MASTING AND INSECT POLLINATION

IN THE DIOECIOUS ALPINE HERB ACIPHYLLA

A thesis submitted in partial fulfilment of the requirements for the

Degree

of Master of Science in Biological Science

in the University of Canterbury

by Laura May Young

University of Canterbury

2006
Table of Contents

List of Tables ............................................................................................................................1

List of Figures ...........................................................................................................................4
Abstract .....................................................................................................................................6
Chapter 1 - Introduction .........................................................................................................8
1.1 Introduction to the genus Aciphylla .........................................................................8
1.1.2 Aciphylla – a unique genus in the family Apiaceae .................................8
1.2 Historical background ............................................................................................10
1.3 Taxonomy ..............................................................................................................10
1.4 Why study the reproductive ecology of Aciphylla? ...............................................11
1.5 Dioecy and sexual dimorphism ..............................................................................11
1.6 Insect pollination ....................................................................................................13
1.6.1 Insect pollination and the importance of floral display ......................................13
1.7 Mast seeding ..........................................................................................................14
1.8 Why do some plant species mast when such high costs are involved? .................16
1.8.1 Proximate versus ultimate reasons for masting ......................................16
1.8.2 Non-evolved hypothesis to explain mast-seeding ...................................16
1.8.3 Evolutionary hypotheses to explain mast-seeding ..................................17
1.8.3.1 Predator satiation .....................................................................18
1.8.3.2 Wind-pollination ......................................................................18
1.8.3.3 Animal fruit dispersal and large seed size ...............................19
1.8.3.4 Accessory costs of reproduction ..............................................19
1.8.3.5 The animal pollination hypothesis ...........................................19
1.9 The bizarre reproductive ecology of Aciphylla ......................................................20
1.10 Research Aims .....................................................................................................21

Chapter 2- Does floral display size affect reproductive success in Aciphylla? .................23
2.1 Introduction ............................................................................................................23
2.1.1 The importance of floral display size at multiple levels in Aciphylla .....23
2.1.2 Measuring floral display at the inflorescence and plant level .................25
2.1.2.1 The importance of resource and pollen limitation in Aciphylla
– does female success depend on individual reproductive effort? .......27

i
 2.1.3 Measuring population-level flowering effort – pollinator efficiency .....29
2.1.4 Objectives ...............................................................................................32
2.1.4.1 Part (A): Floral display at the level of the individual plant .................32
2.1.4.2 Part (B): Population level flowering effort - male to female
flowering densities ...........................................................................................33
2.2 Site and species selection .......................................................................................35
2.2.1 Study species ...........................................................................................37
2.2.1.1 Aciphylla aurea W.R.B. Oliver ................................................37
2.2.1.2 Aciphylla scott-thomsonii Cockayne & Allan ..........................38
2.2.1.3 Aciphylla ‘lewis’ ......................................................................38
2.2.2 Field work ...............................................................................................40
2.3 Methods ..................................................................................................................40
2.3.1 Part (A): Floral display at the level of the individual plant ....................40
2.3.2 Part (B): Population level flowering effort - male to female flowering
densities ...........................................................................................................44
2.3.3 Pollen limitation ......................................................................................46
2.3.4 Other field measurements .......................................................................47
2.3.5 Lab work .................................................................................................47
2.3.5.1 Measuring reproductive success – counting seeds ...................47
2.3.5.2 Germination experiments .........................................................49
2.3.6 Statistical Analyses .................................................................................50
2.4 Results ....................................................................................................................51
2.4 Part (A): Effects of insect-exclusion bags, plant-level floral display, hand-
pollination, and local pollen on seed set ......................................................................51
2.4.1 Mean seed set rates of control inflorescences at all sites ........................51
2.4.2 Effects of bagging on seed set rate .........................................................52
2.4.3 Effects of floral display on seed set ........................................................53
2.4.3.1 Analyses by site .......................................................................53
2.4.3.2 Analyses by treatment and graphical representation to
compare sites ........................................................................................58
2.4 Part (B): Effects of female isolation, hand-pollination and pollen on seed set ......66
2.4.4 Analyses by site ......................................................................................66
2.4.5 Seed germination ....................................................................................70
2.4.6 Summary .................................................................................................70

ii
 2.5 Discussion ..............................................................................................................72
2.5.1 Part (A): Floral display at the plant level ................................................72
2.5.1.1 The importance of insect versus wind as pollen vectors in
Aciphylla ..............................................................................................72
2.5.1.2 The importance of plant-level floral display (and how it is
measured) on reproductive success ......................................................73
2.5.2 Part (B): Population level flowering effort .............................................75
2.5.3 Conclusions .............................................................................................77

Chapter 3 - The effects of floral display and population structure on insect visitation
rates and how this relates to reproductive success .............................................................79
3.1 Introduction ............................................................................................................79
3.1.1 Aims ........................................................................................................80
3.2. Methods ................................................................................................................81
3.2.1 Collecting insects ....................................................................................81
3.2.2 Insect visitation surveys ..........................................................................81
3.2.3 Insect visitation to male inflorescences ..................................................82
3.2.4 Insect visitation rates in response to floral display .................................82
3.2.5 Statistical Analyses .................................................................................83
3.3 Results ....................................................................................................................84
3.3.1 What insects are visiting Aciphylla inflorescences? ...............................84
3.3.2 Are insects attracted to larger floral displays? ........................................85
3.3.2.1 Insect visitation rates and flower removal manipulation .........85
3.3.2.2 Insect visitation and inflorescence height ................................87
3.3.2.3 Insect visitation and inflorescence flowering length ...............89
3.3.3.4 Does bract removal affect insect visitation rates? ....................91
3.3.3.5 Insect visitation and number of inflorescences per plant .........92
3.3.4 Insect visitation to isolated flowering female plants ...............................94
3.3.5 Insect visits to male vs. female inflorescences .......................................94
3.3.6 How do these results relate to results on seed set rate in chapter 2? ......96
3.4 Discussion ..............................................................................................................98
3.4.1 The insect fauna of Aciphylla: a generalist or specialist pollination
system? .............................................................................................................98

iii
 3.4.2 Why are insects more abundant on male inflorescences, and are
females receiving sufficient visits? ..................................................................99
3.4.3 How do the effects of floral display on insect visitation rates relate to
seed set rates? .................................................................................................101
3.4.3.1 Insect visitation and floral display .........................................101
3.4.3.2 Are isolated flowering females reproductively successful? ...102
3.4.4 Conclusions ...........................................................................................102

CHAPTER 4 – Sex-ratios, dioecy and sexual dimorphism in Aciphylla .........................104

4.1 Introduction ..........................................................................................................104
4.1.1 Objectives .............................................................................................107
4.2 Methods ................................................................................................................107
4.2.1 Measuring population plant and inflorescence sex ratios .....................107
4.2.2 Measuring sexual dimorphism in plant and inflorescence traits ...........108
4.3 Results ..................................................................................................................109
4.3.1 Sex ratios ...............................................................................................109
4.3.2 Sexual dimorphism in plant and inflorescence traits ............................111
4.4 Discussion ............................................................................................................114
4.4.1 Male-biased sex ratio in Aciphylla .......................................................114
4.4.2 Apparent vs. real sex ratio ....................................................................114
4.4.3 Other sexually dimorphic traits in Aciphylla ........................................118
4.5 Conclusions ..........................................................................................................119

CHAPTER 5 – Discussion ...................................................................................................120

5.1 Insect pollination, masting and reproductive success in Aciphylla ......................120
5.2 The importance of synchronous flowering and display size ................................121
5.3 Pollen- and resource-limitation and masting .......................................................122
5.4 The effects of sexual dimorphism on reproductive success in Aciphylla ............125
5.5 Further research on the reproductive ecology of Aciphylla .................................126
5.6 Conclusions ..........................................................................................................127

Acknowledgements ..............................................................................................................129
References .............................................................................................................................130
Appendix 1 - Floral morphology in Aciphylla scott-thomsonii .........................................150
Appendix 2 – Insects collected from Aciphylla flowers .....................................................153

iv
List of Tables

Table 2.1 Outline of each predictive term used in this study and the hypothesised directions
of the relationships between the each floral display term and seed set rate .........................34

Table 2.2 Summary of site information for the three study areas used in this research .......36

Table 2.3 Quantitative information for floral display experimental manipulations performed
in the field .............................................................................................................................43

Table 2.4 Quantitative plot information for isolated-female experiment ............................ 46

Table 2.5 Mean seed set rates of bagged versus control inflorescences.............................. 53

Table 2.6 Analysis of Deviance Table for A. aurea at Burkes Pass showing variation in
seed set rate as a response to five predictive terms.............................................................. 54

Table 2.7 Analysis of Deviance Table for A. aurea at Hakataramea Pass showing variation
in seed set rate as a response to five predictive terms.......................................................... 55

Table 2.8 Analysis of Deviance Table for A. scott-thomsonii at Hakataramea Pass showing
variation in seed set rate as a response to five predictive terms .......................................... 56

Table 2.9 Analysis of Deviance Table for A. ‘lewis’ at Lewis Pass showing variation in
seed set rate as a response to four predictive terms .............................................................57

Table 2.10 The effects of removing inflorescence floral bracts on seed set rate................. 57

Table 2.11 Analysis of Deviance Table for isolated female experiment in A. aurea at
Burkes Pass (site 2).............................................................................................................. 66

Table 2.12 Analysis of Deviance Table for isolated female experiment in A. aurea at
Hakataramea Pass (site 2) .................................................................................................... 67

1
Table 2.13 Hypothesised directions of relationships between seed set rate and predictors,
and the actual outcomes from the GLM summary tables .................................................... 71

Table 3.1 Mean insect visitation to female Aciphylla inflorescences with 0, 40 and 80
percent of flowers removed ................................................................................................. 86

Table 3.2 Mean relative insect visitation to female Aciphylla inflorescences with 0, 40 and
80 per cent of flowers removed ........................................................................................... 87

Table 3.3 Insect visitor numbers to female Aciphylla inflorescences in response to variation
in inflorescence height ......................................................................................................... 88

Table 3.4 Insect visitor numbers in response to variation in flowering length.................... 90

Table 3.5 Mean insect visitor numbers per female inflorescence per instantaneous
observation in response to bract removal............................................................................. 92

Table 3.6 Mean insect visitor numbers (per instantaneous observation) on single female
inflorescence on plants with 1,2,3,4 or 8 inflorescences ..................................................... 93

Table 3.7 The mean number of insects per inflorescence per observation in each taxonomic
category on male and female Aciphylla inflorescences according to site and species ........ 95

Table 3.8 Summary of the results from chapters 2 and 3 showing the effects of floral
display at both the individual (flower removal experiment) and population (isolated female
experiment) level on seed set rate and insect visitation rates at all ..................................... 97

Table 4.1 Percentage of male plants for four populations of flowering Aciphylla in
Canterbury ......................................................................................................................... 110

Table 4.2 Mean plant dimensions for male and female Aciphylla plants at three populations
in Canterbury, including mean plant height, mean plant area, and number of rosettes per
plant ................................................................................................................................... 112

2
Table 4.3 Mean inflorescence height and flowering length of male and female Aciphylla
plants at three sites in Canterbury...................................................................................... 112

Table 4.4 A comparison of single-year plant sex-ratio data for six species of dioecious
Aciphylla in the South Island of New Zealand between this study and data published in
Webb and Lloyd, (1980) .................................................................................................... 113

3
List of Figures

Figure 1.1 A rosette of Aciphylla aurea (golden spaniard) showing compound leaves with
spiny tips and a young, developing inflorescence with sharp bracts in the centre of the
rosette .....................................................................................................................................9

Figure 2.1 Aciphylla glaucescens provides a clear example of the variation in floral display
size within an inflorescence during a high-flowering year ..................................................24

Figure 2.2 Expected effects of pollen and resource limitation when supplemental pollen is
added ....................................................................................................................................29

Figure 2.3 Map of the South Island of New Zealand showing the three study site locations
within the wider Canterbury region .....................................................................................35

Figure 2.4 Mid-section of inflorescence stalk of A. scott-thomsonii showing detail of

inflorescence ........................................................................................................................39

Fig 2.5 Diagrammatic representation of isolated-female experimental design, to investigate

the effects of male floral densities on female reproductive success ....................................45

Figure 2.6 Intact developing fruit, or schizocarp, (left) consisting of 2 winged mericarps,
each containing one seed .....................................................................................................48

Fig 2.7 The natural variation in seed set rates amongst sites using control inflorescences .52

Figure 2.8 Box and whisker plots for all sites showing the variation in seed set rates for
each level of flower removal, pollination and bag treatment ...............................................59

Figure 2.9 Interaction plots showing the effects that simultaneous pollen and resource
limitation following flower removal and hand pollination had on seed set rates ................ 61

Figure 2.10 The effect of inflorescence flowering length on seed set rate for hand vs.
natural pollination treatment at all sites ............................................................................... 63

4
Figure 2.11 The effect of inflorescence height on seed set rate at all sites.......................... 65

Figure 2.12 The effect of local pollen abundance (measured with the Distance-Weighted
local Pollen Index, or DWPI) on seed set rate at Burkes and Hakataramea Passes ............ 68

Figure 2.13 The effects of increasing distance to the nearest male inflorescence (i.e. pollen
source) on seed set rate at isolated female sites, where male to female flowering ratios were
experimentally manipulated .................................................................................................69

Figure 3.1 Total number of insect visitors to female Aciphylla inflorescences (summed over
all observation days) in response to inflorescence height ....................................................89

Figure 3.2 Total number of insect visitors to female Aciphylla inflorescences (summed over
all observation days) in response to flowering length .........................................................91

Figure 3.3 Total insect visitor numbers (per instantaneous observation) on single female
inflorescence on plants with 1,2,3,4 or 8 inflorescences .....................................................93

Figure 4.1 Male and female plants of Aciphylla aurea showing variation in display size and
inflorescence number of males and females present within a population ......................... 105

Figure 4.2 Proportion of male and female flowering plants in each population of Aciphylla
species ................................................................................................................................109

Figure 4.3 Proportion of male and female inflorescences, measured as the percentage of
male and percentage of female inflorescences (out of a total of n inflorescences) ........... 110

Figure 4.4 Mean number of inflorescences per male and female plant for all flowering
plants counted within each population ...............................................................................111

5
ABSTRACT

Aciphylla species (wild spaniard/speargrass) are an iconic component of the Australasian

high country flora, but their reproductive system is enigmatic. They are insect-pollinated
dioecious mast seeders (synchronous highly variable seed production), which seems
maladaptive. The resource supply to pollinators is highly variable, yet dioecious plants are
dependent on pollinators, and dioecious masting requires male and female plants to flower
synchronously. Floral display in Aciphylla is relatively large, with tall inflorescences
bearing thousands of flowers, suggesting that plants would not have the resources to
produce such large stalks every year. But why do they have such huge inflorescences in the
first place? I tested whether pollinator attraction is providing an economy of scale which
favours intermittent production of very large inflorescences, by manipulating floral display
size during a high-flowering year and measuring insect visitation rates and seed set (female
reproductive success). Using space-for-time substitution and selective removal of male
inflorescences, I also tested whether female seed set was affected by distance to flowering
male plants (i.e. changes in local pollen availability) to see if flowering asynchrony would
reduce pollination success. Bags were used to exclude pollination by insects and test for
wind pollination, and hand pollination was done to test for pollen limitation. Insect surveys
suggest that Aciphylla has a generalist pollination system (to avoid satiating a specialist
pollinator during ‘mast’ years’). Male inflorescences received significantly more visits than
females, and some seeds were set inside bags (although only 20-30%), suggesting wind
pollination may occur at low levels. Seed set rate was higher for taller inflorescences with
greater flowering length in A. aurea but tall inflorescences with excess flowers led to a
decrease in seed set rates in A. scott-thomsonii. Hand pollination significantly increased
seed set rates although these effects were not as large as expected (e.g. 10% increases from
natural to hand-pollinated inflorescences were typical). There was no evidence for resource
limitation in any species. Female plants in dense flowering populations had higher seed set
rates, and individual floral display size in females was particularly important when females
were ‘isolated’ from males. Insect visitation rates were generally higher on inflorescences
with a larger floral display, suggesting that display size is important for pollinator
attraction. Overall, these results suggest that the pollinator-attraction benefits of such a
large floral display (at both the plant and population level) are possibly providing an
economy of scale, although the relative effects are small.

6
“Few plants are more spectacular than a large
speargrass in flower.
A tussocky hillside where hundreds stand like
great golden candles is a sight not easily
forgotten”.
Moore and Irwin, 1978

7
Chapter 1 - INTRODUCTION

1.1 Introduction to the genus Aciphylla

Aciphylla is a long-lived herbaceous genus in the carrot family Apiaceae, (subfamily

Apioideae). There are over 40 species of Aciphylla endemic to New Zealand (Allan, 1982)
and three endemic to Australia (Pickering & Hill, 2002). Their distribution spans from
coastal to mostly montane and sub-alpine. Aciphylla is one of the more speciose genera of
New Zealand flowering plants (Dawson & Le Comte, 1978). They are an iconic component
of the New Zealand high country grassland/shrubland partly because when in flower, the
inflorescences stand out like “great golden candles” in the tussocky hillsides (Moore &
Irwin, 1978).

1.1.2 Aciphylla – a unique genus in the family Apiaceae

Wardle, (2002) describes Aciphylla as being “the most bizarre plants of mountain
grasslands and herbfields”. Not only is Aciphylla a unique component of the New Zealand
alpine flora, but they are also unusual among the New Zealand (and global) Apiaceae
family. This is because many species have rigid leaf- and bract-segments tipped by needle-
sharp spines, and an unusually large number of reduced compound umbels aggregated into
narrow elongate inflorescences (Dawson, 1971). Also atypical of the family is that they are
dioecious (separate sexes on different plants) (Webb, 1981) and mast-seeding (highly
variable and synchronous reproduction) (Mark, 1970), however, resembling most other
Apiaceae genera, they are pollinated by insects.

Aciphylla has rosettes of rigid spine-tipped leaves and hence are commonly known as ‘wild
spaniard’ or ‘speargrass’ (Mitchell et al., 1998), or in maori, ‘taramea’, literally translated
to mean “sharp leaves”. Because of this, Leonard Cockayne in 1910, dubbed one of the
larger species Aciphylla colensoi, “a most formidable plant with bayonet-like leaves a yard
long” and quoted “early explorers understandably regarded these as enemies, a solid
phalanx of speargrass constituting well-armed opponents, impassable for man or beast”
(Moore & Irwin, 1978). All species have at least one rosette of spiny leaves radiating from
a crown or rootstock with a deeply descending taproot. Leaves are usually all basal, mostly

8
compound, and with close-set imbricated sheaths (Allan, 1982), with the lower part of the
fleshy sheath often buried beneath the soil surface (possibly a way for this nutrient-rich
resource to escape this being eaten by moas, kea, takahe (Atkinson & Greenwood, 1989)
and kakapo (Powlesland et al., 2006)). The genus is exceptional because 16 of the species
have leaves modified to form groups of rigid needle-sharp spines 0.4-1.5 m in length. The
spines are dense and evenly spaced to form a hemispherical tussock-like form (Fig. 1.1).
The leaves are tough and fibrous and also have spiny stipules which together form a collar
of spines around the base of the leaf rosette (Atkinson & Greenwood, 1989). They have one
or more inflorescences of compound, bracted umbels. Bracts have a long sheath with
pinnae and are also extremely spiny. Anyone who has ever been tramping or hunting in the
high country will know all about the injurious consequences that follow walking into an
Aciphylla plant (see below).

Fig. 1.1 A rosette of Aciphylla aurea (Golden spaniard) showing compound

leaves with spiny tips and a young, developing inflorescence with sharp bracts
in the centre of the rosette.

9
1.2 Historical background

Aciphylla is common in open mountain grasslands, and particularly following fire they can
be abundant (Mark, 1970). Pollen and charcoal analysis from south-central South Island
Holocene age deposits indicate that Aciphylla repeatedly increased its distribution and
abundance in southern dryland areas such as the Mackenzie Basin throughout the
Holocene. The genus is an important recolonising species in response to drought and local
fires which naturally occurred during this time (McGlone & Moar, 1998). Around the
period from 800 years BP to 1860 AD, burnt-out Nothofagus forests and Halocarpus scrub
were replaced by Aciphylla-rich tussock grassland, however, from 1860 onwards, Aciphylla
began disappearing, while grasses and Rumex acetosella continually increased with
European pastoralism (McGlone & Moar, 1998). Additionally, European settlers burnt
large areas of Aciphylla to travel through some mountain passes (Moore & Irwin, 1978).
The current distribution of Aciphylla populations in the dryland valleys such as the
Mackenzie Basin is likely to be influenced by a combination of fires and droughts of the
late Holocene, farming and both pastoral grazing and grazing by introduced mammalian
pests.

1.3 Taxonomy

There have been a number of attempts at dividing the genus into groups based on various
aspects of their morphology, size, and growth habit, amongst other features (e.g. Oliver,
1956; Dawson & Le Comte, 1978). Most recently, D. Glenny (pers. comm. 2005) has been
undertaking an extensive, detailed taxonomic revision of the entire genus and has come up
with at least ten groups based on various morphological, physiological and evolutionary
criteria. Although most authors quote the genus Aciphylla as having over 40 species, D.
Glenny’s revision suggests that there are more likely to be approximately 27 species with a
number of varieties. Most groups consist of the smaller and higher-altitude species
(previously Paniculatae – (Oliver, 1956), referring to inflorescences of “compound umbels
on approximate spreading peduncles at the stem apex”). The remaining few groups
constitute the larger and more evolutionarily recent varieties (previously Elongatae –
inflorescences of compound, spiny-bracted umbels arranged along an elongate extension of
the central stem (Oliver, 1956)). The research in this thesis focuses on three of the larger

10
varieties of Aciphylla (within Elongatae), with elongated inflorescences. Hereafter, when
commenting on features and characteristics generally portrayed by the larger, lower-
altitudinal (or montane) species, for simplicity, the term Aciphylla will be used (although it
is uncertain how specific features mentioned apply to the genus as a whole).

1.4 Why study the reproductive ecology of Aciphylla?

Aciphylla has a unique reproductive system in that they are dioecious, insect-pollinated and
mast seeding. This is not a set of components that are usually found associated together
within one particular breeding system (Proctor et al., 1996). As well as this, they are
sexually dimorphic and have been reported to have a male-biased sex ratio (Webb & Lloyd,
1980). Despite having a fascinating ecological background, little is known about
Aciphylla’s breeding system, hence the need for further investigation into their interesting
and bizarre reproductive ecology. The sections below break down each of the
aforementioned components of the breeding system and introduce some of the details about
their biological significance. The final section will thread these individual components
together in relation to Aciphylla in order to explain why this reproductive system is
interesting and unique.

1.5 Dioecy and sexual dimorphism

Dioecy, in which separate individuals perform male and female functions, is a relatively
rare breeding system among the angiosperms (Vamosi & Vamosi, 2004). Dioecy occurs in
about 10% of the world’s flora and is especially common in isolated island floras of New
Zealand and Hawaii (Webb & Kelly, 1993) and amongst tree species in the tropics (Bawa,
1974). The causes of this high frequency of dioecism in the flora of New Zealand (and
islands in general) are not clear. Webb & Kelly, (1993) suggested that the unspecialized
pollinating fauna may have encouraged unisexuality as a means of preventing selfing. (See
Sakai & Weller, (1999) for an in-depth review on the evolution and maintenance of
dioecy). Most dioecious species are insect-pollinated and have on average, smaller flowers
than hermaphrodites that are often green or white and inconspicuous with unspecialized
pollinators (Proctor et al., 1996). Aciphylla fits this description (Pickering, 2000).

11
Many dioecious species, particularly those in largely hermaphrodite families, have vestigial
organs of the others sex present in the flower and the sterile anthers in the females may be
part of the attractant in female flowers; some provide sterile pollen, although normally
nectar is the attractant as most female flowers do not produce pollen (Proctor et al., 1996).
It should be noted that ‘leakage’ in sexuality (“leaky dioecism”) has been reported for some
Aciphylla species. In these cases, female inflorescences may have a small number of
terminal male umbels (D. Glenny, pers. comm. 2005) although it is uncertain as to whether
these are functional and whether this can lead to geitonogamous selfing (Newstrom &
Robertson, 2005). It must be pointed out that this variation in sexuality was not commonly
encountered throughout this research.

Male flowers of dioecious species are often larger and showier than female flowers,
produced in greater quantity, and in larger inflorescences (Bawa, 1980), and therefore
exhibit sexual dimorphism in sex characters (Webb & Lloyd, 1980; Delph, 1999). This
often means that male flowers are more likely to attract the attention of insect pollinators
from a distance so that insects may visit them before the females (Proctor et al., 1996)
(essential for successful pollination). Some Aciphylla species have also been shown to
exhibit these traits (e.g. in Australia: Aciphylla glacialis (Pickering, 2001), and in New
Zealand: A. aurea, A. monroi, A. poppelwellii, and A. scott-thomsonii (Webb & Lloyd,
1980)). While dioecy almost completely eliminates selfing in plants, it also increases the
plant’s reliance on animal (or wind) movements of pollen between individuals of both
sexes for successful reproduction (Bawa, 1980).

The New Zealand flora generally possesses a high number of sexually dimorphic taxa
compared with global levels (Webb & Kelly, 1993). There have also been numerous
reports of dimorphism in terms of male-biased sex ratios in New Zealand alpine plant
species, particularly for Aciphylla (e.g. Lloyd & Webb, 1977; Webb & Lloyd, 1980). This
departure from equality may arise from gender differences in life histories such as costs of
reproduction, and male versus female function (i.e. pollen and ovule production
respectively) (Delph, 1999).

12
1.6 Insect pollination

Pollination is the transfer of pollen from an anther to a stigma. The movement of pollen
between plants by insects (entomophily) is by far the most common pollination system in
the world (Proctor et al., 1996). In many hermaphrodite plants (which make up 80% of all
flowering plant species) this can happen within the same flower or on the same plant, but in
dioecious species, it is fundamental for pollen to be moved between plants, from male to
female. The plant-insect pollination system is a mutualism (DeAngelis & Holland, 2006),
where the insects rely on the flowers for food (pollen and/or nectar), while the plant relies
on the insect to move its pollen between plants in order for successful outcrossing to occur
(Bronstein, 2004). (Note: an insect may be in a mutualistic relationship with a plant in
some ecological circumstances, while it could be parasitic in others, (see Thomson, 2003
and Holland et al., 2004). So, plants provide floral rewards that entice insects to visit their
flowers in order for them to carry pollen away or inwards for successful fertilisation. If the
plant has a specialist pollinator, it needs to provide a reliable food resource through time in
order for the plant-pollinator mutualism to be maintained, and the plant can be susceptible
to pollen limitation if the pollinator becomes rare in the environment (Ladley et al., 1997).
A generalist pollinator system (see Newstrom & Robertson, 2005), however, may allow the
plant to get away with providing more variable rewards in space and time, as the insect
pollinators are not necessarily dependent on it as a food source, although there is the
additional problem of plants competing with each other for generalist pollinators, which
can lead to pollen limitation in some cases (Robertson, 1895; Waser, 1978; Campbell,
1985; Feinsinger & Tiebout (III), 1991).

1.6.1 Insect pollination and the importance of floral display

Visual and olfactory cues are used by plants to signal and attract insect pollinators (Proctor
et al., 1996). For plants with individual flowers which are grouped together in
inflorescences such as Aciphylla, it is the inflorescence rather than the individual flower
that is the effective functional unit for pollinators, enhancing the floral display to attract
pollinators (Proctor et al., 1996). A large floral display attracts insects from a distance,
particularly unspecialised insects which are characteristic as pollinators of dioecious

13
species. The importance of the plant’s floral display for pollinator attraction has been
reported numerous times in the literature (e.g. Willson & Price, 1977; Robertson, 1992;
Johnson et al., 1995; Andersson, 1996; Armstrong, 1997; Kawarasaki & Hori, 1999; Lortie
& Aarsen, 1999; Pickering, 2000; Irwin et al., 2004; Mitchell et al., 2004; Kudo & Harder,
2005; Orellana et al., 2005; Brunet & Sweet, 2006; Buide, 2006). Floral display not only
attracts pollinators at the plant level, but has also been shown to attract pollinators at the
population-level (e.g. Augspurger, 1980; Kato & Hiura, 1990; Jennerston, 1993; Bosch &
Waser, 1999; Lofgren, 2002; Forsyth, 2003; Kirchner et al., 2005). This is commonly
achieved through reproductive synchrony (Crone & Lesica, 2004), and the resulting
economies of scale (Norton & Kelly, 1988; Crone & Lesica, 2004).

1.7 Mast seeding

An important aspect of the reproductive ecology of Aciphylla is that they are mast-seeding,
(also referred to as mast flowering, supra-annual flowering, mass-fruiting or masting
(Kelly, 1994)). Masting has been generally defined as the synchronous highly variable
seed/flower production among years by a population of plants (Kelly, 1994). Masting is a
population phenomenon that results when individual plants within a population synchronise
their reproductive activity (Haase, 1986) by concentrating reproductive effort into a large
floral display during “mast” years at the expense of little or no flowering during “non-
mast” years (Crone et al., 2005).

This masting phenomenon has been witnessed for centuries because of phenomena such as
the mass over-production of acorns some years - a massive food crop affecting mice, birds
and other wildlife, and the virtual absence of these crops in other years, effectively starving
these animals (Koenig & Knops, 2005). These events were not hard for people to miss and
were especially important for farmers who fed acorns to their livestock. Records for this
phenomenon go as far back as the mid-17th century (Koenig & Knops, 2005). The term
“mast” comes from the German word, mǽst, for nuts/acorns of forest trees that have
accumulated on the ground, especially for those used as food for fattening swine (Janzen,
1971; Kelly & Sork, 2002). This German word for “fat” was then incorporated into the Old
English language (Bodsworth, 2003) and took on a similar meaning. It was Janzen, (1971,
1976) however, who was among the first to draw detailed scientific attention to mast

14
seeding (but see also Salisbury, 1942 and Silvertown, 1980), and defined it as the
“synchronous production of seed at long intervals by a population of plants”. He suggested
masting served to reduce levels of loss to seed predators (see section 1.8.3.1). His
definition suggests that there are just two types of years: seeding (mast) years and non-
seeding years - a concept referred to as ‘strict masting’ (Kelly, 1994). However, most
polycarpic plants do not display this bimodal pattern of strict masting, instead falling along
a continuum of variability levels (Herrera, 1998a; Kerkoff & Ballantyne (IV), 2003) from
‘strict’ to ‘normal’ masting (see Kelly, 1994). Hence, in this study, mast-seeding will be
considered in the light of Kelly’s (1994) more general definition: “the synchronous highly
variable seed production among years by a population of plants” because technically most
plant species are not known to be strictly masting (Herrera, 1998a).

Masting occurs most frequently in long-lived plants which are less affected by the costs of
not reproducing in some years (Kelly, 1994). It is a worldwide phenomenon and is found in
plants from many different taxonomic groups and from most parts of the world (Isagi et al.,
1997). Mast-seeding seems to be common in temperate forest trees (Silvertown, 1980) and
herbs (Mark, 1970; Campbell, 1981; Brockie, 1986; Webb & Kelly, 1993; Kelly et al.,
2001) and present in some tropical forest trees (Janzen, 1971). It also appears to be
especially prevalent in the New Zealand flora (Webb & Kelly, 1993; Kelly, 1994). The
ultimate reasons for the high masting frequency in New Zealand are mysterious, but
Schauber et al. (2002) note that “New Zealand’s long history before humans and
mammalian herbivores arrived may have enabled even herbs and grasses to evolve long life
spans, which are necessary for masting to be a viable strategy”. Masting is a marked feature
in several New Zealand alpine plant genera (e.g. Aciphylla, Celmisia, and Chionochloa)
(Mark, 1970; Mark & Adams, 1973; Campbell, 1981) and is more pronounced at higher
altitudes (Webb & Kelly, 1993). The snow tussock Chionochloa (a co-inhabitant with
Aciphylla in New Zealand alpine grasslands) is one of the best-studied examples of a mast-
seeding species, gaining large selective benefits from masting through predator satiation
(see Kelly & Sullivan, 1997; Kelly et al., 2000; Rees et al., 2002).

Masting in itself is paradoxical phenomenon (Kelly & Sullivan, 1997; Koenig et al., 2003)
and the evolution of masting is puzzling because it imposes a number of clear selective
disadvantages (Kelly et al., 2000) including lost opportunities for reproduction (Waller,
1979) and more severe density-dependent mortality among seedlings produced in mast

15
years (Hett, 1971; Augspurger, 1981; Kelly et al., 2001). Masting affects the timing of
flowering and fruiting as well as critical aspects of the plant’s life cycle such as pollination
and dispersal (Tisch, 1996). Therefore, masting has important consequences for the
community and ecosystem, because animals that rely on pulsed resources (such that
masting plants provide) for a food source such as pollen, nectar or seeds, will face an
unreliable food supply (Curran & Leighton, 2000; Ostfeld & Keesing, 2000; Schauber et
al., 2002). Thus it is important to understand plant reproductive patterns in order to
understand the dynamics of a whole system.

1.8 Why do some plant species mast when such high costs are involved?

1.8.1 Proximate versus ultimate reasons for masting

It is important to differentiate between proximate and ultimate causes of masting.

Proximate causes refer to those factors governing the synchrony in flowering/fruiting
between individual plants, and the magnitude and timing of masting events (Sork &
Bramble, 1993; Isagi et al., 1997). It essentially considers the question: how do plants
mast? Factors such as weather and resources for example, are likely to be proximate causes
for masting (Kelly & Sork, 2002). Proximate causes therefore do not drive the selection for
masting, but may provide the initial variation on which selection can act (Waller, 1993).
Proximate reasons for masting, such as spatial synchrony and cues for flowering are not the
focus of this study, but should be an important consideration for future research in this area
because of the likely effects of global warming on flowering variability and its inter-trophic
consequences (see McKone et al., 1998). Ultimate explanations for masting deal with the
question: why mast? There are two general categories that attempt to answer this question.
The first involves a non-evolved response, while the second involves evolutionary
hypotheses to explain this phenomenon.

