Group Title: comparison of island and mainland pollination ecology
Title: A comparison of island and mainland pollination ecology
Full Citation
Permanent Link:
 Material Information
Title: A comparison of island and mainland pollination ecology
Alternate Title: Pollination ecology
Physical Description: v, 71 leaves : ill., map ; 28 cm.
Language: English
Creator: Spears, Edwin Eugene, 1953-
Copyright Date: 1983
Subject: Fertilization of plants   ( lcsh )
Zoology thesis Ph. D
Dissertations, Academic -- Zoology -- UF
Genre: bibliography   ( marcgt )
non-fiction   ( marcgt )
Statement of Responsibility: by Edwin Eugene Spears, Jr.
Thesis: Thesis (Ph. D.)--University of Florida, 1983.
Bibliography: Bibliography: leaves 65-70.
General Note: Typescript.
General Note: Vita.
 Record Information
Bibliographic ID: UF00099354
Volume ID: VID00001
Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
Resource Identifier: alephbibnum - 000450217
oclc - 11446966
notis - ACL1885


This item has the following downloads:

comparisonofisla00spea ( PDF )

Full Text



Edwin Eugene Spears, Jr.





I would like to thank the members of my committee for their

constructive criticisms of this dissertation, and especially Peter

Feinsinger, who did an excellent job of prodding his first doctoral

student. Thanks are also due to Carol Binello, who's timely

assistance was greatly appreciated.


ACKNOWLDEGMENTS .................................. ii

TABLE OF CONTENTS ............................... iii

ABSTRACT ......................................... iv

INTRODUCTION ...................................... 1

METHODS ........................................... 4
Study Species ................................ 4
Study Sites .................................. 6
Breeding Systems ............................. 9
Flowering Phenologies ....................... 10
Pollinators ................................. 11
Reproductive Success ........................ 12
Pollinator Limitation ....................... 13

RESULTS .......................................... 15
Breeding Systems ............................ 15
Phenologies ................................. 20
Pollinator Communities ...................... 25
Reproductive Success ........................ 31
Pollinator Limitation ...................... 42

DISCUSSION ....................................... 51
Breeding Systems ............................ 51
Phenologies ................................. 54
Pollinator Communities ..................... 56
Reproductive Success
and Pollinator Limitation .................. 61
Evolutionary Considerations ................. 64

LITERATURE CITED ................................. 65

BIOGRAPHICAL SKETCH .............................. 71


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



Edwin Eugene Spears, Jr.

December, 1983

Chairman: Dr. Peter Feinsinger
Major Department: Zoology

In this study, I compared male and female reproductive success of

island and mainland populations of two plant species, Opuntia stricta

(Cactaceae) and Centrosema virginianum (Fabaceae) on the west coast of

Florida. During the 1980, 1981, and 1982 breeding seasons, both male

success, as measured by pollen dispersal, and female success, as

measured by fruit or seedset, showed a significant reduction in island

populations relative to mainland populations of these species despite

the close geographic proximity of the study sites. The pollinator

species that visited Opuntia stricta on the island site were a subset

of the mainland pollinator assemblage. A complete shift in

pollinating species of Centrosema virginianum occurred between the

mainland and the island. On the island site, pollination was less

effective for this species and less reliable for both species than on

the mainland sites. Observational and experimental data suggested

that pollinator scarcity was the cause of reduced reproductive success

on islands. On more distant, oceanic islands, these effects may be

even more extreme and could either filter out inappropriate colonists

or act as an agent of directional selection.



Floras of geographically isolated islands differ in many ways

from the floras of mainlands with similar climates (Carlquist 1974,

Rick 1966). The reproductive biology of island plants is often

distinct from mainland plants when examined at the community level. A

greater proportion of island plants is self-fertilizing than of

mainland plants (Baker 1955, Baker and Hurd 1968, Linsley et al.

1966, Rick 1966). Differences also exist in the floral structure of

island plants. The flowers of insect pollinated plants on islands

tend to be unspecialized, open structures and are smaller and less

showy than mainland flowers (Carlquist 1974). Finally, wind

pollination is more common in island communities than in mainland

communities (Whitehead 1969).

These differences in plant reproductive characteristics may have

evolved in response to the depauperate pollinator communities usually

found on islands (Baker 1955, Carlquist 1966, 1974). Island

biogeographic theory predicts reduced diversity on islands for any

guild or community of organisms, relative to mainland sites (MacArthur

and Wilson 1967). Reduced diversity on islands has been found in

insects (Simberloff and Wilson 1969) and birds (Diamond 1970), among

many other taxa (see Carlquist 1974 and MacArthur and Wilson 1967 for


If pollinators follow the general trend of reduced diversity on

islands, a flowering plant that has colonized an island may have a

lower chance of being visited than the same plant growing on the

mainland. Island plants that require animal pollinators might suffer

a decrease in reproductive output if visits are less common or

certain. This would place such plants at a competitive disadvantage

relative to other species or phenotypes that could self-fertilize, use

a wide variety of pollinators, or use wind as a pollen dispersal


Reduction in seed or fruit set, measures of female reproductive

success, has been found in plant species growing in other conditions

where pollinators are uncertain or rare, Examples include herbs

flowering in early spring in temperate woodlands (Schemske et al.

1978) and insect-pollinated species in the Arctic (Kevan 1972).

Pollen dispersal, a measure of male success, depends on pollinator

visitation (Beattie and Culver 1979, Levin and Kerster 1974) and

should be lower if pollinator visit rates are reduced. Linhart and

Feinsinger (1980) demonstrated reduced pollen dispersal in Mandevilla

hirsuta (Apocynaceae) on Tobago compared to a larger island, Trinidad,

and attributed this to erratic bird visitation to Mandevilla flowers

on the smaller island.

Pollinator scarcity could affect island plant communities by

either 1) acting as a selective filter, inhibiting colonization of

islands by plants with floral structures and breeding systems

requiring pollinators, or 2) exerting directional selection pressure

on plant populations established on an island, adapting the floral

structures and breeding systems of these colonists to independence

from pollinators.

Few studies have shown that pollinators are scarcer or less

predictable on islands than on mainland sites. On the Galapagos

Islands, several orders of pollinating insects are represented by very

few species, although the parts of the South American mainland with a

similar climate have a diverse pollinator assemblage (Linsley 1966).

Only a single species of flower-visiting bee is found on all the

islands (Linsley et al. 1966). The diversity and density of

nectarivorous birds are lower on a small, distant island (Tobago) than

on a large, near island (Trinidad) (Feinsinger et al. 1979, 1982,

Linhart and Feinsinger 1980). Even when pollinator scarcity has been

demonstrated, the reproductive cost to the plant species on the island

has almost never been estimated, except for one hummingbird-pollinated

species on Trinidad and Tobago, where both pollen dispersal and seed

set were reduced on the smaller island (Linhart and Feinsinger 1980).

Are the reproductive differences usually observed between island

and mainland plant communities due to pollinator scarcity? To answer

this question requires an examination of the reproductive success of a

plant species with populations on an island and an adjacent mainland.

In this study I examined two insect-pollinated plant species with

populations growing on the mainland and on two off-shore islands, one

smaller and more distant than the other. Three basic questions were

addressed: 1) Are pollinators less abundant and less predictable on

the island sites? 2) Is reproductive success, both male and female,

reduced in island plant populations? 3) If reproductive success is

reduced, can this be shown to result from pollinator limitation?


Study Species

The two study species are Opuntia stricta (Cactaceae) and

Centrosema virginianum (Fabaceae). These species were chosen because

1) they were the most abundant native species growing at the far

island site, where the angiosperm community was least diverse, and 2)

these species represent extremes in floral structural complexity.

Opuntia stricta, the prickly pear cactus, is a large, perennial,

succulent plant. Often over a meter tall, plants can cover as much as

three square meters at my study sites. Most species in the genus

Opuntia undergo extensive vegetative growth. Sections of the stems,

or "pads," are frequently lost from the parent plant (Rympa 1952).

These pads can then grow into a full-sized plant. This form of

vegetative propagation is common among Opuntia stricta growing at my


The flower of Opuntia stricta has many petals and numerous

stamens with a single pistil. The stigma is five-lobed. Each lobe is

tightly appressed to the others when the flower first opens, but these

separate and expose the receptive surfaces by the time the anthers

dehisce. The flower is bright yellow and bowl-shaped, producing

abundant pollen but little or no measurable nectar. At these sites,

there is polymorphism in flower color and size within a study

population. Some individuals had flowers with a noticable orange tint

and flowers that were usually larger than the bright yellow flowers of

other individuals. These individuals may contain characters

transmitted from other species of Opuntia through hybridization, since

the several Florida species apparently readily hybridize (Lyman

Benson, personal communication). To reduce this potential source of

variation, I chose only individuals with identical floral and pad

morphology as study plants.

The open floral structure suggests that a wide array of

flower-visiting insects could utilize Opuntia, but most of the

pollinators that I observed visiting this species were pollen-foraging

hymenoptera. In the field, most flowers last a single day, opening

after sunrise and closing late in the afternoon. Timing of flower

opening, stigma and anther dehiscence, and petal closing are largely

temperature dependent. On cool, cloudy days flowers may open only

partially and remain open for only a short time. Such flowers usually

open again the next day. The inferior ovary of Opuntia stricta is

large and fleshy by the time of flower opening. Mature fruits are

approximately twice the size of the ovary of a newly-opened flower.

At my study sites, Opuntia blooms from late April until early June,

for approximately six weeks.

