New Zealand Journal of Botany, 1999, Vol. 37 : 503–50 9
0028–825X/99/3703–0503
$7 .00 © The Royal Society of New Zealand 199 9
503
Seed production in Festuca novae-zelandiae: the effect of altitud e
and pre-dispersal predation
J . M . LOR D
Botany Department
University of Otag o
P .O . Box 5 6
Dunedin, New Zealan d
Keywords reproduction; seed predation ; altitude ;
Festuca novae-zelandiae ; Poaceae
D . KELLY
Department of Plant and Microbial Science s
University of Canterbury
Private Bag 480 0
Christchurch, New Zealand
The impact of pre-dispersal seed predators on th e
reproductive output of plants has been examined fo r
dozens of species across the majority of life-form s
and biomes (Crawley 1992) . It is clear from suc h
studies that seed predation prior to dispersal can hav e
a dramatic effect on the seed output of plants ;
predators attacking reproductive structures at variou s
stages commonly account for at least 50% of a see d
crop and often up to 80 or 90% (Janzen 1969, 1971 ;
Harper 1977 ; Stephenson 1981) . However, predatio n
levels can also vary enormously between relate d
sympatric species (Grieg 1993), and betwee n
individuals, habitats, and years within specie s
(Janzen 1969, 1971 ; Crawley 1992 ; Kelly et al .
1992) . This is because predation occurs within the
context of the host population's immediat e
environment and the accessibility and food potential
(i .e ., size and number of seeds) of individual hos t
plants . When a plant and its seed predators co-occur
over a range of environments there is no reason to
suppose that both should respond in exactly the sam e
way to changes in environmental conditions . In fact,
the host plant can successfully escape predation i n
space if it grows and reproduces in a wider range o f
sites than the predator . This can have interestin g
results in terms of the plant's seed production
throughout its range. Where the impact of predation
along an environmental gradient has been examined
(e .g ., Louda 1982a, 1982b, 1983 ; Randall 1986) ,
differences in response to environmental factor s
between host and predator tended to decouple seed
production of the host from the environmental
gradient examined . Randall (1986) demonstrate d
clearly that maximum seed production can b e
achieved at what appear to be less favourable site s
simply because seed predators have a more restricted
range than the host plant . Here we are concerned
with the impact of pre-dispersal predation on the
Abstract Festuca novae-zelandiae is a New Zealand endemic bunchgrass that occurs from near sea level to 1400 m . In a study of seed production at a
range of altitudes, flowering plants were found t o
produce between 0 and 967 seeds . On average ,
28 .3% of florets were attacked by invertebrate see d
predators and a further 46 .2% of ovules failed t o
produce a mature seed for reasons other than predation . The most common, identifiable, predators wer e
Dipterid larvae, including the flightless Diplotoxa
moorei. Seed set was most strongly related to the
number of florets initiated, but was also significantly
related to rates of ovule failure and pre-dispersal
predation . Seed set did not vary systematically wit h
altitude, but ovule failure rate showed a significan t
increase and predation rate a significant decreas e
with increasing altitude . Predation rate was als o
positively related to plant density ; altitude and plant
density together accounted for 56% of variation i n
predation rate . Path analysis showed that the nearzero direct correlation between seed set and altitude
was the product of an indirect positive effect, via a
strong negative effect on predation rate, being cancelled out by indirect negative effects via reduce d
floret initiation and increased ovule failure rates wit h
increasing altitude .
B99004
Received 27 January 1999 ; accepted 19 April 1999
INTRODUCTIO N
504
New Zealand Journal of Botany, 1999, Vol . 3 7
seed production of Festuca novae-zelandiae acros s
a range of environments . Particularly, we examine
the importance of predation relative to variation i n
seed production due to factors such as initia l
investment in ovules and failure of ovules to resul t
in mature seeds .
Festuca novae-zelandiae (Hack .) Cockayne is a
New Zealand endemic perennial bunchgrass whic h
recruits mainly via sexual reproduction, althoug h
some vegetative spread probably occurs as a result
of clonal fragmentation (Lord 1993) . Unlike the
mast-seeding Chionochloa bunchgrasses of alpine
grassland, mature, healthy F. novae-zelandia e
individuals flower and produce seed every yea r
(Lord 1998) . The lemma of the mature seed is aroun d
6 mm long . Seed mass is between 0 .8 and 1 .2 m g
and is positively correlated with altitude (Lord 1994) .