1.8.2 Non-evolved hypothesis to explain mast-seeding

One of the earliest explanations for mast-seeding was that the plants had no choice i.e. that
masting was a non-evolved response to a naturally variable environment (Busgen &

16
Munch, 1929). This is now known as the resource matching hypothesis of mast seeding
(Kelly, 1994; Kelly & Sork, 2002), which implies that mast years are simply good years for
all aspects of the plant, and hence growth and reproduction are expected to be positively
correlated with each other (Kelly & Sork, 2002). This hypothesis has largely been ruled out
by the fact that variability in reproductive output is considerably greater than the variability
in environmental factors in many mast- seeding systems (Koenig & Knops, 2000) and by
the switching of resources from growth to reproduction during mast years, (Norton &
Kelly, 1988; Kelly & Sork, 2002; Monks & Kelly, 2006). Also at many sites, masting
species grow side-by-side with species which have relatively constant seed output from
year to year (Campbell, 1981; Webb & Kelly, 1993), indicating that the environment alone
is not responsible for masting behaviour. Therefore, because masting could not merely be a
side effect of resources that vary from year to year, mast seeding appears to be
predominantly an evolved strategy in which there are some overall lifetime reproductive
benefits to the plant of focusing reproduction in some years at the expense of not
reproducing in others (Norton & Kelly, 1988; Koenig & Knops, 2000; Kelly & Sork,
2002), but see Kerkoff (2004).

1.8.3 Evolutionary hypotheses to explain mast-seeding

The alternative ultimate explanations for masting consider the evolutionary selective
pressures that may have led a particular plant species to exhibit masting behaviour. There
are a number of hypotheses that have been put forward as selective forces for plants
producing variable, synchronous seed crops. For all the hypotheses, the common element is
that the selective advantage occurs through an economy of scale (Janzen, 1971; Norton &
Kelly, 1988) whereby large reproductive efforts are more efficient than smaller ones, so
plants reproducing in step with mast years will have higher fitness (Kelly & Sork, 2002).
The benefits of these large synchronous reproductive events accrue by overcoming some
environmental constraint such as pollen limitation or seed predation (Kerkoff, 2004).
Among the many hypotheses for masting, there are two which have by far received the
most support: the predator satiation and the wind-pollination hypotheses. Below I briefly
review six of the most common evolutionary-based hypotheses which have been suggested
as favouring mast seeding, but it is the final two that are of most significance to this study.

17
1.8.3.1 Predator satiation

The predator satiation hypothesis states that large, intermittent seed crops reduce losses to
seed predators (Janzen, 1971, 1978; Silvertown, 1980). Seed consumers are effectively
starved during intermast intervals and swamped during mast events, driving large
fluctuations in consumer abundance and increasing the potential fitness benefit of masting
(Curran & Leighton, 2000; Schauber et al., 2002). Studies show that predator satiation has
been found to occur in a range of plants such as Chionochloa (snow tussocks), (Kelly &
Sullivan, 1997; McKone et al., 1998; Kelly et al., 2000; Rees et al., 2002), oaks, and
beeches (Janzen, 1971) by a variety of animals from insects to mammals (Kelly & Sork,
2002). These, among many other examples, show evidence for a lower seed predation rate
in high-seed years.

1.8.3.2 Wind-pollination

The wind-pollination hypothesis states that masting should be strongly selected in species
that can achieve greater pollination efficiency through synchronised above-average
flowering effort (Nilsson & Wastljung, 1987; Smith et al., 1990; Kelly & Sork, 2002).
Plants that synchronously produce masses of flowers and pollen once every few years, are
likely to experience higher pollination rates and fruit set than plants that produce average
amounts of flowers and pollen each year (Koenig & Knops, 2005). Masting plants showing
improved reproductive success from wind-pollination are many, and include Lodgepole
pine (Pinus contorta) (Smith et al., 1990), rimu (Dacrydium cupressinum) (Norton &
Kelly, 1988), and the beeches (Fagus (Nilsson & Wastljung, 1987) and Nothofagus (Kelly
et al., 2001)). Theoretical models by Satake & Iwasa (2002) suggest that pollen limitation
could be a crucial driving force behind highly variable and synchronous seed production
that is so characteristic of wind-pollinated masting species (but this remains to be
empirically supported (Sork et al., 2002; Koenig & Ashley, 2003)).

18
1.8.3.3 Animal fruit dispersal and large seed size

The following two hypotheses involve selection at the fruit level, where firstly, large fruit
crops could result in wider dispersal by generalist animal dispersers or scatterhoarders
(Christensen & Whitam, 1991) and secondly, larger fruit size could increase variability in
seed number between years (although this does not require synchrony among individuals)
(Kelly, 1994). So far there exists little or no empirical or theoretical support for these
hypotheses and they will not be examined in this study.

1.8.3.4 Accessory costs of reproduction

This hypothesis proposes that small reproductive efforts are energetically inefficient
because of high accessory costs (Kelly, 1994). It is based upon the idea that relatively
indirect costs associated with reproduction decrease (per seed produced) with an increasing
size of the reproductive effort, thus it is energetically more efficient to reproduce less often
but at a higher level (at the individual plant level). High accessory costs could lead to
individual plants producing occasional large reproductive efforts being favoured. This
hypothesis does not require synchrony at the population level, thus accessory costs may
help select for masting in association with more direct selection forces, but cannot explain
masting at the population level (Kelly, 1994).

1.8.3.5 The animal pollination hypothesis

The animal pollination hypothesis suggests that when the majority of individual plants
flower synchronously, they are likely to attract more pollinators than they could
individually and in this way encourage pollination. For wind-pollinated species, it has been
frequently shown that percent fruit set is higher when flowering density is higher (see Kelly
& Sork, 2002). Researchers thus suggest that mast flowering is more likely to evolve in
wind-pollinated species because increased pollen availability enhances pollination success,
while in animal-pollinated species, increasing flowering may saturate the pollinators
(Norton & Kelly, 1988; Smith et al., 1990). This was supported by Herrera (1998) and
Kelly & Sork (2002) who surveyed hundreds of masting datasets and showed that wind-

19
pollinated species were significantly more variable in seed output (higher mean CV’s (co-
efficient of variation)) than animal-pollinated species. Therefore, to date there is almost no
convincing evidence of pollination economies of scale for masting, animal-pollinated
species because the general school of thought is that animals are more likely to be satiated
by large crops (Koenig & Knops, 2005) providing diseconomies of scales (Herrera et al.,
1998b; Kelly & Sork, 2002).

Despite this, if animals are attracted by a large floral display, masting could improve
pollination (Norton & Kelly, 1988; Kelly, 1994). However, if the system involved a
specialised plant-insect pollination mutualism, how would the sole pollinator species
overcome the variability in flowering between years? One would expect that a specialised
plant-pollinator system would be highly co-evolved, as the pollinator would require timing
its reproductive cycle with that of the plant (like specialised seed/flower predators of
Chionochloa spp. which undergo prolonged diapause, see McKone (2001)). A specialist
pollinator is likely to be more affected by variable reproduction of its food source in terms
of satiation during high flowering years and starvation during others. Another possibility
however, is if the plant has a specialist pollinator with a very short generation time that is
able to breed up during a mast event, as seems to happen with thrips-pollinated
Dipterocarps in Southeast Asia (Ashton et al., 1988). On the other hand, if the plant is
pollinated by a suite of generalist animals (common for insect-pollinated plants (Proctor et
al., 1996)), masting and animal pollination would be more likely to work, particularly if
pollinators are attracted to large floral displays. For insect pollination economies of scale to
occur would require a generalist fauna that favour large floral displays and would not
become satiated.

1.9 The bizarre reproductive ecology of Aciphylla

Little is known about the reproductive ecology of Aciphylla, although they are a ubiquitous
and well-known component of the high country (high alpine to lower montane) grassland
flora. What is known, is that they exhibit mast seeding; are insect-pollinated and dioecious
and together this presents a paradoxical situation. Firstly, insect-pollinated plants should
not provide an inconstant resource supply as this could lead to either a break-down of the
mutualism between the pollinator and the plant, caused by pollinator satiation during

20
“mast” events and subsequent starvation during years of little or no flowering. Dioecy
requires synchrony because male and female plants to flower at the same time in order to
achieve successful pollination. This synchrony could either be achieved through constant
average reproduction each year, (which is typical of most insect-pollinated dioecious
systems), or by occasional synchronous large floral displays (i.e. masting) if there exist
other selective forces encouraging the plant to have significantly large floral display size,
such that it can’t be produced every year. The latter is seen in Aciphylla, and it is
hypothesised that in order to avoid pollinator satiation, Aciphylla must overcome this
paradox by having a generalist pollinator fauna, and that pollinator economies of scale
could be one of the important selective forces behind the large floral display size in the first
place.

1.10 Research Aims

This study aims to test the hypothesis that pollinator economies of scale are an important
selective force for the large population-level flowering effort in Aciphylla and that the high
costs of reproduction that come with floral investment explain the reasons for masting in
this genus. This can be tested at two scales: at the individual plant level it is hypothesised
that a larger floral display size will have increased reproductive success due to enhanced
attractiveness to insect pollinators; while at the population level, masting synchrony leads
to more inflorescences (increased population-level reproductive effort), and improved
pollination success due to pollinator economies of scale. Specifically, the outline of this
thesis is as follows:

1) Chapter 2 aims to test the hypothesis that pollinator economies of scale are an
important selective force for the large floral display size in Aciphylla. This chapter
will be separated into two sections: Part (A) tests whether variation in plant-level
floral display size affects seed set rate in female Aciphylla inflorescences using both
natural variation in floral display size and manipulative space-for-time substitution
experiments, and simultaneously tests for pollen and resource limitation. Part (B)
investigates whether variation in population-level flowering effort affects female
seed set, again employing both natural and manipulative experimental approaches.

21
2) Chapter 3 applies the hypotheses outlined above to test whether variation in (A)
plant-level floral display size and (B) population-level flowering effort affect
pollinator visitation rates to female Aciphylla inflorescences, and whether this
relates to seed set rates (Chapter 2). Furthermore, it explores whether pollinators are
more attracted to inflorescences of male or females. Additionally, I provide the first
comprehensive description of the insects associated with flowers of Aciphylla
species.

3) Chapter 4 investigates aspects of sexual dimorphism in the dioecious Aciphylla. I

test whether there is evidence for sexual dimorphism in terms of a male-biased sex
ratio, among other morphological and phenological characteristics in flowering
Aciphylla to determine whether these differences relate to sex-specific pollination
success (Chapter 3).

4) Chapter 5 discusses the implications associated with each of the above chapters and
how they relate, in the context of Aciphylla and in the wider ecological framework.

22
Chapter 2- Does floral display size affect reproductive success in Aciphylla?

2.1 INTRODUCTION

As outlined in Chapter 1, large-leaved, montane Aciphylla species including A. aurea and

A. scott-thomsonii, among others, are long-lived, dioecious (male and females are
individual plants), mast-seeding (Mark, 1970; Campbell, 1981) herbaceous plants, that
have tall showy inflorescences with thousands of flowers. Aciphylla are thought to be
primarily pollinated by insects and wind (Dawson, 1971; Webb, 1986) although the relative
importance of these agents as effective pollen vectors has rarely been tested experimentally
in Aciphylla (but see Brookes & Jesson, 2006). In many genera worldwide, numerous
studies link pollination success to floral display size at both the individual plant- (e.g.
Vaughton & Ramsey, 1998; Orellana et al., 2005; Buide, 2006) and population- (Kato &
Hiura, 1990; Kirchner et al., 2005; Kindlmann & Jersakova, 2006) levels, in both wind-
(e.g. Smith et al., 1990; Kelly & Sullivan, 1997; Kelly et al., 2001) and insect- (e.g. Davila
& Wardle, 2002; Grindeland et al., 2005) pollinated systems. In many flowering plants,
reproductive success depends on the ability to attract animal pollinators (Vaughton &
Ramsey, 1998) and this is frequently achieved through having a large floral display (Kudo
& Harder, 2005). In Aciphylla, biological and/or environmental pressures over time appear
to have selected for large floral display size and synchronous variable reproduction i.e.
masting, (Kelly, 1994). This suggests that the lifetime benefits gained by having a large
floral display and reproducing only some years must outweigh the costs of forfeiting
reproduction during others (Janzen, 1971; Silvertown, 1980) and also must be more
effective than having a constant, but reduced, floral display size each year (Norton & Kelly,
1988; Koenig et al., 2003).

2.1.1 The importance of floral display size at multiple levels in Aciphylla

Floral display in large-leaved, montane Aciphylla species is interesting at multiple levels,

including the inflorescence, plant and population scales. Firstly, inflorescences are
particularly tall with thousands of flowers (e.g. up to 15,300 flowers for a female
inflorescence, and averaging 9,746 flowers for an average size female inflorescence of 1.9
metres for A. scott-thomsonii, see Appendix 1). Within an inflorescence, there is natural

23
variation in the proportion of the inflorescence with flowers present relative to the entire
height of the stalk (Fig. 2.1). There is also variation in inflorescence height with a
continuum of short to tall inflorescences present within a population. On the individual
inflorescence, there are also other structures that may be important for floral display such
as the spiky floral bracts. The role that bracts play during flowering is not certain, but there
is some possibility that they could be influential in pollinator attraction and/or seed
production in some way. Secondly, at the plant level, inflorescence number varies between
plants. Plants have one to many rosettes and each rosette can make a single inflorescence.
One to three inflorescences is common for a female plant during a flowering year, and 1-4
for males, but in some cases there can be up to 10 inflorescences per plant. Thus variation
in plant-level floral display also exists within a population.

Figure 2.1 Aciphylla glaucescens (Cow Creek, Waihopai Valley, Marlborough) provides a clear example of
the variation in floral display size within an inflorescence during a high-flowering year. All inflorescences
here are relatively tall, but on some stalks, flowers are not presented until quite a distance up the rachis (e.g.
front-right), while on others, flowers begin just above the leaves (e.g. centre-left). Photo courtesy of Barry
Hope.

24
Finally, floral display varies prominently from year-to-year in large montane Aciphylla
species. In plants generally, not all species are highly variable in seed production (i.e.
‘mast’ years with most individuals flowering, interspersed with ‘non-mast’ years with little
or no flowering), but instead plant species tend to fall along a continuum of variability
levels (Webb & Kelly, 1993; Kelly, 1994; Herrera, 1998a) from ‘strict masting’ to ‘normal
masting’, i.e. highly variable to nearly constant (Kelly, 1994). This variability in the timing
and effort of flowering differs between populations of Aciphylla depending on species,
altitude, latitude and resource availability (Pickering & Hill, 2002). Therefore, floral
display in terms of population effort also varies between years and different populations at
different sites. Floral display can be examined at these multiple levels simultaneously to
assess the relative importance that the variation in floral display at each level has on
reproductive success in Aciphylla. First however, we must consider the evolutionary and
ecological significance of floral display size and its importance in reproductive success in
plants, particularly in mast-seeding, dioecious, entomophilous breeding systems.

2.1.2 Measuring floral display at the inflorescence and plant level

Animal-pollinated plants are expected to evolve strategies that maximise the efficiency of
pollen transfer to and from the plant. Plants have evolved a number of mechanisms that
increase the probability that visitors will move pollen among plants to ensure outcrossing
(Marr et al., 2000). The evolution of dioecy has probably been the most successful of these
(Lloyd, 1975; Charlesworth & Charlesworth, 1978; Sakai & Weller, 1999), however, this
still requires the plant to employ mechanisms for successful pollinator attraction. One such
strategy may be to develop large showy floral displays that attract more pollinators (Proctor
et al., 1996; Grindeland et al., 2005). There is a wealth of literature providing evidence that
floral display size plays an important role in the reproductive success of many plant species
worldwide, (e.g. Andersson, 1996; Le Corff et al., 1998; Kawarasaki & Hori, 1999). In
particular, many studies have reported a positive relationship between visitation rates of
pollinators and floral display size at the level of the individual plant (e.g. Kato & Hiura,
1990; Grindeland et al., 2005; Kudo & Harder, 2005; Brunet & Sweet, 2006). There is also
evidence for this among species in the Apiaceae family (e.g. Pickering, 2001; Davila &
Wardle, 2002).

25
The timing of pollen availability (Harder & Wilson, 1994) and nectar production (Marden,
1984) as well as the amount of reward offered by a plant can determine both the rate of
visitation and the subsequent behaviour of flower visitors and therefore have consequences
for pollen transfer (Robertson et al., 1999).

Floral display can be measured in a number of ways, including flower number (Robertson
et al., 1999), flower or inflorescence display size (Abe, 2000), or showiness (Ashman et al.,
2004). The notion that large floral displays exist to attract pollinators is referred to as the
pollinator attraction hypothesis (Sutherland, 1987). This is analogous to the animal-
pollinator hypothesis of masting (Kelly, 1994) (see Chapter 1). Pollinator attraction is
critical to the reproductive success of the majority of flowering plants (Kearns et al., 1998)
however, the payment for attracting pollinators can be costly, both in the amount of
resource invested (Pyke, 1991) and by attracting nectar robbers (Irwin et al., 2004).

If a larger floral display can attract more visitors, does this actually lead to an increase in
seed set? Some researchers hypothesise that seed-set on female plants is a direct indicator
of insect attraction (Ortiz-Perez et al., 2004). In reality however, factors such as pollinator
efficiency, pollen carryover and resource availability must also be taken into account.
While larger floral displays may attract more insect pollinators, in self-compatible
hermaphrodite or monoecious plants, a large proportion of visits made by pollinators is
likely to occur between flowers on the same plant, resulting in geitonogamous self-
pollinations rather than cross-pollinations (Robertson, 1992; Mitchell et al., 2004).
Therefore, in bisexual plants, the effect of increased geitonogamy with increased floral
display may be considered an unavoidable detrimental effect of large displays (Hessing,
1988). Given that Aciphylla is completely dioecious (Webb, 1979) the detrimental effects
of geitonogamy from a larger floral display will not be seen. Nevertheless, dioecious
species with showy floral displays do not necessarily escape from pressures that may be
detrimental to their overall reproductive success. Dioecious species often have sexually
dimorphic floral displays, with a common pattern of males being larger and/or showier than
females (Delph, 1996; Eckhart, 1999) (see Chapter 4). If insects are more attracted to
showier floral displays and better floral rewards, and visitation rates are disproportionally
higher on males than females as many studies have shown (e.g. Le Corff et al., 1998;
Vaughton & Ramsey, 1998; Pickering, 2000; Voight et al., 2005), this reduces the chance
of females being pollinated (Bawa, 1980). This could be detrimental for female seed set

26
particularly when pollinator abundance is low (Vamosi & Otto, 2002) due to increased
competition for pollinators. In Aciphylla, male inflorescences are generally showier than
females (Pickering, 2000), therefore this species is an ideal candidate to additionally test
whether inter-sexual competition for pollinators exists and whether this has consequences
for female reproductive success (tested in Chapter 3). Furthermore, in dioecious plants
generally, investment in reproduction is usually considerably higher for females than males
(Webb & Lloyd, 1980; Pickering, 2000; Pickering & Arthur, 2003; Wheelwright & Logan,
2004). This has also been shown for Aciphylla species (Hogan et al., 1998) and is therefore
likely to play an important role in the floral display size of female plants (see Chapter 4).

2.1.2.1 The importance of resource and pollen limitation in Aciphylla – does female
success depend on individual reproductive effort?

In this study, it is hypothesised that larger reproductive effort (increased floral display)
increases reproductive success (i.e. seed set) due to increased pollinator attraction.
However, it is anticipated that these effects may not be so straightforward due to the
opposing effects of resource limitation that are also likely to be acting on display size.
Therefore, these effects must be considered in order to make assumptions about the relative
importance of floral display size, pollinator attraction and resource limitation.

Bateman’s Principle states that males are limited by the ability to find mates due to
competition for mating opportunities while females are limited by resources required for
offspring provisioning (Bateman, 1948). “However, in flowering plants this principle
appears too simplistic because if male searching for mates (including pollen dissemination
via external agents) is not sufficiently successful, then the reproductive success of both
sexes will be limited by the number of matings rather than resources” (Burd, 1994 pp. 83).

Limitations of female success due to inadequate pollen receipt appear to be a common

phenomenon in plants (Vaughton & Carthew, 1993; Burd, 1994; Ashman et al., 2004;
Newstrom & Robertson, 2005). Across species, supplemental pollen often leads to
increased fruit set, especially in plants that are highly self-incompatible such as dioecious
species, suggesting that inadequate pollen receipt is a primary cause of low fecundity rates
in many perennial plants (Hirayama et al., 2005; Knight et al., 2006; Pias & Guitian, 2006).

27
It was mentioned above that, as well as pollen, resources are likely to be playing an
important role in the reproductive success of female plants (Horvitz & Schemske, 1988). In
animal-pollinated plants, resources must be allocated for both pollinator attraction and
provisioned for successful offspring (seed or fruit) production (Ashman et al., 2004). Plant
populations are expected to evolve traits where taller inflorescences with more flowers may
promote increasing rates of pollinator visits and act to optimise fecundity, while resource
limitation may offset the ability for taller inflorescences to set more seeds (Haig &
Westoby, 1988). Therefore floral display size can be thought of as an optimal investment
between pollinator attraction and fecundity.

Thus there exists a dynamic interplay between the two opposing forces, requiring tradeoffs
between large reproductive effort and successful offspring provisioning, i.e. while taller
inflorescences attract more pollinators and may achieve higher rates of seed set, eventually
resources may become limited no longer enabling all ovules that have been fertilised to set
seed (Haig & Westoby, 1988). Because of the conflicting dynamics between the two, how
can we separate them to test for the relative importance of pollinator attraction and resource
limitation?

Resource limitation has been inferred when supplemental pollination does not increase seed
production (Brookes & Jesson, 2006). To simultaneously test for pollen and resource
limitation in this study, flowers are removed (0, 40 and 80% removal) to create three levels
of floral display size within populations of Aciphylla, experimentally imitating natural
variation in floral display size within a population. As previously mentioned, Aciphylla has
a large floral display consisting of inflorescences capable of reaching 3 metres tall with
thousands of flowers. In an important paper by Haig and Westoby (1988), they presented a
graphical model in which ovule fertilisations rise with increased allocation to pollinator
attraction (increased flowering length) while at the same time ovule maturation ability
(seed set rate) declined because enhanced attraction diverts resources from seed and fruit
production. Figure 2.2 depicts the hypothesised effects of pollen and resource limitation
scenarios on seed set rate.

28
 (a) Pollen limitation (b) No pollen or (c) Pollen + resource
resource limitation limitation

10 0 10 0 100
Percent seed set

50 50 50

0 0 0
0 40 80 0 40 80 0 40 80

Percentage of flowers removed

Fig. 2.2 Expected effects of pollen and resource limitation when supplemental pollen is added. Supplemental
pollination (––––––), natural pollination (– – – –) and induced pollen limitation by removing flowers (- - - -).
Under scenario (a) supplemental pollination resulting in an increase in seed set indicates pollen limitation.
Also shown is the effect of inducing pollen limitation by removing flowers. As flower number decreases,
floral attraction is reduced and thus lower seed set is expected. Under scenario (b) there is no increase in seed
set when pollen is added. Under scenario (c) when the plant is simultaneously pollen and resource limited, an
increase of resources to the remaining flowers and supplemental pollination would induce greater increases in
seed set than only being limited by one factor. (Adapted from Brookes & Jesson, 2006).

While this study does not attempt to test for the effects of resource limitation per se by
supplementing resources, by removing flowers it increases the amount of resources
available per seed on inflorescences with fewer flowers. Therefore, this study employs both
supplementary pollination and flower removal experiments to test for both pollen and
resource limitation simultaneously in Aciphylla by testing the above model explicitly. As
well as this, factorially designed manipulative experiments which provide more flexibility
and power to test proposed hypotheses, were used to test for the effects of floral display
size on pollinator attraction (Chapter 3) and reproductive success (this Chapter) at multiple
levels.

2.1.3 Measuring population-level flowering effort – pollinator efficiency

The benefits of having a large floral display at the plant level may scale up to the level of
the population, meaning that population flowering effort (i.e. pollinator economies of scale)

29
could also be important for reproductive success. Because Aciphylla show masting
behaviour, there is a larger number of plants flowering overall in the population during
high-flowering years. Does masting benefit reproductive success of individual females in
Aciphylla by increasing population-level pollinator attraction through economies of scale?

Masting plant species are known to show clear benefits from masting when they are wind
pollinated (Norton & Kelly, 1988; Kelly et al., 2001). Concentration of pollen production
in mast years increases the probability of pollination for wind-pollinated species (Nilsson
& Wastljung, 1987). On the other hand, relatively few species that display masting
behaviour are insect-pollinated (Kelly & Sork, 2002; Crone et al., 2005). The negative
effect of extreme fluctuation in plant reproduction on the stability of populations of
pollinators might help to explain why most plant taxa that mast are wind-pollinated rather
than animal-pollinated (Smith et al., 1990). Many other authors regard animal pollination
to be disadvantageous in a mast-seeding system due to pollinator satiation (see Chapter 1
and review in Kelly, (1994)). Masting has been assumed to swamp specific pollinators by
producing such large temporally-concentrated flowering crops. However, if animals are
more attracted by large floral displays, masting could improve pollination (Kelly, 1994)
and overall lifetime reproductive fitness. This could be achieved by having a diverse and
generalised pollinator fauna that does not rely on the masting species as a sole food
resource, and are not satiated then starved, but is highly attracted to that species in the years
that it does flower.

If flowering effort is so great during mast years that the masting species can ‘out-compete’
other generalist-pollinated plant species in the community for pollinators, then masting
may be a successful method for ensuring overall lifetime reproductive fitness, because the
costs of forfeiting reproduction some years must be outweighed by the benefits of
concentrating flowering effort into certain years.

Because Aciphylla is dioecious, pollination requires male and female plants to flower
synchronously. Individual reproductive success in obligate outcrossers is profoundly
influenced by the presence of conspecifics in space and time, and individual plants may
succeed best when they synchronise with others (Howe & Westley, 1986). Plant population
size and density can influence the interactions between plants and pollinators and affect
reproductive success (Schemske, 1980; Kirchner et al., 2005) by affecting the quantity and

30
quality of pollination services received (Kunin, 1993). Dense stands of a flowering plant
may collectively attract more pollinators as opposed to small patches or low population
size, because small or sparse stands may be less apparent and/or offer low pollen and nectar
rewards (Rathcke & Lacey, 1985; Bosch & Waser, 1999).

The effects of population-level flowering (pollinator attraction effort) on an individual

plant can be observed in rare plant species, isolated plants or populations, naturally sparse
populations or in a naturally low flowering year of masting species. For instance, in low
flowering years, when only a small proportion of individuals flower, and are out of
synchrony with the majority of the population (Smith et al., 1990), this could have negative
consequences for the individual in terms of pollinator attraction and reproductive success if
the flowering population is the “attraction unit” (Groom, 1998; Forsyth, 2003). The effects
of this can be stronger when a masting species has highly self-incompatible individuals
(Forsyth, 2003) such as dioecious Aciphylla. For individual female plants flowering out of
synchrony, these ‘isolation’ effects could be particularly detrimental to reproductive
success when they are effectively isolated from the pollen supply (i.e. male plants) (Groom,
1998). In this scenario, pollen and/or pollinator limitation are expected to be important,
particularly if population flowering effort is important for pollinator attraction (Kato &
Hiura, 1990; Bosch & Waser, 1999; Kirchner et al., 2005).

The effects described above can be studied using long-term datasets over time to measure
changes in flower production between years. However, because of the time frames under
which many studies are conducted, long-term experiments may not be practical. Therefore,
by manipulating the natural flowering effort, flowering density and sex-ratio within a mast
year using space-for-time substitution methods (e.g. Nilsson & Wastljung, 1987) as an
alternative to long-term studies, it is possible to imitate a low-flowering year to test the
effects of variation in floral display on reproductive success.

Many studies have investigated the effects of floral display on reproductive success but few
have tested for the relative importance of floral display size for pollinator attraction and
seed set at multiple levels (but see Kato & Hiura, 1990; Bosch & Waser, 1999; Grindeland
et al., 2005).

31
2.1.4 Objectives

2.1.4.1 Part (A): Floral display at the level of the individual plant

The first part of this chapter attempts to test for the importance of female floral display size
on reproductive success through increased pollinator attraction at the plant-level, while
simultaneously testing for the importance of pollen and resource limitation. It is
hypothesised that an increased floral display will lead to increased female reproductive
success (seed set) through pollinator attraction but only if resources and/or pollen are not
limiting factors. Experiments were conducted during a high-flowering year using both
experimental manipulations and natural measures of floral display size within a population.
More specifically the questions were:

• How is seed set rate affected when bags are applied to inflorescences to exclude
insect pollinators?

• How does variation in floral display size affect rates of seed set in female Aciphylla
inflorescences? How does this vary when display size is measured by each of the
following?

- flowering length
- inflorescence height
- percent flower removal
- number of inflorescences

• Is pollen limitation occurring at the level of the individual plant?

• Are floral bracts playing an important role in reproductive success and is seed set
rate affected when bracts are removed?

32
2.1.4.2 Part (B): Population level flowering effort - male to female flowering densities

The second aim was to investigate the effects of male floral densities (i.e. local pollen
availability) and asynchronous flowering on female reproductive success. ‘Space-for-time’
substitution was used to simulate the reduction of floral densities that would occur during a
low-flowering year and test what would happen to female reproductive success if male to
female flowering synchrony was reduced. It is hypothesised that by removing showy
attractive male inflorescences from the population, and decreasing population level floral
display in Aciphylla, this will lead to a decrease in reproductive success. More specifically:

• How is reproductive success in female inflorescences affected with increasing

distance to the nearest flowering male plants (i.e. distance from pollen source)?

• Is pollen limitation occurring in female inflorescences that are more isolated from
flowering male plants?

Table 2.1 outlines each predictive term used in this study and the hypothesised directions of
the relationships between the each floral display term and seed set rate, and provides an
explanation for each of these predictions based on the hypotheses discussed in the section
above (i.e. pollen and/or pollinators are limited and larger floral displays are more
attractive to pollinators).

33
Table 2.1 Expected relationships between each predictive term and the response variable, seed set rate. The first six terms represent floral display at the level of the individual
plant, followed by an interaction between percent flower removal and pollination treatment, (as performed in statistical models in this chapter testing how each predictor
affects seed set rate in female Aciphylla inflorescences). Explanations for expected direction of relationships are based on the assumption that a bigger display size increases
the number of insect pollinators and subsequently will experience higher seed set. 1 The double signs next to flower removal indicate where there were 3 levels of removal.
The first sign represents the direction of the relationship between the control (0%) and 40% removal and the second between 0% and 80% removal. 2 The interaction between
percent flower removal and pollination treatment is complex and cannot be simply represented by a single symbol (see Fig. 2.2).

Predictive terms Expected

Floral display relationship Explanation

Floral display at the Hand pollination + Hand pollination = higher seed set, natural pollination = lower seed set if pollen limited
individual-plant Flower removal (0,40, 80%) –/–– Smaller floral display, less attractive to insects = lower seed set 1
level Flowering length (cm) 0 Should be cancelled out by the effects that percent flower removal has already explained in the model
Inflorescence height (cm) + Taller inflorescences attract more insects, therefore more pollen received will give higher seed set
Number of inflorescences + More inflorescences on a plant attracts more insects = higher seed set
Floral bract removal – Bract removal may deter pollinators, and/or remove resources required to set seed = lower seed set
Hand pollination x flower- NA 2 Refer to Fig. 2.2 and text in section 2.1.2.1
removal (%) interaction
Floral display at the Local pollen index + Females closer to pollen (males), insect visit males more often, females
population level visited more by insects closer to males if pollen is the main attraction = increase seed set
Increasing distance to males – Females further from males, less local pollen, less insects, fewer visits = lower seed set
Other experiments Pollinator-exclusion bags – Bags prevent insect pollinators from delivering pollen = low or zero seed set

34
2.2 Site and species selection

The number of flowering Aciphylla populations during the summer field season of
2004/2005 limited the number of sites available for study. An initial survey to find
potential flowering populations around the middle of the South Island was carried out in
late October 2004. Sites selected were Lewis Pass (Lat/Long: 42o 22.713 S, 172o 24.021 E),
Burkes Pass (Lat/Long: 44o 5.565 S, 170o 34.792 E) and Hakataramea Pass (Lat/Long: 44o
18.667 S, 170o 34.292 E). The three site locations were at the extreme northern (Lewis
Pass) and southern (Burkes and Hakataramea Passes) limits of the Canterbury region of the
South Island of New Zealand (Fig. 2.3). The sites were all at low alpine passes with easy
access and within a 200 metre altitudinal range of each other. See Table 2.2 for information
on study sites.

Lewis Pass

Burkes Pass
Hakataramea Pass

Fig. 2.3 Map of the South Island of New Zealand

showing the three study site locations. All sites
are located within the wider Canterbury region,
with Lewis Pass at the northern boundary and Burkes
and Hakataramea Pass at the southern boundary.

35
Table 2.2 Summary of site information for the three study areas used in this research including site location and description, elevation in metres above sea level (m.a.s.l.), and a brief
description of climate, geology, soils, vegetation and background/modifications of the general area. Data sourced from Land Environments of New Zealand (LENZ), (Leathwick et al. 2003)
and surveys from Ecological Regions and Districts of New Zealand (McEwen 1987).