Centrosema virginianunm, the butterfly pea, is a perennial,

herbaceous vine. At my study sites, Centrosema plants grow in

intertwined masses, which makes identification of individual vines

impossible. Each vine is several meters in length and produces

abundant flowers. The flower of Centrosema is highly specialized.

Structurally it is an inverted flag flower adapted for pollination by

large hymenopterans (Faegri and van der Pijl 1979). One ?etal is

modified to act as a landing platform for the pollinator. The

remaining petals form a structure surrounding the anthers and pistil,

which must be manipulated by a visitor to transfer pollen and expose

the stigma. Flowers last a single day. After fertilization, the

superior ovary must grow substantially to produce a mature fruit. In

this area, Centrosema usually flowers from early August through

September, for approximately nine weeks.

Study Sites

I chose study sites based on the presence of both plant species

and on similarity of the habitat. The near island, Cedar Key, and the

far island, Seahorse Key, had areas where both species were abundant.

On the mainland, however, Opuntia strict was found only as isolated

individuals; I found no population large enough to include in this

study. Comparisons among Opuntia populations were therefore limited

to the larger, inshore island and the smaller, offshore island. Cedar

Key, the larger, inshore island, is considered to be the mainland for

the comparison of Opuntia stricta populations (similar comparisons

using a large island as "mainland" were made by Diamond 1970, and

Linhart and Feinsinger 1980). I found a suitable mainland site for

Centrosema, so that three populations of this species were compared.

Seahorse Key, the smaller, offshore island, is located

approximately eight kilometers off the west coast of Florida at 290

07'30" (Figure 1). The island is 1.5 kilometers long and varies

between 0.1 and 0.5 kilometers in width. My study transect extended

0.5 kilometers along the south beach, a sandy, low-energy beach

dominated by sea-oats (Uniola paniculata) with live oak (Quercus

virginianum) further inshore. Populations of both study species grow

in the ecotone between the oak-dominated forest and the grassy





c+ o H*
11 0
r* 4 0

Hm 3


(D (D
Q p
0 9

m 0


0 0
o- (D
C .


0 p




:. \ {/ -

=: .-

".' ",..: .." ". -'
. 'i' : : '

." ."'." :" "
:! .'. W ."
.' " "

I ; *. "
II ".I .

I. /\ 7



A (

N "

;`-*N* '. : IN



?~ 4
--= t y% :;

_ I

Cedar Key is located midway between Seahorse Key and the

mainland, connected to the latter by a man-made causeway. Cedar Key

has approximately three times the surface area of Seahorse Key,

approximately 3 kilometers long and between 0.2 and 1.0 kilometers

wide. My Cedar Key transect was the same length as that on Seahorse

Key, also along a south-facing, low-energy beach. Vegetation was

similar to that on Seahorse Key, but appeared to be more strongly

influenced by human habitation and richer in species.

I found a mainland site for Centrosema in 1981, a disturbed

roadside area surrounded by a myrtle oak (Quercus myrtifolia) hammock.

This Centrosema population is less extensive than the two island

populations. As a consequence, my study transect was reduced to

approximately 0.25 kilometers. I attempted to choose habitats as

similar to each other as possible to use as study sites. This was

done to reduce other potential sources of variation in these

interpopulation comparisons so that any differences that I observed

among the populations would be due to an "island effect" and not to

habitat differences.

Breeding Systems

I determined the breeding system of Opuntia stricta by removing

stem sections (pads) with well-developed flower buds and placing then

in a greenhouse. Previous field observations had suggested that

isolated pads of Opuntia would give rise to mature fruits provided

that flower buds were already present. In 1982, approximately one

week before the flowering season began, I removed a pad from each of

twelve plants in each study population and potted the pads in a

greenhouse. Each pad chosen had produced between six and seventeen

flower buds. These pads were watered every other day for

approximately two months. When the flowers opened, I subjected them

to one of three treatments: unmanipulated, self-pollinated, or

cross-pollinated. Fruits were harvested after two months, before

fruit maturation, but after sufficient time for fertilized seeds to

develop fully. I counted all seeds in each experimental fruit and

determined fruit abortion rates. Many seeds were much smaller than

normal and were often discolored. I assumed such seeds were aborted

and did not include them in further analyses.

To determine if the fruit abortion rates that I observed in the

greenhouse were an artifact of growing conditions or the result of the

experimental manipulations, I also bagged flowers in the field, using

mosquito netting to prevent insect visitation. These flowers were

either left unvisited or were cross-pollinated by hand. I compared

abortion rates of these fruits to those of unmanipulated and

hand-crossed flowers in the greenhouse.

Flowering Phenologies

A census of twelve plants of Opuntia stricta was made weekly at

each study site in 1980, 1981, and 1982. Because it was difficult to

distinguish individual plants of Centrosema virginianum, I chose three

discrete patches at each site. If the flowers per patch exceeded 200,

I estimated to the nearest twenty flowers. Centrosema was not chosen

as a study species until the middle of its flowering season in 1980,

and the Cedar Key study site was mowed in the middle of the 1982

Centrosema breeding season, so complete census data for this species

are available only for the 1981 season.


In 1980, 1981, and 1982 I determined pollinator diversity and

abundance by timed floral observations throughout the breeding

seasons. Data for Opuntia stricta in 1980 consisted of hourly, three

minute observations at each of the twelve study plants from the time

of flower dehiscence in the morning until the flowers closed at

approximately 16:00 EST, once a week at each site. A preliminary

ANOVA of these 1980 data revealed no significant variation in visit

frequencies or visitor diversity among the study plants at a given

site. Therefore, I limited visitor censuses for the remaining seasons

to observations of a single individual at each study site. The

observation period was increased to fifteen minutes each hour. For

each weekly visitor census, I chose the individual with the largest

number of open flowers. I feel this technique is valid as a relative

measure because this one plant would often contain from 25-50% of the

total flowers in bloom on all the study individuals. This was a

possible source of bias in determining visit rates; however, there

was little difference among the study sites in the number of flowers

that I observed for a given week during the breeding season. I

determined visit rates and visitor diversity for Centrosema

virginianum by observing the patch with the greatest number of

flowers. As a relative measure of pollinator abundance, I estimated

the number of insect visits / flower'hour based on a 15 minute

observation period each hour the flowers were open. This was not an

absolute measure of pollinator density because floral density strongly

influenced the value. For example, if the number of pollinator visits

that I measured on Seahorse and Cedar Keys were similar, but more

flowers were under observation on Seahorse Key, visit rate would be

lower on Seahorse Key. Number of visits / flower'hour was, however,

a valid relative measure of the visit frequency a given flower in a

study population can expect.

Both study species received insect visits by numerous

lepidopterans and small hymenopterans, which did not contact the

female reproductive structures. I did not include these species in

calculations of either diversity or abundance.

Reproductive Success

Both of these study species produce hermaphroditic flowers,

flowers that produce both pollen and ovules. The male and female

functions of these flowers can have different success under a variety

of environmental conditions (Willson 1979) and both should be

determined when comparing reproductive success. Female success of a

plant can be measured as the number of fruits or seeds produced (Lloyd

1979). Seed set is probably a better measure, since each seed

represents a potential individual in the next generation. To

determine female reproductive success, I measured natural fruitset in

both species and the seeds per fruit in Opuntia stricta. Each week of

the 1981 and 1982 breeding seasons I tagged and left open to natural

visitation 20 flowers of Opuntia strict. For Centrosema virginianum

the weekly sample size was 20 in 1981 and 30 in 1982. Centrosema

fruits mature quickly; I harvested fruits of this species 1-2 weeks

after tagging. Opuntia fruits required a longer development time and

were harvested in September, by which time they had begun to ripen, as

shown by the reddening of the "pear."

T conducted an additional experiment with Opuntia to determine

the relationship between number of pollinator visits and r-sulting

seedset. I bagged flowers of a randomly chosen plant before they

opened, removed the bags when the flowers opened, and rebagged each

flower after a given number of visits.

Male reproductive success is the number of ovules fertilized by

pollen from a given plant (Willson 1979). This is difficult to

measure, however, so male reproductive success is often estimated by

determining a plant's pollen dispersal pattern. I estimated pollen

dispersal with fluorescent dyes, a standard technique (Stockhouse

1976) recently evaluated by Waser and Price (1982). Because floral

density is known to affect pollen dispersal (Levin and Kerster 1974),

I conducted these experiments only during times when flower densities

were similar among the study sites during the 1981 and 1982 breeding

seasons. The anthers of a marked flower were dusted with UV

fluorescent dye early in the morning. At the end of the day, when the

flowers began to wilt, I harvested flowers at 0.5, 1.0, 2.0, 4.0, and

6.0 meters from the dusted flower and examined them for the presence

of dye with a hand lens under UV light. The data that I collected

were percentages of flowers at each distance containing dye on their

reproductive structures.

Pollinator Limitation

I conducted field experiments to determine whether pollination

might be limiting female reproductive success in my study populations.

Between 15 and 60 Centrosema flowers were hand-pollinated during the

1981 and 1982 flowering peaks (late September to early October), the

only times when sufficient flowers were available at all sites for the

experiment. I tagged but left unmanipulated an equal number of

flowers to serve as controls. Whenever possible, the control flowers

were the nearest neighbors of the experimental flowers. I carried out

pollinator limitation experiments on Opuntia stricta only during the


1983 breeding season. Unfortunately, the results of this experiment

cannot be easily compared to visit-rate or reproductive success data

collected during earlier breeding seasons. Each week of the 1983

flowering season, I hand-crossed 13 to 20 Opuntia flowers at each

study site. Again, an equal number of flowers was tagged and left

untouched as controls. I determined fruitset for both species and

seeds per fruit for Opuntia.