F. novae-zelandiae is ideal for a study o f
environmental effects on seed predation as it i s
widespread, occurring from near sea-level to 1400 m
a .s .l . in various habitats from sparsely vegetate d
alluvial surfaces and avalanche tracks to naturall y
tree-less subalpine shrublands and deforeste d
montane rangelands . Also, as all florets, includin g
those which fail to produce a seed, are retained o n
the plant until seed maturity, losses to predispersa l
seed predation are able to be accurately estimated .
METHODS
Seven sites were selected in montane grasslan d
around the University of Canterbury Researc h
Station at Cass, New Zealand (43°02 ' S, 171°45 ' E )
(site altitudes 640-1320 m a .s .l.), with an additional
low altitude site (65 m) in perennial grassland 80 k m
south-east on the Canterbury Plains (Table 1) .
Together these sites covered the bulk of the target
species' natural altitudinal range . In March 1990,
height and mean basal diameter (using a diameter
tape) were measured and all inflorescences collected
from at least 20 randomly selected reproductive
plants per site, for a total of 166 individuals over all
sites .
Of the 166 individuals measured in the field, 11 2
had produced at least 20 florets . Randomly selecte d
florets from these individuals were opened an d
examined under a stereo microscope . All florets
produced by each individual were examined up to a
maximum of 100 florets for any one individual . A
total of 8204 florets were examined . Florets were
scored as (1) healthy, meaning the seed was full sized, hard, and pale gold, (2) failed, includin g
undeveloped or partially developed ovules with no
signs of predation, as well as near full-sized seed s
which were brown and wizened, (3) infected by an
ergot-like fungus, or (4) showing signs of predation,
i .e ., either the predator was present or frass and/or
damage was evident . For plants in which only a
subset of florets was examined, seed crop size wa s
calculated as % seed set x number of florets .
Analysi s
Basal diameter, number of inflorescences, florets ,
and seeds per plant, and percent predation were logtransformed prior to analysis to improve normality.
All other variables approximated a norma l
distribution. Linear regression and ANOVA model s
were used initially to test for relationships among th e
measured variables . Path analysis, followin g
procedures in Sokal & Rohlf (1981, p. 642), was then
used to separate the direct and indirect effects o f
altitude, plant size (basal diameter) and density,
number of florets initiated, and failure and predatio n
rates on crop size . This method is ideally suited to
problems where predictor variables are inter-relate d
Table 1 Components of seed yield for Festuca novae-zelandiae at eight sites along an altitudinal gradient . Density
is plants m-2 . %Reprod is percent of population flowering . Inflorescences per plant (Inf/plt), florets per plant (F1r/pit) ,
and % seed set are mean values for each site.
Altitude
65
640
670
760
820
1060
1260
1320
Density
16 .0
3 .0
11 .4
22 .0
12 .6
7 .7
6 .6
7 .1
%Reprod
37.5
33 .3
37.1
29.1
31 .7
26.0
42 .4
28 .2
Inf/plt
9 .8
10 .5
6 .6
8 .4
5 .9
4 .8
10 .9
4 .1
F1r/pit
623 .2
470 .7
679 .6
575 .0
410 .8
89 .5
617 .2
171 .3
%Seed set
5 .5
33 .8
16.2
26.2
40 .4
9 .8
49.1
17.5
Seeds m2
195 . 8
168 .4
732 . 1
492 . 8
739 .3
105 .9
507 .9
64 .5
Lord & Kelly—Seed production in Festuca novae-zelandiae
50 5
A
and the exact path of causality is uncertain (Soka l
& Rohlf 1981 ; Pedhazur 1982) . Path analysis allow s
Plant size
Florets
the total correlation between variables to b e
t*
decomposed into direct and indirect effects, via th e
calculation of path coefficients, which are eithe r
Altitudes
%Failed – –*– – 'Seeds
standardised partial regression coefficient or, in th e
case of completely independent contributing factors ,
correlation coefficients (Sokal & Rohlf 1981) . Th e
researcher constructs a model, usually in the for m
Plant density%Predation
of a path diagram, of the direct and indirect effect s
of measured variables on each other and on a target
<0 . 1
variable, then calculates coefficients for each path
OA-0 3
in the diagram . The "goodness-of-fit" of the model
03-0 5
can then be tested against correlations in the raw
0 .5-0 .7
:
the
sum
of
the
products
of
all
paths
linking
tw
o
data
> 0 .7
B
variables estimates the correlation between the m
(Sokal & Rohlf 1981) .