Site Location description Elevation (m.a.s.l) Climate Geology and soils Vegetation Historical background

Boggy grassland area Stable streams in valley

either side of road at floors with open
Parent material: mostly Triassic
the Pass (near tarn at grasslands of fescue
Torlesse greywacke & argillite
start of St James Cool, moist, warm tussocks, and bogs on Change from continuous
mountains; gravelly glaical
Walkway). summers, rainfall terraces have beech forest to patchy
Lewis Pass 860 m outwash alluvial soils on river
Conservation status: 2800-4800mm p.a. Dracophyllum, Carex, forest, some burning from
flats, fertility: moderate; drainage:
Forest Park. (higher at the Pass) Phormium, Sphagnum, etc sheep graziers
good

Burkes Pass Scenic Mostly Mesozoic-Paleozoic Mostly modified by

Subhumid hill climate Tussocklands with
Burkes Pass Torlesse greywacke and argillite, repeated burning
Reserve, near with cool to cold silver/fescue tussock,
roadside Tekapo 700 m some schists, tertiary gravels & beginning in Polynesian
winters and mild dry scrub - matagauri,
Highway (SH1) marine deposits, Pleistocene loess- times & grazed by both
summers; rainfall Coprosma, introduced
covered gravels; subsoils pale- domestic and wild
800-1200 mm p.a. ; grasses
coloured, compact and droughty in animals
NW winds prevail
summer

Alongside Landform: rolling to steep

Hakataramea River, Modified and depleted
mountains either side of Pass, tussocklands with heavy
between Grampian parent material: mainly Paleozoic Mostly modified by
weed infestation - repeated burning
and Dalgety Ranges, Semi-arid to sub- Torlesse greywacke and schist -
McKenzie district, hawkweeds & broom. beginning in Polynesian
humid inland climate some older basalts, argillite
Hakataramea Scrub in gullies:
South Canterbury. 900 m with cold winters and conglomerates, alluvium and loess, times & grazed by both
Pass matagauri, fescue/silver domestic and wild
Farmed (merino mild to hot summers; basic volcanic and tertiary rocks,
sheep). tussock, Gaultheria, animals
rainfall less than 600- including limestone; subsoils at Bulbinella, range of
1200 m p.a. Pass pale, compact & droughty;
introduced grasses
fertility: moderate, drainage: near
river - wet and boggy

36
2.2.1 Study species

At these sites, three species were studied: A. aurea at Burkes and Hakataramea Passes, A.
scott-thomsonii at Hakataramea Pass, and A. ‘lewis’ at Lewis Pass. These taxa are
described below.

2.2.1.1 Aciphylla aurea W.R.B. Oliver

Aciphylla aurea (also known as golden spaniard) is abundant mainly east of the main
divide of the South Island (Wardle, 2002) from Marlborough through to Southland. Habitat
ranges from montane to sub-alpine tussocklands and shrublands, usually on well-drained
hill slopes, ridges, outwash surfaces and around rock outcrops (D. Glenny, pers. comm.
2005). Plants form 1-10 rosettes of golden yellow-green tussocks up to 1.3 m in diameter
and up to 70 cm tall.

Inflorescences consist of a tapering rachis (stalk) up to 1.8 m tall. There can be up to 10

inflorescences on a plant, but 1-4 is more common. The rachis supports compound lateral
umbels, each with multiple umbels and umbellules branching from the peduncle. These
lateral branches start from above the leaves, continuing to the apex, subtended by yellow
linear bracts which are long (18-37 cm) and consist of a sheath (29-37 mm length) and long
spiky stipules and leaflets. The rachis is hollow and the lower 1-20 bracts are usually
sterile. Bract and umbel length decreases towards the apex. Bracts at the upper half of the
inflorescence are reflexed at the top of the sheath to point downwards, while the stipules
point upwards to create a “cage” of spines (D. Glenny, pers. comm. 2005). See Fig. 2.4 for
a diagram of an Aciphylla inflorescence.

The winged fruits consist of two mericarps, forming a laterally compressed schizocarp
(Webb & Simpson, 2001). Fruits in A. aurea are dark brown. While the entire female
inflorescence is presented at the same time, flower maturation is basipetal with the top of
the inflorescence maturing first (Brookes & Jesson, 2006). Flowering begins from early to
mid December at montane altitudes.

37
Although formidable, this plant is susceptible to browsing by animals such as hares, rabbits
and livestock (Allan, 1982), along with a suite of invertebrates such as weevils and various
Lepidopteran larvae, many of which are Aciphylla-specific herbivores.

2.2.1.2 Aciphylla scott-thomsonii Cockayne & Allan

Common and widespread in Otago and South Canterbury, the natural habitat of A. scott-
thomsonii is sub-alpine scrub, extending slightly into alpine tussocklands, shrublands and
scrub on south-facing colluvial hill slopes. It is most common beside streams or in flushes
at valley heads (D. Glenny, pers. comm. 2005). A. scott-thomsonii is often found in
association with A. aurea (at damper, more fertile and scrubby habitat) and like numerous
members of the New Zealand Apiaceae, these two species can readily hybridise (Dawson
& Le Comte, 1978). Care was taken to avoid using hybrid specimens for purposes of this
research.

A. scott-thomsonii are typically much larger than A. aurea and form huge tussocks with
long, glaucous, extremely sharp pinnate leaves up to 1.5 m long (Allan, 1982).
Inflorescences can reach an impressive 4 metres tall with yellow-green fruits. Reproductive
arrangement is very similar to A. aurea (see above). Flowering in A. scott-thomsonii begins
early to late December and fruit ripens by mid-February (D. Glenny, pers. comm. 2005).

2.2.1.3 Aciphylla ‘lewis’

The northern limit of A. scott-thomsonii is recorded as Arthur’s Pass in mid-Canterbury

(Lat/Long: 42o 56.633 S, 171o 33.942 E), (Mark & Adams, 1973). However, D. Glenny
(2005), currently revising the taxonomy of this group, has collected plants as far north as
Kowhai Saddle near Kaikoura that are more similar to A. scott-thomsonii than any other
Aciphylla species. The form at Deer Valley near Lewis Pass represents a viable,
reproducing population very similar to A. scott-thomsonii, but with aspects of the next most
similar species, A. colensoi. D. Glenny (2005) suggests that it is difficult to draw a
boundary between A. scott-thomsonii and A. colensoi at Lewis Pass. Differences in leaf and
inflorescence morphology have been noted by D. Glenny (2005). This population will be

38
referred to as A. ‘lewis’ from this point on. Leaf voucher specimens (of all study species)
were collected, pressed and mounted, then deposited in the University of Canterbury
Biological Sciences herbarium. Refer to specimen numbers #38835 to #38838 LMY.

Like A. scott-thomsonii, A. ‘lewis’ grows alongside streams and rivers in extremely wet
conditions amongst other bog-dwelling plants such as flax (Phormium tenax), fescue
tussocks and Sphagnum. The extent of its distribution is not yet confirmed, although is
probably restricted to wet riversides east of the Main Divide around the greater Lewis Pass
area. Plant size and colour is similar to that of A. scott-thomsonii but inflorescences and
bracts are a bright yellow and fruits are yellow-green. Flowering in A. ‘lewis’ typically
begins around mid to late November.

Stipules

Developing
fruits

Bracts

Compound
lateral umbel

Scale 12 cm

Fig. 2.4 Mid-section of an inflorescence stalk of Aciphylla scott-

thomsonii showing detail of spiky bracts with stipules,
subtending compound lateral umbels made up of clusters of
flowers (umbellules) which are beginning to develop into fruits
in the above example.

39
2.2.2 Field work

The experiments were carried out at all three field locations over the summer of 2004/2005,
beginning late November when flowering commenced, through until early March when
fruit and seed set. Burkes and Hakataramea Passes were experiencing a heavy flowering
year therefore those populations were sufficiently large to perform manipulative
experiments in the field. At Lewis Pass, flowering Aciphylla plants were sparser and there
were relatively fewer in the population over a larger area, representing a relatively low
flowering year.

2.3 METHODS

2.3.1 Part (A): Floral display at the level of the individual plant

An experimental area (plot) was roughly measured out at each site. These plots varied in
size between sites because a large enough area to encompass sufficient numbers of
flowering plants for experimentation was needed, as well as leaving enough non-
experimental (control) plants. Aciphylla scott-thomsonii habitat is usually limited to wet
sites near streams and rivers therefore Hakataramea and Lewis Pass sites with A. scott-
thomsonii and A. “lewis” respectively were very long and relatively thin compared to A.
aurea populations (refer to Tables 2.3 & 2.4 for quantitative site information). The
following floral display experiments and the effect that they had on seed set rates all
involved female inflorescences only.

1) How do pollinator-exclusion bags affect seed set?

To see whether seed set was affected when insect pollinators were excluded, fine mesh
curtain bags (approximately 0.5 mm mesh diameter) were used to cover inflorescences.
These were put on at the beginning of the season when inflorescences were small and in an
early developmental stage, with bracts covering unopened flowers. Bags were tied at the
base of the inflorescence as close as possible to the point of attachment to the rosette. As

40
inflorescences grew taller, bags were extended so as to prevent stalks from bending over or
becoming damaged during growth and elongation. Bags were sufficient to exclude all
insects however mesh size was not small enough to exclude pollination by wind. Control
inflorescences were also tagged nearby or on the same plant for comparison with bagged
plants. Thus this method was used to test for seed set by mechanisms such as wind
pollination, selfing or apomixis (although selfing is unlikely given that plants in this study
were always strictly dioecious). At Hakataramea Pass however, one plant consisting of
mostly female flowers but with some male flowers present, was discovered. This plant was
noted as hermaphroditic and bagging was used to test for possible self-fertilisation.

2) Does floral display size affect seed set rate?

This question was approached in four ways. The first employed manipulative field
experimentation where 0, 40 and 80% of flowers were removed from inflorescences to test
whether seed set rate was affected by the varying proportion of flowers along the stem.
Percent flower removal was hence treated as a factor. The second two approaches measure
continuous variation in flowering length i.e. the length of inflorescence with flowers
present, and inflorescence height in relation to seed set rate, i.e. total length from ground to
apex. The fourth measures display size on the plant level as a whole, by using the number
of inflorescences on the plant as an indicator of display size.

a) Flower removal experiment

Three levels of flower removal were chosen to represent variation in floral display size in
each population. These were 0, 40 and 80% removal with flowers being removed from the
lower part of each inflorescence, leaving the flowering portion remaining at the top. To test
for pollen limitation, supplemental pollen was added to some inflorescences in each level
of flower removal treatment. Flower removal and pollen supplementation experiments were
set up factorially creating more flexibility and power to test the proposed questions.

41
Treatments were assigned randomly to female plants that had at least 2 inflorescences to
enable paired design sampling. Within the same plant, one level of flower removal
treatment was performed on both inflorescences (e.g. 40% removal) with one inflorescence
receiving supplemental pollen and the other remaining naturally pollinated. This design
was useful for controlling between-plant effects.

The flowering length of the inflorescence was measured from the lowest bract to the apex
and the appropriate length of flowers to be removed was calculated as 0, 40 or 80% off the
total flowering portion. Flowers were snipped off using fine scissors and care was taken to
avoid damaging the stem. Because inflorescences had not usually grown to full size at the
time of first flower removal with late expansion concentrated at the tip, flowering length
was re-measured between 2 and 3 weeks later and any more flowers that needed removing,
were removed.

b) Flowering length

Flowering length, i.e. the flowering portion of the inflorescence (cm), was measured after
flower removal treatments were performed. This was done in order to run statistical
analyses to test whether the effects of the absolute value of flowering length (rather than
the percentage removed – as above) was a better expression of floral display size and if
there was any effect on seed set rate. Flowering length was recorded at the end of the
season immediately prior to harvest.

c) Inflorescence height

The entire length of the inflorescences from the base to the tip of the apex was referred to
as inflorescence height (cm). Measurements of inflorescence height were recorded at the
end of the field season just prior to harvest.

42
 d) Number of inflorescences

The number of inflorescences on each treatment plant was recorded and used in statistical
analyses to test whether this accounted for any variation in seed set rate.

3) How does the removal of floral bracts affect seed set?

To determine whether the spiny, sterile floral bracts were having any effect on seed set
rates, bracts were removed on some inflorescences. Again, plants with at least 2
inflorescences were randomly selected from the population in order for paired design
sampling to be used. On one inflorescence, all bracts were removed, and the other was left
as a control for comparison. Bracts were snipped off from the entire length of the
inflorescence, leaving about 2 cm beneath the base of each peduncle to avoid damage to the
stem or floral parts. (See Table 2.3 for all manipulative treatments, plot size, number of
replicates, etc).

Table 2.3 Quantitative information for floral display experimental manipulations performed in the field. The
six manipulated flower/pollination treatments include 0, 40 and 80% flower removal each with and without
hand pollination. Bract removal and bag treatments also included controls for each. (* = no hand-pollination
performed at Lewis site due to weather, hence only 3 treatments – 0, 40, and 80% flower removal). In total
252 female inflorescences with various treatments were harvested from the sites below.

Site
dimensions Type of No. of No. of Total no. of
Site Species (m) treatment treatments Replicates inflorescences
Burkes Pass A. aurea 127 x 236 Flower/pollination 6 8 48
Bract removal 2 9 18
Bag 2 6 12
Hakataramea A. aurea 500 x 220 Flower/pollination 6 8 48
Pass Bract removal 2 8 16
Bag 2 7 14
Hakataramea A. scott- 1200 x 50 Flower/pollination 6 6 36
Pass thomsonii Bract removal 2 6 12
Bag 2 5 10
Lewis Pass A. ‘lewis’ 800 x 230 Flower/pollination 3* 8 24
Bract removal 2 3 6
Bag 2 4 8
TOTALS 252

43
4) Does local pollen availability affect seed set rates?

To determine whether local pollen densities were affecting female reproductive success,
the distance between each female plant used in the above experiments and the closest four
flowering male plants was measured. The number of inflorescences on each of the four
nearest male plants was recorded. This information was used to construct a weighted local
pollen index, indicating local densities of pollen from nearby male inflorescences. The
following formula was devised:

s1 s2 s3 s4
Distance-weighted local pollen index (DWPI) = 2
+ 2
+ 2
+ 2
d1 d2 d3 d4

where s is the number of male inflorescences on the four nearest male plants and d is the
distance in metres to the plant (indicated by the subscripts given). The DWPI decreases
with the square of the distance and increases linearly with the number of inflorescences on
each male. A larger value results when females are close to males with many
inflorescences.

2.3.2 Part (B): Population level flowering effort - male to female flowering densities

Isolated-female experiment

The aim of this study was to reduce the local density of flowering male plants to imitate a
low mast year when female and male plants might flower out of synchrony, or are
separated by increasing distance. To investigate the effects of male floral densities (i.e.
pollen availability) on female reproductive success (seed set), floral densities were
manipulated at separate plots at Burkes and Hakataramea Passes by removing male
inflorescences. At Lewis Pass, the population structure was such that there were a number
of flowering female plants that were naturally isolated (i.e. far away from flowering male
plants) therefore no experimental removal of flowering males was carried out.

44
 Isolated female plants

Distance to nearest
flowering male plant

Fig 2.5 Diagrammatic representation of isolated female experimental design to investigate the effects of male
floral densities on female reproductive success. All flowering male inflorescences removed from plot (inner
rectangle), leaving only female plants at varying distances to male plants. Female plants (∆) in the centre of
the plot are more isolated from male plants (○) than females nearer the edge of the plot. Arrows represent
distance between female and nearest flowering male plant outside plot. This distance was measured for each
female.

One plot was marked out at each site, where all developing male inflorescences inside the
plot were removed. Only female inflorescences remained inside the plot and males were
left randomly distributed around the edge of the plot. Female plants were thus at varying
distances to the nearest flowering male (see Fig. 2.5). Within the plot, female plants with 2
or more inflorescences were randomly selected and tagged to enable paired design
sampling (see Table 2.4 for plot information). The position of each female plant was
mapped, as well as the position of each flowering male plant around the outskirts of the
plot. The distance between each experimental female and the nearest flowering male plant
outside the plot was measured. On each experimental plant, one inflorescence was given
supplemental pollen treatment and the other was left naturally pollinated. The hypothesis
was that pollen limitation (hand-pollinated seed set – control seed set) would be stronger in
females further from males. Hand pollination was performed 3 separate times over
approximately 3 weeks (see below).

45
Table 2.4 Plot information for isolated-female experiment. Pollen limitation was tested for by pollen
supplementation i.e. hand pollinated. (* = no hand-pollination performed at Lewis Pass site due to adverse
weather).

Number Number Total number Dist. range to

Plot size hand naturally of nearest male
Site Species (m2) pollinated pollinated inflorescences (m)

Burkes Pass A. aurea 7,650 20 21 41 1-60

Hakataramea Pass A. aurea 42,780 18 18 36 1-200
Lewis Pass A. 'lewis’ 184,000 0* 38 38 1-500

2.3.3 Pollen limitation

To determine whether seed set of female plants in all of the experiments outlined above at
Burkes and Hakataramea Passes was limited by the availability of pollen, supplemental
pollen was applied by hand to one inflorescence from each pair of treatments. Hand
(supplemental) pollination was done by whacking a section of male inflorescence with a
substantial pollen load over the female inflorescence while covering the non-hand-
pollinated stalk with a paper bag to avoid accidental pollination.

Hand pollination was repeated on each inflorescence three separate times over
approximately three weeks during December 2004, to encompass variation in flower
longevity and stigma receptivity over the course of the flowering season. Hand pollination
was performed only on non-rainy days to avoid pollen loss and wastage. Because there
were no days without rain during field work at Lewis Pass, hand pollination was not
performed there.

Male inflorescences used for hand pollination were taken from outside the experimental
plots, 50 to 300 metres away from the boundary to avoid altering the natural structure of
the population inside the plot. Naturally pollinated inflorescences were left untouched.
After male inflorescences were used for hand pollinating females, they were discarded over
300 metres away from the plot.

46
2.3.4 Other field measurements

For each of the 377 experimental inflorescences, a number of measurements were made at
the: 1) inflorescence, and 2) plant level. These were used to compile detailed data matrices
to gain a more in-depth understanding of how both experimentally manipulated and natural
variables affect reproductive success in Aciphylla.

1) All inflorescences were harvested between late February and early March when fruit and
seeds were filled but not ripe enough to be dispersed. The height of each inflorescence was
re-measured and a categorical estimate of aphid densities on inflorescences was recorded
(low, medium or high). Each inflorescence was then carefully cut near the base of the
rachis and placed into separate paper for storage until seed sorting and counting began.
Some were stored in a freezer, but due to limited space, others were dry-stored.

2) Plant-level measurements recorded were plant area, plant leaf height, rosette number and
number of inflorescences. Diameter was measured as the greatest diameter across the
clump formed by all rosettes belonging to a single plant (length, L), and the diameter
perpendicular to that (width, W). To calculate the area of the plant, the formula for
measuring an ellipse was used (L + W / 2 ) π . Plant leaf height was measured using the
2

distance from the leaf base at ground level to the tip of the tallest leaf. These measurements
were used to test the effect of plant size on reproductive success.

Other plant species flowering at each site were also noted along with the weather each day
field work was done. Insect visitation rates to every inflorescence in all of the experiments
were also recorded with multiple observations (see Chapter 3).

2.3.5 Lab work

2.3.5.1 Measuring reproductive success – counting seeds

Flowers that had been pollinated had developed into fruits by the time plants were
harvested. Reproductive success in females was measured by counting the proportion of

47
 fruits that made seeds (‘filled’) vs. the proportion that did not make seeds (‘unfilled’). Each
‘fruit’ is technically described as a laterally compressed schizocarp consisting of 2
mericarps, each containing one seed (see Fig. 2.6). For this study, seed set not fruit set was
measured by counting each mericarp individually, therefore each mericarp had either zero
or one seed (i.e. ‘filled’ or ‘unfilled’ respectively). For every inflorescence a sub-sample of
seeds were sorted and counted by hand (see below for sub-sampling methods). A total of
122,811 seeds were counted. Aside from the 2 categories of whether a fruit was ‘filled’ or
‘unfilled’ with a seed, the proportion of predated seeds (both fully and partially) was also
recorded. Predated seeds were assumed to have once been ‘filled’ before they were eaten
by an invertebrate seed predator, and hence were also counted as ‘filled’.

This measure of reproductive success in female plants slightly differed to methodology

used previously by Brookes & Jesson (2006). Rather than testing each seed for viability to
identify germinability, the simpler categorisation of seed ‘filled’ or ‘unfilled’ was
employed. Proportion of filled seeds will from now on be referred to as ‘percent seed set’.
This method is frequently used for quantifying pollination efficiency and although it is not
a direct measure of pollination success, it does set a lower limit (Tisch, 1996). This method
was used due to large sample sizes and time constraints. Slightly different methods of seed
counting were employed for A. aurea and A. scott-thomsonii as explained below.

Winged
Mericarps

Schizocarp

Carpophores

Fig. 2.6 Intact developing fruit, or schizocarp, (left) consisting of 2 winged mericarps, each containing one
seed. The carpophore is the centre stalk that attaches to the mericarps. The picture on the right shows the fruit
at maturity with mericarps still attached prior to dispersal.

48
Aciphylla aurea and A. ‘lewis’

For each inflorescence, all fruits were removed from the entire length of the inflorescence
by shaking them off the stalk into a paper bag. The contents of the bag were shaken well to
mix them, then a random sub-sample of 250-350 developed and undeveloped fruits were
selected out for counting. The number of seeds that fell into each category (filled, unfilled
and eaten) was recorded. I counted only 250-350 seeds per inflorescence, as trials showed
that the ratio of filled: unfilled did not change with increasing number counted beyond that.

Aciphylla scott-thomsonii

For A. scott-thomsonii, all seeds were counted from every fifth lateral umbel along the axis
of the inflorescence. This method was used to obtain an idea of whether seed set rate varied
along the inflorescence. Only the grand mean seed set rate data for each inflorescence
however, will be presented in this section. See Appendix 1 for more detail of seed set along
the inflorescence and how flower number per umbel can be accurately predicted from its
particular location along the rachis.

2.3.5.2 Germination experiments

Sub-samples of seeds from all species and sites were randomly selected from the
experimental inflorescences in an attempt to germinate some of the seeds. This experiment
began in late August 2005. The purpose of this was to test the assumption that ‘filled’ fruits
were viable and ‘unfilled’ fruits were not. Because the seeds had been dry-stored since
harvest in late-February to early-March, soaking them before sowing was recommended
(Metcalf, 1995). One hundred filled and 100 unfilled fruits were separated out for each
species from each site and soaked in warm water overnight. These were then surface-
sterilised in 50% bleach solution for 10 minutes and a soak/rinse cycle in distilled water
was performed 3 times to clean off the bleach. Seeds were placed into sterilised Petri dishes
lined with damp germination paper, sealed with plastic wrap and refrigerated for 4 weeks to
imitate cold winter field conditions. In mid-October, seeds were removed from cool storage
and Petri dishes were then left at room temperature and kept moistened for ten months.

49
2.3.6 Statistical Analyses

Statistical analyses were performed in the statistical package R version 1.5.1 and in
Microsoft Excel 2003. Generalised Linear Models (GLM's) were used to examine the
influence of the various treatments on rates of seed set. For all GLM’s with the response
variable ‘seed set’, a binomial error distribution with a logit link function was used because
there were two possible outcomes for the fate of a fruit: ‘filled’ or ‘unfilled’. Because the
assumption of independence necessary for use of tests based on the chi-squared distribution
appeared to be breached, the fit of all models was tested using F-ratios.

For each experiment performed at each site, a large multi-variate model was created to test
for the effects of each predictor on seed set rate. Model selection was attempted in multiple
ways, with the aim of fitting a model that could best explain variation in seed set rate and
that would allow for comparison between sites. Model construction was performed by
initial inclusion of all factors and variables of interest. The order of predictors in the model
was varied for each site until a generic order that could be used for all sites was selected.
Predictors lacking significant explanatory power were then removed from the model. For
both the floral display size (A) and isolated-female (B) experiments, five predictors were
used in the final model in the following order: pollination treatment, local pollen
availability (using DWPI or distance to nearest male inflorescences), percent flower
removal, flowering length, and inflorescence height. See Table 2.1 for predicted
relationships of all explanatory predictors and seed set rate (the order of variables are
shown in the GLM tables in section 2.4).

When testing various models to select the most appropriate one to explain variation in seed
set rate, percent flower removal (as a factor) explained little or no significant variation in
the data on its own at most sites. When the variable flowering length was added to the
model after percent flower removal, it usually explained a greater proportion of the
variance while percent flower removal explained little or none. Of the few sites where
percent flower removal explained the most variation, flowering length explained almost
none. Therefore, these two predictors were highly correlated, as expected. Flowering length
was essentially the same predictor as percent flower removal (because most of the variation
in final flowering length was due to experimental removal, rather than variation among

50
unmanipulated inflorescences), but it expresses it as a continuous variable rather than a
factor with three levels of flower removal. Different sites responded to either one or the
other of these two predictors measuring “variation in the amount of flowers on an
inflorescence”, but were both retained in the final model to maintain consistency between
sites.

Apart from analysing the effects of floral display on seed set rate by site and species (for
floral display at multiple levels) using multi-variate models, some analyses were also done
by treatment and presented graphically to allow for more convenient comparison between
sites and species. The effect of the interaction between pollen supplementation and flower
removal (testing for simultaneous pollen versus resource limitation), was presented using
two separate graphs. The first compared seed set rate against percent flower removal as a
factor for supplemental and natural pollination, while the second compared seed set rate
and flowering length (cm) as a continuous variable for supplemental and natural
pollination.

2.4 RESULTS

2.4 Part (A): Effects of insect-exclusion bags, plant-level floral display, hand-
pollination, and local pollen on seed set

2.4.1 Mean seed set rates of control inflorescences at all sites

Seed set rates for control inflorescences varied between sites (Fig 2.7). Overall, Aciphylla
scott-thomsonii at Hakataramea Pass (HP) had the highest mean seed set rate of 87%, while
also at Hakataramea, seed set in A. aurea was relatively high (75% at HP1 and 72% at
HP2). Seed set at Lewis Pass (LP) was the next highest (66%). Aciphylla aurea at both
Burkes Pass (BP) sites experienced the lowest levels of seed set (57% BP1 and 53% at
BP2).

51
 100
90
Mean seed set rate (controls) 80
70
60
50
40
30
20
10
0
A. aurea A. aurea A. aurea A. aurea A. scott A. ‘lewis’

BURKES PASS HAKATARAMEA PASS LEWIS PASS

Fig 2.7 The natural variation in seed set rates amongst sites using control inflorescences, including all
experiments (+/– 1 SE). The two bars for A. aurea at both Burkes Pass and Hakataramea Pass indicate mean
seed set rates for both experiments: floral display size (left) and isolated female (right).

2.4.2 Effects of bagging on seed set rate

Bagged inflorescences excluding insect pollinators experienced a highly significant

decrease in seed set rate, compared with controls, at all sites (Table 2.5; see also Fig. 2.8).
Seed set inside bags was expected to be zero or nearly so, yet in three cases mean seed set
rates inside bags were over 20%, which is surprisingly high, but still significantly less than
non-bagged plants suggesting insects played a vital role in pollen movement.

52
Table 2.5 Mean seed set rates of bagged versus control inflorescences. Statistical values given are for the
effects of bagging on seed set rate from binomial GLM’s. (Abbreviations: BP = Burkes Pass, HP =
Hakataramea Pass and LP = Lewis Pass).

Species Mean seed set rate (%) % Deviance

and sites Control Bag F (d.f) explained P-value

A. aurea BP 52.0 23.8 354.24 (1,11) 50.33 < 0.0001

A. aurea HP 76.8 37.5 430.31 (1,9) 44.96 < 0.0001

A. scott-thomsonii HP 91.4 23.7 3067.4 (1,5) 92.68 < 0.0001

A. 'lewis' LP 79.1 9.2 1320.9 (1,8) 91.99 < 0.0001

2.4.3 Effects of floral display on seed set

2.4.3.1 Analyses by site

Burkes Pass – Aciphylla aurea

At Burkes Pass, all predictive terms had a significant effect on seed set rate. Inflorescence
height was the most significant predictor and accounted for over 13% of the variation in
seed set rate (Table 2.6). As expected, taller inflorescences had a higher rate of seed set.
Local pollen abundance (DWPI) also explained significant variation in seed set rates, and
as expected, an increase in local pollen abundance contributed to a higher seed set rate.
Hand pollinated inflorescences also set significantly more seed than those left openly
pollinated, although this explained only 1.7% of the variation in the model. Flower-removal
treatment significantly decreased seed set rate, suggesting that with an increase in the
number of flowers along the inflorescence stem, comes an increase in seed set rate.
Therefore, results demonstrate that taller inflorescences with more flowers have a higher
rate of seed set than shorter ones at Burkes Pass, and seed set rate is further increased when
there is more local pollen available. The model accounted for 25.7% of the variation in
seed set rates at Burkes Pass (ranging from 33-92%, see Fig. 2.8). There was also a
significant interaction between pollination and percent flower removal (see Fig. 2.9).

53
Table 2.6 Analysis of Deviance Table for A. aurea at Burkes Pass showing variation in seed set rate as a
response to five predictive terms. The terms DWPI (Distance-Weighted local Pollen Index), flowering length
and inflorescence height are non-manipulated variables while the terms pollination (hand vs. natural) and
percent flower removal (0, 40 & 80%) are manipulated factors. Finally shown is the interaction between the
two manipulated factors – pollination and flower removal. The double signs next to flower removal indicate
where there were 3 levels of removal. The first sign represents the direction of the relationship between the
control (0%) and 40% removal and the second between 0% and 80% removal. † = refer to Fig. 2.8 for
pollination x flower removal effects.

Terms added Df F Resid. Df Resid.dev % Dev. expl. P-value Slope

NULL 45 1426.73

Hand pollination 1 23.62 44 1403.11 1.66 <0.001 +

DWPI 1 98.91 43 1304.20 6.93 <0.001 +

Flower removal (%) 2 16.22 41 1271.75 2.27 <0.001 –/––

Flowering length (cm) 1 10.37 40 1261.38 0.73 0.0013 –

Inflorescence height (cm) 1 191.95 39 1069.43 13.45 <0.001 +

Pollination x Flower removal 2 4.68 37 1060.08 0.66 0.0093 †

Hakataramea Pass – Aciphylla aurea

The most significant variable affecting seed set in A. aurea at Hakataramea Pass was
inflorescence height, although the direction of this relationship was not as expected. As
inflorescence height increased, seed set rate significantly decreased (Table 2.7). Percent
flower removal also significantly decreased seed set rate as expected. Hand pollination
significantly increased seed set along with greater local pollen supply (DWPI). A
pollination x flower removal interaction also produced significant effects on seed set rates
(see Fig. 2.9 for detail). Overall, this model (Table 2.7) accounted for 26.3% of the
variation in A. aurea seed set at Hakataramea Pass (which ranged from 42-95%, see Fig.
2.8).

54
Table 2.7 Analysis of Deviance Table for A. aurea at Hakataramea Pass showing variation in seed set rate
as a response to five predictive terms. Terms and symbols are as for Table 2.6.

Terms added Df F Resid. Df Resid.dev % Dev. expl. P-value Slope

NULL 39 811.9

Hand pollination 1 13.83 38 798.07 1.70 <0.001 +

DWPI 1 20.94 37 777.13 2.58 <0.001 +

Flower removal (%) 2 35.04 35 707.05 8.63 <0.001 –/––

Flowering length (cm) 1 0.05 34 707 0.01 0.82 –

Inflorescence height (cm) 1 92.32 33 614.67 11.37 <0.001 –

Pollination x Flower removal 2 8.10 31 598.47 2.00 <0.001 †

Hakataramea Pass – Aciphylla scott-thomsonii

In A. scott-thomsonii at Hakataramea Pass, the most significant outcome was that with an
increase in flowering length there was significant decrease in seed set rate (Table 2.8). The
variable flowering length explained more than 40% of the variation in seed set, accounting
for over four fifths of the variation that the entire model explained (which was 49.4%). The
range in seed set rate was 63-97% in A. scott-thomsonii. Taller inflorescences had a higher
seed set rate, although this explained only 4.7% of the deviance (Table 2.8). Hand
pollination only slightly increased seed set rate (see Fig. 2.8), although this was highly
statistically significant. Local pollen abundance was relatively unimportant at this site, as
was percent flower removal, although results significantly showed that percent seed set
increases with more flowers, as expected. Thus there is strong evidence for inflorescences
that are tall but with fewer flowers having higher seed set in A. scott-thomsonii.

55
Table 2.8 Analysis of Deviance Table for A. scott-thomsonii at Hakataramea Pass showing variation in
seed set rate as a response to five predictive terms. Terms and symbols are as for Table 2.6.

Terms added Df F Resid. Df Resid.dev % Dev. expl. P-value Slope

NULL 30 1965.61

Hand pollination 1 13.65 29 1951.96 0.69 <0.001 +

DWPI 1 5.19 28 1946.77 0.26 0.023 –

Flower removal (%) 2 21.94 26 1902.89 2.23 <0.001 –/––

Flowering length (cm) 1 801.09 25 1101.81 40.76 <0.001 –

Inflorescence height (cm) 1 93.24 24 1008.57 4.74 <0.001 +

Pollination x Flower removal 2 6.88 22 994.8 0.70 0.001 †

Lewis Pass – Aciphylla ‘lewis’

The total variation in seed set rate explained by the Lewis Pass model was just 13.7%, with
seed set values ranging from 2-96%. The most significant explanatory term was flowering
length which accounted for around 8.5% of the variation in seed set in A. ‘lewis’ (Table
2.9). Seed set was greater with an increase in flowering length. Distance to the nearest
flowering male plant (pollen availability) was also highly significant, with isolated female
inflorescences experiencing lower rates of seed set. Also statistically significant was an
increase in seed set rate with increasing inflorescence height. Pollen limitation was not
tested for at Lewis Pass although with mean seed set rates of around 65% (see Figs 2.7 to
2.9) it would have been interesting to test whether seed set rates would increase if
supplementary pollen was applied.

56
Table 2.9 Analysis of Deviance Table for A. ‘lewis’ at Lewis Pass showing variation in seed set rate as a
response to four predictive terms. Terms and symbols are as for Table 2.6. Note: no pollination treatment was
carried out at Lewis Pass.