Breeding Systems

Table 1 compares fruitset among greenhouse plantings of Opuntia

strict from Seahorse and Cedar Keys for each of three treatments.

Fruitset rates for unvisited and hand-crossed flowers in the field are

also presented. These fruitset data were analyzed using a log-linear

model (Sokal and Rohlf 1981) to determine which variables

significantly influenced fruitset. Treatment was the only variable

significantly affecting Opuntia fruitset (X2 =23.98, p<0.001);

neither locale nor growing condition (greenhouse or field) affected

fruitset. There were no significant interactions among the main
effects. A contingency X showed no difference between the fruitset

for greenhouse flowers that were selfed or crossed, but significant

differences did exist among unvisited flowers and the other treatments

(Unvisited vs. Selfed X2=5.20, p<0.05, Unvisited vs. Crossed
X2=16.47, p<0.01). Also, hand-crossed flowers in the field were

significantly more likely to set fruit than were unvisited flowers in

the field.

Table 2 presents mean seedset for the Opuntia fruits from the

greenhouse experiments. An ANOVA of mean seeds per fruit found no

significant locale or among-plant variation. Mean seeds / fruit did

differ significantly among treatments (F=148.6, p<0.001) and a

Duncan's Multiple Range Test revealed that all treatment means in

Table 2 are significantly different. To summarize the Opuntia strict

Table 1: Fruitset in Opuntia stricta under three experimental treatments
from pads growing in a greenhouse and plants in the field.
No between-site differences in fruitset were significant, but
unvisited flowers set significantly fewer fruits than did
crossed or selfed flowers (alpha = 0.05).

Greenhouse Data

Cedar Key
Seahorse Key

Cedar Key
Seahorse Key






Field Data




Table 2: Mean number of seeds/fruit of Opuntia stricta from three
experimental treatments (X + S.E. (coefficient of variation)).
No between site differences in seedset were significant,
however, mean seedset for each of the three treatments were
significantly different (alpha = 0.05).




Cedar Key
Seahorse Key

5.8 + 1.8(189.2)
8.0 2.8(212.3)

66.2 + 5.0(40.0)
61.3 + 5.6(40.8)

73.6 + 3.4(41.4)
73.0 T 4.0(30.2)

breeding system, no differences existed between Seahorse and Cedar Key

populations in either fruit or seedset for my experimental treatments.

Unvisited flowers set fewer fruits and had fewer seeds / fruit than

did either selfed or crossed flowers. Selfed flowers had a similar

fruitset rate to crossed flowers, but had fewer seeds / fruit on the

average than did crossed flowers. Hand-crossed flowers had high rates

of fruitset and the greatest mean number of seeds / fruit of the three


I analyzed only fruitset data for Centrosema virginianum (Table

3). Seeds / fruit ranged between 15 and 23 for all fruits that

reached maturity, averaging approximately 18 seeds / fruit. Many of

the experimental fruits, however, did not reach maturity due to heavy

predation by unidentified herbivores, so I limited Centrosema

reproductive data to fruitset. Pod predation could have been a

possible source of bias in the fruitset data if fruits had been

entirely consumed.. Only if the entire fruit and petiole were missing

was the fruit considered to be aborted. Partially eaten pods were

most abundant on the mainland, which suggested that predation may have

been greater at that site. Because fruitset was predicted to be

greater on the mainland, this bias would have been a conservative one.

I tested for locale and treatment effects on fruitset with a

log-linear model and found treatment effect to be highly significant
(X2=66.08, p<0.001) but found no site effect or any higher order

interactions between these two variables. Fruitset differed

significantly among all three treatments (Unvisited vs. Selfed

X2=19.52, p<0.001, Unvisited vs. Crossed X2=63.93, p<0.001, Selfed

vs. Crossed X2=10.52, p<0.01). The mainland and two island

populations of Centrosema did not differ in their response to a given

Table 3: Fruitset in Centrosema virginianum under three experimental
treatments from plants in the field. No among-site
differences in fruitset were significant, however, each of
the three treatments resulted in significantly different
fruitset (alpha = 0.05).

Unvisited Selfed Crossed

Seahorse Key 0%(N=40) 23.3%(N=30) 30.0%(N=90)
Cedar Key 0%(N=7) 16.7%(N=18) 40.0%(N=50)
Mainland 0%(N=35) 6.7%(N=15) 48.4%(N=45)

experimental treatment. Unvisited flowers did not set fruits, and

selfed flowers produced significantly fewer fruits than did crossed



Figures 2 and 3 represent the flowering phenologies of Opuntia

stricta and Centrosema virginianum. I used a Kolmogorov-Smirnov

2-Sample Test (Siegel 1956) to compare the phenologies of Opuntia

populations at the two locales for 1980 and 1981; the 1982 census was

incomplete due to the destruction of several individuals at the Cedar

Key study area. The flowering distributions of the two sites differed

significantly for both years (p<0.01). The Kolmogorov-Smirnov

2-Sample Test could not be used to compare three study populations;

instead, I used the X2 test for k-independent samples (Siegel 1956),

which tests for differences in frequency distributions, to compare

Centrosema populations. The three sites differed significantly in

flowering phenologies for the 1981 season. The 1980 and 1982 seasons

appeared to have less among site variation, but the data were too

limited for statistical analysis.

Different flowering phenologies between sites could be the result

of individual plants flowering for a longer time at one locale, or the

differences in phenology could be caused by a greater variance in

blooming times among individual plants at a given site. To

distinguish between these two possibilities, I calculated a "flowering

diversity" value for each individual of Opuntia stricta. This was not

possible for Centrosema. Simpson's C-inverse (Simpson 1949) was used,

Figure 2: Flowering phenologies of Opuntia stricta for the 1980 (A),
1981 (B), and 1982 (C) flowering seasons. Flowering season
extended from late April until early June.

200 -










60 -



1 T T I


r \

io ,

1- - o
0,I I 0 8
y-^' i ? i- '^ 0



/ \



I I i I I I I I I I i I
0 2 3 4 5 6 7 8 9 10 II 12





-C -


00 -




Figure 3: Flowering phenologies of Centrosema virginianum for
the 1980 (A), 1981 (B), and 1982 (C) flowering seasons.
Flowering season extended from early August through






40 -

\ "


\ A
b '
1 I I ; I

I I i I I


O '

t \

n -rre

-I - i - i i



/ '















I i I I I I I i i I I
0 2 3 4 5 6 7 8 9 10 II 12



I I 1 I I I I I I


C inv.= __>__



where Pi equals the number of open flowers observed during week i

divided by the total number of open flowers observed for that

individual. These values incorporate the number of weeks an

individual was in bloom and the evenness of flowering over a season,

i.e. how concentrated the flowering was within the breeding season.

I log-transformed these individual "flower diversity" values and

analyzed by ANOVA to determine which factors affected flowering

diversity. Between year variation was highly significant (F=24.91,

p<0.001). Between site variation was not significant. These data

suggest that the observed differences in Opuntia phenology between the

two study sites were due to greater among individual variance in

blooming times on Seahorse Key, and not to a longer blooming time for

each plant.

Pollinator Communities

Tables 4 and 5 present the pollinators of Opuncia stricta and

Centrosema virginianum and the relative frequencies of their visits.

I observed three species transferring pollen on Seahorse Key, although

one species, Bombus pennsylvanicus, was found only one week in 1981.

Cedar Key had a richer pollinator community than Seahorse Key for both

species. The mainland population of Centrosema virginianum was

visited almost exclusively by Bombus pennsylvanicas. Two of the

pollinators that I observed visiting these two plant species are

Table 4: The relative abundance of pollinators of Opuntia strict at
both study sites. Relative abundance of a species was
determined for each breeding season by dividing the number
of flower visits observed for that species by the total
number of observed visits. Number of observed visits to
Opuntia stricta by each species is in parenthesis. Visitors
that did not carry pollen were not included.

Seahorse Key 1980 1981 1982

Agarostemon splendens 56.5%(26) 46.4%(116) 95.0( 57)
Megachile brevis 43.5%(20) 47.6%(119) 5.0%(3)
Bomnus oennsyl;anicus -- 6.0%(15) ---

Cedar Key

Aasaostemon solendens 7.8%(20) 42.1%(128) 21.L5(36)
Meiachile brevis 30.8%(79) 23.0(70) 7.75(13)
Apis mellifera 50.8%(130) 24.7%(75) 69.05(116)
Melissoides sp. 10.2%(26) 6.6%(20) 1.8%(3)
Campsomeris auadrimaculatus 0.4( 1) 3.65(11) ---

Table 5: The relative abundance of pollinators of Centrosema virginianum
at all study sites. Relative abundance of a species was
determined for each breeding season by dividing the number of
flower visits observed for that species by the total number of
observed visits. Number of observed visits to Centrosema
virginianum by each species is in parenthesis. Visitors that
did not carry pollen were not included.

