Content uploaded by David L. Gorchov
Author content
All content in this area was uploaded by David L. Gorchov
Content may be subject to copyright.
Oecologia (2004) 139: 359–375
DOI 10.1007/s00442-004-1518-2
POPULATION ECOLOGY
Kara E. Miller
.
David L. Gorchov
The invasive shrub,
Lonicera maackii
, reduces growth
and fecundity of perennial forest herbs
Received: 15 August 2003 / Accepted: 22 January 2004 / Published online: 16 March 2004
# Springer-Verlag 2004
Abstract Effects of invasive plant species on native plant
species are frequently assumed or inferred from compar-
isons, but rarely quantified experimentally. Such quanti-
fication is important to assessing risks and impacts of
invasives. We quantified the effects of Lonicera maackii,
an exotic shrub invasive in many eastern North American
forests, on survival, growth, and reproduction of three
perennial herbs: Allium burdickii, Thalictrum thalic-
troides, and Viola pubescens. We predicted that the spring
ephemeral, A. burdickii, would be most impacted, due to
early leaf expansion of L. maackii. Field experiments were
carried out in two deciduous forest stands, one (Gregg’s
Woodlot, GW) disturbed and the other (Western Woods,
WW) relatively undisturbed. In each stand, individual
herbs were transplanted into a blocked design of 60 plots
where L. maackii was present, absent, or removed, and
monitored for 5 growing seasons. Lonicera maackii did
not affect survival of transplants, but reduced growth and
final size of individuals of all three species. For two of the
species, A. burdickii and V. pubescens, L. maackii reduced
the proportion of live plants flowering in both stands, and
reduced the seed or fruit number per flowering individual
in GW. For T. thalictroides the proportion flowering was
not affected, but seed number per flowering plant was
reduced by L. maackii in both stands. For all three species,
cumulative seed production over the course of the study
was reduced by L. maackii. Overall, effects on the spring
ephemeral, A. burdickii, were similar to effects on the
other herbs. Because mortality of these established
individuals was not affected, short-term studies might
conclude forest herbs are unaffected by invasive shrubs.
However, the growth and reproduction impacts documen-
ted here suggest that populations are impacted in the long-
term.
Keywords Allium burdickii
.
Competition experiment
.
Exotic plants
.
Thalictrum thalictroides
.
Viola pubescens
Introduction
Biological invasion by non-native species is considered
one of the major threats to the global environment
(Vitousek et al. 1996; Mack et al. 2000), and invasive
plants in particular (Cronk and Fuller 1995) are considered
to have major impacts on plant populations, communities,
and ecosystems. In the United States, exotic plant species
cost approximately $34 billion in damage and control each
year (Pimentel et al.44">2000).
However, assessing the impact of an invasive plant
species requires quantifying the per-capita effects of
individual plants (Parker et al. 1999), and this has rarely
been done (D’Antonio and Kark 2002). Effects of invasive
plants on native species are often inferred from compara-
tive studies; relatively few experiments (Witkowski 1991;
Midgley et al. 1992; Dillenburg et al. 1993; Huenneke and
Thomson 1995; McCarthy 1997; Meekins and McCarthy
1999; Gould and Gorchov 2000; Gorchov and Trisel
2003) have been done to quantify impacts. Experiments
involving multiple species are particularly desirable as
they will enable us to assess whether the impacts of an
invasive species on different native species are similar, as
would be predicted from neutral models (Hubbell 2001;
Davis 2003), or whether native species with niches similar
to the invasive are more strongly affected.
One of the many species of exotic shrubs invading
forests in the eastern United States is Lonicera maackii
(Rupr.) Herder, Amur Honeysuckle. This robust, multi-
stemmed deciduous shrub is native to northeast Asia and
was introduced to the USA in 1898. Subsequently it was
the subject of cultivar improvement and promoted for
landscaping and wildlife benefits by the USDA (Luken
K. E. Miller
.
D. L. Gorchov (*)
Department of Botany, Miami University,
Oxford, OH 45056, USA
e-mail: GorchoDL@muohio.edu
Fax: +1-513-5294243
Present address:
K. E. Miller
Applied Ecological Services Inc.,
120 W. Main St.,
West Dundee, IL 60118, USA
and Thieret 1995). Lonicera maackii has currently escaped
cultivation in 30 eastern and central states (Trisel and
Gorchov 1994; USDA, NRCS 2001).