Terms added Df F Resid. Df Resid.dev % Dev. expl. P-value Slope

NULL 20 1611.49

Nearest male 1 58.05 19 1553.44 3.60 <0.001 –

Flower removal (%) 2 14.65 17 1538.79 0.91 <0.001 –/+

Flowering length (cm) 1 136.53 16 1402.26 8.47 <0.001 +

Inflorescence height (cm) 1 10.9 15 1391.36 0.68 <0.001 +

The effect of bract removal on seed set

Inflorescences with bracts removed had a significantly higher seed set on A. aurea at
Burkes Pass, A. scott-thomsonii at Hakataramea Pass and on A. ‘lewis’ at Lewis Pass. Bract
removal on A. aurea inflorescences at Hakataramea Pass however, had a weak but
significant negative effect on seed set (Table 2.10).

Table 2.10 The effects of removing inflorescence floral bracts on seed set rate. Statistical values given are for
the effects of bract removal (as a factor) on seed set rate, using binomial GLM’s. (Abbreviations: BP =
Burkes Pass, HP = Hakataramea Pass and LP = Lewis Pass).

Species Mean seed set rate (%) % Deviance

and sites Control Bract removal explained F (d.f) P-value
A. aurea BP 51.96 58.79 1.7 14.7 (1,15) <0.001
A. aurea HP 75.43 71.1 2.1 4.2 (1,11) 0.04
A. scott-thomsonii HP 86.87 89.17 1.3 5.7 (1,10) 0.02
A. 'lewis' LP 66.14 83.99 12.3 64.7 (1,13) <0.0001

57
2.4.3.2 Analyses by treatment and graphical representation to compare sites

The following figures depict graphically, the effect of each predictive term on its own, by
site. The graphs are based on the same datasets from which the above generalised linear
models were created, but linear models were fitted to show the trend from each main effect.
The graphs do not take into account the variation explained by the other variables in the
multivariate models, but they are useful for showing overall trends and making site
comparisons. The above GLM’s should be referred to for statistical values.

Flower-removal, pollination, and bagging effects on seed set

Overall, seed set rate was relatively high across all sites with the exception being for bag
treatments where seed set rates were consistently lower than all other treatments (Fig. 2.8).
Seed production in open-pollinated inflorescences was, in most cases, significantly less
than in hand-pollinated inflorescences (Tables 2.6 – 2.9), although the magnitude of this
effect was small. Across sites, this pattern was evident at the 0% and 40% flower removal
levels, but inflorescences with 80% flower removal were not affected (Figs. 2.8 and 2.9).

58
 Burkes Pass – A. aurea Hakataramea Pass – A. aurea
Burkes - A.aurea Hakataramea - A.aurea

80

80
Seed Set Rate

Seed Set Rate

60
60

40
40

20
20

0h 0n 40h 40n 80h 80n bag 0h 0n 40h 40n 80h 80n bag
Treatment Treatment

Hakataramea
Hakataramea -–A.scott-thompsonii
A. scott-thomsonii Lewis
LewisPass
Pass–- A. ‘lewis’
A.'lewis'

80
80
Seed Set Rate

Seed Set Rate

60
60

40
40

20
20

0h 0n 40h 40n 80h 80n bag 0 40 80 bag

Treatment Treatment

Fig. 2.8 Box and whisker plots for all sites showing the variation in seed set rates for each level of flower
removal, pollination and bag treatment. The middle horizontal line within the box shows the median seed set
rate. Also shown are the upper and lower quartiles (lines either side of the median), and whiskers
encompassing 95% range of the data (designated by dashed lines), with outliers that fall beyond the 95%
range/whiskers (dots). See Tables 2.6 to 2.9 for statistical tests.

Simultaneously testing for the effects of resource and pollen limitation on seed set

According to the multi-variate site GLM’s, for A. aurea at Burkes and Hakataramea Passes,
the factor percent flower removal was a stronger predictor of seed set rate than the
continuous variable flowering length, while flowering length was a significantly stronger
predictor for A. scott-thomsonii than percent flower removal (Tables 2.6, 2.7 and 2.8

59
respectively). Both terms fundamentally represent the same variable but each was
measured in a different way (see Methods section). This suggests that Fig 2.9 is more
appropriate for interpreting the relative importance of simultaneous pollen and resource
limitation for A. aurea while Fig. 2.10 is more appropriate for A. scott-thomsonii.

Percent flower removal and hand pollination

Figure 2.9 depicts the effects of the interactions between hand pollination and flower
removal. If both pollen and resources simultaneously limited seed set, hand pollination of
plants with 80% flower removal treatment would result in an increase in seed set compared
to the control (see Fig. 2.2). Overall, there was no evidence of resource limitation in any of
the Aciphylla species in this experiment because there was no significant increase in seed
set rates for hand pollinated inflorescences with flowers removed (i.e. inflorescences with
more resources per flower). This suggests that the biological importance of resource
limitation was not particularly relevant within one flowering year when measured using
percentage of flowers removed.

To some degree, results for A. aurea at Hakataramea Pass conformed to the expected
effects that the pollinator attraction hypothesis would have on seed set rates, i.e.
inflorescences with more flowers had higher seed set rates (Fig. 2.9). This outcome did not
agree with the predictions made by the pollen and resource limitation model (see Fig. 2.2c)
as there was no evidence for resource limitation, because inflorescences with 40 or 80% of
the flowers removed did not have higher seed set rates relative to the controls. This
suggests that other factors, not tested for here, were possibly affecting the ability of
inflorescences with flowers removed to set more seed despite theoretically having more
resources allocated per flower to develop seeds.

60
 Burkes Pass – A. aurea Hakataramea – A. scott-thomsonii
100 100

95 95

90 90

85 85

80 80

75 75

70 70

65 65

60 60

55 55

50 50
0 40 80 0 40 80
Percent seed set

Hakataramea – A. aurea Lewis Pass – A. ‘lewis’

100
100
95
95
90
90
85
85
80
80
75
75
70
70

65 65

60 60

55 55

50 50
0 40 80 0 40 80

Percent flower removal

Fig. 2.9 Interaction plots showing the effects that simultaneous pollen and resource limitation following
flower removal and hand pollination (± 1 SE) had on seed set rates. Dark triangles (▲) represent hand
pollinated inflorescences and light squares (■) represent naturally pollinated inflorescences. Note: no hand
pollination performed at Lewis Pass due to weather. See Tables 2.1 to 2.4 for statistical significance values.

Consistent among most sites however, was evidence for pollen limitation, although in
biological terms, this effect was weak. Natural pollen limitation can be inferred when
inflorescences with zero percent of flowers removed have higher rates of seed set when
supplementary pollinated. This was evident among all three sites where inflorescences

61
were hand pollinated (Fig. 2.9) but the effects were very small, particularly in A. scott-
thomsonii (see Table 2.8).

Flowering length and hand pollination

Flowering length (after manipulations) varied between sites and species: Aciphylla aurea
BP ranged from 5 cm to 90 cm ( x = 43 cm ±2.9 SE, n = 62), A. aurea HP ranged from 6
cm to 110 cm ( x = 50 cm ±3.3 SE, n = 55), A. scott-thomsonii ranged from 21 cm to 152
cm ( x = 50 cm ±5.1 SE, n = 50) and A. ‘lewis’ LP ranged from 13 cm to 128 cm ( x =
60.6 cm ±5.6 SE, n = 31). Overall, the effects of increasing flowering length on seed set
rates in A. aurea and A. ‘lewis’ were very small with the direction of the relationship
differing between sites.

The effects of hand pollination were also small, although as expected, adding pollen did
increase seed set compared with open-pollinated inflorescences (Fig. 2.10). For A. aurea at
Burkes Pass there was a slight negative effect of flowering length on seed set, while in A.
‘lewis’, an increase in flowering length was important in contributing to higher seed set
rates. However, the direction this relationship at Lewis Pass appears to be strengthened by
the outliers at the bottom-left of the graph. These were inflorescences that had stunted
growth and relatively low seed set compared with the rest of the Lewis Pass population.
There was considerable scatter in the A. aurea data at Hakataramea with no significant
effect of flowering length on seed set rates (Table 2.7). In the multivariate GLM (Table
2.8), flowering length was important in explaining a significant amount of the variation in
seed set rate in A. scott-thomsonii, however, Fig. 2.10 shows the relationship between
flowering length and seed set to be quite weak in contrast. In the multivariate GLM,
flowering length is probably having the main effect on seed set, as the table allows for
explanation by other variables. In A. scott-thomsonii the effects of hand pollination slightly
decreased with increasing flowering length, again suggesting alternative factors may have
been influencing the ability for inflorescences with flowers removed to set more seed as
would be expected (Fig. 2.10).

62
 Burkes – A. aurea Hakataramea – A. scott-thomsonii
Burkes - A. aurea Hakataramea - A. scott-thomsonii

100

100
80

80
Seed set rate

Seed set rate

60

60
40

40
20

20
0

0
0 20 40 60 80 100 0 50 100 150

Flowering length (cm) Flowering length (cm)

Hakataramea – A. aurea Lewis Pass – A. ‘lewis’

Hakataramea - A. aurea Lewis Pass - A. lewis
100

100
80

80
Seed set rate

Seed set rate

60

60
40

40
20

20
0

0 20 40 60 80 100 0 50 100 150

Flowering length (cm) Flowering length (cm)

Fig. 2.10 The effect of inflorescence flowering length on seed set rate for hand vs. natural pollination
treatment at all sites. Dashed lines (----) and solid symbols represent hand-pollinated inflorescences and
unbroken lines (–––) lines and hollow symbols represent control inflorescences. Note: x-axis ranges are
standardised between the two A. aurea sites, and between A. scott-thomsonii and A. ‘lewis’ sites. No hand
pollination was performed at Lewis Pass due to weather. See Tables 2.6 to 2.9 for statistical significance
values.

63
The effect of inflorescence height on seed set

Mean inflorescence height varied between sites and species: A. aurea BP ranged from 80
cm to 155 cm ( x = 122.2 cm ±2.3 SE, n = 62), A. aurea HP ranged from 62 cm to 190 cm
( x = 141.5 cm ±3.4 SE, n = 55), A. scott-thomsonii HP ranged from 90 cm to 280 cm ( x =
212.4 cm ±5.5 SE, n = 50) and A. ‘lewis’ LP ranged from 74 cm to 164 cm ( x = 128 cm
±4.7 SE, n = 31). The relationship between inflorescence height and seed set rate is shown
in Figure 2.11. Seed set rate increased with inflorescence height in all sites except A. aurea
at Hakataramea. This relationship was strongest for A. aurea at Burkes Pass and positive,
though weakly significant in A. scott-thomsonii and A. ‘lewis’ (see Tables 2.6 to 2.9). A
small number of outliers are evident in the bottom two graphs of Figure 2.11. There are
occasional inflorescences with very low seed set, for unknown reasons. The outlier present
at the A. aurea site at Hakataramea Pass was an inflorescence subjected to intense
caterpillar damage.

64
 Burkes -A.
Burkes A.aurea
aurea Hakataramea
Hakataramea -- A.
A.scott-thomsonii
scott-thomsonii

100

100
80

80
Seed set rate

Seed set rate

60

60
40

40
20

20
0

0
60 80 100 120 140 160 100 150 200 250
Inflorescence height (cm) Inflorescence height (cm)

Hakataramea
Hakataramea–-A.
A. aurea
aurea Lewis
LewisPass
Pass- -A.
A.‘lewis’
lewis
100

100
80

80
Seed set rate

Seed set rate

60

60
40

40
20

20
0

50 100 150 200 60 80 100 120 140 160 180

Inflorescence height (cm) Inflorescence height (cm)

Fig. 2.11 The effect of inflorescence height on seed set rate at all sites. See Tables 2.1 to 2.4 for statistical
significance values. Axes are standardised to the same intervals to allow for comparison between sites.

65
2.4 Part (B): Effects of female isolation, hand-pollination and pollen on seed set

2.4.4 Analyses by site

Burkes Pass site 2 – Aciphylla aurea

For the isolated female experiment at Burkes Pass, where local pollen supply was
experimentally manipulated by removing male inflorescences, close to 40.5% of the total
variation in seed set rate of A. aurea could be accounted for by the model (Table 2.11).
There was no experimental removal of flowers within an inflorescence at this site or at
Hakataramea site 2 (also A. aurea). Seed set rates ranged between 18-81% at Burkes Pass
site 2. Natural (un-manipulated) flowering length was the most significant variable,
explaining almost 35% of the variation in seed set. Also significant was distance to male
inflorescences (an indicator of the local pollen abundance). As expected, the further from
flowering males, the lower the seed set rate. Hand pollination only slightly increased seed
set rates and inflorescence number had a very weak negative effect on seed set.

Table 2.11 Analysis of Deviance Table for isolated female experiment in A. aurea at Burkes Pass (site 2).
All terms added in to the GLM are non-manipulated variables apart from hand pollination which was a
manipulated treatment.

Terms added Df F Resid. Df Resid.dev % Dev. expl. P-value Slope

NULL 37 1622.18

Hand pollination 1 19.23 36 1602.95 1.19 <0.001 +

Nearest male 1 75.33 35 1527.62 4.64 <0.001 –

No. inflorescences 1 6.5 34 1521.12 0.40 0.011 –

Flowering length (cm) 1 554.89 33 966.24 34.21 <0.001 +

Inflorescence height (cm) 1 0.43 32 965.81 0.03 0.51 +

66
Hakataramea Pass site 2 – Aciphylla aurea

The most significant variable affecting seed set rate in A. aurea at Hakataramea Pass site 2 was
inflorescence number. Plants with more inflorescences achieved higher rates of seed set (Table
2.12) as expected. Inflorescence height significantly decreased seed set rate, while flowering
length significantly increased seed set, but only weakly. Hand pollination did not further
improve seed set, and distance to nearest flowering male plant had a very weak but significant
negative effect. Seed set rates ranged between 49-92% and the model accounted for only
16.4% of the variation observed in seed set rates in A. aurea at Hakataramea Pass.

Table 2.12 Analysis of Deviance Table for isolated female experiment in A. aurea at Hakataramea Pass
(site 2). All terms added in to the GLM are non-manipulated variables apart from hand pollination which was
a manipulated treatment.

Terms added Df F Resid. Df Resid.dev % Dev. expl. P-value Slope

NULL 32 709.6

Hand pollination 1 0.35 31 709.25 0.05 0.55 +

Nearest male 1 5.7 30 703.55 0.80 0.02 –

No. inflorescences 1 64.01 29 639.55 9.02 <0.001 +

Flowering length (cm) 1 5.38 28 634.17 0.76 0.02 +

Inflorescence height (cm) 1 40.6 27 593.57 5.72 <0.001 –

Distance to males was also highly significant for A. ‘lewis’ at Lewis Pass, with isolated
females having lower seed set rates. This result was presented in section 2.4.3.1 (see Table
2.9)

Are isolated females pollen limited? The effects of local pollen abundance on seed set

The Distance-Weighted local Pollen Index (DWPI) value gets bigger with increasing
proximity to males with more inflorescences, therefore this represents a more abundant
local pollen supply. In A. aurea at both Burkes and Hakataramea Passes, DWPI had a

67
significant positive effect on seed set rates at sites of un-manipulated male and female
densities, indicating that proximity to males (and hence synchronous flowering) is
important in A. aurea, with females not being too isolated from a pollen source to ensure
adequate seed set (Fig. 2.12).

Burkes – A. aurea Hakataramea – A. scott-thomsonii

Burkes - A. aurea Hakataramea - A. scott-thomsonii
100

100
80

80
Seed set rate

Seed set rate

60

60
40

40
20

20
0

0
0 2 4 6 8 0.0 0.5 1.0 1.5 2.0 2.5 3.0

Local pollen abundance (DWPI) Local pollen abundance (DWPI)

Hakataramea – A. aurea
Hakataramea - A. aurea
100
80
Seed set rate
60
40
20
0

0 5 10 15 20

Local pollen abundance (DWPI)

Fig. 2.12 The effect of local pollen abundance (measured with the Distance-Weighted local Pollen Index, or
DWPI) on seed set rate at Burkes and Hakataramea Passes. Larger DWPI values indicate higher local pollen
availability. Dashed lines (----) and solid symbols represent hand-pollinated inflorescences and unbroken
lines (–––) lines and hollow symbols represent control inflorescences. See Tables 2.6 to 2.8 for statistical
significance values.

68
This is supported by the manipulated isolated female experiment results for A. aurea where
seed set rate decreased with increasing distance from the nearest flowering male plants at a
larger scale (Fig. 2.13). The same trend was also present for A. ‘lewis’, with isolated
females experiencing lower seed set rates. Local pollen abundance did not improve rates of
seed set in A. scott-thomsonii.

Burkes expt.
Burkes - A.2aurea
– A. aurea Hakataramea
Hakatarameaexpt.
expt.22 –- A.
A. aurea
aurea
100

100
80

80
Seed set rate

Seed set rate

60

60
40

40
20

20
0

0 10 20 30 40 50 60 0 50 100 150 200

Distance to nearest male (m) Distance to nearest male (m)

LewisPass
Lewis Pass– -A.
A. ‘lewis’
lewis
100
80
Seed set rate
60
40
20
0

0 10 20 30 40 50 60 70

Distance to nearest male (m)

Fig. 2.13 The effects of increasing distance to the nearest male inflorescence on seed set rate at isolated
female sites, where male to female flowering ratios were experimentally manipulated. Dashed lines (----) and
solid symbols represent hand-pollinated inflorescences and unbroken lines (–––) lines and hollow symbols

69
represent control inflorescences. Note the difference in degree of isolation between the same species at
different sites. No hand pollination was performed at Lewis Pass due to weather.
2.4.5 Seed germination

Although seed germination tests were conducted to determine the germination rates of
‘filled’ seeds and to see whether any of the seeds classified as ‘un-filled’ germinated, there
was insufficient time to yield such results. Only two A. aurea seeds from Hakataramea and
one A. aurea seed from Burkes Pass germinated during the 8 month period, all of ‘filled’
seeds. There was no germination of ‘unfilled’ seeds. Aciphylla seeds are usually known to
remain dormant for a substantial period of time and often germinate over a 3-4 year period
(Metcalf, 1995).

2.4.6 Summary

Overall, there were many statistically significant effects of floral display, hand pollination
and local pollen abundance on seed set rates. Many of these effects were however,
biologically weak. The directions of the effects were mostly as expected (see Table 2.2),
but there were notable exceptions.

Table 2.13 presents the overall direction of the significant relationships between seed set
rates and predictive variables compared to the expected direction of the relationships. Most
consistently, hand pollination increased seed set rate at all sites. Also as expected, seed set
rate nearly always increased with either more local pollen (DWPI) or decreased with
increasing distance to male inflorescences, i.e. female plants closer to males improved seed
set rate. This suggests that pollinator attraction is important in A. aurea and A. ‘lewis’, as
an increase in local pollen supply lead to higher reproductive success. The only exception
to this pattern was A. scott-thomsonii at Hakataramea Pass where seed set was not
significantly affected by local pollen supply in biological terms.

In most cases, removing flowers negatively affected seed set rates, meaning that
inflorescences with more flowers remaining had higher seed set rates with the exception
again being A. scott-thomsonii (Table 2.13). In sites where flower removal was not
performed or where percent flower removal did not explain any variation in seed set rate,

70
an increase in natural flowering length was significantly associated with an increase in seed
set rate, reiterating that floral display size is important for pollinator attraction. Figures 2.9
and 2.10 provided no convincing evidence for resource limitation at any site. Also
relatively consistent among sites and in accordance with the expected direction, increasing
seed set rates were significantly associated with taller inflorescences. Both A. aurea sites at
Hakataramea Pass were the exception to this trend however.

Table 2.13 Hypothesised directions of relationships between seed set rate and predictors, and the actual
outcomes from the GLM summary tables. 0 represents a non-significant result for a particular term and blank
spaces indicate terms that do not apply at certain sites. Lewis Pass was used for both the flower removal and
isolated female experiment. This was a non-manipulated isolated female experiment site (i.e. females were
naturally far from males). The double signs next to the term ‘percent flower removal’ indicate where there
were 3 levels of removal. The first sign represents the direction of the relationship between the control (0%)
and 40% removal and the second between 0% and 80% removal. Since percent flower removal and flowering
length are representing essentially the same variable (length of inflorescence with flowers on), and only one
usually explained significantly more variation than the other in each site model, the stronger effect is
highlighted. Site and species abbreviations: BP = Burkes Pass, HP = Hakataramea Pass, LP = Lewis Pass, A.
scott-th = A. scott-thomsonii, and 1 and 2 refer to experiments 1 and 2 (floral display size and isolated female
experiments respectively).

Expected Flower removal experiment

direction Isolated female experiment
Predictive terms of A. aurea A. aurea A. scott-th A. lewis A. aurea A. aurea
relationship BP HP HP LP BP 2 HP 2

Hand pollination + + + + + +
DWPI + + + –
Nearest male (m) – – – –
No. of inflorescences + – +
Flower removal (%) –/–– –/–– –/–– –/–– –/+
Flowering length (cm) 0 (or +) – 0 – + + +
Inflor. height (cm) + + – + + + –

In summary, pollen provisioning and close proximity to natural pollen sources increased
seed set across almost all species and sites. Moreover, seed set rates increased with taller

71
inflorescences, and overall seed set also increased with greater flowering length in A. aurea
and A. ‘lewis’ but not in A. scott-thomsonii.
2.5 DISCUSSION

In terms of the reproductive strategy employed by Aciphylla i.e. irregular mass flowering,
overall high seed set rates seem to suggest that this reproductive strategy works well for
female plants at least in this high-flowering year. Bagged inflorescences had low seed set
rates, confirming that insects are important pollen vectors in Aciphylla. In most cases,
inflorescences with a larger floral display (either taller in height and/or with more flowers
along the rachis) had higher rates of seed set (although the intricacies of these results are
discussed below). Hand pollinated inflorescences with fewer flowers did not have a
significantly higher seed set rate relative to hand pollinated inflorescences with no flowers
removed, therefore the hypothesis that flower removal would increase resources available
per flower for seed set was not supported. Hence there was no evidence for resource
limitation when defined in this sense. Pollen limitation was evident across all sites because
seed set improved with hand pollination, although the increase was generally small.

2.5.1 Part (A): Floral display at the plant level

2.5.1.1 The importance of insect versus wind as pollen vectors in Aciphylla

Results provide strong evidence that Aciphylla species are pollinator-dependant as the
addition of insect-excluding mesh bags significantly reduced seed set rates to an average of
23%. This is an important finding, as it was unclear in the literature what the relative
importance of wind versus insects as pollinator was (Dawson, 1971; Webb, 1986; Mitchell
et al., 1999). In A. aurea at Hakataramea Pass however, mean seed set inside bags was over
37%. This was unexpectedly high but a possible explanation for high bag seed set rates at
this site is that because this was the last site to be set up, some ovules may have already
been fertilised (due to the more developed stage of some of the flowers) before the bags
were put on. On the other hand, Lewis Pass was the first site where bags were set up, and
this is reflected by the very low levels of seed set that occurred (ca. 9%). The reasons for
the 23% of seeds that were made inside bags at the other sites are unclear but there are a
number of likely explanations. Plants could potentially make seeds inside bags due to

72
either: wind pollination (because mesh bags almost certainly excluded insects but possibly
not wind-blown pollen); apomixis (reproduction by seeds formed without sexual fusion); or
by putting bags on after some flowers were receptive and ovules had already been
fertilised. Note that I checked that leaky dioecy (some male flowers inside the female bags)
was not occurring. These possibilities should be explored in more detail to determine
whether any wind pollination or apomixis is occurring, as this could have important
consequences for reproductive assurance of Aciphylla species in pollinator-limited
environments.

2.5.1.2 The importance of plant-level floral display (and how it is measured) on

reproductive success

According to the pollinator attraction hypothesis, increased floral display size could
enhance female fitness as more offspring could be produced (Davila & Wardle, 2002;
Grindeland et al., 2005). If resources are not limiting, female reproductive success can be
increased through greater allocation to floral display (Ehlers et al., 2002) although there
may be a threshold size for females over which increases in size do not increase seed set
due to the costs of offspring provisioning (Vaughton & Ramsey, 1998). This study showed
that overall, decreasing the size of the floral display in terms of reducing flowering length,
decreased seed set. This was true for all species and sites apart from A. scott-thomsonii at
Hakataramea Pass, where seed set was significantly increased with a decrease in flowering
length. Before discussing the importance of floral display for reproductive success in
Aciphylla, we must first consider the ways in which we define and measure floral display.

Results varied slightly when floral display was measured using different methods. For A.
aurea, the factor percent flower removal was a better measure of variation in the amount of
flowers along the stem than the continuous variable flowering length, as it explained
significantly more variation in seed set. On the other hand, variation in seed set for A. scott-
thomsonii and A. ‘lewis’ was explained significantly by flowering length and not percent
flower removal. In some cases, inflorescence height was the best predictor of seed set,
while unexpectedly, number of inflorescences on a plant was rarely a significant predictor
of seed set, hence its exclusion from the site models for experiment 1 (flower removal).
This study therefore emphasises the importance of the methodology and statistical analyses

73
employed in this type of research, where results and outcomes may vary depending on how
the variable of interest is defined and approached.
With the exception of A. scott-thomsonii, these results suggest that decreasing floral display
(shorter inflorescences and smaller flowering length) reduces attractiveness to insect
pollinators (tested in Chapter 3). A greater floral display has been associated with
increasing pollinator visits in both hermaphroditic and dioecious plants (Andersson, 1996;
Donnelly et al., 1998) but for self-incompatible and dioecious plants, this is particularly
important given their obligate requirement for outcross pollen. Aciphylla being dioecious
requires out-crossing and this appears to be achieved effectively by having a large floral
display. In A. aurea and A. ‘lewis’, taller inflorescences with more flowers experienced
greater seed set during this high-flowering year. The effects of this were further amplified
when the local natural pollen abundance was greater.

Aciphylla scott-thomsonii on the other hand was more successful when inflorescences were
taller but with fewer flowers. This provides an ideal explanation for the common
occurrence of inflorescences within populations that have extraordinarily tall inflorescences
while often only having a low proportion of the inflorescence bearing flowers (see Fig.
2.1). This appears to be a clever strategy on the plant’s behalf because while being taller
was almost always associated with having higher seed set rates, being taller and having
more flowers was not. Therefore the may plant effectively get away with advertising for
pollinators by simply being taller, avoiding the costs of having to put un-necessary
(expensive) effort in to ovule production. The problem here however, is that accessory
costs per flower (or seed) could potentially become very high for stalks with few flowers,
and it may rapidly become uneconomic if flowering length is reduced and the stalk is large.
This is probably overcome by the fact that the stalk is largely made of carbon, while the
more expensive reproductive structures (i.e. seeds) require more nitrogen (Hogan et al.,
1998). Furthermore, inflorescence structures on Aciphylla, including carpels (in females),
spiny bracts, and stalks (males and females) can contribute to their own carbon and energy
costs by photosynthesising (Hogan et al., 1998). Therefore for species with extremely large
inflorescences such as A. scott-thomsonii and A. glaucescens, because seed set is typically
very high (see Hogan et al., 1998)), this natural tendency for plants to produce tall stalks
with fewer flowers is probably an ideal adaptation for pollinator attraction and is an
optimal trade-off between display size and fecundity.

74
These results additionally suggest that there are negative effects on seed set rates in A.
scott-thomsonii on some inflorescences with excess flowers (i.e. there was no further
increase in seed set beyond a certain flowering length, even when hand pollinated).
Alternative reasons for having a high flower-to-fruiting ratio exist and will be discussed in
Chapter 5.

Resource availability, pollen availability, predation, adverse climatic conditions and

genetic conditions may all act to constrain plant fertility (Charlesworth, 1989). Of these
constraints, resource availability and pollen quantity and quality are generally considered
the principal determinants of among-plant variation in female reproductive success (Haig &
Westoby, 1988; Pias & Guitian, 2006). Some authors contend that pollen-limitation is
common (Bierzychudek, 1981; Ackerman & Montalvo, 1990) while others contend that
reproduction in most species is resource limited (Howe & Westley, 1986). Haig and
Westoby (1988) argued that if floral display size is an optimal investment between
pollinator attraction and fecundity, then a plant should be simultaneously limited by both
resources and pollen. Because seed set rates in treatments with flowers removed (i.e.
increased resources) was not greater than seed set in controls when supplemental pollen
was added, there was no support found for simultaneous resource and pollen limitation in
Aciphylla aurea, A. ‘lewis’ or A. scott-thomsonii. However, seed production in open-
pollinated inflorescences was in most cases, significantly less than in hand-pollinated
inflorescences, suggesting that pollinators are failing to supply maximal pollination
services. By providing supplementary pollen, mean seed set rate was (statistically)
significantly higher, although in most cases, this was probably not biologically as
significant, (i.e. mean seed set rate for hand-pollinated inflorescences was never more than
10% greater than the natural rate in open-pollinated inflorescences). This will be discussed
further in Chapter 5.

2.5.2 Part (B): Population level flowering effort

At the population level, males were removed from a large area, leaving female
inflorescences at varying distances from the nearest pollen source. This tested whether
female plants were sufficiently attractive to encourage visitors from further distances and

75
effectively aimed to understand the effects on reproductive success of individuals flowering
out of synchrony during a low-flowering (‘non-mast’) year.

Isolation, measured by the distance to the nearest flowering conspecific male, was a good
predictor of pollen limitation at isolated female experiment sites, because seed set of
individual female plants decreased with isolation. In the site GLM for the isolated female
experiment in A. aurea at Burkes Pass, nearest male was the second most significant
predictor of seed set rate after while natural flowering length. Hand pollination also
significantly increased the seed set rate of isolated females at this site.

At Hakataramea Pass, inflorescence number per plant was a significant predictor of seed
set rate for A. aurea. Female isolation was also important here but much less so than at
Burkes Pass and surprisingly, hand pollination did not significantly increase seed set. It is
likely that the effects of inflorescence number were significant at this site because of the
relatively low flowering densities in that particular area, suggesting that floral display at the
plant-level may play a more important role in pollinator attraction when population
densities are lower. Also at Hakataramea Pass, overall seed set rates for A. aurea were
higher than at Burkes Pass. Because seed set rates were extremely high in A. scott-
thomsonii at Hakataramea, this suggests they were highly attractive to insects. Perhaps by
flowering in synchrony with a highly successful con-generic, A. aurea is reaping the
benefits from being close-by. Alternatively, these differences in fecundity between the two
nearby sites could be attributed to variation in site productivity (but this was not tested
here).

At Lewis Pass, where there was natural female isolation due to the sparse nature of the
population, seed set also significantly decreased with isolation. This was the second-most
important variable after flowering length, in predicting seed set rate. Overall, taller
inflorescences with more flowers were more successful, and there were increased benefits
to this when females were isolated from males.

These results show that the negative effects of isolation and asynchronous flowering are
evident, but remain small in this study. The scale of the experimental design was most
likely too modest to reflect the magnitude of the outcomes likely to occur during natural
low-flowering years. High-flowering densities of Aciphylla remaining outside the

76
experimental area possibly also influenced the overall outcome of these experiments.
Natural low-flowering years of course vary in their degree of flowering density, female
isolation and sex-ratio, but the overall scale would generally be larger. The outcomes of
this study however, are indicative of what would occur during a natural low-flowering year
with isolated individuals occurring naturally at much greater distances from conspecifics,
and provide a helpful insight into the negative effects on individuals if they flowered
during a ‘non-mast’ year. The overall conclusion is that the insect pollinators of Aciphylla
appear to move readily over the scales used in this experiment (up to several hundred
metres), albeit with some decrease in presumed pollen delivery rate (equivalent to rates of
seed set), tested in Chapter 3.

Monitoring the effects of pollen limitation on naturally low-flowering years remains to be

tested but would provide a better insight into whether this is an important selection pressure
for masting in Aciphylla species. Ashman et al. (2004) suggest that a larger response to
experimental pollen supplementation should occur in populations or species that experience
greater variance in pollen acquisition or have smaller costs, therefore in masting species
like Aciphylla, this should be expected between years.

2.5.3 Conclusions

The overall importance of floral display and densities for female reproductive success

There is much evidence for floral display size playing an important role in the reproductive
success of many plant species worldwide, (e.g. Andersson, 1996; Kawarasaki & Hori,
1999). Most studies have considered the importance of display size at the plant level, but
there are only a small number that test for the relative importance of floral display at both
the plant level and on a larger scale at the population and community levels (e.g. Kato &
Hiura, 1990; Bosch & Waser, 1999; Grindeland et al., 2005). To date, there have been
almost no studies showing that reproductive success can be achieved through economies of
scale in masting, insect pollinated, dioecious species (but see Forsyth, 2003). Therefore this
research provides one of the first insights that this paradoxical breeding system works
effectively by testing whether both plant and population level factors influence
reproductive success in Aciphylla.

77
Haig and Westoby’s (1988) argument that plants should evolve to optimally allocate
resources towards pollinator attraction and seed maturation seems to hold true to a certain
degree for the Aciphylla populations studied here. The way they appear to have achieved
this is through masting. A number of insect-pollinated plant species with large floral
displays suffer pollen limitation (e.g. Wagenius, 2005) but masting may be a strategy that
has evolved to mitigate the effects of this. During this high-flowering year, there were no
effects of resource limitation seen, while the effects of pollen limitation were small.
(Brookes & Jesson, 2006) found similar patterns for Aciphylla squarrosa, another large-
leaved montane relative. At the same time, results from the isolated female experiment
suggested that when female plants are flowering out of synchrony or are isolated, the
effects on reproductive success are negative. If reproductive effort is concentrated into
certain years, at the expense of sacrificing reproductive opportunities during others, there
should theoretically be plentiful resources allocated to pollinator attraction, pollen
dissemination (in males) and offspring provisioning (in females) that particular year, and
the effects of resource and/or pollen limitation should be negligible.