Plant size
Floret s
To examine the relative contributions of ou r
t*
measured variables to seed production in Festuc a
novae-zelandiae, we constructed two path diagram s
%Failed
– – – -> Seed s
and tested their goodness-of-fit to the observe d Altitude
*
correlations in the dataset . In both models we
allowed the number of florets initiated per plant, and
rates of predation and ovule failure, to have direc t Plant density
%Predation
*
effects on the number of seeds per plant, wherea s
altitude, plant size, and plant density were allowe d
to only indirectly affect seed production .
Fig. 1 Path diagrams representing alternative hypothIn the first model (Fig . lA) we allowed site eses concerning the relationship between environmental
altitude to have an indirect effect on seed productio n and reproductive factors and seed production in Festuca
. Line thicknesses relate to the magnifirstly via its effect on the number of florets initiated . novae-zelandiae
tude of path coefficients . Positive coefficients are indiThis represents the simplest way in which see d cated by solid lines, negative coefficients by dashed lines .
production might be affected by an altitudina l Asterisks indicate values significantly different from 0
gradient, via changes in initial investment in flowers. (P < 0.05) . A, model A ; B, model B .
We also allowed altitude to affect predation rate, as
predators of F. novae-zelandiae are likely to vary in
abundance and behaviour with altitude, as has been models to correlations in the data was the n
found in previous studies of pre-dispersal predation compared. SAS version 6 .03 (SAS Institute 1988 )
(Randall 1986 ; Kelly et al . 1992) . Plant size wa s was used for all linear models and for calculatin g
allowed to affect seed production indirectly via a n correlations and standardised regression coefficients .
effect on the number of florets initiated . Plant density
was allowed to influence predation rates but not see d
production directly .
RESULT S
In the second model (Fig . 1B) we allowed altitude
to also affect plant size, plant density, and failur e Festuca novae-zelandiae density varied from 3 to 2 2
rates . A correlation between altitude and failure o f plants m-2 among sites . On average, only 33 .2% o f
ovules to result in mature seeds is biologicall y a population was reproductive (Table 1), but n o
reasonable as a decline in habitat favourability coul d reproductive individual was more than 1 .5 m from
conceivably affect the ability of an individual t o a potential pollen donor . Reproductive plant s
provision embryos, or the shortened growing seaso n initiated between 24 and 2994 florets (mean 491 .9
at higher altitudes might reduce the possibility of lat e ± 596 .1, Table 1) . Log number of florets per plan t
season ovules being pollinated . Plant density was was significantly related to log basal diameter (R 2 =
also allowed to affect failure rates as might be th e 0 .13, d .f. = 104, P < 0 .001) and differed significantly
case if seed set was pollen limited. The fit of these between sites (one-way ANOVA, F= 3 .73, d .f. = 7 ,
506
New Zealand Journal of Botany, 1999, Vol . 3 7
Fig. 2 Trends in A, mean percent of ovules that failed
to produce a seed for reasons other than predation ; B, mean
percent of florets damaged by predation ; C, mean number
of seeds produced per plant, by Festuca novae-zelandiae
individuals along an altitudinal gradient . Error bars are
standard deviations .
Chionochloa (Poaceae) seeds (White 1975 ; Kelly et
al . 1992 ; Kelly & Sullivan 1997) . Mature adults o f
Diplotoxa moorei (Diptera, Chloropidae), an
uncommon fly with nonfunctional wings (Spencer
1977), were also found in collection bags and in
pupae inside florets . A small number of florets
contained unidentified juvenile Hemipterids .
Between 44% and 99% of florets showing signs of
predation (visible internal or external damage, frass )
no longer contained the predator.
Total losses to pre-dispersal predation range d
from 45 .6% to 95 .8% of florets produced per plan t
at the lowest altitude site, and from 8 .82% to 44 .0%
at the highest altitude site . This decrease in predation
with increasing altitude was statistically significan t
(R 2 = 0 .50, d .f. = 109, P < 0 .001, % predation log transformed ; Fig . 2B). Log percent predation wa s
also significantly positively related to the density of
reproductive plants at each site (R 2 = 0 .42, d.f. = 109 ,
P < 0 .01) . Together, altitude and reproductive plant
density accounted for 56% of the variation in
predation levels among individuals (d .f. = 109, P <
0 .001) .
Percent seed set ranged from 0 to 79% of florets
initiated, with a mean across all sites of 23 .3% (Table
1), and increased significantly with increasin g
altitude (R 2 = 0 .10, d.f. = 119,P < 0.001) . Individuals
produced from 0 to 967 seeds . Log number of seeds
produced was strongly related to log number o f
florets initiated (R 2 = 0 .52, d.f. = 86, P < 0 .001), and
differed significantly between sites (one-wa y
ANOVA : F = 3 .73, d .f. = 7, P < 0 .01) but showe d
no trend with altitude (Fig . 2C) .