The ability of L. maackii to dominate both disturbed
areas and forest understories has been attributed to plastic
branch architecture and biomass allocation patterns
(Luken 1988; Luken et al. 1995, 1997). Early expansion
(March) and late senescence (November) of leaves (Trisel
1997) give L. maackii a photosynthetic advantage over
native deciduous shrubs and trees. This extended leaf
display decreases the amount of light reaching the forest
floor, and in turn may reduce photosynthetic rates and
growth of smaller shrubs, herbs, saplings and seedlings.
Gould and Gorchov (2000) hypothesized that L. maackii
most negatively affects species most dependent on
irradiance before canopy leaf-out, a special case of the
niche overlap hypothesis. Support for this hypothesis
comes from experiments on three forest annuals; L.
maackii reduced survival to reproductive age only for
the two earlier species, although it reduced fecundity of
survivors for all three (Gould and Gorchov 2000).
Among forest stands in southwest Ohio forests,
abundance of L. maackii is negatively correlated with
density and species richness of tree seedlings and with
herb cover (Hutchinson and Vankat 1997). Within stands,
areas with L. maackii have lower density and species
richness of native shrubs (Medley 1997) and tree seedlings
(Collier et al. 2002), and lower cover and species richness
of herbs (Collier et al. 2002). These patterns suggest a
negative effect of L. maackii, but could alternatively be
due to greater recruitment of L. maackii in sites that are
depauperate for some reason, such as past disturbance.
However, removal of L. maackii has been shown to
increase survival of tree seedlings (Gorchov and Trisel
2003), in addition to the above cited effects on forest
annuals. Perennial herbs might be less responsive to
invasive shrubs because of their tolerance to low resource
conditions, limited seedling establishment, or slow vege-
tative reproduction (McCarthy 1997).
In order to test whether perennial herbs are suppressed
by invasive shrubs in deciduous forests, we investigated
the response of three herb species to removal of L. maackii
over a 5-year period. Specifically, we quantified how L.
maackii affects the survival, growth, and fecundity of
individuals of these three species. Furthermore, to
investigate whether impact was related to dependence on
irradiance before canopy leaf-out, we examined whether
the response of the herbs was related to leaf phenology,
specifically whether the spring ephemeral was more
negatively affected than the two full-season perennials.
Materials and methods
Study sites
Field experiments were carried out in two deciduous forest stands
near Oxford, Ohio (39°30′N, 84°45′ W). Western Woods (WW) is a
40 ha relatively undisturbed stand dominated by Quercus rubra,
Fraxinus spp., Acer saccharum, and Fagus grandifolia with a tree
basal area (BA) of 24.9 m
2
/ha (Gould and Gorchov 2000). Lonicera
maackii was the most common shrub, with a density of 0.3 shrubs/
m
2
(Gould and Gorchov19">2000). Gregg’s Woodlot (GW) is a 7 ha
stand dominated by Carya ovata, Fraxinus spp., C. laciniosa, and
Q. rubra. Compared to WW, BA was less (21.4 m
2
/ha) and L.
maackii density was greater (0.7 shrubs/m
2
; Gould and Gorchov
2000), perhaps due to greater anthropogenic disturbance. Selective
cutting, livestock grazing, and burning were practiced in the first
half of the twentieth century, but the forest has remained untouched
since 1970 (T. Gregg, personal communication).
Study species
We studied three perennial herbs differing in leaf phenology. Allium
burdickii (Hanes) A.G. Jones (Liliaceae), is a spring ephemeral,
found in rich, upland woods in the upper Midwest and Plains states
(Jones 1979). This taxon is considered A. tricoccum Solander var.
burdickii Hanes by McNeal and Jacobsen (2002). It produces two to
three ephemeral leaves that senesce before the scape and umbellate
inflorescence develop in early June. Genetic individuals (genets) can
reproduce vegetatively via bulb division to form genetically
identical daughter ramets.