It was anticipated that in this study, space would be a good analogue for time, in order to
observe fitness gains to individuals in a ‘mast’ (high flowering density) situation, compared
with a ‘non-mast’ situation (asynchronous/isolated flowering female plants), but the overall
effects were generally very small. For example, at Burkes Pass, for the A. aurea isolated
female experiment, the predicted seed set decreased from 59% to 48% over the range of
experimentally induced low-flowering distances (to the nearest male) of 2 to 55 metres.
When this is compared to the size of the huge benefits from masting known to occur in
other systems, through predator satiation effects (e.g. seed set in Chionochloa pallens at
Mount Hutt increased from 4% to 94 % during mast years (Kelly & Sullivan, 1997)), or
wind pollination (e.g. in Nothofagus solandri, seed set is known to increase from <10% to
60% during mast years (Kelly et al., 2001)), it is harder to see the observed increases in
seed set through pollination economies of scale as being large enough to offset the potential
costs of masting in the Aciphylla populations studied here. However, one limitation is that
the range of variation in seed set could only be measured over the range of naturally
produced inflorescence heights (as inflorescence height could not be manipulated). I
predict that we might see more pronounced negative effects in these larger montane
Aciphylla species if inflorescences were significantly smaller (through decreased pollinator

78
attraction), or more isolated (through decreased pollinator effectiveness) in space and time,
than the experimental ranges in this study.

Chapter 3 - The effects of floral display and population structure on insect visitation
rates and how this relates to reproductive success

3.1 INTRODUCTION

The reproductive biology of the New Zealand flora has many unusual features: flowers are
often small, structurally simple and not showy. Separate sexes (dioecy and monoecy),
fleshy fruit and mast seeding are unusually common (Webb & Kelly, 1993), and pollination
systems have been characterised as unspecialised, imprecise and entomophilous (Newstrom
& Robertson, 2005). This leads to combinations of reproductive strategies that are unusual
and uncommon worldwide (Newstrom & Robertson, 2005) such as Aciphylla, which has a
mast seeding (Webb & Kelly, 1993), largely dioecious and sexually dimorphic (Webb et
al., 1999), insect-pollinated reproductive system. This is an unusual and interesting system
that on the face seems it would not work well. For a masting species to be dioecious
imposes important constraints on minimum levels of synchrony among the sexes. A
masting species also provides a clumping of rewards in space and time with large
flower/seed crops some years but little or none during other years. This presents an
unpredictable and patchy food supply for the insect pollinators that rely on Aciphylla as a
resource.

For a masting plant population to be successful, individuals need to attain sufficiently high
seed set rates during a mast year. To achieve this requires a pollination system that is
resistant to fluctuations in flower production, and this is usually achieved by wind
pollination (Kelly, 1994). Insect pollinators must be either: a) generalists that are not solely
reliant on Aciphylla for food each year and can forage on a range of other plant species but
are still attracted to Aciphylla when it flowers; or b) specialised insect pollinator syndromes
that have an emergence pattern that coincides with the flowering of Aciphylla. Little is
known about the pollinator guild of Aciphylla and the degree of specialist or generalist
plant-pollinator interactions. Insect visitation rates can be correlated with pollination
success (Proctor et al., 1996). If a large floral display size gives a disproportionate increase

79
in pollinator attraction, occasional large flowering years will lead to a higher average
visitation rate (and pollination success), than regular small ones.

Masting species that can attract more insects during a high-flowering year (e.g. through
increasing the size of floral display) would theoretically gain a selective advantage in doing
so through economies of scale. To better understand the causes and consequences of mast
seeding in Aciphylla, we need to know more about the pollination system and the
importance (if any) of having a large floral display.

3.1.1 Aims

Overall, this chapter investigates the relationship among insect visitation rates, floral
display size and seed set rates. It endeavours to elucidate the pollination system in large
montane Aciphylla species. Specific questions are:

• What types of insects are visiting Aciphylla and are they potential pollinators?
• Are insects pollinators attracted to larger floral displays? Is this correlated with any
variation in seed set rate in response to floral display size?
• Does the removal of floral bracts affect visitation rates?
• How does plant-level floral display size (i.e. number of inflorescences per plant)
affect visitation rates?
• Are insects attracted to isolated female inflorescences? Is insect visitation rate
higher when males and females are flowering in closer proximity to each other, and
is this related to plant reproductive success?
• Does insect visitation rate differ among male and female inflorescences?
• How do results from the above questions relate to seed set rate (reproductive
success) in Aciphylla? (Chapter 2).

80
3.2 METHODS

3.2.1 Collecting insects

Insects were collected from male and female Aciphylla plants at all sites in order to obtain
information about what some of the insect visitors and potential pollinators to Aciphylla
were. This was carried out during peak flowering time (December and January 2004/2005)
and collections were made on two to four days throughout this period. Insects collected
were those seen either on male or female flowers, yet it was difficult to determine a priori
whether these insects were pollinators or not. Insects observed on flowers were caught
either by direct trapping into glass vials or with a pooter catcher. Samples were preserved
in 70% ethanol solution for subsequent photography, labelled and kept separate by site.
Identification to the lowest possible taxonomic level was carried out.

3.2.2 Insect visitation surveys

Insect observations were conducted on all female plants within the floral display size, bract
removal, and female ‘isolation’ experiments performed in Chapter 2 (see Chapter 2 for full
details of the experimental design). To record insect visits to inflorescences, visitors that
were potential pollinators were grouped into 6 taxonomic categories that could be
distinguished visually from a distance: big flies, hoverflies, small flies, beetles, weevils,
and moths. This was done for practicality and ease of analysis, but does not necessarily
depict accurately the relative roles or efficiencies of the different pollinator groups. Other
visitors including ladybirds and spiders were also recorded. For all experiments at every
site (excluding Lewis Pass due to weather), insect visits to each inflorescence were
recorded.

Insect visitation is very difficult to measure on inflorescences with large numbers of small
flowers such as Apiaceae (Dafni, 1992). For the purposes of this study, an attractiveness
index (see Dafni, 1992) was employed to measure insect visitation to all the treatment

81
inflorescences specified in Chapter 2. For each inflorescence, an ‘instantaneous’ insect
count was used to measure the degree of ‘attractiveness’ to insects. This involved an
observer and a recorder moving between inflorescences, stopping near each only for long
enough at each to record all the insects within each of the six taxonomic categories on that
inflorescence at that particular point in time. Because of the large sample size of plants in
each experiment, and the amount of time taken to move between each one, this method was
preferred above measures of insect visitation rates such as ten-minute counts. Start and
finish times of the instantaneous visitation counts were noted, as well as the date, weather
conditions and any other relevant information such as other plants flowering at the time.
Instantaneous insect visitation observations were performed 3-4 times at each site over a 3
week period during peak flowering season in late December/early January 2004/2005.

3.2.3 Insect visitation to male inflorescences

Male reproductive success can be difficult to measure, especially for plants such as
Aciphylla where each inflorescence can have tens of thousands of flowers. At Burkes and
Hakataramea Pass for A. aurea, and Hakataramea Pass for A. scott-thomsonii, an average of
8 male inflorescences were randomly selected and tagged. Using the instantaneous insect
visitation method outlined above, 3-4 observations were recorded for the male
inflorescences over a 3 week period during peak flowering. These data were compared to
control treatment female inflorescences within the same sites to see whether insect visits
differed between male and female inflorescences.

3.2.4 Insect visitation rates in response to floral display

This insect visitation information was used to test whether differences in pollinator
attraction (i.e. number of insect visits per inflorescence per instantaneous count) were
related to variation in floral display size and flowering densities in Aciphylla. As in Chapter
2, the effect of floral display was assessed at both the individual and the population level.
Specifically, the following comparisons were made:

82
1) Total insect visitation rate per inflorescence for female inflorescences vs. flowering
length, where the number of flowers was manipulated in three treatment levels (0,
40 and 80 % removal).

2) Relative insect visitation rate (adjusted to reflect insects per flower) for female
inflorescences vs. flowering length, where the number of flowers was manipulated
in three treatment levels (0, 40 and 80 % removal).

3) Insect visitation rate vs. flowering length (as a continuous variable) for female
inflorescences, including both un-manipulated plants and the manipulated plants
used in (1).

4) Insect visitation rate to female inflorescences vs. height of inflorescence, using all
plants as in (3).

5) Insect visitation rates to female inflorescences in response to bract removal using a

paired design (one inflorescences with all bracts removed and one un-manipulated
inflorescence per plant).

6) Insect visitation rates to female inflorescences in response to the total number of

inflorescences on a plant (as a measure of plant level floral display), using all plants
as in 3.

7) Insect visitation rate to female inflorescences as a function of the distance to nearest

males (a measure of population level floral display) using un-manipulated areas.

8) Insect visitation rate to female inflorescences as a function of the distance to nearest

males (a measure of population level floral display) using areas where male spikes
were removed to create low male density areas.

9) Differences in insect visitation rates to male vs. female inflorescences using

unmanipulated male and female plants

83
3.2.5 Statistical Analyses

The insect count data were analysed using single-factor Generalised Linear Models
(GLMs) with a poisson error distribution and a log link function (in R version 1.5.1). For
comparison (2) above, relative insect visitation rates were calculated by multiplying the
number of insect visits to 0, 40 and 80 percent flower removal by 2, 5 and 10 respectively.
These values were chosen to retain integer values for poisson analysis. For all comparisons,
a repeated measures design was used in order to avoid pseudo-replication on the
inflorescence level and from the multiple observation dates. This required F and P values to
be recalculated using inflorescence deviance as the residual deviance, in order to obtain the
appropriate degrees of freedom. To calculate percentage deviance explained (as an
indicator of the biological significance, equivalent to r-squared) the predictor deviance was
divided by the null residual deviance then multiplied by 100.

Local availability of male inflorescences (measured using the Distance Weighted local
Pollen Index (DWPI), see Chapter 2 methods) was tested as a predictor for variation in
insect visitation rates to females in response to female plant-distance to flowering male
plants. Secondly, data from isolated female manipulative experiments were also tested to
see how insect visitation rates varied when female inflorescences were relatively far
(isolated) from the closest flowering males. This was tested using distance to nearest
flowering male. Inflorescence effect and insect visitation round were also considered in the
model therefore, a nested design was employed in this analysis.

Each of the variables were tested using their own individual GLMs (as opposed to large
models including multiple predictive terms in Chapter 2) due to the necessity for the nested
repeated measures analyses to account for variation between each insect observation round.

3.3 RESULTS

3.3.1 What insects are visiting Aciphylla inflorescences?

A large and diverse insect fauna was associated with Aciphylla during flowering. A
summary of all taxa collected from inflorescences during December 2004 and January

84
2005 (peak flowering time) is presented in Appendix 2. At all three sites, there were seven
orders of insects found including many species of fly (Diptera) from thirteen families,
several moth (Lepidoptera) species from at least three families, at least two species of
cranefly (Plecoptera) from two families, native solitary bee and wasp species
(Hymenoptera), a number of Hemipterans and a large number of beetles (Coleoptera) from
at least eleven families. Many of these insect taxa could potentially act as pollen vectors in
Aciphylla.

Other insects of significant interest (although possibly not contributing to pollination due to
different life history strategies and morphological structures) were shield bugs
(Heteroptera), at least four aphid species (Homoptera), which are sap-suckers and
potentially seed predators, and four ladybird (Coccinellid) species.

Two giant weevil species (Coleoptera: Curculionidae) were also identified, one being the
endangered speargrass weevil Lyperobius huttoni and the other being an entomologically
significant rediscovery of the “extinct” (category ‘X’ - IUCN endangered species list)
Canterbury knobbled weevil Hadramphus tuberculatus. This was the first sighting of this
species since 1922.

3.3.2 Are insects attracted to larger floral displays?

3.3.2.1 Insect visitation rates and flower removal manipulation

In A. aurea at Burkes Pass and A. scott-thomsonii, insect visitation rate varied significantly
between flower removal treatments (Table 3.1). Un-manipulated inflorescences received at
least 12 times more visitors than inflorescences with fewer flowers in A. aurea at Burkes
Pass, with flower removal explaining over 36% of the variation observed in insect
visitation, suggesting there was a large biological effect. Although there was significant
variation between the 3 levels of flower removal in A. scott-thomsonii at Hakataramea, the
inflorescences with 40% of flowers removed had unexpectedly more insect visitors than
those with no flowers removed. Experimental manipulation of floral display size (0, 40 &
80% removal) had no significant effect on insect visitor numbers in A. aurea at

85
Hakataramea Pass. In all three sites, the 80% flower removal treatment consistently
received the least number of insect visitors.

Inflorescences of A. scott-thomsonii consistently had at least five times as many insect

visitors as A. aurea, both on average and in total (Table 3.1). Inflorescences of A. scott-
thomsonii are much larger than those of A. aurea (see Chapter 2) and consequently require
more visits to fertilise the greater number of flowers, therefore it makes sense that this
species receives more insect visits.

Table 3.1 Mean insect visitation to female Aciphylla inflorescences with 0, 40 and 80 per cent of flowers
removed (in units of insects per inflorescence per instantaneous count). The total number of insects seen in
each category over the total sampling period is given in parentheses. Statistical values given are for flower
removal effects on insect visitor numbers from poisson GLMs with flower removal as a factor (percent
variance explained, F statistic, (d.f.), and P value). (Abbreviations: BP = Burkes Pass, HP = Hakataramea Pass
and A. scott-th. refers to A. scott-thomsonii).

Study species Percent Flower Removal % Deviance

and site 0% 40% 80% explained F (d.f.) P-value

A. aurea BP 0.64 (27) 0.04 (2) 0.02 (1) 36.73 27.77 (2,43) <0.0001

A. aurea HP 0.41 (16) 0.42 (19) 0.12 (5) 7.107 2.436 (2,39) 0.1007

A. scott-th. HP 5.97 (167) 9.58 (230) 3 (72) 18.7 10.985 (2,35) 0.0002

Even after the data were adjusted for flower removal to obtain relative visitation rates (~
visits per flower), mean seed set still remained significantly different among the three
levels of flower removal for A. aurea at Burkes Pass and A. scott-thomsonii at Hakataramea
Pass (Table 3.2). Insect visits to A. aurea at Hakataramea Pass remained non-significant
between flower removal categories. For A. aurea at Burkes Pass, there was still a
significant positive relationship, with more insect visitors to inflorescences with the most
flowers, however percent deviance explained decreased, suggesting the biological
significance of this was slightly weaker. For A. scott-thomsonii differences became more
significant and percent deviance explained increased relative to un-adjusted values (Table
3.1) to over 25%. There was an interesting variation in the proportional insect visitation

86
values in response to flower removal treatments for A. scott-thomsonii and A. aurea at
Hakataramea pass, with the highest visitor numbers being observed on the 40% removal
treatment. This trend was not observed for A. aurea at Burkes Pass.

Table 3.2 Mean relative insect visitation to female Aciphylla inflorescences with 0, 40 and 80 per cent of
flowers removed, (scaled for the percentage of flowers remaining on the inflorescence). For raw insect counts
see Table 3.1. Statistical values given are for relative flower removal effects on insect visitor numbers from
poisson GLMs with flower removal as a factor (percent variance explained, F statistic, (d.f.), and P value).
(Abbreviations: BP = Burkes Pass, HP = Hakataramea Pass and A. scott-th. refers to A. scott-thomsonii).

Study species Percent Flower Removal % Deviance

and sites 0% 40% 80% explained F (d.f.) P-value

A. aurea BP 1.29 0.21 0.21 17.72 8.55 (2,43) 0.00075

A. aurea HP 0.82 2.11 1.19 4.08 1.30 (2,39) 0.28

A. scott-th. HP 11.93 47.92 30 25.18 16.66 (2,35) 0.00001

3.3.2.2 Insect visitation and inflorescence height

Inflorescence height significantly influenced the number of insect visitors to A. aurea at

Burkes Pass and to A. scott-thomsonii at Hakataramea Pass (Table 3.3). The effect was not
significant for A. aurea at Hakataramea. Taller inflorescences tended to attract higher
visitor numbers to both species at all sites, although this was more distinct in A. scott-
thomsonii (Fig. 3.1).

87
Table 3.3 Insect visitor numbers to female Aciphylla inflorescences in response to variation in inflorescence
height (cm). Statistical values given are for the effects of inflorescence height on insect visitor numbers from
single-factor poisson regressions with inflorescence height as the sole explanatory variable (percent deviance
explained, F statistic, (d.f.), and P value). Slope and intercept values from the model are given in loge units,
with the fitted model being in the form of insect numbers = e(slope*inflorescence height + intercept). (Abbreviations: BP =
Burkes Pass, HP = Hakataramea Pass and A. scott-th. refers to A. scott-thomsonii).

%
Study species Mean inflorescence height
Deviance
and sites ± 1 standard dev. Slope Intercept explained F (df) P-value

A. aurea BP 120.52 +/- 18.6 0.029 -4.85 6.83 8.46 (1,64) 0.005

A. aurea HP 139.56 +/- 25.52 0.012 -2.92 2.24 2.14 (1,57) 0.149

A. scott-th. HP 209.79 +/- 38.63 0.010 -0.50 10.34 11.31 (1,51) 0.001

Aciphylla scott-thomsonii overall received higher visitor numbers than A. aurea. Because
the effect inflorescence height played on visitor numbers in A. scott-thomsonii and A. aurea
at Burkes Pass was highly significant, it is possible to estimate the mean number of insects
likely be present on an inflorescence of a particular height during any particular
instantaneous observation (from the equations in Table 3.3). For the mean A. scott-
thomsonii inflorescence height of 209.8 cm, an average of 4.9 visitors could be expected
per instantaneous observation. For an inflorescence of 248.4 cm (+ 1 standard deviation of
the mean), we would expect 7.2 insects, and a shorter inflorescence of 171.2 cm (- 1 std.
dev.), around 3.4 insect visitors. For the mean A. aurea (Burkes Pass) inflorescence height
of 120.5 cm, an average of 0.26 visitors could be expected per instantaneous observation.
For an inflorescence of 139.1 cm (+ 1 std. dev. of the mean), we would expect 0.44 insects,
and a shorter inflorescence of 101.9 cm (- 1 std. dev.), around 0.15 insect visitors.

88
 Burkes – A. aurea Hakataramea - A. scott-thomsonii
Burkes - A. aurea Hakataramea - A. scott-thompsonii

35
Total insect visits per inflorescence

Total insect visits per inflorescence

6

30
5

25
4

20
3

15
2

10
1

5
0

0
0 50 100 150 0 50 100 150 200 250 300

Inflorescence height (cm) Inflorescence height (cm)

Hakataramea
Hakataramea -- A.
A. aurea
aurea
Total insect visits per inflorescence
6
5
4
3
2
1
0

0 50 100 150 200

Inflorescence height (cm)

Fig. 3.1 Total number of insect visitors to female Aciphylla inflorescences (summed over all observation
days) in response to inflorescence height. The relationship was significantly positive for A. aurea at Burkes
Pass and A. scott-thomsonii at Hakataramea Pass but non-significant for A. aurea at Hakataramea Pass,
although the direction of the relationship remained positive (see table 3.3).

3.3.2.3 Insect visitation and inflorescence flowering length

Flowering length (the length of the inflorescence stalk with flowers present) was also a
strong predictor of insect visitor numbers to A. aurea at Burkes Pass and A. scott-thomsonii

89
at Hakataramea Pass (Table 3.4). The greater the length of the stem with flowers present,
the more visitors there were in total. For A. aurea at Hakataramea, results were non
significant, but at Burkes Pass, flowering length was a stronger predictor of insect visits
than inflorescence height as it had the larger percentage deviance explained (Tables 3.3,
3.4). Figure 3.2 graphically demonstrates the positive relationship between length of
inflorescence with flowers and number of insect visitors.

Table 3.4 Insect visitor numbers in response to variation in flowering length (cm). Statistical values given are
for the effects of length of inflorescence with flowers on insect visitor numbers from poisson GLMs with
length of inflorescence with flowers as a variable (percent deviance explained, F statistic, (d.f.), and P value).
Slope and intercept values from the model are given in loge units, with the fitted model being in the form of
insect numbers = e(slope*flowering length + intercept). (Abbreviations: BP = Burkes Pass, HP = Hakataramea Pass and
A. scott-th. refers to A. scott-thomsonii).

Study species Mean flowering length % Deviance

and sites ± 1 standard dev. Slope Intercept explained F (df) P-value

A. aurea BP 43.63 +/- 21.95 0.048 -3.74 23.66 43.47 (1,64) <0.0001

A. aurea HP 51.56 +/- 23.41 0.011 -1.89 2.1 2.01 (1,57) 0.162

A. scott-th. HP 87.42 +/- 34.51 0.008 0.95 6.64 6.72 (1,51) 0.012

90
 Burkes – A. aurea Hakataramea - A. scott-thomsonii
Burkes - A. aurea Hakataramea - A. scott-thompsonii

35
Total insect visits per inflorescence

Total insect visits per inflorescence

6

30
5

25
4

20
3

15
2

10
1

5
0

0
0 20 40 60 80 100 0 50 100 150

Flowering length (cm) Flowering length (cm)

Hakataramea - A. aurea
Hakataramea - A. aurea
Total insect visits per inflorescence
6
5
4
3
2
1
0

0 20 40 60 80 100

Flowering length (cm)

Fig. 3.2 Total number of insect visitors to female Aciphylla inflorescences (summed over all observation
days) in response to flowering length (length of inflorescence with flowers on). The relationship was
significantly positive for A. aurea at Burkes Pass and A. scott-thomsonii at Hakataramea Pass but non-
significant for A. aurea at Hakataramea Pass (see table 3.5).

3.3.3.4 Does bract removal affect insect visitation rates?

Bract removal had no significant effect on insect visitor numbers to female inflorescences
at any site (Table 3.5). The trend for A. aurea at Hakataramea Pass (P = 0.093) was
negative – bract removal resulted in a (non-significant) decrease in the numbers of insect

91
visits. Although non significant, this trend was supported in A. aurea at Burkes Pass but not
in A. scott-thomsonii.

Table 3.5 Mean insect visitor numbers per female inflorescence per instantaneous observation in response to
bract removal. Statistical values (percent deviance explained, F statistic, (d.f.), and P value) are from poisson
GLMs comparing insect visitor numbers between bract removal treatments and control inflorescences.
(Abbreviations: BP = Burkes Pass, HP = Hakataramea Pass and A. scott-th. refers to A. scott-thomsonii).

Study species Bract removal % Deviance

and site Control No bract explained F (df) P-value
A. aurea BP 0.81 0.56 1.48 0.66 (1,14) 0.431

A. aurea HP 0.39 0.095 11.79 3.39 (1,11) 0.093

A. scott-th. HP 5.75 6.58 0.55 0.098 (1,8) 0.763

3.3.3.5 Insect visitation and number of inflorescences per plant

There was no significant difference in mean visitation rates to inflorescences situated on

plants varying in inflorescence number, although an upward trend was apparent for A.
aurea at Hakataramea Pass and A. scott-thomsonii (Table 3.6, Figure 3.3). Aciphylla scott-
thomsonii inflorescences had a greater mean number of visitors in every category than A.
aurea (Table 3.6).

92
 Burkes – A. aurea Hakataramea - A. scott-thomsonii
Burkes - A. aurea Hakataramea - A. scott-thompsonii

30
Total insect visits per inflorescence

Total insect visits per inflorescence

5

25
4

20
3

15
2

10
1

5
0

0
1.0 1.5 2.0 2.5 3.0 1 2 3 4 5 6 7 8

Number of inflorescences on plant Number of inflorescences on plant

Hakataramea - A. aurea
Hakataramea - A. aurea
Total insect visits per inflorescence
4
3
2
1
0

1.0 1.5 2.0 2.5 3.0 3.5 4.0

Number of inflorescences on plant

Figure 3.3 Total insect visitor numbers (per instantaneous observation) on single female inflorescence on
plants with 1,2,3,4 or 8 inflorescences. Lines are fitted from the single factor poisson GLMs in Table 3.6,
with inflorescence number as a variable. All the relationships are not significant (Table 3.6).

93
Table 3.6 Mean insect visitor numbers (per instantaneous observation) on single female inflorescence on
plants with 1,2,3,4 or 8 inflorescences. The sample size for each category (i.e. the total number of
instantaneous observations) is listed adjacently in parentheses. Statistical values presented are the output from
single factor poisson GLMs with inflorescence number as a variable (percent deviance explained, F statistic,
(d.f.), and P value). (Abbreviations: BP = Burkes Pass, HP = Hakataramea Pass and A. scott-th. refers to A.
scott-thomsonii).

Species Number of inflorescences on plant % Dev.

and sites 1 2 3 4 8 explained F (df) P-value

A. aurea BP 0.38 (60) 0.20 (148) 0.33 (64) NA NA 0.051 0.062 (1,65) 0.804

A. aurea HP 0.14 (48) 0.29 (108) 0.33 (21) 0.42 (12) NA 2.031 1.982 (1,61) 0.164

A.scott-th. HP 3.5 (22) 2.68 (22) 7.79 (34) 4.13 (8) 6.58 (12) 3.234 2.458 (1,47) 0.124

3.3.4 Insect visitation to isolated flowering female plants

Unexpectedly, there was no significant difference in insect visitor numbers to female

inflorescences at varying distances from flowering males (using the Distance Weighted
local Pollen Index (DWPI)) for unmanipulated male flowering densities in A. aurea and A.
scott-thomsonii at both study sites. A. aurea – Burkes Pass poisson GLM: F(1,33) = 0.03, P =
0.870, A. aurea – Hakataramea Pass poisson GLM: F(1,34) = 0.3, P = 0.59 and A. scott-
thomsonii – Hakataramea Pass poisson GLM: F(1,21) = 0.12, p = 0.74.

Where flowering plant distributions were manipulated by removing males and effectively
isolating female plants to represent a ‘non-mast’ year, results were also unexpectedly non
significant. A. aurea – Burkes Pass, poisson GLM: F(1,39) = 0.149, P = 0.701, and A. aurea
– Hakataramea Pass, poisson GLM: F(1,34) = 0.139, P = 0.711.

3.3.5 Insect visits to male vs. female inflorescences

At all sites there were significant differences between insect visits to male compared with
female inflorescences, with males receiving a significantly greater average number of visits

94
per inflorescence per observation. The most common insect visitors to A. aurea at both
sites were big flies, small flies and weevils and these groups were almost always
significantly more common on male inflorescences (Table 3.7a and 3.7b). Total insect
numbers were particularly high on A. scott-thomsonii males, especially big flies and
hoverflies (Table 3.7c).

Table 3.7 The mean number of insects per inflorescence per observation in each taxonomic category on male
and female Aciphylla inflorescences according to site and species. The total number of insects seen in each
category over the entire sampling period is given in parentheses. Statistical values given are for plant sex
effects within each taxonomic group (as well as the total number of insect pollinators) from single-factor
poisson GLMs (percent variance explained, F statistic (d.f.), and P value). NA values indicate too few data to
run test.

a) Burkes Pass – Aciphylla aurea

Small
Plant sex Big flies Hoverflies flies Beetles Weevils Moths Total

Male 0.83 (40) 0 (0) 1.13 (54) 0.02 (1) 2.75 (132) 0.06 (3) 19.17 (230)

Female 0.69 (33) 0 (0) 0.17 (8) 0.08 (4) 0 (0) 0.04 (2) 3.92 (47)

% var. expl. 0.4 NA 16.73 NA 38.56 NA 24.9

F (d.f. 1,22) 0.27 NA 11.16 NA 32.28 NA 9.08

P-value 0.61 NA 0.003 NA <0.001 NA <0.001

b) Hakataramea Pass – Aciphylla aurea

Small
Plant sex Big flies Hoverflies flies Beetles Weevils Moths Total

Male 2.5 (75) 0.3 (9) 0.47 (14) 0.1 (3) 13.07 (392) 0 (0) 16.43 (493)

Female 0.1 (3) 0 (0) 0.1 (3) 0.03 (1) 0.03 (1) 0 (0) 0.27 (8)

% var. expl. 23.39 26.95 8.37 4.28 3.43 NA 38.53

F (d.f. 1,18) 9.03 9.54 2.95 1.29 15.67 NA 15.54

P-value 0.008 0.006 0.1 0.27 0.009 NA <0.001

95
c) Hakataramea Pass – Aciphylla scott-thomsonii

Small
Plant sex Big flies Hoverflies flies Beetles Weevils Moths Total

Male 31.56 (568) 24.11 (434) 0.28 (5) 0.11 (2) 5.56 (100) 0 (0) 61.6 (1109)

Female 3.45 (76) 1.77 (39) 0.45 (10) 0.09 (2) 0.05 (1) 0 (0) 5.8 (128)

% var. expl. 65.98 56.1 1.4 0.19 22.31 NA 65.16

F (d.f. 1,18) 45.51 29.27 0.39 0.05 37.48 NA 39.98

P-value <0.001 <0.001 0.01 0.83 <0.001 NA <0.001

The ratio of mean number of insect visitors per inflorescence per observation for males,
divided by the mean number of insect visitors per inflorescence per observation for females
(Table 3.7) was varied considerably between sites. A. aurea at Burkes Pass had the lowest
ratio (4.9 visitors per male for every one visitor per female), followed by A. scott-thomsonii
(10.6) and A. aurea at Hakataramea (60.9).

3.3.6 How do these results relate to results on seed set rate in chapter 2?

Overall, Table 3.8 shows some significantly similar trends occurring between a number of
variables/factors representing floral display and both reproductive success (seed set) and
insect visitation rates. In general, seed set rates showed similar patterns to that of insect
visitation rates at the majority of sites (for floral display size), in terms of the direction and
significance of the relationships in question (Table 3.8). For instance, a decrease in the
level of floral display (0, 40, 80% flower removal experiment) elicited a decrease in both
insect visitation and seed set rates (to varying extents), while taller inflorescences had a
significant positive effect on both visitation rates and seed set, as did greater flowering
length. There was no significant effect of female isolation on insect visitation rates, but
there was a significant negative effect on seed set rates at most sites (Table 3.8). These
results will be discussed further in Chapter 5.

96
Table 3.8 Summary of the results from Chapters 2 and 3 showing the effects of floral display at both the individual (flower removal experiment) and population (isolated
female experiment) level on seed set rate (Chapter 2) and insect visitation rates (this Chapter) at all sites during the summer of 2004/2005. There are six predictive terms.
Where there are gaps, this means that a particular term does not apply to a certain species and/or site. BP refers to Burkes Pass, HP to Hakataramea Pass and LP to Lewis
Pass. Positive and negative signs indicate the direction of the relationship between each predictor and response, with expected direction provided in the first column based on
initial hypotheses. A zero refers to a statistically non-significant result. For percent flower removal, signs on the left side of the dash explain the direction of the relationship
from 0 to 40% and the right side is from 0 to 80% removal. A double sign on one side of the dash refers to the flower removal treatment that shows that strongest effect.

Expected Flower removal experiment

Predictive direction of ‘Isolated’ female experiment

terms relationship A. aurea - BP A. aurea - HP A. scott-thom. - HP A. ‘lewis’ - LP A. aurea - BP A. aurea - HP

Seed set Visitation Seed set Visitation Seed set Visitation Seed set Visitation Seed set Visitation Seed set Visitation

rate rate rate rate rate rate rate rate rate rate rate rate

Hand pollination + + + + + +

Pollen index + + 0 + 0 – 0

Nearest male – – – 0 – 0

% Flwr removal –/–– –/–– –/– –/–– 0 –/–– +/– –/+

Flowering length + – + 0 0 – + + + + + +

Inflor. height + + + – 0 + + + + 0 – 0

97
3.4 DISCUSSION

3.4.1 The insect fauna of Aciphylla: a generalist or specialist pollination system?

A large number and diversity of insects were collected on flowering Aciphylla, yet without
further study, it is difficult to ascertain the relative roles and pollination efficiencies of the
different groups. Pollinator spectra for many plant species range across orders of insects
(beetles, flies, bees, moths and butterflies) classes of animal (insects, birds and mammals)
and pollinations agents (wind, water and animal) (Herrera, 1996; Ollerton, 1996). Based on
the ready availability of nectar and pollen, and on the large numbers and great diversity of
insect visitors, species in the Apiaceae have been labelled promiscuous (Lindsey, 1984),
which also seemed to hold true for Aciphylla aurea and A. scott-thomsonii in this study.
The distinction between floral visitors and effective pollinators, however, is extremely
important and is rarely discerned (Lindsey, 1984; Newstrom & Robertson, 2005). Although
it appears that a range of insects may pollinate New Zealand Apiaceae members such as
Aciphylla, this may not be the case. Some taxonomic or functional groups may serve to be
more effective and efficient than others.

This study provides the first recorded comprehensive description of the insect fauna
associated with flowers of lower montane Aciphylla species (but see Primack, 1983), but is
not an exhaustive list as collections were limited to diurnal visitors over a short period
during only one season, at three sites, using only one trapping technique.

At all three sites, there were seven orders of insects found. By far the most abundant and
ubiquitous of these were the Dipterans (flies) with a range of species belonging to thirteen
different families collected. This was probably the most significant pollinator group due to
their extremely large numbers at each site, and because many flies are known to be
opportunistic feeders on pollen and/or nectar and are important pollinators worldwide
(Newstrom & Robertson, 2005). Important pollinator families in New Zealand are bristle
flies (Tachinidae) and hoverflies (Syrphidae) (Newstrom & Robertson, 2005), which can
carry significant pollen loads on their gena (or cheeks) and perhaps other families such as
Bibionidae and Calliphoridae (Primack, 1983), all of which were collected on Aciphylla
during this study. Among other insect groups found that are likely to contribute to pollen

98
movement in Aciphylla were: moths, craneflies, native solitary bee species, native wasps,
and a large number of beetles, although there have been very few studies of the
effectiveness of beetles as pollinators. They are less mobile and may not have as important
a role as bees, flies and moths. Godley (1979) listed weevils as pollinators of Aciphylla but
the effectiveness of other small beetles has not been investigated (Newstrom & Robertson,
2005).

Although many of these insects are potential pollinators, it is possible that only a small
proportion of these species are significantly contributing to the movement of pollen from
male to female plants. Lindsey (1984) tested insect floral visitation for plants of nine
populations in a comparative study of three species of the closely related Apioid genera,
Thaspium and Zizia. Results showed that, despite a high diversity of insect visitors,
generally 1-4 species accounted for at least 74% of the pollinations in all populations.