Seahorse Key 1980 1931 1982

Mebachile brevis 100C(210) 100%(20) 100%(68)

Cedar Key

Megachile brevis 43.4%(29) 47.4%(36) 20.2( 50)
Melissoides sp. 56.7%(38) 51.3%(39) 52.C3(129)
Bombus pennsylvanicus --- 1.3(1) ---
Cammsomeris ouadriaculaus -- -- 27.3%(69)


Bombus oennsylvanicus No Site 95.4 (902) 10C5(296)
Aois mellifera 0.1(1) ---
Camisomeris quadrimaculatus 4.42(42) ---

colonial, Bombus pennsylvanicus and Apis mellifera. These forage for

both nectar and pollen. All the remaining species are solitary; all

but the scoliid wasp Campsomeris quadrimaculatus forage for pollen

only. A relatively small subset of species was shared by the two

plant species.

To quantify the apparent differences among the study sites, I

calculated weekly Shannon-Weaver species diversity values (H'= Pi In

Pi, where Pi=proportion of total individuals in species i),

log-transformed these values to normalize them (Sokal and Rohlf 1981),

and calculated an ANOVA. For Opuntia stricta, mean species diversity

values were 1.49 for Seahorse Key and 2.49 for Cedar Key. This

difference was highly significant (F=10.45, p<0.005). Opuntia

pollinator species diversity values did not significantly vary from

week-to-week within a breeding season or among breeding seasons. I

found similar results with Centrosema; site significantly affected

pollinator diversity (F=8.54, p<0.01) but among or within year

variation was not significant. A Duncan's Multiple Range test

indicated that Cedar Key had a significantly greater pollinator

diversity (SD = 1.56) than did Seahorse Key (SD=0.73) or the mainland

(SD=1.07); the latter two did not differ.

Tables 6 and 7 show mean weekly visit rates to flowers of Opuntia

strict and Centrosema virginianum. When analyzed for variance, among

year variation (F=8.74, p<0.01) and between site variation (F=39.25,

p<0.001) were significant for Opuntia; neither was significant for

Centrosema. Although the ANOVA found no significant locale effect

(p=0.17) a Duncan's Multiple Range test of mean visit rates of

Centrosema indicated that Seahorse Key visit rates (X=0.532) differed

significantly from mairiland visit rates (X=4.245), though neither

Table 6: Mean pollinator visit rates to Opuntia stricta in number of
visits/flower/hour (X + S.E. (coefficient of variation)).
Among year and between site differences in visit rates are
all significant (p<0.01).

Seahorse Key

0.17 + 0.07 (97.2)
1.28 + 0.61 (12).7)
1.40 + 1.14 (140.6)

Cedar Key

0.60 0.20 (68.0)
.22 . 0.54 (25.5)
3.73 + 0.51 (25.2)



Table 7: Mean pollinator visit rates to Centrosema virginianum in
number of visits/flower/hour (X + S.E. (coefficient of
variation)). Among-year differences were not significant,
but a year-locale interaction was significant (alpha = 0.01).
When each year was analyzed separately, the only significant
difference was between the mainland site and the two island
sites in 1981 (alpha = 0.01).

Year Seahorse Key Cedar Key Mainland

1980 0.71 + 0.19(54.9) 0.63 + 0.22(49.4) No Site

1981 0.19 + 0.18(231.0) 0.61 + 0.18(65.4) 5.27 + 1.76(66.8)

1982 1.69 (N=l)

6.81 + 1.48(30.8) 2.20 + 0.89(57.2)

differed significantly from Cedar Key (X=1.996).

Not only are visit rates lower for small island populations of

both species, but Centrosema visit rates also show greater variation

on the island sites than on the mainland (Tables 6 and 7). Dawkins

and Dawkins (1973) describe a method to determine if significant

differences exist in the variation around sample means that requires

only the CV values and the numbers of observations. The resulting

C-statistic is compared to a t-distribution to test for significance.

By this measure, the CV for the visit rates of the Seahorse Key

population of Centrosema virginianum is greater than the CV for the

Cedar Key population (C=2.37, p<0.05) and the mainland (C=2.32,

Opuntia populations are not significant, probably due to the small

sample size. In all years, however, the CV's for Opuntia visit rates

are greater on the far island.

To analyze one source of variation in visit rates, I examined the

relationship between weekly visit rates and flowering phenologies for

the study species. Data for 1981 were more complete and are presented

in Figures 4 and 5. The populations of both species on Seahorse Key

have extremely low visit rates during the peak of flowering, and

higher rates when flower densities are low. Neither Cedar Key nor

mainland populations show a similar pattern.

Reproductive Success

Tables 8 and 9 show field fruitset data for the two study

species. I analyzed these data with log-linear models. Neither

locale nor year significantly affected Opuntia fruitset. Both year

and locale had a significant effect on Centrosema fruitset (locale

X2=12.38, p<0.01, year X2=3.92, p<0.05), and there was a significant

Figure 4: 1981 flowering phenologies and weekly pollinator visit
rates of Opuntia stricta at (A) Seahorse Key and (B)
Cedar Key. Note thedifferent scales for visit rates
for each site.





o <






0 1 2 3 -4 5 6 7 8


Figure 5: 1981 flowering phenologies and weekly pollinator visit
rates of Centrosema virginianum at (A) Seahorse Key,
(B) Cedar Key, and (C) the mainland. Note the different
scales for visit rates for the three sites.

250 9 2.50

200 - VISIT RATE -2.00

150- -1.50

100- / 1.00

50- .50

Sr T I
0 I 2 3 4 5 6 7 8 9 10 II 12

100- \ -I.0
M (I 0 o

S -- FLOWERS / 80 L

_ 60- .60
LL -
o --
LL 40- .40

6 20- -.20
z 02

0 I 2 3 4 5 6 7 8 9 10 11 12

100- 10.0
80- .80

60- -.60
40- \.40

20- -.20

0 I 3 Ii I i i I 9 0 I
0 I 2 3 4 5 6 7 8 9 10 II 12


Table 8: Fruitset in Opuntia stricta from tagged, unmanipulated
flowers in the field. Neither between year nor between site
differences in fruitset were significant.

Year Seahorse Key Cedar Key

1981 59.2%(N=135) 58.8%(N=80)




Table 9: Fruitset in Centrosema virginianum from tagged, unr.anipulated
flowers in the field. Both year and site significantly
affected Centrosema fruitset.

Year Seahorse Key Cedar Key Mainland

1981 4.0% (N=100) 18.8% (N=80) 30.0% (N=60)
1982 2'.3% (N=60) 36.7% (N=60) 18.9s (N=90)

year-locale interaction (X2=19.45, p<0.01) indicating that the

fruitset pattern changed within the two year study period. In 1981

fruitset on Seahorse Key was significantly lower than on Cedar Key

(X2=10.24, p<0.01) and the mainland (X2=21.27, p<0.01). In 1982,

fruitset on the mainland was significantly lower than on Cedar Key
(X2=5.70, p<0.05) but no other sites showed significant differences.

Data from the breeding system experiments indicated that seeds /

fruit was a more sensitive indicator of reproductive success than

fruitset in Opuntia stricta. An ANOVA of Opuntia seeds / fruit

determined that several factors had a highly significant effect on the

number of seeds / fruit; locale (F=17.57, p<0.001), year (F=20.44,

p<0.001), week of flowering within the breeding season (F=21.44,

p<0.001), and individual plant at the site (F=4.92, p<0.01) all

significantly affected the number of seeds / fruit. Table 10

summarizes these data. The coefficients of variation for seeds /

fruit differ significantly between the two populations for each of the

three breeding seasons during which I collected fruits (1980 C=A.64,

1981 C= 4.66, 1982 C=3.84, all p<0.01). The number of seeds / fruit

varied more on Seahorse Key than on Cedar Key. Mean seeds / fruit for

each study individual ranged from 27.3 + 4.0 to 86.2 + 6.9 on Seahorse

Key, and from 58.2 10.6 to 99.0 + 8.0 on Cedar Key.

Figure 6 shows the relationship between number of pollinator

visits and resulting seedset in Opuntia strict. A single visit was

enough to ensure a high level of seedset, much higher than was found

in fruits of unvisited flowers. Mean seedset increased with

increasing visitation, up to twelve visits, but the rate of increase

was much lower after the first visit. This relationship between

number of pollinator visits and the resulting number of seeds / fruit

Table 10: Mean number of seeds/fruit from tagged, unmanipulated
Opuntia stricta flowers in the field (YX S.E. (coefficient
of variation)). Both site and year significantly affected
Opuntia seedset.

Seahorse Key

Cedar Key

60.6 + 3.7 (44.8)
65.3 4.8 (65.6)
67.5 + 4.5 (41.7)

74.3 + 2.2 (23.0)
86.5 + 4.5 (35.7)
89.6 3.1 (19.3)


4I- a)`

0d C)


4- Cd-




4' (D

o m

0 '0
m 0p
'l C)

0)l- C)

c 'o
'H p 4k
o k

'0 ~
.c c-p

0- o U








0 0 0 0 0 a 0

ilfliA/sGB~s ]o *ON NVEM

suggested that one source of the variation observed in mean seeds /

fruit was the number of pollinator visits that a flower receives. The

significantly higher variation in Opuntia seedset on Seahorse Key may

have been a result of a greater variation in the number of pollinator

visits that Seahorse Key Opuntia flowers received.

Figures 7 and 8 present the data for pollen dispersal of Opuntia

stricta and Centrosema virginianum, i.e. the index to male

reproductive success used in this study. A contingency X2 for each

distance showed that the patterns of dispersal did not differ

significantly between 1981 and 1982 at any of the study sites. Sample

sizes for each distance were between 15 and 30 flowers for each

season. These figures represent pooled samples for both years. I

calculated contingency X2's for the number of flowers receiving

fluorescent dye at each site for each of the five distance categories.

