persistence
of forest birds in the Costa Rican agricultural countryside. Conservation Biology.
21, 482-494.
Selwood K, MacNally R, Thomson JR. 2009. Native bird breeding in a chronosequence
of revegetated sites. Oecologia 159: 435-446.
Sih A. 1992. Prey uncertainty and the balancing of antipredator and feeding needs. Am
Nat. 139:1052-1069.
Smith LA, Burke DM, Nol E, Elliott KA. 2006. The effects of partial cutting on the Rosebreasted Grosbeak: abundance, food availability and nest survival. Canadian
Journal of Forest Research 36: 1087-1098.
Smith RJ, Moore FR, May C. 2007. Stopover habitat along the shoreline of
Lake Huron, Michigan: emergent aquatic insects as a food resource for spring
migrating landbirds. Auk 124:107–121.

110

Sodhi NS, Posa MRC, Lee TM, Bickford D, Koh LP, Brook BW. 2009. The state
and conservation of Southeast Asian biodiversity. Journal of Biodiversity and
Conservation. DOI 10.1007/s10531-009-9607-5.
Stiles GF, Skutch AF. 1989. A guide to the birds of Costa Rica. Cornell University
Press, Ithaca, NY.
Strong AM, Sherry TW, Holmes RT. 2000. Bird predation on herbivorous insects:
indirect effects of sugar maple saplings. Oecologia 125:370-379.
Suhonen J. 1993. Predation risk influences the use of foraging sites by tits. Ecology.
74:1197-1203.
Sullivan KA. 1984. Information exploitation by Downy Woodpeckers in mixed species
flocks. Behaviour 91:294-311.
Terborgh JL, Lopez L, Nuñex PV,Rao M, Shahabuddin G, Orihuela G, Riveroz M,
Ascanio R, Adler GH, Lambert T, Balbas L. 2001. Ecological meltdown in
predator-free forest fragments. Science 294:1923–1926.
Thorson JM, Morgan RA, Brown JS, Norman JE. 1998. Direct and indirect cues of
predatory risk and patch use by fox squirrels and thirteen-lined ground squirrels.
Behav Ecol. 9:151-157.
Thorstrom R. 2000. The food habits of sympatric forest falcons during the breeding
season in Northeastern guatemala. J Raptor Research. 34:196-202.
Van Beal SA, Brawn JD, Robinson SK. 2003. Birds defend trees from
herbivores in a Neotropical forest canopy. Proceedings of the National Academy
of Sciences 100: 8304-8307.
Van Bael SA,Brawn JD. 2005. The direct and indirect effects of insectivory by birds in
two contrasting Neotropical forests. Oecologia 143:106-116.
Van Bael SA, Bichier P, Greenberg R. 2007. Bird predation on insects reduces foliage
damage to cocoa trees (Theobroma cacao) in western Panama. Journal of
Tropical Ecology 23:715-719.
Van Bael SA., Philpott S, Greenberg R, Bichier P, Barber N, Mooney K, Gruner D.
2008 Birds as predators in tropical agroforestry systems: a meta-analysis.
Ecology 89: 928-934.
Vesk PA, Nolan R, Thomas JR, Dorrogh JW, MacNally R. 2008. Time lags in
provision of habitat resources through revegetation. Biological Conservation
141:174-186.

111

Vitz AC, Rodewald AD. 2005 Can regenerating clearcuts benefit mature-forest
songbirds? An examination of post-breeding ecology. Biological Conservation
127: 477-486.
Wagner JH. 1981. Visibility and bias in avian foraging behavior. Condor 83: 263-264.
Watson M, Aebischer NJ, Cresswell W. 2007. Vigilance and fitness in grey partridges
Perdix perdix: the effects of group size and foraging-vigilance trade-offs on
predation mortality. Journal of Animal Ecology 76: 211–221.
Whittingham MJ, Butler S, Quinn JL, Cresswell W. 2004. The effect of limited visibility
on vigilance behaviour and speed of predator detection: implications for the
conservation of granivorous passerines. Oikos 106: 377-385.
Wijdeven SMJ, Kuzee ME. 2000. Seed availability as a limiting factor in forest
recovery processes in Costa Rica. Restoration Ecology 8:414–424.
Williams-Guillén K, Perfecto I, Vandermeer J. 2008. Bats limit insects in a Neotropical
agroforestry system. Science 320: 70.
Wolda, H.1978. Seasonal fluctuations in rainfall, food and abundance of tropical insects.
Journal of Animal Ecology 47:369–381.
Wolff JO, Van Horn T. 2003. Vigilance and foraging patterns of American elk during the
rut in habitats with and without predators. Canadian Journal of Zoology 81: 266–
271.
Woodland DJ, Jaafar Z, Knight ML. 1980. The pursuit deterrent function of alarm
signals. America Naturalist 115:748-753.
Wunderle JM Jr, Waide RB. 1993. Distribution of overwintering Nearctic
migrants in the Bahamas and Greater Antillies. Condor 95:904-33.
Wunderle JM. 1997. The role of animal seed dispersal in accelerating native forest
regeneration on degraded tropical lands. Forest Ecology and Management 99:
223–235.
Wunderle, Jr. JM, Latta SC. 1998 Avian resource use in Dominican shade coffee
plantations. Wilson Bulletin 110:271-281.
Yarranton GA, Morrison RG. 1974. Spatial Dynamics of a Primary Succession:
Nucleation. The Journal of Ecology 62: 417-428.
Ydenberg RC, Dill LM. 1986. The economics of fleeing from predators. Advances in the
Study of Behavior. 16:229–249.

112

Zahawi RA, Augspurger CK. 2006. Tropical forest restoration: Tree islands as
recruitment foci in degraded lands of Honduras. Ecological Applications16: 464478.
Zanette L, Doyle P, Tremont SM. 2000. Food shortage in small fragments:
Evidence from an area sensitive passerine. Ecology 81:1654-1666.
Zahawi RA, Holl KD. 2009. Comparing the performance of tree stakes and
seedlings to restore abandoned tropical pastures. Restoration Ecology. DOI
10.1111/j.1526-100X.2008.00423.x
Zangerl AR, Hamilton JG, Miller TJ, Crofts AR, Oxborough K, Berenbaum MR, de
Lucia EH. 2002. Impact of folivory on photosynthesis is greater than the sum of
its holes. Proceedings of the National Academy of Science 99: 1088–1091.

113