The wide variety of potential pollinators found on Aciphylla in this study suggests that
these species are adapted to pollination by a suite of generalists. However, further work
should be done to test the relative importance of insect pollinators in Aciphylla, as the
presence of specialist pollinator(s) cannot be ruled out from the findings in this study.

3.4.2 Why are insects more abundant on male inflorescences, and are females
receiving sufficient visits?

All experiments so far have been largely concerned with female reproductive success, i.e.
proportion of seed set. “Success” in males however can be difficult to define and measure
(Sutherland & Delph, 1984). Methods of measuring male reproductive success include
measuring the amount of pollen being carried away from the flowers, or frequency and
duration of pollinator visits to flowers. On species such as Aciphylla with tens of thousands
of flowers on a single male inflorescence, this can be difficult therefore a rough measure of
insect visitation to male flowers was employed here. One thing that did become evident
however was that on average male inflorescences received a much greater number of insect
visitors during the survey than females.

99
So why is there an apparent bias towards insects visiting male rather than female
inflorescences? Male bias in visitation has been reported for several dioecious and
monoecious species (Bierzychudek, 1987; Le Corff et al., 1998). The ability of pollinators
to discriminate between male and female flowers has been observed by many, (e.g. Bawa,
1977; Bierzychudek, 1987; Armstrong, 1997; Le Corff et al., 1998). Baker (1976) suggests
that visits to female flowers are often made only by mistake, and predicted that
discrimination against female flowers should be common amongst species with unisexual
flowers. In this case, female flowers presumably offer less reward to insects than do males,
because they are usually incapable of producing pollen. Male flowers in this case would
produce not only copious quantities of pollen, but may also produce nectar, offering twice
the reward (Holland & Fleming, 2002; Orellana et al., 2005) to species that collect pollen.
Therefore, under these circumstances, it seems logical to assume that insects may be more
attracted to male inflorescences, especially if their floral display is generally larger,
showier, and more rewarding. Realistically however, insect visitation rates between male
and female flowers may depend on the quality of the rewards offered and on the insect’s
preferences which varies greatly between plant species and insect species. Some species
produce different volumes of nectar by sex, e.g. Fuchsia excorticata (Delph & Lively,
1989) and consequently that sex would ultimately be more attractive to nectar feeders but
not pollen feeders.

Possible reasons for the discrepancy between insect visitor numbers to male versus female
inflorescences in Aciphylla could be attributed to the female floral reward. Primack (1983)
reported that insects gather nectar from staminate (male) flowers, as well as pollen in A.
scott-thomsonii, however whether pistillate (female) flowers in Aciphylla offer nectar
rewards remains to be testified. Presuming pistillate flowers of Aciphylla do offer nectar
rewards, the timing and availability of nectar production and quality are not known. Nectar
is not a product of a plants sexual system but a reward offered to a foraging agent (Dafni,
1992), i.e. a secondary sex characteristic (Delph, 1999). There may be daily variation in
nectar production, due to weather or temperature changes, or production at a certain time of
the day such as morning (Proctor et al., 1996). If in Aciphylla, nectar is being produced by
the stylopodium only at a certain time of day, for example, during early morning, nectar-
feeding insects are likely to be more attracted to these flowers during peak production
times (Vaughton & Ramsey, 1998; Kudo & Harder, 2005). It is possible that the timing of
insect observations conducted during this study (which were usually between 10 am and 2

100
pm) may not have encompassed peak nectar production times, and hence attracted fewer
insects, potentially causing the male-bias in visitation rates. Incorporating temporal
variation in observation times should be an important point of consideration for future
research in this area.

Overall, it seems that in Aciphylla, males probably offer both pollen and nectar and females
are presumed to only produce nectar. This suggests that because males offer a larger variety
of rewards to pollinators, they consequently attract a larger diversity of insect feeders that
have different feeding preferences, and this could explain the male-bias insect visitation
rates obtained in this study.

3.4.3 How do the effects of floral display on insect visitation rates relate to seed set
rates?

3.4.3.1 Insect visitation and floral display

Although insect visitation rates were significantly influenced by floral display size, it
depended on the measure of floral display and methodology used to measure visitation. In
both A. aurea (Burkes Pass) and A. scott-thomsonii (Hakataramea Pass), floral display
affected insect visitation rates when it was measured using inflorescence height, flowering
length (the length of stem with flowers on), and flower removal (0, 40 & 80 % flower
removal experiment). These significant outcomes did not always conform to the expected
direction of the relationship however. At Hakataramea, visitation rates were significantly
higher for 40% flower removal – an inflorescence of intermediate floral display. At this
site, insect observations were only performed twice due to weather and time limitations.
The most statistically significant outcome was for A. aurea at Burkes Pass where visitor
numbers significantly increased with increasing floral display size. Observations were
carried out four times at this site, and results suggest that at Burkes Pass, floral display
plays an important role in attracting pollinators, and this subsequently affected seed set.
Studies supporting an increased floral display size increases pollinator attraction are many
(e.g. Donnelly et al., 1998; Kawarasaki & Hori, 1999; Lortie & Aarsen, 1999; Abe, 2000;
Grindeland et al., 2005). The magnitude of this effect is relatively small in this study

101
however, and further research is needed to test visitation rates in more detail to see if the
effect of inflorescence height on pollinator visits per flower is evident.

Visitation rates did not vary significantly when floral display was represented by number of
inflorescences per plant, although there was a suggestive trend towards an increase in
inflorescence number attracting more insects. If insects were visiting taller flowers
significantly more often, one would also expect insects to be more attracted to plants with
many inflorescences due to the benefits gained from minimising foraging efforts for a
concentrated food source (Collevatti et al., 1998). This was not found during this study, as
insect visitors were generally few and inflorescences were usually growing in dense
clumps, probably not limiting insect preferences for localised resources.

3.4.3.2 Are isolated flowering females reproductively successful?

For animal pollinated species, relationships between plant density and pollination rates
depend on the functional and behavioural responses of pollinator visitation to plant density
(Kelly, 1994). Results showed that the density of male inflorescences around females
played an important role in determining seed set rate, with the isolated females receiving
significantly lower levels of seed set than those flowering closer to males. However, the
frequency of insect visits to inflorescences at varying distances to males was not
significantly different. A number of studies show similar findings (e.g. Roberts & Boller,
1948; Grindeland et al., 2005). This would suggest that the reduction in seed set with
increasing distance to nearest male is a result of a decline in the efficiency of the
pollinators (Robertson, 1992; Grindeland et al., 2005) in successfully cross-pollinating
Aciphylla, rather than a decline in the numbers of pollinator visits as such (see Chapter 5
for further discussion).

3.4.4 Conclusions

Overall, seed set rates were high and insect visitation rates were low in this study.
However, having for instance, one insect on the inflorescence at any one time could be
enough if the flowers have a 2 week life and each insect crawls over 25 flowers Therefore,

102
the numbers look low but may be sufficient for pollination. This remains to be examined in
detail. In many natural populations, seed set can be limited by the effectiveness of
pollinators (Bierzychudek, 1987). It seems unlikely, given the naturally high seed set rates
found in this study, and only minor or no increases in seed set when provisioned with
supplementary pollen, that pollinators are ineffective. This is of course, if the reproductive
system in Aciphylla is solely dependant on insects as pollinators. If there are other methods
by which Aciphylla are pollinated and can set seed, then this may help explain why when
visitor numbers are low, seed set can still be high. The questions remaining to be answered
therefore, are: how do Aciphylla manage to achieve such a high seed set rate when insect
visits appear to be low, and why do inflorescences in bags still have a greater than zero
seed set? Further research should be directed towards understanding whether seeds can still
be made without fertilisation, i.e. by apomixis, and should test for the effectiveness of wind
pollination. The implications of this will be discussed further in chapter 5.

103
CHAPTER 4 – Sex-ratios, dioecy and sexual dimorphism in Aciphylla

4.1 INTRODUCTION

Dioecy exists in about 6% of the world’s angiosperm flora and has arisen independently in
many families (Sakai & Weller, 1999). In the Apiaceae, almost all genera are
hermaphroditic (Webb, 1981) and dioecy is unusual for the family; yet some New Zealand
Apioid genera – Aciphylla, Anisotome and Lignocarpa - have developed separate sex plants
(Webb, 1979) (along with two Australian Aciphylla species (Pickering, 2000)). There are a
number of hypotheses that exist to explain why dioecy evolved from hermaphroditism, (not
the focus of this research but see Charlesworth, 1999; Sakai & Weller, 1999). Dioecious
breeding systems usually exhibit sexual dimorphism (Delph, 1996; Eckhart, 1999). New
Zealand possesses a high number of sexually dimorphic taxa, along with Hawaii, (Webb &
Kelly, 1993) - yet another plant trait that appears particularly prevalent in island floras
(Webb et al., 1999). The frequency of sexually dimorphic genera is estimated at 18%,
higher than for most continental areas (Lloyd, 1985; Renner & Ricklefs, 1995).

Sexual dimorphism in plants refers to differences in primary and secondary sex

characteristics among plants that are dimorphic in gender (Delph, 1999) such as dioecious
species. Primary sex characteristics are traits such as gonads and copulatory organs that are
essential for reproduction (Darwin, 1877), (which are necessarily dimorphic), while
secondary sex characters are traits that are not directly connected with reproduction (Geber,
1999). The latter include reproductive sex characters such as features of flowers or
inflorescences (e.g. size and number) that affect male and female mating success or fertility
(Eckhart, 1999), as well as ecological life history traits (e.g. flowering phenology, age at
first flowering, and mortality) (Delph, 1999), and vegetative morphology, physiology and
competition (Lloyd & Webb, 1977; Bawa, 1980; Dawson & Geber, 1999; Delph, 1999;
Geber, 1999). Variation in any of the aforementioned traits between male and female plants
can be considered as sexual dimorphism, and there is considerable evidence for this in
dioecious plants in both reproductive and ecological secondary sex characters (Lloyd &
Webb, 1977). Sexually dimorphic traits have been reported for Aciphylla species (e.g.
Godley, 1964; Lloyd & Webb, 1977; Dawson & Le Comte, 1978; Webb, 1979; Webb &
Lloyd, 1980; Pickering, 2000, 2001; Pickering & Hill, 2002; Pickering & Arthur, 2003).

104
See Figure 4.1 for morphological variation in floral display between male and female
Aciphylla aurea plants.

Fig. 4.1 Male (left) and female (right) plants of Aciphylla aurea showing typical variation in display size,
inflorescence number and showiness of males and females present within a population. This represents an
example of sexual dimorphism in floral display between the sexes, observed in most species of this dioecious
genus Aciphylla.

Sexual dimorphism can arise and be maintained in populations through natural selection
and sexual selection (Darwin, 1877; Lloyd & Webb, 1977; Bawa, 1980; Delph, 1999).
Sexual selection involves the differential effect on individuals of selection for male
function (pollen production and dissemination) and female function (seed production and
dissemination) (Bateman, 1948; Lloyd & Webb, 1977; Bawa, 1980; Delph, 1999). Sexual
selection theory proposes that males of dioecious plants are limited by the number of mates
(Bateman, 1948) resulting in intra-sexual competition among males (Eckhart, 1999), while
for females there is a trade-off between allocating resources to attracting mates and seed
production (Haig & Westoby, 1988) (see Chapter 2). This is thought to lead to sexual
dimorphism in secondary sex characters such as floral display and other life history traits
(Lloyd & Webb, 1977; Dawson & Geber, 1999).

105
Pickering (2000) suggests specific examples of traits thought to be the result of sexual
selection: male-biased sex ratio, younger flowering of male plants, earlier flowering of
male plants, larger floral display in males, and allocation of more resources to reproduction
at the time of flowering in males, and the overall allocation of more resources to
reproduction by females (see Pickering, (2000) and references therein).

In sexually dimorphic dioecious flowering plants, male-biased sex ratios have been
reported for a wide variety of species (Darwin, 1877; Godley, 1964; Lloyd & Webb, 1977).
There have also been a number of hypotheses proposed to explain this deviation from
equality. These include gender-specific patterns of mortality due to different costs of
reproduction in females versus males (Lloyd & Webb, 1977), local mate competition (de
Jong et al., 2002), differential herbivory of the sexes (Ågren et al., 1999), selfish genetic
elements, (Delph, 1999), site productivity and environmental gradients (Pickering & Hill,
2002), among others (see Lloyd & Webb, 1977; Delph, 1999; Pickering & Hill, 2002).

It is thought that as a result of resource limitation, alpine dioecious species may exhibit
sexual dimorphism in life history traits, and as a result of this, populations of alpine
dioecious species may display a male-biased sex ratio (Pickering, 2000). As mentioned
above, Aciphylla populations have been previously reported to exhibit sexual dimorphism
in the floral display of male and female inflorescences in New Zealand (Webb & Lloyd,
1980) and Australia (Pickering, 2000). Webb and Lloyd (1980) measured 53 populations of
20 predominantly-alpine New Zealand apioid Apiaceae species, and reported that of the six
species of Aciphylla studied, the larger species A. scott-thomsonii and A. aurea exhibited a
significantly male-biased sex ratio and had significantly more inflorescences per plant. This
chapter further explores work carried out by Webb & Lloyd (1980) on the sex ratios of
Aciphylla, and aims to quantify some of the sexually dimorphic characters of three New
Zealand large-leaved montane Aciphylla species, focussing on four specific sites at similar
altitudes within the wider Canterbury region of New Zealand.

106
4.1.1 Objectives

The previous two chapters of this study investigated the ecological intricacies of a large
floral display in Aciphylla and the paradoxical combination of mast-seeding, dioecy and
insect-pollination, with particular focus on female fecundity. This chapter aims to explore
the evolutionary impetus for sexual dimorphism and floral display in Aciphylla.
Specifically the aims of this study were:

1) To determine the plant- and inflorescence- sex ratio of dioecious Aciphylla

populations during a high-flowering year for multiple sites and species

2) To investigate and quantify some of the sexually dimorphic characteristics present

in Aciphylla by investigating whether certain plant or inflorescence traits varied
with plant gender.

3) The above measurements are made during only a single flowering season, capturing
a snap-shot of population sex ratios and sexually dimorphic traits. It has been
hypothesised that costs of reproduction are greater for females (Webb & Lloyd,
1980), therefore if male plants can flower more often than females, as this
hypothesis suggests, we are more likely to overestimate the number of males in a
population. The effects of single-year versus long-term sampling to obtain
population estimates of sex ratio will be discussed.

4.2 METHODS

4.2.1 Measuring population plant and inflorescence sex ratios

Sex ratios were recorded in a single year for four populations of three large-leaved montane
Aciphylla species at three different sites in Canterbury, New Zealand (see Chapter 2 for
details on site and species information). Counting was carried out during the summer of
2004/2005 during a relatively high-flowering year (except for Lewis Pass). Twenty-by-
twenty metre plots were set out at 2 or 3 random locations within each population to obtain

107
sample sizes of at least 200 plants within each site. Only flowering individuals were
counted, because the sex-ratio of the remaining plants at each site without inflorescences
could not be determined (as non-flowering Aciphylla plants are impossible to sex in the
field). The proportion of flowering to non-flowering plants was not counted. Determining
the sex of each flowering plant was not difficult during flowering or fruiting in dioecious
Aciphylla populations, therefore counts were made later in the season to include all plants
that flowered that summer. The methods outlined above were carried out at all sites except
for Lewis Pass, where the population was naturally sparser with fewer plants. Instead, the
sex ratio of all flowering plants of A. ‘lewis’ in the entire Lewis Pass population was
obtained.

In addition to the sex of the plants, the number of inflorescences per plant was counted.
From these data, the average number of inflorescences per plant (un-weighted mean of
male and female averages) was obtained, along with the inflorescence sex ratio of the
population. Population plant and inflorescence sex ratios were obtained by calculating the
percentage of males and the percentage of females in the population (of the plants that
flowered). Chi-squared tests of equality were performed to examine whether sex ratios
deviated significantly from the expected ratio of 50:50.

4.2.2 Measuring sexual dimorphism in plant and inflorescence traits

Approximately eight individual plants from both sexes were selected randomly from each
site (excluding Lewis Pass), and plant and inflorescence measurements were made. These
included plant height, plant area, rosette number per plant, inflorescence height (total
length of inflorescence from base to apex) and flowering length (length of inflorescence
with flowers on). These data were used to calculate whether there were differences in plant-
level traits (sexual dimorphism) between sexes, species and sites. Analysis of Variance
tests (ANOVA) were used to compare mean plant trait values between male and female
plants. Inflorescence traits such as flower number per inflorescence and perianth size were
not measured quantitatively during this study, but general observations of these were made
and an extensive literature search was carried out in order to examine whether there was
evidence for sexual dimorphism in other Aciphylla species.

108
 4.3 RESULTS

4.3.1 Sex ratios

There was a preponderance of flowering male plants in all populations studied (Fig. 4.2),
resulting in male-skewed plant sex ratios. For Aciphylla scott-thomsonii at Hakataramea
Pass, the plant sex ratio was significantly more male-biased than the other sites (Table 4.1),
with almost 68% of the flowering plants being male and 32% female. There were also
significantly more male than female plants in the A. aurea population at Hakataramea. The
sex ratios for A. ‘lewis’ at Lewis Pass and A. aurea at Burkes Pass were not significantly
different from 1:1 at these sites (Fig. 4.2 and Table 4.1), however, the trend was still
towards male-bias.

1.0
Proportion of male and female plants

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.0
A. aurea BP A. aurea HP A. 'lewis' LP A. scott-th. HP
Species and site
Species and site

Fig. 4.2 Proportion of male (■) and female (□) flowering plants in each population of Aciphylla species. BP

= Burkes Pass, HP = Hakataramea Pass and LP = Lewis Pass, and A. scott-th. refers to A. scott-thomsonii. (n
= total number of plants counted at each site).

The total inflorescence sex ratio for most populations was even more male-biased than for
plants (Fig. 4.3 and Table 4.1). Again, this was most prominent in A. scott-thomsonii at

109
 Hakataramea Pass with over 75% of the inflorescences sampled within the population
being male. This was also particularly significant for A. aurea at Hakataramea Pass, with
more than 63% male inflorescences and was also significant for A. aurea at Burkes Pass.
Again, the trend towards male-bias was present but not significant at Lewis Pass.

Table 4.1 Percentage of males for four populations of flowering Aciphylla in Canterbury. Population-level
plant and inflorescence sex ratios are shown for each site, with sample size, percentage of males, Chi-square
statistic and P-values showing significant deviation from 50:50 sex ratios. BP = Burkes Pass, HP =
Hakataramea Pass and LP = Lewis Pass, and A. scott-th. refers to A. scott-thomsonii.

Plant counts Inflorescence counts

2
Species and site Sample size % Males χ P-value Sample size % Males χ2 P-value
A. aurea BP 252 54.37 1.92 0.17 362 57.73 8.66 0.003
A. aurea HP 213 57.28 4.51 0.03 357 63.03 24.23 <0.0001
A. scott-th. HP 236 67.8 29.9 <0.0001 503 75.35 129.27 <0.0001
A 'lewis' LP 53 52.83 0.08 0.77 77 55.41 0.86 0.35
Proportion of male and female inflorescences

1.0

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.0
A. aurea BP A. aurea HP A. 'lewis' LP A. scott-th. HP

Species and
Species andsite
site

Fig. 4.3 Proportion of male (■) and female (□) inflorescences, measured as the percentage of male and

percentage of female inflorescences (out of a total of n inflorescences). BP = Burkes Pass, HP = Hakataramea

Pass and LP = Lewis Pass, and A. scott-th. refers to A. scott-thomsonii.

110
There were a significantly greater number of inflorescences per male plant than for females
for A. aurea at Burkes Pass, (ANOVA: F (1,250) = 4.39, P = 0.04) and at Hakataramea
(ANOVA: F (1,211) = 6.35, P = 0.01), and for A. scott-thomsonii at Hakataramea Pass
(ANOVA: F (1,234) = 10.72, P = 0.001) (Fig. 4.4). At Lewis Pass, inflorescence number was
not significantly different between sexes (ANOVA: F (1,51) = 0.43, P = 0.51).

3
Mean no. of inflorescences per plant

0
A. aurea BP A. aurea HP A. 'lewis' LP A. scott-th. HP
Site

Fig. 4.4 Mean number of inflorescences per male (■) and female (□) plant (± 1 SE) for all flowering plants

counted within each population (± 1 SE). BP = Burkes Pass, HP = Hakataramea Pass and LP = Lewis Pass,
and A. scott-th. refers to A. scott-thomsonii.

4.3.2 Sexual dimorphism in plant and inflorescence traits

There were almost no significant differences between mean male and female plant
dimensions, including plant height, plant area, and the number of rosettes per plant at any
site (Table 4.2). There was one significant but weak exception, with female plant area
being greater than males for A. aurea at Hakataramea Pass.

111
Table 4.2 Mean plant dimensions for male and female Aciphylla plants at three populations in Canterbury,
including mean plant height (cm), mean plant area (cm2) and number of rosettes per plant. BP = Burkes Pass,
HP = Hakataramea Pass and LP = Lewis Pass, and A. scott-th. refers to A. scott-thomsonii. Summary
statistics from Analysis of Variance (ANOVA’s) performed for each site and each variable. Significant P-
values (at 0.05) are highlighted in bold.

Plant height (cm) Plant area (cm2) Rosettes per plant

A. aurea A. aurea A. scott A. aurea A. aurea A. scott A. aurea A. aurea A. scott

Plant sex BP HP HP BP HP HP BP HP HP
Female 45.8 45.23 79.71 5654 6358 18123 2.23 2.63 3.64
Male 47.75 42.67 77.5 7847 3946 17102 3.75 1.67 4.25
% var. expl. 0.4 1.4 0.3 5.45 8.1 0.1 7.68 0.06 0.76
F (d.f.) 0.18(1,43) 0.61(1,44) 0.09(1,30) 2.47(1,46) 4.04(1,43) 0.03(1,30) 3.49(1,42) 2.61(1,44) 0.21(1,27)
P-value 0.67 0.44 0.76 0.12 0.05 0.86 0.07 0.11 0.65

There were no significant differences in mean inflorescence height between male and
female plants at all sites, although male inflorescences were consistently taller on average
than females (Table 4.3). Male plants on average did however have significantly greater
flowering length than females, with the effect being strongest for A. scott-thomsonii at
Hakataramea Pass.

Table 4.3 Mean inflorescence height (cm) and flowering length (cm) of male and female Aciphylla plants at
three sites in Canterbury, BP = Burkes Pass, HP = Hakataramea Pass and LP = Lewis Pass, and A. scott-th.
refers to A. scott-thomsonii. Summary statistics from Analysis of Variance (ANOVA’s) performed for each
site and each variable. Significant P-values (at 0.05) are highlighted in bold.

Inflorescence height (cm) Flowering length (cm)

Plant sex A. aurea BP A. aurea HP A. scott-th HP A. aurea BP A. aurea HP A. scott-th HP
Female 74.11 84.19 128.78 56.5 60.12 69.08
Male 80.78 94.33 141.5 73.22 73.58 92.63
% var. expl. 0.85 3.88 4.03 9.85 11.93 26.61
F (d.f.) 0.66 (1,78) 3.15 (1,78) 1.39 (1,33) 8.2 (1,75) 8.13 (1,60) 11.60 (1,32)
P-value 0.42 0.08 0.25 0.005 0.006 0.002

112
Table 4.4 compares plant sex ratios found during this study to Lloyd and Webb’s (1980)
results from one-year sex ratio data from six populations of dioecious Aciphylla species in
the South Island of New Zealand. All species except A. pinnatifida exhibit a preponderance
of male plants, while it is mostly the larger species: A. aurea and A. scott-thomsonii which
exhibit a significantly male-biased sex ratio (Table 4.4).

Table 4.4 A comparison of single-year plant sex-ratio data for six species of dioecious Aciphylla in the South
Island of New Zealand between this study and data published in Webb and Lloyd, (1980). Only plants
flowering were counted due to the inability to sex non-flowering plants. *(P<0.05), **(P<0.01),
***(P<0.001), calculated using Chi-square tests on hypothesis of equality.

Species Site Sample size % Males Source

A. monroi Robert Ridge, Nelson 117 61.8** Webb & Lloyd, 1980
A. poppelwellii Rock & Pillar Range, Otago 103 55.3 Webb & Lloyd, 1980
A. pinnatifida Old Man Range, Otago 202 49.5 Webb & Lloyd, 1980
A. aurea Mt. St. Patrick, Canterbury 152 55.6* Webb & Lloyd, 1980
Mt. St. Patrick, Canterbury 135 51.9 Webb & Lloyd, 1980
Burkes Pass, Canterbury 252 54.8 This study
Hakataramea Pass, Canterbury 213 57.3* This study
A. scott-thomsonii Old Man Range, Otago 157 61.8** Webb & Lloyd, 1980
Hakataramea Pass, Canterbury 236 67.8*** This study
A. 'lewis' Lewis Pass, Canterbury 53 52.8 This study

Further non-quantitative observations made during the flowering season suggested that
male flowers had markedly whiter and larger petals than female flowers. Male flowers
were ‘showy’ and more apparent from a distance than the more drab-coloured females (see
Fig. 4.1), a characteristic especially prominent in A. aurea. It was also observed during the
summer field season of 2004/2005 that male plants commenced and completed flowering
mostly before female plants suggesting that flowering phenologies are also different
between the sexes.

113
4.4 DISCUSSION

4.4.1 Male-biased sex ratio in Aciphylla

There was a significant male bias in plant sex ratio in A. aurea and A. scott-thomsonii
populations at Hakataramea Pass, and a similar (though non-significant) trend towards this
at Burkes Pass and Lewis Pass. Population inflorescence sex ratios were also significantly
male-biased at all sites apart from Lewis Pass. Aciphylla scott-thomsonii showed the most
marked predominance of males compared to females for both plant and inflorescence sex
ratio within the Hakataramea Pass population, while A. ‘lewis’ exhibited a sex ratio closer
to 50:50. There is much evidence for male-biased sex ratios in dioecious species and it is
significantly more common than equal or female-biased ratios (of the species for which sex
ratio data is available) (Delph, 1999).

4.4.2 Apparent vs. real sex ratio

Is this simply an apparent sex ratio that we observe during one flowering year amongst
only the flowering individuals, or does this reflect the real sex ratio of all plants in a
population? Other researchers have attempted to determine sex ratios in Aciphylla in the
past and the answer to this question still remains unclear. Some researchers may have
underestimated the frequency of male plants within Aciphylla populations; e.g. Godley
(1964) reported a female-biased sex ratio during one summer for A. aurea. After anthesis in
large Aciphylla species, male inflorescences fall over soon after flowering whereas female
inflorescences remain erect and are more easily observed (Webb & Lloyd, 1980), and when
sex-ratio counts are performed late in the season, it would be easy to underestimate the
number of males. Others however, have similarly reported significant male-biased sex
ratios for Aciphylla populations (e.g. Lloyd & Webb, 1977; Webb & Lloyd, 1980). Under
the “higher costs of reproduction for females” hypothesis, it is easy to predict that males
should be more “optimistic” about flowering than females, because males can take a
chance of making some flowers in a low year in case there is a female to pollinate. Females
should be selected to more strongly concentrate their flowering in high years when
pollination is assured. Therefore, it can be predicted that low-flowering years should have
strong male-bias, but the highest flowering years should mirror the actual sex ratio of plants

114
on site, or even show an apparent female bias. Hence why single-season counting
techniques should be replaced with long-term studies to map and eventually sex all
individual plants within a population to test whether the male bias in sex ratio is apparent
or real.

A number of researchers have shown from one-year experiments, strongly male-biased sex
ratios for American Holly Ilex opaca (Colville, 1932; Barton & Thornton, 1947; Ackerman
& Creech, 1965). These were largely in populations where less than 50% of the plants were
flowering. Nearing (1947), from observation, suggested that the preponderance of male
plants probably originated from studies of young seedling populations of which many of
the female plants had not yet flowered. Clark & Orton (1967) demonstrated with a seven
year study that in fact, male American Holly plants did flower at a younger age than
females, and as seedling age increased, flowering male and female plants occurred at a
more equal frequency. This may be true for a cohort of even-age young plants reaching
first-flowering age, however, in mixed-age populations in the field, the possible higher
mortality of female plants may come to give an increasing (and real) skew to the
population. This example demonstrates the importance of long-term studies, waiting for all
plants within the population to flower, in order to determine actual population structure and
real sex ratios. A similar situation is likely to be the case for Aciphylla as male plants
flower more often (Webb & Lloyd, 1980) due to the lower costs of reproduction. The
higher costs of reproduction in female Aciphylla, compared with males, was not directly
investigated in this study, however the fact that female inflorescences tended to have fewer
flowers (for a given height) suggests that the costs of reproduction was higher for females
than for males (discussed further in Chapter 5).

Lloyd & Webb (1977) provided evidence for different survival of the sexes for A. scott-
thomsonii with a male-biased sex ratio, where - among smaller plants the sex ratio was
close to 50:50, and that there were many small plants that did not flower that year. Webb &
Lloyd (1980) suggest that these non-flowering plants may account for part of the male-
biased ratio in the whole population, because more of them are likely to be females than
males since females flower less often and are usually larger at first flowering (Lloyd &
Webb, 1977). Furthermore, among larger plants that all flower, there is a distinct male-bias
– either indicating that males grow more quickly than females, or they survive longer
(Webb & Lloyd, 1980).

115
There is much evidence for real male-biased sex ratios in dioecious species and it is
significantly more common than equal- or female-biased ratios (of the species for which
long-term sex ratio data is available) (Delph, 1999). There have been a number of
hypotheses proposed to explain this deviation from equality. A few of them are briefly
described below.

Local mate competition

Sexual dimorphism can evolve through intersexual interactions such as competition or mate
choice in dioecious taxa, e.g. competition for limiting resources within and between the
sexes can cause the sexes to specialise on different forms or habitats. Some empiricists
have suggested that this form of intersexual character displacement explains the widespread
occurrence of spatial segregation of the sexes and of sex ratio biases by habitat in dioecious
plants (see review in Geber, 1999).

Effective pollination

Godley (1964) suggested that male-biased sex ratios may have arisen to compensate for
inefficiencies of pollination in situations where seed production is pollen-limited. However,
Webb (1979), Lloyd et al., (1980) and Webb & Lloyd (1980) provide evidence for female
plants of dimorphic Apiaceae being very effectively pollinated even when sex ratios show
little or no male bias. This study did however demonstrate significant evidence for (weak)
pollen limitation at most sites. This will be discussed further in Chapter 5.

Site productivity

Some researchers suggest that where recognisable habitats are found, sex ratios are often
more male-biased in stressful habitats and female-biased in favourable habitats (Geber,

116
1999). For example, productivity declines with altitude and it has been shown that within a
species, male-bias in sex ratio becomes more pronounced at higher altitudes, and it is
hypothesised that males may gain a reproductive advantage in low resource environments
(Geber, 1999; Pickering & Hill, 2002). Pickering & Hill’s (2002) findings arose from only
single-year sex ratio, and may be simply exhibiting an apparent sex ratio bias.

Different costs of reproduction between males and females

Sex-specific life histories may be a consequence of sex-differential patterns of resource

allocation, which are in turn caused by differences in requirements for disseminating pollen
and maturing fruit (Delph, 1999). This may also include the timing of allocation to various
activities as well as differences in the amount of resources allocated. In a seminal paper by
Lloyd & Webb (1977), several predictions were made concerning how sexual dimorphism
in reproductive effort would lead to life-history differences between the sexes. Most
important to note is that the costs of reproduction between the sexes are significant
evolutionary determinants of sexual dimorphism in life histories (Lloyd & Webb, 1977).
They postulated that the energy cost to females of producing ovuliferous flowers and
maturing fruit is higher than the cost to males of producing polliniferous flowers, and that
many observed secondary sex characters in plants (including differential survival of the
sexes), are probably a consequence of this difference in reproductive effort between males
and females (Lloyd & Webb, 1977). It is important to consider sex ratio at other levels of
floral display, beyond the plant-level. The average number of inflorescences per plant is
usually higher among males than among females in dioecious species (Webb & Lloyd,
1980). This was also found for most species in this study. Webb & Lloyd (1980) propose
that males invariably produce more flowers per plant on average than do females in
dioecious species because the number of flowers per inflorescence is greater in males. The
relative number of flowers per inflorescence is important in dioecious populations (Lloyd
& Webb, 1977), and often contributes most to the relative number of pistillate (female) and
staminate (male) flowers in any population, as Opler & Bawa (1978) found for dioecious
tropical forest trees. Results from this study parallel the findings outlined above.

The greater resource allocation to reproduction by females can result in females needing to
accumulate more resource reserves prior to flowering, affecting seasonal and lifetime

117
flowering phenologies (Lloyd & Webb, 1977; Delph, 1999). Because resource constraints
involved with flowering are supposedly greater for females, female plants may be under-
represented in populations because of factors such as age of first flowering and/or
flowering less frequently (Lloyd & Webb, 1977; Meagher & Antonovics, 1982; Delph,
1999; Eckhart, 1999). Also the high costs to females of producing seed and fruit are likely
to increase female mortality more than male, leading to male-biased sex ratios (Lloyd &
Webb, 1977; Delph, 1999). In some species, females must de-fray the costs of reproduction
by delaying reproductive maturity, or reproducing less frequently than males. Also,
differences in the investment per inflorescence may make it advantageous for males to
produce more inflorescences than do females (Pickering & Arthur, 2003), possibly
contributing to a male-biased sex ratio. (The link between costs of reproduction, sex ratios
and mast-seeding will be further discussed in Chapter 5).

4.4.3 Other sexually dimorphic traits in Aciphylla

Pickering & Hill (2002) reported that for A. simplicifolia, male plants exhibited a
significantly greater number of flowers per inflorescence, number of lateral umbels,
number of umbellules per lateral umbel, number of flowers per umbellule on lateral
umbels, number of umbellules on terminal umbel, and overall greater floral display area
(cm2) than females. These dimensions were not quantified for Aciphylla species in this
study, but similar results are expected judging from observation together with findings
from the literature.