Pollen dispersal of Opuntia was significantly greater on Cedar Key

than on Seahorse Key at every distance except the furthest distance,

six meters. For Centrosema I found no difference in pollen dispersal

between the two islands; however, the mainland had greater pollen

dispersal at all distances except six meters when compared to the two


Pollinator Limitation

Tables 11 and 12 present fruitset data from the pollinator

limitation experiments on the two study species. Opuntia fruitset

showed no difference between the hand-crossed and unmanipulated

flowers at either site. Mean seeds / fruit of Opuntia appeared

reduced in open flowers when compared to hand-crossed flowers on

Seahorse Key (Table 13), but this difference is not significant

(F=2.51, one-tailed p=0.056). A log-linear model indicated that both




to +' '0
- 01
*H r-I

B) a)
H *Hrl
CO 01

O fV
-P C)

C0 *
S*H >
0 *H

*H 0O

*H -P 0 01

HO 01
A 0 CC-P

'H a 0
QJ *


-O -

-i r
X0 L

Oco (o I
S3 I M -

0 0 0 0 00
o 1 / ;-

0 ( 0
/ ^ -"Eo

/ '' Q
// i
/^ //
,^---- -0'

^^ ------ I---- !-------- -

co ID

to a pd
0 C0 () r 4
O )
1i r O) OC

4- H
0 0 v

C) 0 rl C
0' 0C u
* *1H *) >
* fl f- a

bM*H C-

oM ukH 'C
C) 0C bC
0 b *sH a

' a o m C

*H C

re i r-

09 -(


-O -

-, [ O


I I o
o 0~C

S'/ /



0 0 0 0 o 0
0 co N -

3A0 HUIM Sy3MO1i %

Table 11: Fruitset in Opuntia stricta from the 1983 pollinator
limitation experiment. Handcrossed and unmanipulated
ODuntia flowers had similar fruitset rates.

Treatment Seahorse Key Cedar Key

Crossed 90% (N=81) 89% (N=72)

Open 89% (N=81) 85% (N=72)

Table 12: Fruitset in Centrosema virginianum from 1981 and 1982
pollinator limitation experiments. Only on Seahorse Key
in 1981 did fruitset rates differ significantly between hand-
crossed and open flowers (alpha = 0.05)

Year Treatment Seahorse Key Cedar Key Mainland

1981 Crossed 30% (N=30) 45% (N=20) 67% (N=15)
Open 0% (N=40) 40% (N=20) 55% (N=20)

1982 Crossed 32% (N=60) 37% (N=30) 40% (N=30)
Open 23% (N=60) 43% (N=30) 23% (N=30)

Table 13: Mean number of seeds/fruit from the fruits of the 1983
Opuntia stricia pollinator limitation experiment (X + S.E.).
Differences between handcrossed and open flowers were not
significant at p = 0.05.

Treatment Seahorse Key Cedar Key

Crossed 84.2 3.5 (N=75) 73.1 + 2.4 (N=64)

75.9 + 3.9 (N=74)

71.5 + 2.8 (N=61)


treatment (X2=6.31, p<0.02) and locale (X2=15.94, p<0.01)

significantly affected fruitset in Centrosema. Contingency X2 for

each locale showed a significant difference in fruitset between open

and crossed flowers only on Seahorse Key (X2=6.34, p<0.02).

In addition to the experimental evidence for pollinator

limitation of female reproductive success, correlative data also

suggest pollinator limitation. I found a highly significant Spearman

rank correlation coefficient (r=0.898, p=0.015) between weekly visit

rates and the resulting fruitset from these flowers for the Seahorse

Key population of Centrosema virginianum. This correlation was not

significant at Cedar Key or the mainland. Neither weekly fruitset nor

weekly mean seeds / fruit was correlated with pollinator visit rates

for the Opuntia populations.


Breeding Systems

Island plant populations can be examined in the larger context of

plant species expanding into new areas (Rick 1966). Island plant

populations should, therefore, share characteristics with other plant

populations at the limit of the species' range, or populations

introduced into foreign soil. The breeding biologies of many such

plant populations have been examined previously.

Baker (1955) noted that both animal and plant species capable of

long-distance dispersal tend to be self-compatible or hermaphroditic.

This restricts recombination in colonizing organisms, but new gene

combinations will not be as important to individual fitness in new,

unoccupied territory as they will in stable, established populations

(Grant 1975). In a study of the breeding biology of exotic weeds in

Canada, Mulligan (1972) found that most weed species were

self-pollinating; the few species requiring cross-pollination were

long-lived perennials. Rick (1966) examined the reproductive biology

of 16 species of angiosperms on the Galapages Islands and found that

13 of these species were automatically self-pollinating, requiring no

external agent. Rick also noted that many plant families that were

obligate outcrossers on the nearest mainland were missing from the

island flora. Along a California transect, Moldenke (1975) found a

decrease in the proportion of plants requiring insect visits as

environmental severity increased. At the climatically least

predictable site, a coastal island, 70% of the flora habitually

selfed. In a more detailed examination of the breeding biology of a

single species (Gilia achilleifolia), Schoen (1982) found increased

autogamy in the northernmost populations in this species' range where

pollinators were rarer. Colonizing populations are, then, more likely

to be autogamous than are established populations, or populations in

equable habitats.

An obligately outcrossing angiosperm faces two obstacles as a

potential colonist; 1) conspecifics must simultaneously be present

and 2) some vector must be available for pollen transport. The

autogamous colonist faces neither of these barriers. The most obvious

advantage that autogamous plants have over obligate outbreeders is the

ability to establish a new population without the presence of a

conspecific (Baker 1955). An additional advantage exists: an

individual can set seeds in the absence of specialized pollinating

agents (Baker 1955, Baker and Hurd 1968, Carlquist 1974, Rick 1966,

Schoen 1982, Solbrig and Rollins 1977). Those successful colonizing

species that do out-cross tend to have open, generalized flowers,

which can be visited by a wide variety of pollinators (Carlquist 1966,

1974, Mulligan 1972).

According to the above argument neither Opuntia stricta nor

Centrosema virginianum is an ideal colonist. Both species have

greatly reduced reproductive success when their flowers are not

visited by pollinators. Unvisited Opuntia flowers produce some seeds

and fruits, although the frequency of both is low relative to

animal-visited flowers. This species appears to be largely

self-compatible. Unvisited Centrosema flowers were never observed to

set fruit. This species appears to be partially self-compatible; the

seeds / fruit from selfed flowers were lower than the number found in

crossed flowers. The highly specialized flower of Centrosema is

probably another handicap to colonization. Without a large

hymenopteran species present, sexual reproduction in Centrosema should

be severely limited. The persistence of these species on Seahorse Key

is not surprising, however, considering the island's proximity to the

mainland. Immigration rates of the plants and potential insect

pollinators from the mainland are presumably high (MacArthur and

Wilson 1967). Furthermore, once established, individuals of these

perennial species are likely to persist and spread through vegetative

propagation alone. In plant species that have the option of

vegetative propagation, this type of asexual reproduction is often

favored when the population density is low, as during early

colonization of a newly invaded site (Abramson 1975). The success of

these two species as colonists is probably due to this propagation.

The Seahorse Key populations of both species achieve relatively high

levels of sexual reproduction and, presumably, outcrossing, so

effective pollination takes place even on the far island site.

Perhaps due to frequent gene flow from the nearby mainland,

neither Seahorse Key population has differentiated in floral biology

from Cedar Key or mainland populations. There has been no increase in

the potential for autogamy of the island populations. Had evolved

differences existed at the population level, objective comparisons of

natural female reproductive success would have been difficult. As it

is, however, any reproductive differences found among these

populations can be attributed to ecological effects that vary among

the populations, and not to evolved differences.


The flowering seasons of plants on oceanic islands are often

longer than their mainland counterparts (Carlquist 1966, 1974). The

explanation usually offered for this extended reproductive period is

the milder maritime climate found on islands. Alternative

explanations exist, however. The flowering phenologies of many

mainland plant communities are thought to be highly structured by

interspecific competition among plants for pollination (Heinrich

1975a, 1975b, 1976, Reader 1975, Thomson 1980).. The more diverse a

plant community is, the greater the probability of competition for

pollination among the flowering plant species. Plants flowering

before and after a particular species are competing with that species

for pollination. Individuals that flower out of synchrony with the

majority of conspecifics will be surrounded by flowers of other

species and will receive few visits, or will be unlikely to receive

conspecific pollen from and transmit pollen to a conspecific. This

risk becomes greater as the number of species competing for

pollination increases. Thus, a strong stabilizing selection on

flowering time is expected in diverse plant communities (Thomson

1980). Phenological studies of alpine (Mosquin 1976, Pleasants 1980,

Thomson 1980), northern temperate (Heinrich 1975b), and bog (Heinrich

1976, Reader 1975) plant communities have attributed the apparently

non-random flowering patterns to displacement of flowering times

through competition for pollination.