Thalictrum thalictroides (L.) Eames and Boivin (Ranunculaceae),
also called Anemonella thalictroides (L.) Spach, is a full-season
perennial widely distributed across the eastern United States in dry
or moist woods (Gleason and Cronquist 1991). Because of its early
leaf expansion (March), Lubbers and Christensen (1986) considered
T. thalictroides a spring ephemeral, but we considered it a full-
season perennial because it senesces its leaves from July to
September, as much as 3 months after A. burdickii and other spring
ephemerals. Flowering occurs in April and May, and the single-
seeded achenes mature beginning in late May. T. thalictroides
produces only basal leaves the first growing season.
Viola pubescens Aiton var. pubescens. (Violaceae) is a full-
season, widespread perennial found in woods and meadows
(Gleason and Cronquist 1991, Ballard 1994). Leaf expansion begins
in late March or early April and the leaves last through summer
(unpublished data). Like most Viola species, V. pubescens produces
both chasmogamous (CH) flowers and cleistogamous (CL) flowers.
CH flowers are capable of both outcrossing and selfing and are
larger and showier than CL flowers, which can only self (Mattila
and Salonen 1995). In our population CH flowers opened in late
April while CL flowers developed from late May until the end of the
growing season. CH fruits mature in May and early June, whereas
CL fruits mature from June to late summer. The seeds develop
within capsules and have elaiosomes; they are likely ant-dispersed
given evidence that ants disperse the seeds of V. pensylvanica
(Culver and Beattie 1978), a species now included in V. pubescens
(Ballard 1994).
Experimental design
We used a blocked design for field experiments at each site. At WW
we used 20 blocks established by Gould and Gorchov (2000)in
November 1994; each block contained one plot where L. maackii
was present, one where L. maackii was absent, and one where L.
maackii was removed (n =60 plots). Within each block, plots were
placed 2–10 m apart to minimize differences in slope, drainage, and
canopy cover. The
L. maackii -absent plot was placed where there
was no evidence of live or dead L. maackii, and each of the two
other plots was placed where there were at least two L. maackii
individuals >1 m tall within a 1.5 m radius. One of the latter plots
was randomly assigned to the removal treatment; L. maackii shrubs
were removed by excavating the meristematic burl. Overhanging
branches and resprouting stems of L. maackii were trimmed
annually in the removal plots.
At GW we used 30 blocks, each with a L. maackii -present plot
and a L. maackii -removed plot 2–10 m away, originally selected by
360
Gould and Gorchov (2000) from a pool of 60 present and 60
removal plots established at the site in 1992. Each plot was initially
centered on a large L. maackii shrub and randomly assigned to
treatment. Removal of L. maackii began in 1992 when stems of the
central L. maackii were cut to 25 cm in height (Gorchov and Trisel
2002). Continued removal of resprouting stems from 1992 to 1995
led to the death of most of these central shrubs. After 1995 sprouts
from the few survivors were trimmed annually, and all other L.
maackii within 2 m were removed. No L. maackii -absent plots were
included due to the high density of L. maackii at GW.
Individuals of each herb species were collected from local
populations. The underground parts were separated from the stems
and leaves, rinsed, air dried, weighed, and numbered. For each herb
species, four individuals were randomly assigned for planting into
each of the 60 treatment plots at both sites (n =240 individuals/
species/site). Bottomless poultry-wire planting cages
(83 cm×60 cm×60 cm), which protected the herbs (but not L.
maackii) from mammalian herbivores, were placed to the north of
the center of each treatment plot under the shade of the central L.
maackii individual. Placement of individuals within the cage was
determined randomly using a 4×4 Latin Squares method. Individuals
were transplanted within 1–3 days of collection; A. burdickii in June
1995, T. thalictroides in May 1996, and V. pubescens in September
1997. There was no significant treatment difference in the initial
weight of transplants for any species at either site (one-way
ANOVA; all GW df =1; all WW df =2; all P >0.05).