P < 0 .01), but within-site variation ("error" )
contributed 79% of total variation in floret
production . Log florets per plant showed a weak
decrease with increasing altitude (R2 = 0 .05, d .f. =
104, P < 0 .05), as did log inflorescences per plant ,
and florets per inflorescence (R2 = 0 .03, d .f = 165 ,
P < 0 .05 ; R 2 = 0.08, d . f. = 104,P < 0 .01, respectively,
Table 1) . The proportion of ovules failing to resul t
in mature seeds ranged from 3 .7% to 100% of floret s
produced, and increased significantly with altitud e
(R2 = 0 .42, d .f. = 111,P < 0 .001 ; Fig. 2A) . Log basal
diameter was not significantly related to altitude .
The most commonly found predator was a n
unidentified orange larva (Diptera, Cecidomyiidae)
1-1 .5 mm long, which we were unable to rear t o
maturity, but which appears similar to, but smalle r
than, an important, undescribed predator of
Path analyses
Although several variables related to the number o f
seeds produced per plant also showed a significant
relationship to site altitude, individual see d
production was not itself directly related to altitude .
However, path analysis allows for indirect effects to
be calculated . In our first path diagram (Fig . IA) w e
allowed altitude to have only indirect effects on seed
production via an effect on floret production an d
predation rate, leaving ovule failure rate as a n
independent predictor. The estimated correlation s
between seed production and the other variable s
derived from this model did not closely matc h
correlations calculated from the data (Table 2) .
In our second model we allowed altitude to als o
affect plant size, plant density, and ovule failur e
rates . Plant density was allowed to indirectly affect
seed production, also via an effect on ovule failur e
rate . Correlations estimated from this model com e
80 -
A
0
•
20 0
80 60 40 20 -
400 •
a.
300 •
0
200
400
600 800
1000 1200
140 0
Altitude (m )
Lord & Kelly—Seed production in Festuca novae-zelandiae
closer to correlations calculated from the data (Tabl e
2) . The number of florets produced had the stronges t
direct effect on seed production, confirming the
results of linear regression . Predation was more
important than ovule failure as a direct determinan t
of seed production. The strongest indirect effect o f
altitude on seed production was via its effect o n
predation rate ; altitude positively affected see d
production by depressing predation either directly,
or indirectly via a reduction in plant density . The
negative indirect effects of altitude via a reductio n
in florets produced, and an increase in ovule failur e
rates, were half the strength of the indirect effect via
predation rates, but together were sufficient to bring
about an overall near-zero correlation betwee n
altitude and seed production .
DISCUSSION
It is a basic principle of ecology that the abundanc e
or performance of a species can be limited by quit e
different factors in different parts of its range . For
example, in a study of seed production and predatio n
in Juncus squarrosus, Randall (1986) found that see d
density was limited by investment in ovules at hig h
altitudes but by predation at intermediate altitudes .
Similarly, Louda (1982a, 1982b, 1983) found that
recruitment of two species of Happlopappus was
limited by seed predators at coastal sites and b y
lower reproductive output and/or higher seedlin g
mortality at inland sites .
In this study, the most important measure d
predictor of seed production in F. novae-zelandia e
was, not surprisingly, floret production . In general ,
variation in flower number is primarily a product of
507
differences in plant size rather than environmentall y
induced changes in allocation patterns (Bazzaz &
Ackerly 1992) . This generalisation is borne out b y
the relative magnitudes of the path coefficients in ou r
model ; the effect of plant size on floret productio n
was twice that of altitude . However, the significant
effect of altitude on floret production independen t
of plant size suggests that F. novae-zelandiae i s
adopting different resource allocation strategies i n
different environments ; specifically, it is investing
less in reproduction with increasing altitude, whic h
is a common pattern among plants (Harper 1977 ;
Louda 1982a, 1982b, 1983 ; Randall 1986; Hara e t
al . 1988) .
Despite a reduction in floret production with altitude, no direct relationship was found between al titude and seed production at the sites studied .