MacArthur and Wilson (1967) predicted and empirical studies have

demonstrated (see Carlquist '974 for references) reduced plant

diversity on islands. Such a reduction in potential competitors

should reduce the selective pressure on individuals for synchronous

flowering, which might lead to "ecological release" of flowering

season length. In this study, I did not determine overall plant

species richness for the study sites. I noted, however, that

Centrosema and Opuntia appeared to be the most abundant

insect-pollinated flowers at the far island site, Seahorse Key. The

study species were members of a more diverse assemblage of flowering

plants at the Cedar Key site, and the mainland site had many

conspicious, animal-pollinated species in flower in addition to

Centrosema. Opuntia flowered longer on Seahorse Key than on Cedar

Key. Individual plants from both sites had flowering times of the

same duration, so the differences in phenological patterns that I

observed were probably due to more among-plant variation at Seahorse

Key. I doubt, however, that this difference can be ascribed to a

reduced competitive pressure on Seahorse Key. Jackson (1966) has

shown that small microclimatic changes can significantly affect

flowering phenology. My study transect on Seahorse Key cut across a

greater variety of microhabitats than did the Cedar Key transect. For

example, some of the Seahorse Key Opuntia plants that I censused were

exposed to full sunlight; others were almost completely shaded. The

Cedar Key Opuntia were all at least partially shaded. The pattern for

Centrosema suggests a similar length of breeding season for the three

study populations, but the peak of flowering is slightly later on the

two island sites. Without clearer evidence, including greenhouse

experiments (Lack 1982), the phenological differences among the sites

must be considered a reflection of differences in the habitats, and

not a response to reduced pollinator competition on the island.

Again, frequent gene flow from the nearby mainland may be responsible

for the absence of conspicious divergence.

Pollinator Communities

Plants colonizing a new habitat risk leaving their usual

pollinators behind (Baker and Hurd 1968, Carlquist 1974). Often, the

usual pollinators are replaced by new species of visitors. Exotic

weeds that have invaded Canada did not bring their usual pollinators

with them, rather, those that require animal visitors are pollinated

by native, Canadian insects (Mulligan 1972). The introduced tomato

has left its pollinators in the Andes, despite this species' extensive

range in North America (Rick 1950). Self-sterile tomatoes planted in

California had 1-2% fruitset, while a similar density of plants

growing in Peru had 41-43% fruitset, a difference that Rick (1950)

attributed to lowered pollinator efficiency. Tomatoes are "buzz"

pollinated, which requires specialized behavior of visitors (Buchmann

1974, 1983, Thorpe and Estes 1975). Visitors unable to manipulate

such flowers correctly are ineffective at pollen transfer.

Even colonization over short distances can result in a shift in

pollinating species, as when a species colonizes a recently disturbed

site. Beattie (1969) found an increased number of visitors to Viola

at disturbed sites, but most were generalized foragers that deposited

a great deal of foreign pollen on Viola stigmas. Even though the

usual pollinators of a plant are often replaced by new species in a

newly-invaded habitat, the colonizing plant may suffer from a

reduction in pollination efficiency.

A plant colonizing a new habitat on the mainland may face a new,

less well-adapted array of visiting species, but colonization of an

island entails the additional problem of a depauperate fauna from

which to recruit new pollinators. On the Galapagos Islands, the

assemblage of pollinators includes only one bee species, seven

butterflies, twelve hawkmoths, and a few species of flies and beetles.

In contrast, these taxa and other pollinator taxa are very diverse on

nearby mainland areas of similar climate (Linsley 1966). Hummingbirds

were an important component of the pollinating community on the large

island, Trinidad, but diversity was much reduced on Tobago (Feinsinger

et al. 1982). Tobago populations of Erythrina pallida, a plant

species highly adapted to hummingbird visitation, had fewer species of

visitors and a lower visit frequency than did Trinidad populations

(Feinsinger et al. 1979).

Data presented here demonstrate both of the phenomena mentioned

above: a shift in pollinating species, and a reduction in pollinator

diversity and frequency on the distant island. Centrosema virginianum

growing on the mainland was visited almost exclusively by Bombus

pennsylvanicus, a pollinator to which the flower is well adapted.

Every Bombus visit operates the floral mechanism, such that

reproductive structures consistently contact the insect's back (for

details of this type of pollination, see Macior 1967). On Cedar Key,

Centrosema loses its Bombus pollinators and acquires a more diverse

assemblage of visitors. Of the several visitor species, only one, the

scoliid wasp Campsomeris quadrimaculatus, operates the floral

mechanism. This species does not appear to be a reliable visitor to

Centrosema; Campsomeris was observed visiting the Cedar Key

population during only one of the three breeding seasons for which I

made observations. During the 1980 and 1981 seasons, the only visits

on Cedar Key came from two genera of solitary bees that rarely

operated the flowers properly, although they did remove, and possibly

did transfer pollen, which these genera are known to utilize as a

larval food source (Linsley 1978, Mitchell 1962). Only one solitary

species,'Megachile brevis, was observed to visit Centrosema on the far

island, Seahorse Key. Opuntia stricta was visited by fewer species on

Seahorse Key than on Cedar Key. The Seahorse Key pollinator

assemblage was a subset of the Cedar Key pollinator community.

Eusocial species of pollinators were present on both the mainland

(Bombus pennsylvanicus) and Cedar Key (Apis mellifera) but are absent

on Seahorse Key. On Seahorse Key I saw a single Bombus foraging at

Opuntia on one occasion during the 1981 flowering season. Why are

colonial species missing from the Seahorse Key pollinator fauna?

I suggest that the energy required for colony maintainance

renders island sites unattractive to eusocial apids. Colonial species

require a rich and dependable food source. Observations on Agave

(Schaffer et al. 1979), Cassia (Johnson and Hubbell 1975). and

cultivated blackcurrant, raspberry, and strawberry (Free 1968) all

indicate that eusocial species of pollinators concentrate their

attention on rich floral resources and often ignore unprofitable,

scattered, or unpredictable (Real 1981) resources. The more highly

eusocial a species is, the more concentrated the food resources must

be to attract foragers (Schaffer et al. 1979). The degree of

coloniality exhibited by a species may affect that species' ability to

exist in a resource-limited environment. One hypothesis explaining

the rapid spread of the Africanized honeybee is the ability of this

genotype to exploit a lower level of floral resources than can the

European genotype (A.B. Bolten, personal communication). In European

colonies, many Apis mellifera remain at the nest when resources are

too low to warrant recruitment, unlike colonies of Africanized bees,

that may switch to foraging individually at times of resource

scarcity. Therefore, in marginal habitats with low or unpredictable

resource levels, European honeybees cannot maintain colonies, leaving

these habitats open to colonization by the Africanized bees.

My only measure of floral resource availability is a subjective

one. As I noted earlier, flower diversity appeared to be much lower

on Seahorse Key than at either of the other sites. At many times of

the year I noticed no animal-pollinated plants in bloom along my

Seahorse Key study transects, unlike the Cedar Key and mainland

transects. On Seahorse Key, resource shortages for flower visitors

may occur sporadically throughout the year. These shortages would

place severe stress on a colony of eusocial insects, which must feed

and maintain large numbers of developing young. Highly eusocial

species, such as Apis mellifera, are able to store resources to

provision against periods of resource scarcity, but if these periods

occur frequently such stored reserves may not be sufficient. A queen

of a colonial species reaching a marginal site such as Seahorse Key

would probably be unable to establish a successful colony.

The depauperate plant community on an island can affect the

pollinator community of the island by filtering out those pollinators,

such as most eusocial bees, that require a rich and consistent

resource base. Solitary flower-visitors apparently can exist on

Seahorse Key. Solitary bees can forage profitably at lower floral

densities than can colonial species (Johnson and Hubbell 1975,

Schaffer, et al. 1979). Males and females use flowers as nectar

sources, flower species that may be different from their pollen

sources (Linsley 1978). Females gather pollen to provision a larval

cell and make as many cells as pollen and nest-site availability allow

(Linsley 1978). Despite the abilities of some species .to utilize a

wide array of food sources, the life cycles, and especially emergence

times of solitary bees are often closely tied to their floral

resources (Linsley 1978, Linsley and MacSwain 1958, Hurd et. al.

1971). A severe flower shortage may limit population sizes and adult

life-spans of the solitary species but should not completely exclude

these species from a resource-poor habitat, such as Seahorse Key.

These data suggest that population densities of solitary

pollinators may be more limited at Seahorse Key than at the other

sites. Visit frequency to flowers is greatest at the beginning and

end of the flowering curves (Figures 4 and 5), and lowest when flowers

are most abundant. This drop in visit frequency with increasing

abundance is less pronounced or absent on Cedar Key and the mainland.

The increased variation in visit frequency on Seahorse Key may be

another indirect effect of the reduced plant diversity presumed to

occur there. If pollinators can utilize several plants as floral

resources, then the pollinator populations may be "buffered" from

extreme fluctuations in size between seasons. If, however, only a few

flowering species are available to a forager, then any changes in one

species' abundance or time of flowering may seriously affect the

population size of the pollinator for some time. A reduced pollinator

population can, consequently, cause much between and within season

variation in visit frequency to the plant, which may result in reduced

pollinator service and lower reproductive success for plant species

found on islands. In effect, a negative feedback cycle is generated;

a depauperate plant community reduces pollinator diversity and numbers

by excluding social species and limiting the population sizes of

solitary species. This may, in turn, inhibit establishment of new

plant species on an island since the appropriate visitors are less

likely to be available to pollinate the plant.

Reproductive Success and Pollinator Limitation

The perfect plant colonist would be completely autogamous;

reproductive success for an individual growing on an island would not

differ from that for a mainland conspecific, provided the physical

environment was equally suitable. Both species that I studied,

however, have low seed and/or fruitset unless they are visited by an

appropriate pollinator. Sexual reproduction in these species depends

Ao a large extent on an external agent. Another source of potential

variation in the sexual success of the different populations is the

habitat. I have already suggested that some environmental differences

occur among the sites (see Phenologies section). Are the significant

among-site differences in sexual success of these populations due to

variation in pollination service or environmental variation?