The herbs were monitored weekly for survival, growth, and
fecundity from April to September each year beginning the spring
following transplanting. Other plants emerging within the planting
cages were removed by hand (generally only aboveground parts) at
each census to prevent aboveground competition from masking a
treatment effect. Growth was determined by annual changes in size,
measured as leaflet number for T. thalictroides, total leaf number
(basal leaves + stem leaves, which correlated strongly with number
of stems) for V. pubescens, and total leaf width, which was shown to
correlate strongly with ramet biomass in the closely related A.
tricoccum (Nault and Gagnon 1993), for A. burdickii.
Flower production for T. thalictroides and V. pubescens was
recorded as number of flowers. The small, numerous flowers of A.
burdickii could not be counted non-destructively, so only the
presence or absence of a scape (inflorescence) was noted.
Fruits of A. burdickii were collected once seed maturation began
in the population. Fruits were counted in the laboratory and opened
to obtain seeds. Seeds were dried to determine mean seed weight per
individual.
Thalictrum thalictroides
flower and seed (achene) counts were
made in the field. Seeds were not collected for weighing because
achenes from a single flower do not mature and dehisce
synchronously and it was difficult to determine when they were
mature.
Fruits of V. pubescens were counted in the field, but seed counts
and weights were difficult to obtain because seeds are released
following capsule dehiscence. To optimize seed collection, Coleman
lantern mantles were tied around the peduncles of fruits as they
matured. Bagging depended on fruit type and length. Few CL fruits
were bagged because they often have short peduncles (<10 mm).
Rather than bagging, CL fruits were monitored closely, and the
peduncles were pinched off when the capsules matured. At most,
50% of the fruits at a site were collected due to difficulties with
bagging and collecting fruits before they dehisced. Seeds of V.
pubescens were not collected in 1998 because few plants were
reproductive their first year. In 1999 and 2000, seeds were counted
and weighed from all collected fruits. The number of seeds per fruit
was only analyzed for fruits with complete seed counts. Only
mature, full seeds were counted and weighed. In 1999, the seeds
were oven-dried before weighing, but in 2000 seeds were only air-
dried because a subset was needed for another experiment.
As a measure of cumulative growth, we obtained biomass of each
surviving individual. Individuals were harvested, separated into
below- and above-ground parts, dried at 55°–65°C, and weighed.
Vegetative A. burdickii individuals were harvested 15–30 May 2001;
flowering individuals were harvested 1 week later. For A. burdickii
bulb mass but not root mass was included in total biomass because
roots were incompletely harvested. T. thalictroides was harvested
from 14 to 22 June 2001. Before drying, the number of tubers was
counted for each individual. V. pubescens was harvested from GW
26–29 April and from WW 10–17 May 2002.
Statistical analysis
Treatment effects on survivorship were analyzed using the
LIFETEST procedure in SAS version 8.01 (Fox 1993; Allison
1995). Because mortality of the herbs during their first year was
likely due to transplant shock rather than treatment, the initial
number of individuals alive was set as the number alive in the spring
after transplanting.
Analysis of variance (ANOVA) was used to detect treatment
effects on growth and reproduction (except proportion flowering)
separately for each species and year using the GLM (general linear
model) procedure in SAS. The same analysis was used for harvest
biomass of each species, and for average weight per tuber and
number of tubers per T. thalictroides plant. The blocked experi-
mental design required a mixed model approach in which treatment
effects were fixed and block and interaction effects were random.
We used the TEST option in the RANDOM statement to calculate
expected mean squares, the combinations of these that were
appropriate for the F tests, and the appropriate degrees of freedom
(SAS 1989). Most variables were log-transformed to improve
normality and ensure homogeneous variance. Post hoc comparisons
of means among present, absent, and removal treatments at WW
were done using Bonferroni multiple pairwise comparisons.
In any given year, analyses of growth included all live plants. For
A. burdickii, daughter ramets produced vegetatively were not
considered separate individuals. As in many studies involving
plant reproduction, analyses of fecundity variables were more
complicated due to hierarchical variables that are sequential in their
appearance (Mitchell 1993). Thus, only plants flowering in a given
year were analyzed for numbers of flowers, fruits, and seeds.
To detect treatment differences in the number of live individuals
flowering each year we constructed 2×2 contingency tables for GW
and 3×2 tables for WW, and used likelihood ratio chi-square (G
2
)
tests (SAS PROC FREQ). When the treatment effect was significant
at WW, we constructed separate 2×2 tables for each pair of
treatments and carried out separate G
2
tests on each. G
2
tests were
also used to determine whether treatment effected vegetative
reproduction in A. burdickii, by comparing the number of harvested
plants that had 1 versus >1 bulbs (ramets).