However, both ovule failure and predation rate s
varied significantly with altitude . By simultaneousl y
examining the importance of these factors, while
controlling for plant size effects, we found evidenc e
that site altitude has strong indirect effects on see d
production . The overall near-zero correlation between altitude and seeds per plant appears to be th e
product of a strong positive indirect effect on see d
production, via a reduction in predation with altitude,
being cancelled out by a reduction in florets initiate d
and an increase in ovule failure rates with increasing altitude . The positive relationship between fail ure rate and altitude is arguably due to environmenta l
factors limiting the ability of individuals to provision ovules to maturity, rather than to problems o f
pollen availability and quality . Although failure o f
ovules to produce mature seeds is often a consequence of pollen limitation (Stephenson 1981), thi s
is unlikely to be a factor affecting seed productio n
Table 2 Pearsons correlation coefficients for attributes of both individual plants and study sites ,
and seed production (seeds/plant) for 166 Festuca novae-zelandiae plants, and estimated correlation s
from two path analysis models (Fig . IA, B) . Altitude is site altitude in metres a .s .l., Florets is lo g
number of florets per plant, Diameter is log basal diameter, %Failed is % of ovules that failed t o
produce a seed for reasons other than predation, %Damaged is log % florets damaged by predation ,
and Density is plants m-2 . *, P < 0 .05 ; ns, nonsignificant correlation .
Actual correlation with
seeds/plant
Altitude
Florets
Diameter
%Failed
%Damaged
Density
0 .002 ns
0 .719*
0 .361*
-0 .046 ns
-0 .351*
-0 .046 ns
Estimated correlation with seeds/plant
Model B : Fig . 1 B
Model A : Fig. IA
0 .063
0 .688
0 .285
0 .295
0 .475
-0 .357
0 .00 5
0 .73 1
0 .34 8
-0 .16 9
-0 .34 6
-0 .133
508
here ; F. novae-zelandiae is wind-pollinated and all
reproductive plants used for the study were withi n
1 .5 m of a potential pollen donor .
Path analysis indicated that predation was twic e
as important as ovule failure rates in determinin g
seed production in F. novae-zelandiae when all site s
were considered. However, the effect of altitude o n
both predation and ovule failure meant that th e
relative importance of these two factors to see d
production altered with increasing altitude . At the
lowest altitude site, predation was far more importan t
in limiting seed set than the failure of ovules to resul t
in mature seeds, whereas at higher altitudes ovule
failure was the more important factor (Fig . 2A, B) .
Of course, the weakness of our study is that w e
examined seed production in only one year, but it i s
unlikely that such general trends in relation t o
altitude would vary significantly between years . One
could imagine, however, that in warmer year s
predation may be more important in limiting seed set ,
regardless of altitude .
Although the seed predator responsible for mos t
of the damage to florets was not identified, the presence of orange Cecidomyiid larvae and Diplotoxa
adults points to strong parallels with the matrix perennials of tall-tussock grassland in the genu s
Chionochloa . Chionochloa seeds are larger than
those of F. novae-zelandiae and are attacked by orange Cecidomyiid larvae and larvae of D. similis,
which are larger than those of D. moorei (Kelly e t
al . 1992 ; Kelly & Sullivan 1997) . Both males an d
females of D. moorei have reduced, nonfunctional
wings (Spencer 1977), so the species is presumabl y
very limited in mobility . The most widely accepte d
explanation for flightlessness in insects relates t o
stability in abiotic and biotic (food, mates, etc .) environmental parameters (Wagner & Liebherr 1992) .
Mature, healthy F. novae-zelandiae individuals appear to flower most years and there is n o
syncronisation of "on" and "off' years withi n
populations (Lord 1998), so F. novae-zelandiae provides a reliable food supply for D. moorei . In contrast, Chionochloa species flower very irregularly
(Kelly et al . 1992), but D . similis has retained normal-sized wings so can potentially move away during non-flowering years (although it does not fly
readily, Kelly et al. 1992) . This study represents the
first record of a food plant for D. moorei, which was
previously thought to be a parasite or scavenger in
ant nests (Spencer 1977) . Subsequent to the completion of this study, D . moorei adults were also hatched
from florets of Poa cita, the other dominant matrix
perennial of short-tussock grassland (J . Lord unpub -
New Zealand Journal of Botany, 1999, Vol . 3 7
lished data) . Poa cita also appears to flower yearl y
and occupies a similar altitudinal range to F. novaezelandiae.
ACKNOWLEDGMENTS
Thanks to C . J . Burrows, M. I. Dunlop, D. A . Norton, and
J . R . Spence for helpful discussion and comments o n
earlier drafts, and to P . Johns of the Department of Zoology, University of Canterbury, for identifying insec t
specimens. The New Zealand Department of Conserva tion provided permission to sample on conservation land .