The flowers of Opuntia and Centrosema are hermaphroditic. Both

male and female reproductive success must be considered, as the two

functions may not always be correlated (Bertin 1982a, Charnov 1979,

Janzen 1977, Lloyd 1979, Willson 1979). My measure of male success

was the frequency of flowers receiving dye, and presumably pollen, at

various distances from a donor marked flower. Pollen transfer in both

species is insect-mediated. The differences in pollen-dispersal

patterns for both species must be a direct result of reduced

pollinator service to the island populations.

Female reproduction in plants, as in animals, is more

energetically expensive than male reproduction (see Charnov 1979,

Lloyd 1979, and Willson 1979 for references). It is, therefore, more

subject to environmental influences than is male reproduction

(Stephenson 1981). Resource availability is known to limit female

reproductive success in a number of species (Lee and Bazzaz 1982a,

Sutherland unpublished m.s., Stephenson 1979, 1982, Udovic and Aker

1981, Wyatt 1976). Artificial reduction (Stephenson 1980) or

enhancement (Lee and Bazzaz 1982a) of resources available to the

maternal parent has been shown to influence fruitset. Other species

appear to have abundant resources, though, and the number of fruits /

plant appears to be pollinator limited (Bertin 1982a, Bierzychudek

1981, Dafni and Ivri 1979, Willson and Schemske 1980). It is probably

incorrect to think of plants as being either pollinator or resource

limited; the two factors may both play a role in limiting female

reproductive success. Abortion of fruits or seeds is seldom random in

plants that appear to be primarily resource-limited (however, see

Casper and Wiens 1981), but may be selective (Janzen 1977). The

number of fertilized ovules in a fruit (Lee and Bazzaz 1982b) or even

the paternity of the developing fruit (Bookman 1983, Bertin 1982b,

Wyatt 1976) influences its probability of abortion. Under most

circumstances, pollination probably affects female reproductive

success of a plant, although numbers of fruits and seeds produced may

be too gross a measure to detect this effect.

Even using the unsubtle measures of fruit and seedset, female

reproductive success was detectably lower at the far island site than

elsewhere for both study species during at least one of the breeding

seasons for which I made observations. My data suggest pollinator

limitation was the cause of the lowered sexual success. Field

hand-crossings resulted in similar fruit and seedset for both study

populations of Opuntia stricta and resulted in similar fruitset among

the three Centrosema virginianum populations although there was a

trend in this latter species that suggested increasing resource

limitation along the mainland-far-island gradient. When both species

were given identical pollination treatments (Unvisited, Selfed and

Crossed), I found no significant variation among the study sites, yet

natural Centrosema fruitset was significantly lower on Seahorse Key

during the 1981 study season. Centrosema flowered earlier in 1982 and

both pollinator visit rates and fruitset showed a different pattern in

that season, with the mainland and far island both showing relatively

low reproductive success compared to Cedar Key. Fruits of Opuntia

growing on Seahorse Key contained a significantly lower and more

variable number of seeds / fruit than did fruits from Cedar Key during

the three breeding seasons. The pollinator limitation experiments,

where hand-crossing was done with control populations of open flowers,

demonstrated pollinator limitation for the Seahorse Key population of

Centrosema in 1981 and suggested pollinator limitation for Opuntia

growing on that island in 1983. Therefore, I conclude that Seahorse

Key plant populations suffer reductions in both male and female

reproductive functions, relative to Cedar Key or mainland populations,

and that this is due to reduced pollinator service.

A cautionary note must be added to this conclusion. Time

constraints and the relative scarcity of mainland populations of

theses species growing in a comparable habitat limited the number of

populations that I was able to examine. The differences that I

observed among these populations in visit rates and reproductive

success are assumed to be due to an "island effect" and, indeed, the

differences are all in the predicted directions. However, I was

unable to determine the interpopulation variation among mainland

plants of these study species. This weakness is mitigated -o some

extent by the duration of the study, but, just as temporal variation

was often a significant factor, spatial variation within a study site

could affect these results.

Evolutionary Considerations

Seahorse Key populations of Centrosema virginianum and Opuntia

stricta have not differentiated from the mainland populations in

phenology or in breeding biology. Because these populations are

separated by short distances, immigration rates to the island are

probably high enough to prevent evolutionary divergence of these

island populations. Nonetheless, even across such short distances,

there is a difference in reproductive success. My data implicate the

reduced abundance and reduced effectiveness of island pollinators as

the cause of this difference. If such differences can be detected

over distances as short as eight kilometers, the differences in

pollinator effectiveness and reproductive success of entomophilous

plant species may be much greater on distant, oceanic islands. This

study suggests that pollinator limitation can exert a strong selective

pressure on island plants, a pressure that eventually, through

filtration of inappropriate colonists, or through directional

selection operating on island plant populations, shapes the plant

communities of islands.


Abramson, W.G. 1975. Reproductive strategies in dewberries. Ecology
56. 721-726.

Augspurger, C.K. 1980. Mass flowering of a tropical shrub (Hybanthus
prunifolius): Influence on pollinator attraction and movement.
Evolution 34: 475-488.

Baker, H.G. 1955. Self-compatibility and establishment after
"long-distance" dispersal. Evolution 9: 347-348.

Baker, H.G. and P.D. Hurd, Jr. 1968. Intrafloral ecology. Annual
Review of Entomology 13: 385-414.

Beattie, A.J. 1969. Studies in the pollination ecology of Viola. I.
The pollen content of stigmatic cavities. Watsonia 7: 142-156.

Beattie, A.J. and D.C. Culver. 1979. Neighborhood size in Viola.
Evolution 33: 1226-1229.

Bertin, R.I. 1982a. Floral biology, hummingbird pollination and
fruit production of trumpet creeper (Campsis radicans). American
Journal of Botany 69: 122-134.

1982b. Paternity and fruit production in trumpet creeper
(Campsis radicans). American Naturalist 119: 694-709.

Bierzychudek, P. 1981. Pollinator limitation of plant reproductive
effort. American Naturalist 117: 838-840.

Bookman, S.S. 1983. Costs and benefits of flower abscission and
fruit abortion in Asclepias speciosa. Ecology 64: 264-273.

Buckmann, S.L. 1974. Buzz pollination of Cassia quiedondilla
(Leguminosae) by bees of the genera Centris and Melipona.
Bulletin of the Southern California Academy of Sciences 73:

S1983. Buzz pollination in angiosperms. in Handbook of
experimental pollination ecology. pp. 73-113. ed. C.E. Jones
and R.J. Little. Scientific and Academic Editions, New York,
New York. USA.

Carlquist, S. 1966. The biota of long-distance dispersal. IV.
Genetic systems in the floras of oceanic islands. Evolution 20:

1974. Island biology. Columbia University Press. New York,
New York. USA.

Casper, B.B. and D. Wiens. 1981. Fixed rates of random ovule
abortion in Cryptantha flava (Boraginaceae) and its possible
relation to seed dispersal. Ecology 62: 866-869.

Charnov, E.L. 1979. Simultaneous hermaphroditism and sexual
selection. Proceedings of the National Academy of Sciences 76:

Dafni, A. and Y. Ivri. 1979. Pollination ecology of and
hybridization between Orchis coriophora L. and 0. collina Sol.
ex Russ. (Orchidaceae) in Israel. New Phytologist 83: 181-87.

Dawkins, R. and M. Dawkins. 1973. Decisions and the uncertainty of
behaviour. Behaviour 45: 83-103.

Diamond, J.M. 1970. Ecological consequences of island colonization
by southwest Pacific birds: II. The effect of species diversity
on total population density. Proceedings of the National Academy
of Sciences 67: 1715-1721.

Faegri, K. and L. van der Pijl. 1979. The principles of
pollination ecology. Third edition. Pergamen Press. Oxford,

Feinsinger, P., Y.B. Linhart, L.A. Swarm, and J.A. Wolfe. 1979.
Aspects of the pollination biology of three Erythrina species on
Trinidad and Tobago. Annals of the Missouri Botanical Garden 66:

Feinsinger, P., J.A. Wolfe, and L.A. Swarm. 1982. Island ecology:
Reduced hummingbird diversity and the pollination biology of
plants, Trinidad and Tobago, West Indies. Ecology 63: 494-506.

Free, J.B. 1968. The foraging behavior of honeybees (Apis mellifera)
and bumblebees (Bombus spp.) on blackcurrent (Ribes nigrum),
raspberry (Rubus idaeus) and strawberry (Fragaria X ananassa)
flowers. Journal of Applied Ecology 5: 157-168.

Grant, V. 1975. Genetics of flowering plants. Columbia University
Press, New York, New York, USA.

Heinrich, B. 1975a. Energetics of pollination. Annual Review of
Ecology and Systematics 6: 139-170.

S1975b. Bee flowers: An hypothesis on flower variety and
blooming times. Evolution 29: 325-334-

S1976. Flowering phenologies: Bog, woodland, and disturbed
habitats. Ecology 57: 890-899.

Hurd, P.D.,Jr., E.G. Linsley and T.W. Witaker. 1971. Squash and
gourd bees (Peponapis, Xenoglossa) and the origin of the
cultivated Cucurbita. Evolution 25: 218-234.

Jackson, M.T. 1966. Effects of microclimate on spring flowering
phenology. Ecology 47: 407-415.