It was also observed during the summer field season of 2004/2005 that male Aciphylla
plants commenced and completed flowering mostly before female plants finished. Clark &
Clark (1987) also found in Zamia skinneri, a tropical rainforest cycad, that males began
releasing pollen before the first female was receptive. This makes sense in that by
flowering first, males encourage pollinators initially so they can then deliver pollen on to
females (Proctor et al., 1996). However, the viability of pollen grains after such a period of
time may be affected, and in this case it would not make sense for the male plants in a
population to have their flowers finish and die before female flowers are receptive. Clark &
Clark (1987) also found for Zamia, that individual males were reproductively active for
longer periods than individual females were receptive. It is possible that the female flowers

118
are able to collect (and store or use) pollen earlier than seems obvious from visual
observation. This would make sense of the male timing, and also perhaps of the high seed
set inside bags (Chapter 2). These findings have merely been observational in this study
and should be explored further in future studies as this can have further consequences for
both male and female reproductive success in terms of pollen wastage, and pollen
limitation for females (Marden, 1984; Harder & Wilson, 1994).

4.5 Conclusions

Sexual dimorphism, specifically male-biased sex ratios, was found to occur in nearly all the
Aciphylla populations studied here. The male biased sex ratio and male bias in other
reproductive traits (such as inflorescence size) are entirely consistent with the literature that
states that male-biased sex ratios should be more prominent in dioecious, alpine flora, and
is most likely due to the differential costs of reproduction between male and female plants.
Male bias during a high flowering year indicates that sexual dimorphism is an important
feature of the reproductive ecology of Aciphylla. It still remains to be differentiated
whether there is:

1) an apparent male bias in a single year because a higher percentage of males flower
in any given year (or in all but the highest flowering years);
2) a real male bias in surviving plants due to slower growth and higher mortality of
females, even though the seed sex ratio might be 50:50;
3) Real male bias at all stages due to other skewing factors which means more male
seeds than female are produced.

119
CHAPTER 5 – Discussion

5.1 Insect pollination, masting and reproductive success in Aciphylla

Previously little was known about the reproductive ecology of Aciphylla, although they are
a ubiquitous and prominent component of the high country grassland flora. They have a
unique breeding system that is characterised by being mast seeding, insect-pollinated,
dioecious and sexually dimorphic in a number of traits. Together this results in a
paradoxical reproductive system, because all these individual features are not usually
associated together within one particular breeding system. Theoretically, insect-pollinated
species should not provide a pulsed resource supply (a consequence of masting) if the
pollination system in highly specialised, as this could lead to a break-down of the
mutualism between the pollinator and the plant through pollinator satiation (Kelly, 1994).
On the other hand, if the pollination system of a masting species is generalised, the plant
may compete for pollinators with other flowering plants in the community (Rathcke &
Lacey, 1985). Therefore, how does Aciphylla escape from the apparent disadvantages of
being both masting and insect pollinated?

Chapter 3 showed that Aciphylla are most probably pollinated by a range of insect taxa
from a large number of families and hence have a suite of generalised pollinators. Although
this study did not test for the relative effectiveness of each taxon as pollinators, it is
anticipated that by having more than one pollinator, Aciphylla have probably overcome the
effect of satiating (and starving) pollinators with an unpredictable food supply, (an
important corollary of mast seeding). However, as a consequence of being visited by a
number of insects, Aciphylla must then have to compete for insect pollinators with a wide
range of other plant species flowering at the same time (Rathcke & Lacey, 1985) (assuming
pollinators are a limiting factor in reproductive success (Burd, 1994; Pias & Guitian,
2006)).

It is possible that masting behaviour in Aciphylla may serve to out-compete other plants for
pollinators by providing the biggest and best rewards during high flowering years.

120
Evidence exists for some plant species when in full flower; insects may solely target that
species, in preference to other flower species (Rathcke & Lacey, 1985). In oldfields in
Michigan, honeybees visit Solidago graminifolia only after other plant species ceased
flowering; consequently, earlier flowering clones set fewer seeds (Gross & Werner, 1983).
Also in gardens, competition for honeybee visits commonly reduces seed set among crop
varieties (Evans, 1980). Bees have been shown to search more efficiently for flowers that
are similar to the flower they had just left, directing most of their foraging visits in a single
trip to one flower species, ignoring other rewarding species that they encounter (Waser,
1986; Chittka et al., 1999). Bees also show higher constancy to plants that are locally
abundant (Kunin, 1993; Chittka et al., 1997). It is hence possible, that some insect taxa that
pollinate Aciphylla, are also demonstrating this preference for one species. Therefore it is
likely that in Aciphylla, generalist insect pollinators feed on other (more constant)
flowering species when Aciphylla provide few resources (i.e. during low-flowering years)
and during high-flowering years, insects switch food resources to Aciphylla. Because of
their large floral display size and the provision of huge quantities of pollen and probably
nectar, with dense populations in localised patches (which is likely to be more efficient for
insects in terms of foraging bouts), they offer a highly attractive food source during the
years when they do flower. In this way they may be able to overcome the negative effects
of offering an unpredictable food reward, (through masting) to their pollinators and still
achieve high seed set as observed in this study.

5.2 The importance of synchronous flowering and display size

The isolated female experiment tested for not only flowering asynchrony (i.e. individuals
flowering in a ‘non-mast’ year), but for female asynchrony, because if in dioecious species,
males and females do not flower in synchrony, this could have detrimental consequences
for the reproductive success of both sexes (Crone et al., 2005). This study focussed mostly
on female fecundity, because it is more readily measurable than male success. In Chapter 2
it was demonstrated at some sites that, with increasing distance to flowering males, females
experienced lower seed set rates. Although these effects were relatively small, they
nevertheless demonstrated that even at a relatively small scale, the need for synchronous
flowering between females and males is important. Also important was display size
because individual inflorescences mostly showed increased benefits through having taller

121
inflorescences with more flowers (except A. scott-thomsonii). Larger floral displays were
especially important when females were isolated. In a true non-mast year, when only a
small fraction of the individuals are flowering, it is expected that the effects of pollen
limitation will be considerably greater, and therefore there should be selection against very
small population flowering efforts and small floral display size in Aciphylla. This could
explain the emergence towards having such a large and spectacular floral display size in
large montane Aciphylla species.

It was expected in this study that these negative effects of isolation seen in females (i.e.
lower seed set) would correlate with a decrease in pollinator attraction (i.e. lower average
insect visitation rates to isolated females); however, this was not the case. Isolated females
experienced lower seed set rates than non-isolated females, but did not experience lower
insect visitation rates. The most likely explanation for this is that in areas with only female
inflorescences, insect pollinators move among female stalks just as frequently as in mixed-
sex areas, but are less effective at delivering pollen as they have fewer opportunities to
recharge their pollen loads on males. Another contribution to this effect could be the
possible role of wind pollination, which should vary with distance to males independent of
insect visitation rates. Mixed pollination systems, including generalist insect and wind, are
not uncommon in plants (Mahy et al., 1998). In areas with low insect abundance, such as
alpine communities (Primack, 1983), predominantly insect-pollinated plant species can also
be pollinated by wind, and this has been shown to provide reproductive assurance in some
species, e.g. Linanthus parviflorus (Polemoniaceae), (Goodwillie, 1999). Wind pollination
has been reported for Aciphylla by Dawson (1971) and Mitchell et al., (1999), but they did
not report any tests to confirm whether this was the case. This study points to wind
pollination being possible, but probably not as important as insect pollination for Aciphylla.

5.3 Pollen- and resource-limitation and masting

Pollen limitation is described as an inadequate quantity or quality of pollen that reduces

plant reproductive success (Ashman et al., 2004). Over the past 2 decades, determining
whether seed production is pollen limited has been an area of intense empirical study
(Burd, 1994), however there have only been weak attempts to explain the ecological and
evolutionary causes and consequences of pollen limitation, or evaluate how this influences

122
plant species coexistence, community structure and ecosystem functioning. For example,
plant life histories and mating systems may all influence or be associated with the
probability or strength of pollen limitation (Kunin, 1993; Ashman et al., 2004).

This study tested for pollen limitation to females in Aciphylla populations, using hand
pollination techniques and assessed whether decreased flowering densities and
inflorescence display size limited pollinators. Moreover, it also tested how severe the
effects of pollen limitation were by simultaneously testing for resource limitation. Results
demonstrated no evidence for resource limitation when measured by removing flowers (see
Chapter 2), i.e. with increasing resources per seed to inflorescences with flowers removed.
Brookes & Jesson (2006) tested for simultaneous pollen and resource limitation in A.
squarrosa by removing flowers (comparable to methods used in this study) and obtained
similar results. They also found that resource reduction (removing leaves) and fertiliser
addition did not significantly influence fruit set. Both of these studies suggest that
reproduction in these species was limited by other unknown factors. The effects may not be
seen in a one-year study because masting involves switching resources from growth to
reproduction during flowering years (Kelly & Sork, 2002; Monks & Kelly, 2006). Pollen
was significantly limited however, implying that plants invested sufficient resources for
maximum seed provisioning during this high- flowering year. However, inflorescences at
Burkes and Lewis Passes had relatively lower seed set overall compared with Hakataramea
Pass, leaving many ovules that didn’t (or couldn’t) make seed. This in itself suggests that
there was some level of resource limitation occurring at the plant-level because otherwise
supplemental pollination would have increased seed set rates in a more pronounced
fashion, bringing it nearer to 100%. The effects of pollen and resource limitation were
almost non-existent for A. scott-thomsonii, probably because natural seed set rates were
extremely high (averaging around 90%). In saying that, tall inflorescences with greater
flowering length had slightly negative effects on seed set rate, with more ovules probably
representing a resource cost (Vaughton & Ramsey, 1998).

Some plants exhibit high flower to fruit ratios (Sutherland & Delph, 1984; Sutherland,
1987; Holland et al., 2004), however, in female flowers, the potential advantage of
increased gamete production is unclear. Extra ovules in large flowers are often not
converted into seeds, and because they usurp resources that could be used to increase
fecundity during current or future reproductive events, increased ovule production in

123
excess of what can be ripened may be disadvantageous (Vaughton & Ramsey, 1998).
Newstrom and Roberston (2005) suggest that low fruit-to-flower ratios may be a normal
part of several alternative plant strategies to maximise fitness from the excessive flowers or
to allow a bet-hedging strategy that maximises fruit set when pollinator service is
unreliable. Alternatively, increased ovule production may allow seeds of higher genetic
quality to be produced without reducing seed production if inferior embryos are aborted,
(Vaughton & Carthew, 1993; Newstrom & Robertson, 2005).

Examples of plants that exhibit high flower to fruit ratios are not uncommon (Sutherland,
1987) and include orchids which regularly have low capsule set, but compensate by
producing thousand of tiny seeds per capsule (Nilsson, 1992). However, Aciphylla develop
a maximum of 2 seeds in one schizocarp (fruit) and while thousands of flowers may be
presented along an inflorescence, some are not converted into seed. It remains to be
determined whether this occurs due to the ovule not being fertilised due to the lack of
pollen delivery, or whether the plant was not capable of making seed even if the ovule was
fertilised. However in this study, because supplemental pollen almost always increased
seed set to some degree, (although at Burkes Pass it never increased as much as expected),
this suggests that both of these factors (pollen and resources) play an important role in the
reproductive success of Aciphylla, and their relative effects are probably site-specific.

Some plants may incorporate pollen limitation as an important mediator of selection

(Ashman et al., 2004). Bet-hedging (excess flower or ovule supply relative to the average
pollen load received) may be a common strategy that is employed by plants to attract some
pollen in a stochastic pollinator environment. The fitness gains from the conversion of
occasional “jackpot” visits into extra seeds may outweigh the costs of unutilised ovules or
flowers (Ashman et al., 2004). This may be part of the reason for lower seed set rates in A.
aurea, as extra resources (per flower) did not increase seed set as much as expected.

If reproduction is pollen limited in animal-pollinated, dioecious plants in a stochastic

pollinator environment, it could be argued that under any of the following scenarios: (a)
during a low-flowering year; (b) when females are isolated from males (i.e. pollen source);
(c) when the flowering population is sparse, or; (d) when individual plants flower out of
synchrony with the majority of the population (as shown by naturally isolated Lewis Pass
individuals), that the effects of pollen limitation would be more severe (Kunin, 1993;

124
Forsyth, 2003). Therefore, one could expect selection for reproductive strategies that
minimise the effect that factors such as pollen limitation have on reproductive success,
(unless this is part of an alternative adaptive strategy that maximises fitness from excessive
flowers (Newstrom & Robertson, 2005)). The best example of this is probably seen in A.
scott-thomsonii where natural seed set rates are high, resources are not significantly
limited, and excess flowers lead to a slight fitness loss (see Figure 2.10).

5.4 The effects of sexual dimorphism on reproductive success in Aciphylla

Chapter 4 demonstrated that sexual dimorphism was evident in a number of traits in

Aciphylla. Firstly, although largely observational in this study, it was found that life
histories such as flowering phenology and inflorescence longevity differed between male
and female plants. Male plants generally begin flowering before females and male flower
stalks fell over soon after flowering ceased, while female stalks remain standing often for
more than a year after flowering. (Primack, 1983) reported that individual male flowers are
open for a longer period of time than females. Sexual dimorphism was also evident in a
number of inflorescence level traits, with floral display in males being showier (see Fig.
4.1), with greater flowering length, and more inflorescences per plant. Similar findings
have been reported for various New Zealand Aciphylla species (e.g. Dawson & Le Comte,
1978; Webb et al., 1999) as well as for Australian Aciphylla species (Pickering, 2000,
2001). Additionally, inherent in primary sex characteristics, is that males produce pollen as
well as nectar, offering twice the variety of reward than female flowers. In this study, most
insect groups responded to this dimorphism in inflorescence traits and visited males more
frequently than females. Males also received a greater diversity of insect visitors than
females did. These sex-specific traits therefore, probably lead to the discrepancy in visits
between male and female inflorescences via pollinator attraction.

The greater resource allocation to reproduction by females most likely results in female
plants needing to accumulate more resources prior to flowering (Lloyd & Webb, 1977).
This can sometimes take years in masting species, leading to the ability for males to flower
more often than females (Meagher & Antonovics, 1982; Delph, 1999; Eckhart, 1999).
Consequently, this is probably an important reason for the apparent male-biased sex ratios
reported in this study, and elsewhere for Aciphylla (e.g. Pickering & Hill, 2002).

125
Nonetheless, during a high-flowering year (which is particularly important for the
reproductive success of females), a male-biased sex ratio can pose a number of important
consequences on a population. For instance, the sex ratio is indicative of local pollen
densities. Having more male plants and/or inflorescences in a population may positively
affect female fecundity, because as this study suggests, males attract significantly more
insects overall. If Aciphylla do compete for generalist pollinators then males could provide
the resource rewards, and visits to female flowers would also occur more often (even if this
is by chance). On the other hand, male-biased sex ratios could have negative consequences
for female visitation rates if insects consciously differentiate between male and female
inflorescences and display a real preference for males. Overall, males appear to be
attractive to insects regardless of size, while females generally benefit from being taller
and/or having more flowers. Perhaps by being more “attractive” and more abundant, males
are helping the females out by attracting more insects to the Aciphylla population (which
indirectly benefits males by ensuring there will be more seeds produced which each male
can potentially sire).

If reproduction costs relatively less for males in Aciphylla, the pressures towards needing to
mast are not so great for males. If males do in fact make up the majority of the plants
flowering during low-flowering years, whereas females only flower during high-flowering
years (Hogan et al., 1998) then maybe selection for masting has been stronger in females
than males. Hence females may be driven to display masting behaviour as the costs of
reproduction are so great that they can’t do it every year and males may then be forced to
follow suit (although they still flower more often in low-flowering years, presumably on
the grounds that male prediction of when females will flower is imperfect so there is
always a chance of siring some seeds in low-flowering years).

5.5 Further research on the reproductive ecology of Aciphylla

Also observed during this study was that seed predation levels were high on many
inflorescences. Aphids were particularly abundant, and it was discovered that they also
prey upon seeds (as well as sucking sap from foliage). Many plant species are known to
show clear benefits from masting by satiating seed predators during high-flowering years,
e.g. Chionochloa (Kelly & Sullivan, 1997; McKone et al., 1998). Because Chionochloa

126
and Aciphylla species often coexist in high-country grasslands, it is possible they share
some of the same seed predators. The effects of seed predation on reproductive success in
Aciphylla should not be ruled out and remains to be tested in relation to masting.

Monitoring the effects of pollen limitation on naturally low-flowering years also remains to
be tested as long-term studies would provide a better insight into whether this is an
important selection pressure for masting in Aciphylla species. The pollination aspect of this
system was not studied in great detail and it was not possible to gain a clear understanding
of the movement of pollen, the relative effectiveness of the pollinator taxa, and the floral
rewards offered. Obtaining such detailed information about the pollination biology of
Aciphylla was beyond the scope of this research, but detailed analyses would greatly aid in
the understanding of the pollination system of Aciphylla and the myriad interactions
between the plant and its extensive suite of visitors. This is important, because in order to
fully understand the reproductive system of these species, we need to know more about the
role of floral attractiveness to pollinators.

The conservation implications of this research are also noteworthy. A decline in the
diversity and abundance of insect pollinators, for example, caused through habitat
fragmentation (McGlone & Moar, 1998), can lead to a decrease in pollination rate
(Wilcock & Neiland, 2002) and have detrimental consequences on fecundity. This effect
could be particularly threatening to plant species that are self-incompatible and rely on
insects for pollination, such as Aciphylla. Moreover, if plant populations become
fragmented, it becomes more important on a spatial scale as pollinator movement between
more distant populations would be increasingly unlikely (Farwig et al., 2004) and hence
fail to promote genetic diversity within a species (Mitchell et al., 1999). For species such as
Aciphylla subflabellata, which is now becoming rare and sparse on the Canterbury plains
and foothills due to habitat loss, small populations with isolated individuals flowering far
from nearest conspecific neighbour could be affected by loss of pollinator movement and
the likelihood of decreased transfer of conspecific pollen.

127
5.6 Conclusions

For Aciphylla, overall high seed set rates in the mass-flowering summer studied suggests
that this is a reproductive strategy that works well for female plants. Although masting is
probably a necessary outcome of having a large floral display size, benefits must be greater
than the high costs of masting in order for this reproductive system to be successful.
Results demonstrated that Aciphylla are generalist insect-pollinated and probably compete
for insects through having a large and attractive floral display during high-flowering years.
In most cases, inflorescences with a larger floral display had higher reproductive success.
There was no evidence for resource limitation, while pollen limitation was evident across
all sites (although this effect was relatively small). Hence, for Aciphylla, being a bigger
inflorescence in dense, synchronous flowering populations was generally better for overall
reproductive success. Hence masting may be a necessary strategy resulting from the need
to have large, synchronous flowering to achieve pollinator economies of scale.

128
Acknowledgements

To my supervisors: Prof. Dave Kelly - thanks for being so approachable, encouraging,

helpful and highly knowledgeable. It was an honour to work with one of the world’s best in
this field. Dr Jo Ward – it was wonderful to discuss all my ideas with you, thank you
especially for this (among everything else). You were both fantastic supervisors. Sincere
thanks to the Miss E.L. Hellaby Indigenous Grasslands Research Trust for the funding for
this research, and to Alan Mark in particular for his long-term interest in Aciphylla.

To the owners of Dalgety and Grampian Stations in the Mackenzie country – thank you for
letting me work on your land where I had fantastic views of Mt Cook each day (sorry I
wasn’t doing Spaniard biocontrol though!) Thanks also to DoC for permits and allowing
me to have field sites in great locations around Canterbury. Thanks also to South Island
DoC field staff and CTC members for proving useful information on flowering Aciphylla
populations.

For help in the field I thank Sarah Raoulia Luxton, Robbie Holdaway, Angie Merkel, and
Ronny Groenteman. Maggie Tisch and Jenny Ladley – thank you for everything you did
for me – you were both integral parts of this project in many ways. Thanks to David
Glenny (Landcare Research) for sharing your considerable knowledge of Aciphylla with
me, whether it be during a rogaine or a trip to the hills. Thanks also to Dave Condor for
putting up my smelly, spiky plants in storage for so long at the glasshouse.

Jon Sullivan and Richard Holdaway – thanks for helpful advice and insightful discussions,
and Matt Walters (not only for your excellent photography skills) but for very important
discussions about bush-bashing trips in the Kahurangi Wilderness. I greatly appreciate all
the help from entomologists Peter Johns, John Dugdale, Steve Pawson, Ian McLellan, and
John Marris – I learnt a great deal from you all. Big thanks in particular to Alison Evans
(DoC) for providing the impetus for the insect side of this project in the first place and
teaching me how to deal with the media. Thanks to the Conservation Minister, the Hon
Chris Carter for the lovely letter of appreciation after finding the “extinct” weevil
Hadramphus tuberculatus. (I can’t really write to him and thank him personally, so this is
the best place to do it in case he ever reads my thesis!)

Huge thanks to my family – Warren and Jennifer Young for support, Alan and Janette
Holdaway for reading drafts and useful comments, and my wonderful group of friends in
and around the ecology “lab” for reading drafts, help, support, encouragement, and fun
times. Deb, Michal, Chuckie, Merodie, Mel, Ronny, et al., it has been great to go through
this with all of you – you all know what it’s like and you have been vital in helping me
bring it all together whether it be in the lab, office, field or recreationally. Finally I would
like to thank my wonderful husband Robbie for being there for me in almost every way
possible. Thanks for your help in the field, statistical advice, extremely useful discussions,
listening to me get excited about strange spiky plants, for making me dinner and for being a
big part of it all.

129
References

Abe, T. (2000) Flowering phenology, display size, and fruit set in an understory dioecious
shrub, Aucuba japonica (Cornaceae). American Journal of Botany, 88, 455-461.

Ackerman, J.D. & Montalvo, A.M. (1990) Short- and long-term limitations to fruit
production in a tropical orchid. Ecology, 74, 263-272.

Ackerman, W.L. & Creech, J.L. (1965) Progeny analysis of interspecific crosses between
Ilex cornuta Lindl. and Ilex ciliospinosa Loes. Proceedings of the American Society
of Horticultural Science, 86, 741-749.

Ågren, J., Danell, L., Elmquist, T., Ericson, L., & Hjälten, J. (1999). Sexual dimorphism
and biotic interactions. In Gender and sexual dimorphism in flowering plants (eds.
M. A. Geber, T.E.Dawson, and L. F. Delph), pp. 217-246. Springer-Verlag, Berlin,
Germany.

Allan, H.H. (1982) Flora of New Zealand Government Printer, Wellington.

Andersson, S. (1996) Floral display and pollination success in Senecio jacobaea

(Asteraceae): Interactive effects of head and corymb size. American Journal of
Botany, 83, 71-75.

Armstrong, J.E. (1997) Pollination by deceipt in nutmeg (Myristica insipida,

Myristicaceae): Floral display and beetle activity in male and female trees.
American Journal of Botany, 84, 1266-1274.

Ashman, T.L., Knight, T.M., Steets, J.A., Amarasekare, P., Burd, M., Campbell, D.R.,
Dudash, M.R., Johnston, M.O., Mazer, S.J., Mitchell, R.J., Morgan, M.T., &
Wilson, W.G. (2004) Pollen limitation of plant reproduction: ecological and
evolutionary causes and consequences. Ecology, 85, 2408-2421.

130
Ashton, P.S., Givnish, T.J., & Appanah, S. (1988) Staggered flowering in the
Dipterocarpaceae: new insights into floral induction and the evolution of mast
fruiting in the aseasonal tropics. American Naturalist, 132, 44-66.

Atkinson, I.A.E. & Greenwood, R.M. (1989) Relationships between moas and plants. New
Zealand Journal of Ecology, 12, 67-96.

Augspurger, C.K. (1980) Mass-flowering of a tropical shrub (Hybanthus prunifolius):

influence of pollinator attraction and movement. Evolution, 34, 475–488.

Augspurger, C.K. (1981) Reproductive synchrony of a tropical shrub: experimental studies

on the effects of pollinators and seed predators on Hybanthus prunifolius
(Violaceae). Ecology, 62, 755–788.

Baker, H.G. (1976). 'Mistake' pollination as a reproductive system, with special reference
to the Caricaceae. In Tropical trees: variation, breeding and conservation (eds. J.
Burley, Styles, B.T.), pp. 161-169. Academic Press, London, U.K.

Barton, L.V. & Thornton, N.C. (1947) Germination and sex populations studies of Ilex
opaca Ait. Contributions from Boyce Thompson Institute, 14, 405-410.

Bateman, A.J. (1948) Intra-sexual selection in Drosophila. Heredity, 2, 349-368.

Bawa, K.S. (1974) Breeding systems of tree species of a lowland tropical community.
Evolution, 28, 85-92.

Bawa, K.S. (1977) The reproductive biology of Cupania guatemalensis Radlk.

(Sapindaceae). Evolution, 31, 52-63.

Bawa, K.S. (1980) Evolution of dioecy in flowering plants. Annual Review of Ecology and
Systematics, 11, 15-39.

Bierzychudek, P. (1981) Pollinator limitation of plant reproductive effort. American

Naturalist, 117, 838-840.

131
Bierzychudek, P. (1987) Pollinators increase the cost of sex by avoiding female flowers.
Ecology, 68, 444-447.

Bodsworth, J. (2003) An Anglo-Saxon Dictionary. In The Anglo-Saxon Dictionary Project,

(based on the manuscript collections of the late Joseph Bodsworth D.D.F.R.S).

Bosch, M. & Waser, N.M. (1999) Effects of local density on pollination and reproduction
in Delphinium nuttallianum and Aconitum columbianum (Ranunculaceae).
American Journal of Botany, 86, 871-879.

Brockie, R.E. (1986) Periodic heavy flowering of New Zealand flax (Phormium,
Agavaceae). New Zealand Journal of Botany, 24, 381-386.

Bronstein, J.L. (2004) The costs of mutualism. Integrative and Comparative Biology, 41,
825-839.

Brookes, R.H. & Jesson, L.K. (2006) No evidence for simultaneous pollen and resource
limitation in Aciphylla squarrosa: A long-lived, masting herb. In Press.

Brunet, J. & Sweet, H.R. (2006) Impact of insect pollinator group and floral display size on
outcrossing rate. Evolution, 60, 234-246.

Buide, M.L. (2006) Pollination ecology of Silene acutifolia (Carophyllaceae): floral traits,
variation and pollinator attraction. Annals of Botany, 97, 289-297.

Burd, M. (1994) Bateman's Principle and plant reproduction: the role of pollen limitation in
fruit and seed set. The Botanical Review, 60, 83.

Busgen, M. & Munch, E. (1929) The structure and life of forest trees, 3rd edn. Chapman
and Hall, London.

Campbell, A.D. (1981) Flowering records for Chionochloa, Aciphylla, and Celmisia
species in the Craigeburn Range, South Island, New Zealand. New Zealand Journal
of Botany, 19, 97-103.

132
Campbell, D.R. (1985) Pollinator sharing and seed set of Stellaria pubera: competition for
pollination? Ecology, 66, 544-553.

Charlesworth, B. (1999). Theories of the evolution of dioecy. In Gender and sexual

dimorphism in flowering plants (eds. M. A. Geber, T.E. Dawson, and L. F. Delph),
pp. 34-60. Springer-Verlag, Berlin.

Charlesworth, B. & Charlesworth, D. (1978) A model for the evolution of dioecy and
gynodioecy. American Naturalist, 112, 975-997.

Charlesworth, D. (1989) Why do plants produce so many more ovules than seeds? Nature,
338, 21-22.

Chittka, L., Gumbert, A., & Kunze, J. (1997) Foraging dynamics of bumblebees: correlates
of movements within and between plant species. Behavioural Ecology, 8, 239-249.

Chittka, L., Thomson, J.D., & Waser, N.M. (1999) Flower constancy, insect psychology,
and plant evolution. Naturwissenschaften, 86, 361-377.

Christensen, K.M. & Whitam, T.G. (1991) Indirect herbivore mediation of avian seed
dispersal in pinyon pine. Ecology, 72, 534-542.

Clark, D.A. & Clark, D.B. (1987) Temporal and environmental patterns of reproduction in
Zamia skinneri, a tropical rain forest cycad. Journal of Ecology, 75, 135-149.

Clark, R.B. & Orton Jr, E.R. (1967) Sex ratio in Ilex opaca Ait. Horticultural Science, 2,
115.

Collevatti, R.G., Campos, L.A.O., & Da Silva, A.F. (1998) Pollination ecology of the
tropical weed Triumfetta semitriloba Jacq. (Tiliaceae), in the South-Eastern Brazil
Revista Brasileira de Biologia, 58, 383-392.

Colville, P. (1932) Growing Christmas Holly on the farm. U.S. Department of Agriculture
Farmers Bulletin, 1693, 21.

133
Crone, E.E. & Lesica, P. (2004) Causes of synchronous flowering in Astragalus
scaphoides, an iteroparous perennial plant. Ecology, 85, 1944-1954.

Crone, E.E., Polansky, L., & Lesica, P. (2005) Empirical models of pollen limitation,
resource aquisition, and mast seeding by a bee-pollinated wildflower. The American
Naturalist, 166, 396-408.

Curran, L.M. & Leighton, M. (2000) Vertebrate responses to spatio-temporal variation in

seed production of mast fruiting Dipterocarpaceae. Ecological Monographs, 70,
101-128.

Dafni, A. (1992) Pollination Ecology: A practical approach IRL Press at Oxford

University Press, Haifa, Isael.

Darwin, C. (1877) The different forms of flowers on plants of the same species John
Murray, London.

Davila, Y.C. & Wardle, G.M. (2002) Reproductive ecology of the Australian herb
Trachymene incisa subsp. incisa (Apiaceae). Australian Journal of Botany, 50, 619-
626.

Dawson, J.W. (1971). Relationships of the New Zealand Umbelliferae. In The Biology and
Chemistry of the Umbelliferae (ed V.H. Heywood), Vol. 1. Linnean Society of
London by Academic Press, London.

Dawson, J.W. & Le Comte, J.R. (1978) Research on Aciphylla - a progress report. Tuatara,
23, 49-67.

Dawson, T.E. & Geber, M.A. (1999). Sexual dimorphism in physiology and morphology.
In Gender and sexual dimorphism in flowering plants (eds. M. A. Geber,
T.E.Dawson, and L. F. Delph), pp. 176-215. Springer-Verlag, Berlin.

de Jong, T.J., Van Batenburg, F.H.D., & van Dijk, J. (2002) Seed sex ratio in dioecious
plants depends on relative dispersal of pollen and seeds: an example using a
chessboard simulation model. Journal of Evolutionary Biology, 15, 373-379.

134
DeAngelis, D.L. & Holland, J.N. (2006) Emergence of ratio-dependent and predator-
dependent functional responses for pollination mutualism and seed parasitism.
Ecological Modelling, 191, 551-556.

Delph, L.F. (1996) Sexual dimorphism in flower size. American Naturalist, 148, 299-320.

Delph, L.F. (1999). Sexual dimorphism in life history. In Gender and sexual dimorphism in
flowering plants (eds. M. A. Geber, T.E.Dawson, and L. F. Delph), pp. 150-173.
Springer-Verlag, Berlin.

Delph, L.F. & Lively, C.M. (1989) The evolution of floral colour change: pollinator
attraction versus physiological constraints in Fuchsia excorticata Evolution, 43,
1252-1262.

Donnelly, S.E., Lortie, C.J., & Aarsen, L.W. (1998) Pollination in Verbascum thapsus
(Scrophulariaceae): the advantage of being tall. American Journal of Botany, 85,
1618-1625.

Eckhart, V.M. (1999). Sexual dimorphism in flowers and inflorescences. In Gender and
sexual dimorphism in flowering plants (eds. M.A. Geber, J.W. Dawson & L.F.
Delph), pp. 123-148. Springer-Verlag, Berlin.

Ehlers, B.K., Olesen, J.M., & Agren, J. (2002) Floral morphology and reproductive success
in the orchid Epipactus helleborine: regional and local across-habitat variation.
Plant Systematics and Evolution, 236, 1820-1831.

Evans, L.T. (1980) The natural history of crop yield. American Scientist, 68, 388-397.

Farwig, N., Randrianarina, E.F., Voight, F.A., Kraemer, M., & Bonning-Gaese, K. (2004)
Pollination ecology of the dioecious tree Commiphora guillauminii in Madagascar.
Journal of Tropical Ecology, 20, 307-316.

Feinsinger, P. & Tiebout (III), H.M. (1991) Competition among plants sharing
hummingbird pollinators: laboratory experiments on a mechanism. Ecology, 72,
1946-1952.

135
Forsyth, S.A. (2003) Density-dependent seed set in the Haleakala silversword: evidence for
an Allee effect. Oecologia, 136, 551-557.

Geber, M.A. (1999). Theories of the evolution of sexual dimorphism. In Gender and sexual
dimorphism in flowering plants (eds. M. A. Geber, T. E. Dawson, and L. F. Delph),
pp. 97-122. Springer-Verlag, Berlin.

Godley, E.J. (1964) Breeding systems in New Zealand plants 3. Sex ratios in some natural
populations. New Zealand Journal of Botany, 2, 205-212.

Godley, E.J. (1979) Flower biology in New Zealand. New Zealand Journal of Botany, 17,
441-466.

Goodwillie, C. (1999) Wind pollination and reproductive assurance in Linanthus

parviflorus (Polemoniaceae), a self-incompatible annual. American Journal of
Botany, 86, 948-954.

Grindeland, J.M., Sletvold, N., & Ims, R.A. (2005) Effects of floral display size and plant
density on pollinator visitation rate in a natural population of Digitalis pupurea.
Functional Ecology, 19, 383-390.

Groom, M.J. (1998) Allee effects limit population viability of an annual plant. American
Naturalist, 151, 487-496.

Gross, R.S. & Werner, P.A. (1983) Relationships among flowering phenology, insect
visitors, and seed set of individuals: experimental studies on four co-occurring
species of goldenrod (Solidago: Compositae). Ecological Monographs, 53, 95-117.