For V. pubescens, we tested for treatment effects on CH and CL
flowers and fruits separately, but only CL variables were normally
distributed. To determine the effect of treatment on CH flower and
fruit production, we used non-parametric generalized linear models
(PROC GENMOD, SAS version 8.01; see McKenzie and Halpern
1999; Berg and Redbo-Torstensson 1999; and Guisan and Theurillat
2000 for use in ecology). Specifically, we fitted a Poisson
distribution and used a non-linear logarithmic function to link the
response and predictor variables. To correct for overdispersion, the
d-scale option in PROC GENMOD was used to change the scale
parameter from 1 to the square root of the deviance divided by the
degrees of freedom. The models included treatment and block
predictor variables as in the ANOVA models used for the other
demographic variables. Pairwise comparisons for WW were built
into the models using contrast statements.
Before testing treatment effects on the number of seeds per V.
pubescens fruit we tested whether seed number differed between CH
and CL fruits. There were no significant differences at GW in 1999
(ANOVA; F =0.54) and 2000 (F =0.07), or at WW in 1999 (F
=0.44), but CH fruits produced more seeds per fruit than CL fruits at
WW in 2000 (F =4.73; P <.05) necessitating separate analyses. In
2000 only, average seed weight of V. pubescens was analyzed
separately for CH and CL fruits, because seed collection was
complete enough to determine that CH seeds were 1.4 times heavier
than CL seeds at both sites (two-way ANOVA; F =6.23 P <.05 at
361
GW; F =12.62; P <0.001 at WW). In analyses of both seed number
and weight, there were not enough blocks with values from all
treatments, so the block effect was removed from the ANOVA
model.
Cumulative fecundity per individual over the study was
quantified as the total number of seeds for A. burdickii and T.
thalictroides and total number of fruits (CH, CL, and total) for V.
pubescens. Cumulative fruit and seed counts could not be analyzed
by standard parametric tests because of large proportions of zeros, a
form of overdispersion. For this reason we used a zero inflated
Poisson (ZIP) model, which attributes some fraction (1−π) of the
observed zeros to a “zero-inflation” component, and the remainder
(π) to a Poisson distribution [Po(x,μ)] (Böhning et al. 1997; Ridout
et al. 1998). Thus, for each treatment our model [f(x; π , μ)=(1−π)
(Po(x,0)+π Po(x, μ)] fit two parameters: a zero inflation probability
(1−π) and a Poisson mean (μ). Block effects were not included in
the models. We tested whether π and μ differed between treatments
using SAS code (PROC NLMIXED) written by M. Hughes. For
WW, all pairwise comparisons were tested, with Bonferroni
significance adjustment. A generalized coefficient of determination
(R
2
) was calculated by comparing the log-likelihood of the fitted
model [L(β)] with that of the null model (with only an intercept, and
no treatment effect) [L(0)], following the formula of Nagelkerke
(1991): R
2=
{1−[L(0)/L(β)]
2/n
}/max (R
2
), where n = sample size
and max (R
2
)=1−L(0)
2/n
.
Results
Allium burdickii
Survival of A. burdickii did not differ significantly among
treatments at either study site (Fig. 1a, b).
Among survivors at GW, individuals in the removal
treatment were significantly larger than in the present
treatment in all study years except 1996 (Fig. 2a, Table 1).
By 2000, the sum of leaf widths on individuals in the
removal treatment was over 1.5 times that of individuals in
the present treatment. Final harvest biomass was sig-
nificantly greater for individuals in the removal treatments
(Fig. 3a, Table 1). In part this was due to greater vegetative
reproduction; 17% of survivors in the removal treatment
Fig. 1 The proportion of
individuals alive each year out
of the number alive the first
spring following transplantation
for each treatment for Allium
burdickii at a GW and b WW,
Thalictrum thalictroides at c
GW and d WW, and Viola
pubescens at e GW and f WW.