Janzen, D.H. 1977. A note on optimal mate selection by plants.
American Naturalist 111: 365-371.

Johnson, L.K. and S.P. Hubbell 1975. Contrasting foraging
strategies and coexistance of two bee species on a single
resource. Ecology 56: 1398-1406.

Kevan, P.G. 1972. Insect pollination of high Arctic Flowers.
Journal of Ecology 60: 831-848.

Lack, A.J. 1982. The ecology of flowers of chalk grassland and their
insect pollinators. Journal of Ecology 70: 771-790.

Lee, T.D. and F.A. Bazzaz. 1982a. Regulation of fruit and seed
production in an annual legume, Cassia fasciculata. Ecology 63:

1982b. Regulation of fruit maturation pattern in an annual
legume, Cassia fasciculata. Ecology 63: 1374-1388.

Lehner, P.N. 1979. Handbook of ethological methods. Garland STPM
Press, New York, New York, USA.

Linhart, Y. and F. Feinsinger. 1980. Plant-hummingbird
interactions: Effect of island size and degree of specialization
on pollination. Journal of Ecology 68: 745-760.

Linsley, E.G. 1966. Pollinating insects of the Galapagos Islands.
Pages 225-232 in R.I. Bowman, editor. The galapagos.
University of California Press, Berkeley, California, USA.

1978. Temporal patterns of flower visitation by solitary
bees, with particular reference to the southwestern United
States. Journal of the Kansas Entomological Society 51:

Linsley, E.G. and S. MacSwain 1958. The significance of floral
constancy among bees of the genus Diadasia (Hymenoptera,
Anthrophoridae). Evolution 12: 219-223.

Linsley, E.G. C.M. Rick, and S.G. Stephens. 1966. Observations on
the floral relationships of the Galapagos carpenter bee.
Pan-Pacific Entomologist A2: 1-18.

Lloyd, D.G. 1979. Parental strategies of angiosperms. New Zealand
Journal of Botany 17: 595-606.

MacArthur, R.H. and E.O. Wilson. 1967. The theory of island
biogeography. Princeton Univ. Press, Princeton, New Jersey,

Macior, L.W. 1967. Pollen foraging behavior of Bombus in relation to
pollination of nototribic flowers. American Journal of Botany
54: 359-364.

Moldenke, A.R. 1975. Niche specialization and species diversity
along a California transect. Oecologia 21: 219-242.

Mosquin, T. 1971. Competition for pollinators as a stimulus for the
evolution of flowering time. Oikos 22: 398-402.

Mulligan, G.A. 1972. Autogamy, allogamy, and pollination in some
Canadian weeds. Canadian Journal of Botany 50: 1767-1771.

Pleasants, J.M. 1980. Competition for bumblebee pollinators in Rocky
Mountain plant communities. Ecology 61: 1446-1459.

Reader, R.J. 1975. Competitive relationships of some bog ericads for
major insect pollinators. Canadian Journal of Botany 53:

Real, L.A. 1981. Uncertainty and pollinator-plant interactions: The
foraging behavior of bees and wasps on artificial flowers.
Ecology 62: 20-26.

Rick, C.M. 1950. Pollination relationships of Lycopersicon
esculentum in native and foreign regions. Evolution 4: 110-122.

1966. Some plant-animal relations on the Galapagos Islands.
Pages 215-224. in R.I. Bowman, editor. The galapagos.
University of California Press, Berkeley, California, USA.

Rympa, R.B. 1952. Vegetative reproduction in Opuntia leptocaulis.
Texas Journal of Science 1: 92-94.

Schall, B.A. 1978. Density dependent foraging on Liatris
pycnostachya. Evolution 32: 452-454.

Schaffer, W.M., D.B. Jensen, L.E. Hobbs, J. Gurevitch, J.R. Todd,
and M.V. Schaffer. 1970. Competition, foraging energetic, and
the cost of sociality in three species of bees. Ecology 60:

Schemske, D.W., M.F. Willson, M.N. Melampy, L.J. Miller, L.
Verner, K.M. Schemske, and L.B. Best. 1978. Flowering ecology
of some spring woodland herbs. Ecology 59: 351-366.

Schoen, D.J. 1982. The breeding system of Gilia achilleifolia:
Variation in floral characteristics and outcrossing rates.
Evolution 36: 357-360.

Siegel, S. 1956. Ncnparametric statistics for the behavorial
sciences. McGraw-Hill, New York, New York, USA..

Simberloff, D.S. and E.O. Wilson 1969. Experimental zoogeography of
empty islands. Ecology 50: 278-295.

Simpson, E.H. 1949. Measurement of diversity. Nature 163: 688.

Sokal, R.R. and F.J. Rohlf. 1981. Biometry. 2nd edition. W.H.
Freeman and Company, San Francisco, California, USA.

Solbrig, 0.T. and R.C. Rollins. 1977. The evolution of autogamy in
species of the mustard genus Leavenworthia. Evolution 31:

Stephenson, A.G. 1979. An evolutionary examination of the floral
display of Catalpa speciosa (Bignoniaceae). Evolution 33:

1980. Fruit set, herbivory, fruit reduction, and the fruiting
strategy of Catalpa speciosa (Bignoniaceae). Ecology 61: 57-64.

1981. Flower and fruit abortion: Proximate causes and
ultimate functions. Annual Review of Ecology and Systematics 12:

Stockhouse, R.E. 1976. A new method for studying pollen dispersal
using micronized fluorescent dusts. American Midland Naturalist
96: 241-245.

Thomson, J.D. 1980. Skewed flowering distributions and pollinator
attraction. Ecology 61: 572-579.

1981. Spatial and temporal components of resource
assessment by flower-feeding insects. Journal of Animal Ecology
50: 49-59.

Thorpe, R.W. and J.R. Estes 1975. Intrafloral behavior of bees on
flowers of Cassia fasciculata. Journal of the Kansas
Entomological Society 48: 175-184.

Udovic, D. and C. Aker. 1981. Fruit abortion and regulation of
fruit numbers in Yucca whipplei. Oecologia 49: 245-248.

Waser, N.M. and M.V. Price. 1982. A comparison of pollen and
fluorescent dye carry-over by natural pollinators of Ipomopsis
aggregate (Polemoniaceae). Ecology 63: 1168-1172.

Whitehead, D.R. 1969. Wind pollination in the angiosperms:
Evolutionary and environmental considerations. Evolution 2:

Willson, M.F. 1979. Sexual selection in plants. American Naturalist
113: 777-790.


Willson, M.F. and D.W. Schemske. 1980. Pollinator limitation,
fruit production, and floral display in pawpaw (Asimina triloba).
Bulletin of the Torrey Botanical Club 107: 401-408.

Wyatt, R. 1976. Pollination and fruit-set in Asclepias: A
reappraisal. American Journal of Botany 63: 845-851.

Zimmerman, M. 1980. Reproduction in Polemonium: Competition for
pollinators. Ecology 61: 497-501.


Edwin Eugene Spears, Jr.

Edwin Eugene Spears, Jr. was born September 27, 1953, in

Asheville, North Carolina, to Edwin E. and Sue R. Spears. He was

followed at regular, non-random intervals by a brother, Jack R., and a

sister, Susan A. Spears. He completed his undergraduate degree in

biology at the University of North Carolina at Asheville in 1975, his

Master of Science degree in zoology at the University of Florida in

1978 and his PhD in zoology at the University of Florida in 1983. His

experiences in pollination biology began at an early age when he

captured what he thought was a very small hummingbird with a glass jar

as it was visiting a cultivated Malavaceae growing in his family's

backyard. It was not a hummingbird, but a day-flying sphinx moth,

that rapidly sufficated in the small jar because of its high metabolic

rate. This early experience left Gene with two impressions that

remain with him to this day: 1) the ephemeral, transitory nature of

beauty, and 2) the pleasure of sneaking up on bugs and catching them

in glass jars.

I certify that I have read this study and that in my opinion
it conforms to acceptable standards of scholarly presentation and
is fully adequate in scope and quality, as a dissertation for the
degree of Doctor of Philosophy.

Peter Feinsinger, Chairman
Associate Professor of Zoology

I certify that I have read this study and that in my opinion
it conforms to acceptable standards of scholarly presentation and
is fully adequate in scope and quality, as a dissertation for the
degree of Doctor of Philosophy.

Thomas C. Emmel
Professor of Zoology

I certify that I have read this study and that in my opinion
it conforms to acceptable standards of scholarly presentation and
is fully adequate in scope and quality, as a dissertation for the
degree of Doctor of Philosophy.

Carmine A. Lanciani
Professor of Zoology

I certify that I have read this study and that in my opinion
it conforms to acceptable standards of scholarly presentation and
is fully adequate in scope and quality, as a dissertation for the
degree of Doctor of Philosophy. //

ana G.Griffin
Professor of Botany

I certify that I have read this study and that in my opinion
it conforms to acceptable standards of scholarly presentation and
is fully adequate in scope and quality, as a dissertation for the
degree of Doctor of Philosophy.


Martha L. Crump
Associate Professor of Zoology

This dissertation was submitted to the Graduate Faculty of the
Department of Zoology in the College of Liberal Arts and Sciences
and to the Graduate School, and was accepted for partial fulfillment
of the requirements of the degree of Doctor of Philosophy

Dean for Graduate Studies and Research

University of Florida Home Page
© 2004 - 2010 University of Florida George A. Smathers Libraries.
All rights reserved.

Acceptable Use, Copyright, and Disclaimer Statement
Last updated October 10, 2010 - - mvs