Haase, P. (1986) Flowering records of some subalpine trees and shrubs at Arthurs Pass,
New Zealand. New Zealand Journal of Ecology, 9, 19-23.

Haig, D. & Westoby, M. (1988) On limits to seed production. American Naturalist, 131,
757-759.

136
Harder, L.D. & Wilson, W.G. (1994) Floral evolution and male reproductive success:
optimal dispensing schedules for pollen dispersal by animal-pollinated plants.
Evolutionary Ecology, 8, 542-559.

Herrera, C.M. (1996) Floral traits and plant adaptation to insect pollinators: A devils
advocate approach. Chapman and Hall, New York.

Herrera, C.M. (1998a) Population-level estimates if interannual variability in seed

production: what do they actually tell us? Oikos, 82, 612-616.

Herrera, C.M., Jordano, P., Guitian, P., & Travaset, A. (1998b) Annual variability in seed
production by woody plants and the masting concept: reassessment of principles
and relationship to pollination and seed dispersal. American Naturalist, 152, 576-
594.

Hessing, M.B. (1988) Geitonogamous pollination and its consequences in Geranium

caespitosum. American Journal of Botany, 75, 1324-1333.

Hett, J.M. (1971) A dynamic analysis of age in sugar maple seedlings. Ecology, 52, 1071-
1074.

Hirayama, K., Ishida, K., & Tomaru, N. (2005) Effects of pollen shortage and self-
pollination on seed production of an endangered tree, Magnolia stellata. Annals of
Botany, 95, 1009-2005.

Hogan, K.P., Garcia, M.B., Cheeseman, J.M., & Loveless, M.D. (1998) Inflorescence
photosynthesis and investment in reproduction in the dioecious species Aciphylla
glaucescens. New Zealand Journal of Botany, 36, 653-660.

Holland, J.N., Bronstein, J.L., & DeAngelis, D.L. (2004) Testing hypotheses for excess
flower production and low fruit-to-flower ratios in a pollinating seed-consuming
mutualism. Oikos, 105, 633-640.

137
Holland, J.N. & Fleming, T.H. (2002) Co-pollinators and specialization in the pollinating
seed-consumer mutualism between senita cacti and senita moths. Oecologia, 133,
534-540.

Horvitz, C.C. & Schemske, D.W. (1988) A test of the pollinator limitation hypothesis for a
neotropical herb. Ecology, 69, 200-206.

Howe, H.F. & Westley, L.C. (1986) Ecology of pollination and seed dispersal. Plant
Ecology Blackwells Scientific, Oxford.

Irwin, R.E., Adler, L.S., & Brody, A.K. (2004) The dual role of floral traits: pollinator
attraction and plant defense. Ecology, 85, 1503–1511.

Isagi, Y., Sugimura, K., Sumida, A., & Ito, H. (1997) How does masting happen and
synchronize? Journal of Theoretical Biology, 187, 231-239.

Janzen, D.H. (1971) Seed predation by animals. Annual Review of Ecology and
Systematics, 2, 465-492.

Janzen, D.H. (1978). Seeding patterns of tropical trees. In Tropical trees as living systems
(eds. P.B. Tomlinson & M.H. Zimmerman), pp. 83-128. Cambridge University
Press, Cambridge.

Jennerston, O. (1993) Insect flower visitation frequency and seed production in relation to
patch size of Viscia vulgaris (Carophyllaceae). Oikos, 68, 283.

Johnson, S.G., Delph, L.F., & Elderkin, C.L. (1995) The Effect of Petal-Size Manipulation
on Pollen Removal, Seed Set, and Insect-Visitor Behavior in Campanula
americana. Oecologia, 102, 174-179.

Kato, E. & Hiura, T. (1990) Fruit set in Styrax obassia (Styracaeae): The effect of light
availability, display size, and local floral density. American Journal of Botany, 86,
495-501.

138
Kawarasaki, S. & Hori, Y. (1999) Effect of flower number on the pollinator attractiveness
and the threshold plant size for flowering in Pertya triloba (Asteraceae). Plant
Species Biology, 14, 69-74.

Kearns, C.A., Inouye, D.W., & Waser, N.M. (1998) Endangered mutualisms: The
conservation of plant-pollinator interactions. Annual Review of Ecology and
Systematics, 29, 83-112.

Kelly, D. (1994) The evolutionary ecology of mast seeding. Trends in Ecology and
Evolution, 9, 465-470.

Kelly, D., Harrison, A.L., Lee, W.G., Payton, I.J., Wilson, P.R., & Schauber, E.M. (2000)
Predator satiation and extreme mast seeding in 11 species of Chionochloa
(Poaceae). Oikos, 90, 477-488.

Kelly, D., Hart, D.E., & Allen, R.B. (2001) Evaluating the wind pollination benefits of
mast seeding. Ecology, 82, 117-126.

Kelly, D. & Sork, V.L. (2002) Mast seeding in perennial plants: Why, How, Where?
Annual Review of Ecology and Systematics, 33, 427-447.

Kelly, D. & Sullivan, J.J. (1997) Quantifying the benefits of mast seeding on predator
satiation and wind pollination in Chionochloa pallens (Poaceae). Oikos, 78, 143-
150.

Kerkoff, A.J. (2004) Expectation, explanation and masting. Evolutionary Ecology

Research, 6, 1003-1020.

Kerkoff, A.J. & Ballantyne (IV), F. (2003) The scaling of reproductive variability in trees.
Ecology Letters, 6, 850-856.

Kindlmann, P. & Jersakova, J. (2006) Effect of floral display on reproductive success in

terrestrial orchids. Folia Geobotanica, 41, 47-60.

139
Kirchner, F., Luijten, S.H., Imbert, E., Riba, M., Mayol, M., Gonzalez-Matinez, S.C.,
Mignot, A., & Colas, B. (2005) Effects of local density on insect visitation and
fertilization success in the narrow-endemic Centaurea corymbosa (Asteraceae).
Oikos, 11, 130-142.

Knight, T.M., Steets, J.A., & Ashman, T.L. (2006) A quantitative synthesis of pollen
supplementation experiments highlights the contribution of resource reallocation to
estimates of pollen limitation. American Journal of Botany, 93, 271-277.

Koenig, W.D. & Ashley, M.V. (2003) Is pollen limited? The answer is blowin' in the wind.
Trends in Ecology and Evolution, 18, 157-159.

Koenig, W.D., Kelly, D., Sork, V.L., Duncan, R.P., Elkinton, J.S., Peltonen, M.S., &
Westfall, R.D. (2003) Dissecting components of population-level variation in seed
production and the evolution of masting behaviour. Oikos, 102, 581-591.

Koenig, W.D. & Knops, J.M.H. (2000) Patterns of annual seed production by Northern
Hemisphere trees: a global perspective. The American Naturalist, 155, 59-69.

Koenig, W.D. & Knops, J.M.H. (2005) The mystery of masting in trees. American
Scientist, 93, 340-347.

Kudo, G. & Harder, L.D. (2005) Floral and inflorescence effects on variation in pollen
removal and seed production among six legume species. Functional Ecology, 19,
245-254.

Kunin, W.E. (1993) Sex and the single mustard: population density and pollinator
behaviour effects on seed set. Ecology, 74, 2145-2160.

Ladley, J.J., Kelly, D., & Robertson, A.W. (1997) Explosive flowering, nectar production,
breeding systems, and pollinators of New Zealand mistletoes (Loranthaceae). New
Zealand Journal of Botany, 35, 345-360.

140
Leathwick, J., Wilson, G., Rutledge, D., Wardle, P., Morgan, F., Johnston, K., McLeod,
M., Kiekpatrick, R. (2003) Land Environments of New Zealand (LENZ). Nga Taio
O Aotearoa, Wellington.

Le Corff, J., Agren, J., & Schemske, D.W. (1998) Floral display, pollinator discrimination,
and female reproductive success in two monoecious Begonia species. Ecology, 79,
1610-1619.

Lindsey, A.H. (1984) Reproductive Biology of Apiaceae. I. Floral Vistors to Thaspium and
Zizia and Their Importance in Pollination American Journal of Botany, 71, 375-
387.

Lloyd, D.G. (1975) The maintenance of gynodioecy and androdioecy in angiosperms.

Genetica, 45, 325-339.

Lloyd, D.G. (1985) Progress in understanding the natural history of New Zealand plants.
New Zealand Journal of Botany, 23, 707-722.

Lloyd, D.G. & Webb, C.J. (1977) Secondary sex characters in plants. Botanical Review,
43, 177–216.

Lloyd, D.G., Webb, C.J., & Primack, R.B. (1980) Sexual strategies in plants II. Data on the
temporal regulation of maternal investment. New Phytologist, 86, 81-92.

Lofgren, A. (2002) Effects of isolation on distribution, fecundity, and survival in the self-
incompatible Achillea millefolium (L.). Ecoscience, 9, 503-508.

Lortie, C.J. & Aarsen, L.W. (1999) The advantage of being tall: Higher flowers receive
more pollen in Verbascum thapsus L. (Scrophulariaceae). Ecoscience, 6, 68-71.

Mahy, G., De Sloover, J., & Jacquemart, A.L. (1998) The generalist pollination system and
reproductive success of Calluna vulgaris in the Upper Ardenne. Canadian Journal
of Botany, 76, 1843–1851.

141
Marden, J.H. (1984) Remote perception of floral nectar by bumblebees. Oecologia, 64,
232-240.

Mark, A.F. (1970) Floral initiation and development in New Zealand alpine plants. New
Zealand Journal of Botany, 8, 67-75.

Mark, A.F. & Adams, N.M. (1973) New Zealand alpine plants A.H. & A.W. Reed Ltd,
Wellington.

Marr, D.L., Leebens-Mack, J., Elms, L., & Pellmyr, O. (2000) Pollen dispersal in Yucca
filamentosa (Agavaceae): the paradox of self-pollination behavior by Tegeticula
yuccasella (Prodoxidae). American Journal of Botany, 87, 670-677.

McKewen, W.M. (1987) Ecological Regions and Districts of New Zealand. 3rd ed. DSIR
Publishing, Wellington.

McGlone, M.S. & Moar, N.T. (1998) Dryland Holocene vegetation history, Central Otago
and the Mackenzie Basin, South Island, New Zealand. New Zealand Journal of
Botany, 36, 91-111.

McKone, M.J., Kelly, D., Harrison, A.L., Sullivan, J.J., & Cone, A.J. (2001) Biology of
insects that feed in the inflorescences of Chionochloa (Poaceae) in New Zealand
and their relevance to mast seeding. New Zealand Journal of Zoology, 28, 89-101.

McKone, M.J., Kelly, D., & Lee, W.G. (1998) Effects of climate change on mast-seeding
species: frequency of mass flowering and escape from specialist insect seed
predators. Global Change Biology, 4, 591-596.

Meagher, T.R. & Antonovics, J. (1982) The population biology of Chamaelirium luteum, a
dioecious member of the lily family: life history studies. Ecology, 63, 1690-1700.

Metcalf, L.J. (1995) The Propogation of New Zealand Native Plants Godwit Publishers
Ltd.

142
Mitchell, A.D., Hogan, K.P., & Chapman, H. (1999) Genetic variation in Aciphylla
glaucescens (Apiaceae). New Zealand Journal of Ecology, 23, 61-67.

Mitchell, A.D., Webb, C.J., & Wagstaff, S.J. (1998) Phylogenetic relationships of species
of Gingidia and related genera (Apiaceae, subfamily Apioideae). New Zealand
Journal of Botany, 36, 417-424.

Mitchell, R.J., Karron, J.D., Holmquist, K.G., & Bell, J.M. (2004) The influence of
Mimulus ringens floral display size on pollinator visitation patterns. Functional
Ecology, 18, 116-124.

Monks, A. & Kelly, D. (2006) Testing the resource-matching hypothesis in the mast
seeding tree Nothofagus truncata (Fagaceae) Austral Ecology, 31, 366-375

Moore, L.B. & Irwin, J.B. (1978) The Oxford Book of New Zealand Plants Oxford
University Press, Wellington.

Nearing, G.G. (1947) Masculine Holly. In Readers Letters to the Editor, Horticulture, Vol.
25, pp. 97.

Newstrom, L.E. & Robertson, A.W. (2005) Progress in understanding pollination systems
in New Zealand. New Zealand Journal of Botany, 43, 1-59.

Nilsson, L.A. (1992) Orchid pollination biology. Trends in Ecology and Evolution, 7, 255–
259.

Nilsson, S.G. & Wastljung, U. (1987) Seed predation and cross-pollination in mast-seeding
beech (Fagus sylvatica) patches. Ecology, 68, 260-265.

Norton, D.A. & Kelly, D. (1988) Mast seeding over 33 years by Dacrydium cupressinum
Lamb. (rimu) (Podocarpaceae) in New Zealand: the importance of economies of
scale. Functional Ecology, 2, 399-408.

Oliver, W.R.B. (1956) The genus Aciphylla. Transactions of the Royal Society of New
Zealand, 84, 1-18.

143
Ollerton, J. (1996) Reconciling ecological processes with phylogenetic patterns: the
apparent paradox of plant-pollinator systems. Journal of Ecology, 84, 767-769.

Opler, P.A. & Bawa, K.S. (1978) Sex ratios in tropical forest trees. Evolution, 32, 812-821.

Orellana, M.R., Rovira, A.M., Blanche, C., & Bosch, M. (2005) Pollination and
reproductive success in the gynodioecious endemic Thymus loscosii (Lamiaceae).
Canadian Journal of Botany, 83, 183-193.

Ortiz-Perez, E., Cervantes-Martinez, I.G., Wiley, H., Hanlin, S.J., Horner, H.T., Davis,
W.H., & Palmer, R.G. (2004) Phenotypic recurrent selection for increased
pollinator attraction to produce hybrid soybean seed American Society of Agronomy
Abstracts, 1029.

Ostfeld, R.S. & Keesing, F. (2000) Pulsed resources and community dynamics of
consumers in terrestrial ecosystems. Trends in Ecology and Evolution, 15, 232-237.

Pias, B. & Guitian, P. (2006) Breeding system and pollen limitation in the masting tree
Sorbus aucuparia L. (Rosaceae) in the NW Iberian Peninsula. Acta Oecologica, 29,
97-103.

Pickering, C.M. (2000) Sex-specific differences in floral display and resource allocation in
Australian alpine dioecious Aciphylla glacialis (Apiaceae). Australian Journal of
Botany, 48, 81-91.

Pickering, C.M. (2001) Size and sex of floral displays affect insect visitation rates in the
dioecious Australian alpine herb, Aciphylla glacialis. Nordic Journal of Botany, 21,
401-409.

Pickering, C.M. & Arthur, J.M. (2003) Patterns of resource allocation in the dioecious
alpine herb Aciphylla simplicifolia. Austral Ecology, 28, 566-574.

Pickering, C.M. & Hill, W. (2002) Reproductive ecology and the effect of altitude on sex
ratios in the dioecious herb Aciphylla simplicifolia (Apiaceae). Australian Journal
of Botany, 50, 289-300.

144
Powlesland, R.G., Merton, D.V., & Cockrem, J.F. (2006) A parrot apart: the natural history
of the kakapo (Strigops habroptilus), and the context of its conservation
management. Notornis, 53, 3-26.

Primack, R.B. (1983) Insect pollination in the New Zealand mountain flora. New Zealand
Journal of Botany, 21, 317-333.

Proctor, M., Yeo, P., & Lack, A. (1996) The natural history of pollination, London.

Pyke, G.H. (1991) What does it cost a plant to produce floral nectar? Nature, 350, 58 - 59

Rathcke, B. & Lacey, E.P. (1985) Phenological patterns of terrestrial plants. Annual Review
of Ecology and Systematics, 16, 179-214.

Rees, M., Kelly, D., & Bjornstad, O.N. (2002) Snow tussocks, chaos and the evolution of
mast seeding. American Naturalist, 160, 44-59.

Renner, S.S. & Ricklefs, R.E. (1995) Dioecy and its correlates in the flowering plants.
American Journal of Botany, 82, 596-606.

Roberts, A.N. & Boller, C.A. (1948) Pollination requirements of English Holly, Ilex
aquifolium. American Society of Horticultural Science, 52, 501-509.

Robertson, A.W. (1992) The relationship between floral display size, pollen carryover and
geitonogamy in Myosotis colensoi (Kirk) Macbride (Boraginaceae). Biological
Journal of the Linnean Society, 46, 333-349.

Robertson, A.W., Mountjoy, C., Faulkner, B.E., Roberts, M.V., & Macnair, M.R. (1999)
Bumble bee selection of Mimulus gattatus flowers: the effects of pollen quality and
reward depletion. Ecology, 80, 2594-2606.

Robertson, C. (1895) The philosophy of flower seasons, and the phenological relations of
the entomophilous flora and the anthophilous insect fauna. American Naturalist, 29,
97-117.

145
Sakai, A.K. & Weller, S.G. (1999). Gender and sexual dimorphsim in flowering plants: a
review of terminology, biogeographic patterns, ecological correlates, and
phylogenetic approaches. In Gender and Sexual Dimorphism in Flowering Plants
(eds. M.A. Geber, T.E. Dawson & L.F. Delph). Springer-Verlag, Berlin Heidelburg.

Salisbury, E.J. (1942) The Reproductive Capacity of Plants. Bell

Satake, A. & Iwasa, Y. (2002) The synchronised and intermittent reproduction of forest
trees is mediated by the Moran effect, only in association with pollen coupling.
Journal of Ecology, 90, 830-838.

Schauber, E.M., Kelly, D., Turchin, P., Simon, C., Lee, W.G., Allen, R.B., Payton, I.J.,
Wilson, P.R., Cowan, P.E., & Brockie, R.E. (2002) Masting by eighteen New
Zealand plant species: the role of temperature as a synchronizing cue. Ecology, 83,
1214-1225.

Schemske, D.W. (1980) Floral ecology and hummingbird pollination of Combretum

farinosum in Costa Rica. Biotropica, 12, 169-181.

Silvertown, J.W. (1980) The evolutionary ecology of mast seeding in trees. Biological
Journal of the Linnean Society, 14, 235-250.

Smith, C.C., Hamrick, J.L., & Kramer, C.L. (1990) The advantage of mast years for wind
pollination. The American Naturalist, 136, 154-166.

Sork, V.L. & Bramble, J.E. (1993) Prediction of acorn crops in three species of North
American oaks: Quercus alba, Q. rubra and Q. velutina. Annals of Scientific
Forestry, 50, 128-136.

Sork, V.L., Davis, F.W., Smouse, P.E., Apsit, V.J., Dyer, R.J., Fernandez-M, J.F., & Kuhn,
B. (2002) Pollen movement in declining populations of California Valley Oak
Quercus lobata: where have all the fathers gone? Molecular Ecology, 11, 1657-
1668.

146
Sutherland, S. (1987) Why hermaphroditic plants produce many more flowers than fruits:
experimental tests with Agave mckelveyana. Evolution, 41, 750-759.

Sutherland, S. & Delph, L.F. (1984) On the importance of male fitness in plants: patterns of
fruit-set. Ecology, 65, 1093-1104.

Thomson, J. (2003) When Is It Mutualism? American Naturalist, 162, S1-S9.

Tisch, P.A. (1996) Masting: A Test of the Pollination Efficiency and Predator Satiation
Hypotheses, MSc Thesis, University of Canterbury, Christchurch.

Vamosi, J.C. & Otto, S.P. (2002) When looks can kill: the evolution of sexually dimorphic
floral display and the extinction of dioecious plants. Proceedings of the Royal
Society of London, 269, 1187-1194.

Vamosi, J.C. & Vamosi, S.M. (2004) The role of diversification in causing the correlates of
dioecy. Evolution, 58, 723-731.

Vaughton, G. & Carthew, S.M. (1993) Evidence for selective abortion in Banksia spinulosa
(Proteaceae). Biological Journal of the Linnean Society, 50, 35-46.

Vaughton, G. & Ramsey, M. (1998) Floral display, pollinator visitation and reproductive
success in the dioecious perennial herb Wurmbea dioica (Liliaceae). Oecologia,
115, 93-101.

Voight, F.A., Jung, S., Farwig, N., & Bohning-Gaese, K. (2005) Low fruit set in a
dioecious tree: pollination ecology of Commiphora harveyi in South Africa. Journal
of Tropical Ecology, 21, 179-188.

Wagenius, S. (2005) Scale dependence of reproductive failure in fragmented Echinacea

populations. Ecology, 87, 931–941.

Waller, D.M. (1979) Models of mast fruiting in trees. Journal of Theoretical Biology, 80,
223-232.

147
Waller, D.M. (1993) How does mast fruiting get started? Trends in Ecology and Evolution,
8, 122-123.

Wardle, P. (2002) Vegetation of New Zealand The Blackburn Press, New Jersey.

Waser, N.M. (1978) Competition for hummingbird pollination and sequential flowering for
two Colorado wildflowers. Ecology, 59, 934-944.

Waser, N.M. (1986) Flower constancy: definition, cause and measurement. American
Naturalist, 127, 593-603.

Webb, C.J. (1979) Breeding systems and the evolution of dioecy in New Zealand Apioid
Umbelliferae. Evolution, 33, 662-672.

Webb, C.J. (1981) Andromonoecism, protandry and sexual selection in Umbelliferae. New
Zealand Journal of Botany, 19, 335-338.

Webb, C.J. (1986). Breeding systems and relationships in Gingidia and related Australasian
Apiaceae. In Flora and Fauna of Alpine Australasia: Ages and Origins (ed B.A.
Barlow), pp. 383-399. CSIRO, Melbourne.

Webb, C.J. & Kelly, D. (1993) The reproductive biology of the New Zealand flora. Trends
in Ecology and Evolution, 8, 442-447.

Webb, C.J. & Lloyd, D.G. (1980) Sex ratios in New Zealand apioid Umbelliferae. New
Zealand Journal of Botany, 18, 121-126.

Webb, C.J., Lloyd, D.G., & Delph, L.F. (1999) Gender dimorphism in indigenous New
Zealand plants. New Zealand Journal of Botany, 37, 119-130.

Webb, C.J. & Simpson, M.J.A. (2001) Seeds of New Zealand Gymnosperms and
Dicotyledons Manuka Press, Christchurch.

148
Wheelwright, N.T. & Logan, B.A. (2004) Previous-year reproduction reduces
photosynthetic capacity and slows lifetime growth in females of a neotropical tree.
Proceedings of the National Academy of Sciences, 101, 8051-8055.

Wilcock, C. & Neiland, R. (2002) Pollination failure in plants: why it happens and when it
matters. Trends in Plant Science, 7, 270-277.

Willson, M.F. & Price, P.W. (1977) The evolution of inflorescence size in Asclepias
(Asclepiadaceae). Evolution, 31, 495-511.

149
Appendix 1 - Floral morphology in Aciphylla scott-thomsonii

Aims: To assess whether a trend in flower number at certain locations on an inflorescence

was detectable among inflorescences of all sizes, and whether seed set rate was related to
this.

Methods: In early December prior to fruiting, while flowers were still present on A. scott-
thompsonii, eight inflorescences were selected from the Hakataramea Pass field site. Four
different sizes were chosen to represent the natural variation in inflorescence height present
in the population, from very tall to relatively short. Two inflorescences of each height class,
(approximately the same size), were selected from different plants. In the lab, all flowers on
every fifth lateral umbel, starting from the apex downwards along the inflorescence, were
counted. This was carried out in order to see whether a trend in flower number at certain
locations on an inflorescence was detectable among inflorescences of all sizes. Measured at
every fifth umbel was distance (cm) from the stem apex; diameter of the stem directly
beneath the umbel; number of umbellules (flower clusters); and number of flowers.
Regression analysis was used to test which variable was the best predictor of flower
number along the stem. The two variables consistently returning the best fit were distance
and diameter at every fifth umbel. This information could then be used to predict the
number of flowers on a compound umbel at any given position along any inflorescence of a
specified size. Later in the season when inflorescences were harvested, seed set was
measured on all 58 A. scott-thomsonii inflorescences by counting all seeds on every fifth
umbel along the stem to see if seed set rates varied with position on the inflorescence. Only
grand means were presented in Chapter 2 results, but examples of how seed set rate (and
flower number per umbel) varied with position along the stem are presented here. Results
are only presented for one inflorescence but provide a good example of what the trends
from most of the other selected inflorescences showed.

Results: Results show that umbels near the middle section of the flowering length of the
inflorescence had the most number of flowers per umbel and higher mean seed set rates
(Fig.1). Most consistently among all inflorescences sampled, stem diameter returned the
higher r2 value for flower number. The trends conformed to a negative polynomial
regression fit, with r2 values ranging from 0.79 to 0.96. Diameter along stem was also well

150
 correlated with seed set rate, but generally with lower r2 values. Distance along stem was
also a highly powerful predictor for both flower number and seed set rate, with all r2 values
ranging between 0.85 and 0.97 for flower number, and r2 values of around 0.2 to 0.6 for
seed set. Either one of these values would be useful for predicting flower number at any
given point along the inflorescence. See Figure 1 for representative inflorescences showing
the polynomial correlations between distance/diameter and flower number per umbel/seed
set rate.

(a) (b)

200
Flower no. per umbel

200
150 150
100 100
50 50
0 0
0 50 100 150 0 2 4
Distance along stem (cm) Diameter at every 5th umbel (cm)

(c) (d)

100 100
Percent seed set

90 90
80 80
70 70
60 60
50 50
0 50 100 150 0 1 2 3 4
Distance along stem (cm) Diameter at every 5th umbel (cm)

Fig.1 Polynomial regression analyses for A. scott-thomsonii at Hakataramea Pass, showing correlations
between flower number per umbel with distance and diameter along the stem for one inflorescence (a) and (b)
early in the flowering season, and between seed set rate and distance/diameter along the stem for one
inflorescence (c) and (d) post harvest. Graph (a) r2 = 0.96, y = - 0.03x2 + 4.01x + 14.99; graph (b) r2 = 0.90, y
= -31.83x2 + 157.71x – 40.28 ; graph (c) r2 = 0.59, y = -0.004x2 + 0.56x + 75.67; graph (d) r2 = 0.55, y = -
7.19x2 + 29.88 + 65.13.

151
Overall, strong relationships exist between the above variables and these can be useful for
estimating flower number or seed set depending on position along the inflorescence.
However, to predict flower number per umbel accurately, it must be taken into account that
this can vary depending on the size of the inflorescence. Fig. 2 demonstrates that
inflorescences of varying flowering lengths all exhibit the same strong negative polynomial
correlation, with inflorescences in the middle portion of the stem having a higher number
of flowers per umbel.

250

200
Flower no. per umbel

R2 = 0.9178

150 R2 = 0.8938

100 R2 = 0.8905
R = 0.8489
2

50

0
0 20 40 60 80 100 120 140 160
Distance along stem (cm)

Fig. 2 Polynomial regression analyses for A. scott-thomsonii showing correlations between flower number per
umbel with distance along stem for four A. scott-thomsonii inflorescences of varying sizes at Hakataramea
Pass. R2 values for each inflorescence are shown.

152
Appendix 2 Insects from the order: Diptera (flies) found on flowers of Aciphylla aurea, A. scott-thomsonii and A. ‘lewis’ at four sites from North to South Canterbury, South Island
of New Zealand (BP = Burkes Pass, HP = Hakataramea Pass, LP = Lewis Pass). Insects were caught during December 2004 and January 2005. Highlighted in bold are families that
are likely to be important as pollinators of Aciphylla.

Stage Location Known to

Family (common name) Genus/species found collected feed on Notes
Agromyzidae (leafminer flies) several spp. adult HP Family - leaf grazers, vary small (<3mm)
Bibionidae (march flies) several spp. adult BP, HP nectar
Emdpididae (danceflies) at least one spp. adult BP, HP nectar family - insect predators, long proboscis
Calliphoridae (blowflies) Calliphora quadrimaculata adult BP, HP pollen/nectar family - flower visitors
Muscidae (houseflies) 3 spp. adult BP, HP nectar family - frequent visitors to Apiaceous flowers
Phoridae (humpbacked flies) several spp. adult BP, HP family - flower visitors/breed on decaying matter
Simulidae (sandflies/midges) Austrosimulium sp. adult HP nectar
Stratiomyidae (soldierflies) Eulialia sp. puparium HP Eulalia adults frequent visitors to Apiaceae, good pollinators
Beris sp. adult HP nectar
Syrphidae (hoverflies) Syrphus sp. adult BP, HP pollen/nectar family - mostly smooth bodies, larvae predators of aphids/scale insects
Eristalis tenax adult BP, HP pollen/nectar Eristalis common, important pollinators of many flowers
larvae BP, HP
Helophilus sp. adult BP, HP pollen/nectar Helophilus hairy/large
Melangyna novaezelandiae adult BP, HP pollen/nectar
Melangyna sp. adult BP, HP pollen/nectar
Tachinidae (parasitoid flies) Montanarturia dimorpha adult HP nectar/pollen parasite of Lepidoptera caterpillars
Procissio cana group adult BP, HP nectar/pollen parasites of larval grass grub (Costelytra zelandica)
Protohystricia sp. adult BP, HP nectar/pollen short proboscis, common on open/flat flowers, abundant in grasslands
larvae BP, HP parasitises porina (hepialid) caterpillars
Pales sp. adult BP, HP nectar/pollen
Tephritidae (fruit flies) at least one spp. adult BP, HP nectar family unlikely to be significant pollinators
Therevidae (stilleto flies) Anabarhynchus sp. adult BP, HP nectar larva in branches and rotting logs
Tipulidae (craneflies) Discobola sp. adult LP
Holocusia sp. adult LP
Zelandotipula sp. adult LP nectar inhabits bogs
larvae LP root feeders (generally), larva subaquatic

153
Appendix 2 continued… Insects from the orders: Lepidoptera (flies) and Coleoptera (beetles) found on flowers of Aciphylla aurea, A. scott-thomsonii and A. ‘lewis’ at four sites
from North to South Canterbury, South Island of New Zealand (BP = Burkes Pass, HP = Hakataramea Pass, LP = Lewis Pass). Insects were caught during December 2004 and
January 2005.

Stage Location
Order/Family Genus/species found coll Food Notes
Lepidoptera:
Crambidae Orocrambus flexuosellus caterpillar BP/HP NZ's most ubiquitous moth
adults BP/HP nectar visit flowers, largely nocturnal but easily disturbed
Geometridae Dasyuris partheniata caterpillar BP/HP leaflets noctural, drop to the centre of plant where they are safe
Noctuidae Agrotis ipsilon (greasy cutworm) BP market gardens/lucerne paddocks Sth Canterbury
Graphania nullifera * caterpillar BP/HP leaves deep ovoid notching of leaf petioles
Coleoptera:
Anthicidae (ant-like flower beetles) at least one sp. adult BP pollen

Cerambycidae (flower longhorn

beetles) Zorion sp. guttigerum adult HP pollen Zorion most strikingly and colouful beetles in NZ fauna
"Blue longhorn"
Coccinellidae (ladybirds) Coccinella leonina adult BP/HP/LP aphids possible contribution to pollination
C. undecimpunctata aphids
Adalia bipunctata aphids
unindentified sp. aphids
Curculionidae Hadramphus tuberculatus adult BP leaves/pollen Specific to Aciphylla, only known site is BP
"Canterbury knobbled weevil"
Lyperobius huttoni adult HP leaves/pollen Specific to Aciphylla, larvae subterranean and feed on roots
"Hutton's speargrass weevil"
Helodidae unknown adult HP ?
Leiodidae (small carrion beetles) unknown adult BP ? probably feed on decaying matter, fungi, carrion
Melryidae (flower beetles) Dasytes sp. adult HP pollen mainly herbivorous, found on flowering shrubs
Oedemerdiae (lax beetles) Selenopalpus aciphyllae adult BP Live in vegetation, common on flowers or herbage
Scarabaedae (scarabaed beetles) Prodontria sp. adult BP/HP Root feeders
Pyronota festiva "manuka beetle" adult BP Folivores
Staphylinidae (rove beetles) unknown adult BP/HP/LP ? live inside stalks of Aciphylla
Tenebrionidae (darkling beetles) Mimopeus convexus adult BP ? only found at a few localised sites in Mckenzie Basin

154
Appendix 2 continued… Insects from the orders: Hemiptera, Hymenoptera and Plecoptera found on flowers of Aciphylla aurea, A. scott-thomsonii and A. ‘lewis’ at four sites from
North to South Canterbury, South Island of New Zealand (BP = Burkes Pass, HP = Hakataramea Pass, LP = Lewis Pass). Insects were caught during December 2004 and January
2005.

Stage Location
Order/Family Genus/species found coll Food Notes
Hemiptera:
Acanthosomatidae (shield bugs)
Oncacontius vittatus adult LP leaves Live in dead flowerstalks of Aciphylla

Porters
Aphididae (aphids) Euschizaphus (undescribed) Pass sap-suckers undescribed and rare
Cavariella aegopodii Sth Island sap-suckers at least five species of introduced aphids found on Aciphylla
Macrosiphum euphorbiae sap-suckers
Brachycaudus helichrysi sap-suckers
Myzus persicae sap-suckers
Hymenoptera:
Colletidae (solitary bees) Leioproctus fulvescens adult HP pollen important pollinator, many hairs, short-tongued, endemic solitary bee
Cicadidae (cicadas) Kikihia sp. adult LP
Plecoptera: (stoneflies)
Gripopterygidae Zelandoperla fenestrate adult LP pollen inhabit boggy areas
Zelandobius confuses adult LP pollen inhabit boggy areas

* Only identifiable taxa have been reported in the tables above. Many other insects were found but have not yet been identified. Aciphylla has many specific herbivores
including the common Lepidopteran Graphania nullifera, which were not found on the flowers of Aciphylla during this study (but are still reported above due to the
significant damage they have on the roots in particular.

NOTES: All insects collected during this study were given to Alison Evans (Department of Conservation, DoC) under the requirements of the collection permit.
Hadramphus tuberculatus was deposited in the Canterbury Museum and its conservation status is currently being revised to change it from “extinct” to “nationally
endangered” or “nationally critical” by DoC.

155