Sample sizes for each treatment
and statistics for survival analy-
sis (SAS PROC LIFETEST) are
provided on each graph; for GW
df =1, for WW df =2
362
were comprised of 2 or more ramets (bulbs), compared to
5% of those in the present treatment (G
2
=9.09, df =1, P
=0.0026). At WW individuals in the present treatment
were generally smallest, but a significant treatment
difference was detected only in 2000 (Fig. 2b, Table 1).
That year, leaves were 1.4 times wider on individuals in
the removal treatment than in the present treatment, but
neither of these treatments differed significantly from the
absent treatment. Final biomass was significantly greater
for individuals in the removal and absent treatments than
for those in the present treatment at WW (Fig. 3b,
Table 1). There was no significant difference among
treatments in the proportion of survivors comprised of >1
ramet (G
2
=3.08, df =2).
At GW, surviving individuals in the removal treatment
were more likely to flower than those in the present
treatment in 1997 (G
2
=9.63, P =0.0019), 1999 (G
2
=11.29, P =0.0008), and 2000 (G
2
=6.57, P =0.0100, all df
=1), but there was no treatment effect in 1996 (G
2
=0.01)
or 1998 (G
2
=0.20) (Fig. 4a). At WW, fewer individuals
flowered in 1999 where L. maackii was present than where
it was absent (G
2
=4.83, df =1, P =0.0280) or removed (G
2
=5.20, df =1, P =0.0226) (Fig. 4b). Treatments did not
differ in proportion flowering in other years (1996:
G
2
=0.19; 1997: G
2
=0.57; 1998: G
2
=2.12; 2000:
G
2
=1.26; all df =2).
Fruit production per flowering individual was signifi-
cantly higher where L. maackii was removed for every
year except the first (1996) at GW, but there was no
treatment effect on fruit production at WW (Table 1).
Flowering individuals at GW matured more seeds in the
removal treatment in each of the last four study years
(Fig. 5a, Table 1). At WW there was a treatment effect on
seed number only in 1998, but that year there were no
significant pairwise contrasts (Fig. 5b, Table 1). The
Fig. 2 Mean (+SD) size of
surviving individuals in each
treatment for each year at GW
(a, c, e) and WW (b, d, f). The
measure of size for Allium
burdickii (a, b) is the sum of leaf
widths, for Thalictrum thalic-
troides (c, d) it is the number of
leaflets, and for Viola pubescens
(e, f ) it is the number of leaves.
For each year×site, different
letters indicate significantly dif-
ferent (P <0.05) treatment
means based on analysis of
variance and, for WW, Bonfer-
roni multiple comparison test
(Tables 1, 2, 3)
363
average weight of seeds was significantly higher for
individuals in the removal treatment in 1996, 1999 and
2000 at GW, but at WW there were no significant
differences among treatments in any year (Table 1).
Over the 5-year experiment, individuals in the removal
treatment produced 1.6 times as many seeds as individuals
in the present treatment at GW (Fig. 6a); the distributions
of seed counts in each treatment differed little from a
Poisson distribution, and the Poisson mean was signifi-
cantly higher in the removal treatment (Table 5). At WW
cumulative seed production was 1.4× greater in the
removal treatment than in the present treatment, with the
absent treatment intermediate (Fig. 6b). These three
treatments had significantly different Poisson means, but
did not differ in zero-inflation probabilities (Table 5).
Thalictrum thalictroides
Survival of T. thalictroides did not differ among treatments
at either site, although there was a trend (P =0.0644) at
GW for higher survival where L. maackii was present
(Fig. 1c, d).
At GW, survivors in the removal treatment were larger
than those in the present treatment in the last two years for
which complete leaflet counts were obtained, averaging
nearly twice as many leaflets in 1999, and more than 2× in
2000 (Fig. 2c, Table 2). At harvest, individuals in the
removal treatment had significantly (over 3×) greater
biomass (Fig. 3c, Table 2), due both to significantly more
tubers and significantly greater average tuber weight
(Table 2). However, at WW, treatment only affected size in
1997, when individuals in the removal treatment had
significantly more leaflets than those in the present
treatment (Fig. 2d, Table 2). At harvest the overall
treatment effect was not significant (Fig. 3d, Table 2),
but the Bonferroni test indicated that removal treatment
plants had significantly greater biomass than those in the
present treatment. Average tuber weight was higher in the
removal treatment than in absent or present treatments, but
the average number of tubers was not significantly
affected (Table 2).