JL was financially supported by the Miss E. L . Hellaby
Indigenous Grassland Research Trust and by a Univer sity Grants Committee Postgraduate Scholarship .
REFERENCES
Bazzaz, F. A . ; Ackerly, D. D . 1992 : Reproductive allocation and reproductive effort in plants. In : Fenner,
M . ed. Seeds . The ecology of regeneration i n
plant communities . Wallingford, UK, CAB Inter national . Pp. 1-26 .
Crawley, M . J. 1992 : Seed predators and plant populatio n
dynamics . In : Fenner, M . ed . Seeds. The ecolog y
of regeneration in plant communities . Wallingford,
UK, CAB International . Pp . 157-191 .
Greig, N. 1993 : Predispersal seed predation on five Pipe r
species in tropical rainforest . Oecologia 93 : 412420 .
Hara, T. ; Kawano, S . ; Nagai, Y. 1988 : Optimal reproduc tive strategy of plants, with special reference t o
the modes of reproductive resource allocation .
Plant Species Biology 3 : 43-59.
Harper, J . L. 1977: The population biology of plants .
London, Academic Press .
Janzen, D . H . 1969 : Seed-eaters versus seed size, number ,
toxicity and dispersal . Evolution 23 : 1-27 .
Janzen, D . H . 1971 : Seed predation by animals . Annual
Review of Ecology and Systematics 2 : 465-492 .
Kelly, D.; Sullivan, J . J . 1997 : Quantifying the benefits o f
mast seeding on predator satiation and wind pol lination in Chionochloa pallens (Poaceae) . Oikos
78: 143-150 .
Kelly, D . ; McKone, M . J. ; Batchelor, K. J . ; Spence, J. R .
1992 : Mast seeding of Chionochloa (Poaceae )
and pre-dispersal seed predation by a specialist
fly (Diplotoxa, Diptera :Chloropidae) . New Zealand Journal of Botany 30: 125-133 .
Lord, J . M . 1993 : Does clonal fragmentation contribute
to recruitment in Festuca novae-zelandiae? Ne w
Zealand Journal of Botany 31 : 133-138 .
Lord & Kelly—Seed production in Festuca novae-zelandiae
Lord, J . M . 1994 : Variation in Festuca novae-zelandiae
(Hack .) Cockayne germination behaviour wit h
altitude of seed source . New Zealand Journal of
Botany 32 : 227-235 .
Lord, J . M . 1998 : Effect of flowering on vegetative growt h
and further reproduction in Festuca novaezelandiae. New Zealand Journal of Ecology 22 :
25-32 .
Louda, S . M . 1982a: Distribution ecology : variation i n
plant recruitment over a gradient in relation t o
insect predation. Ecological Monographs 52: 25 41 .
Louda, S . M . 1982b : Limitation of the recruitment of th e
shrub Happlopappus squarrosus (Asteraceae) by
flower and seed-feeding insects . Journal of Ecology 70 : 43-53 .
Louda, S . M . 1983 : Seed predation and seedling mortality in the recruitment of a shrub, Happlopappus
venetus (Asteraceae), along a climatic gradient .
Ecology 64 : 511-521 .
Pedhazur, E . J . 1982 : Multiple regression in behavioura l
research . New York, Holt, Rinehart and Winston .
50 9
Randall, M . G . M . 1986 : The predation of predispersed
Juncus squarrosus seeds by Colephora alticolella
(Lepidoptera) larvae over a range of altitudes i n
northern England . Oecologia 69 : 460-465 .
SAS Institute. 1988 : SAS/STAT User's Guide, Release
6 .03 . Cary, North Carolina, SAS Institute .
Sokal, R. R . ; Rohlf, F . J . 1981 : Biometry. 2nd ed. New
York, W . H . Freeman & Co .
Spencer, K . A. 1977 : A revision of the New Zealan d
Chloropidae . Journal of the Royal Society of New
Zealand 7 : 433-472 .
Stephenson, A . G. 1981 : Flower and fruit abortion : proxi mate causes and ultimate functions . Annual Review of Ecology and Systematics 12: 253-279 .
Wagner, D . L . ; Liebherr, J . K . 1992 : Flightlessness in
insects . Trends in Ecology and Evolution 7 : 216220 .
White, E. G . 1975 . An investigation and survey of insect
damage affecting Chionochloa seed production
in some alpine grasslands. New Zealand Journal
of Agricultural Research 18 : 163-178 .