Fig. 3 Mean (+SD) dry bio-
mass at harvest for individuals
of Allium burdickii, Thalictrum
thalictroides, and Viola pubes-
cens in each treatment at GW (a,
c, e) and WW (b, d, f). Different
letters indicate significantly dif-
ferent (P <0.05) treatment
means based on analysis of
variance and, for WW, Bonfer-
roni multiple comparison test
(Tables 1, 2, 3)
364
Treatment did not affect the likelihood of flowering at
GW (1997: G
2
=0.04; 1998: G
2
=0.03; 1999: G
2
=0.06;
2000: G
2
=0.10; 2001: G
2
=0.00; all df =1) (Fig. 4c). At
WW individuals in the removal treatment were more likely
to flower in 2000 than those in the present (G
2
=4.50, df
=1, P =0.0340) or absent treatments (G
2
=6.01, df =1, P
=0.0142) (Fig. 4d). However, there were no treatment
effects on proportion flowering in other years (1997:
G
2
=1.51; 1998: G
2
=0.87; 1999: G
2
=0.12; 2001: G
2
=3.08;
all df =2).
At GW reproductive individuals in the removal treat-
ment averaged significantly more flowers in 1999, 2000,
and 2001 (Table 2), and more seeds in 1999 and 2000 (no
data for 2001) than those in the present treatment (Fig. 5c,
Table 2). Treatment effects on reproduction at WW were
confined to the last 2 years of the study. In 2000 both
flower and seed production were significantly higher in the
Table 1 Two-factor mixed
GLM analysis of variance re-
sults for size and reproduction
variables measured for Allium
burdickii individuals grown in
different treatments and blocks.
At GW there were 30 blocks
and the treatments were Loni-
cera maackii present and re-
moval; at WW there were 20
blocks, and a third treatment, L.
maackii absent, was included.
Italicized variables were log-
transformed
* P=<0.05
** P=<0.01
*** P=<0.001
§
P=<0.06
Site Year Dependent variable n Treatment Block Treatment×Block
df F df F df F
GW 1996 Total leaf width 236 1, 29.1 0.30 29, 29 0.74 29, 176 1.64*
Fruits 139 1, 35.8 3.91
§
29. 25.6 2.59** 27, 81 0.47
Seeds 139 1, 33.1 3.32 29, 26 2.20* 27, 81 0.68
Mean seed weight 131 1, 74 12.24*** 29, 74 1.63* 26, 74 0.65
1997 Total leaf width 233 1, 29.5 6.02* 29, 29 1.26 29, 173 1.50
§
Fruits 156 1, 34.2 22.95*** 29, 26.6 1.30 26, 99 0.94
Seeds 156 1, 32.7 13.87*** 29, 26.5 1.26 26, 99 1.15
Mean seed weight 151 1, 94 0.21 29, 94 1.91* 26, 94 1.00
1998 Total leaf width 231 1, 29.5 5.33* 29, 29 2.95** 29, 171 1.85**
Fruits 194 1, 32.1 10.30** 28, 28 1.61 28, 136 1.10
Seeds 194 1, 30.6 4.76* 28, 28 1.31 28, 136 1.71*
Mean seed weight 167 1, 31.6 0.55 28, 25.8 3.59*** 26, 111 1.22
1999 Total leaf width 220 1, 30.7 31.87*** 29, 29 1.96* 29, 160 1.53
§
Fruits 103 1, 22.2 6.62* 26, 19.4 1.35 20, 55 3.00***
Seeds 103 1, 22.7 10.42** 26, 19.2 1.58 20, 55 2.44**
Mean seed weight 98 1, 52.0 9.14** 26, 52.0 0.63 18, 52 1.03
2000 Total leaf width 198 1, 32.7 26.10*** 29, 29 2.83** 29, 138 1.31
Fruits 161 1, 40.4 28.37*** 29, 26.9 2.94** 26, 104 0.62
Seeds 161 1, 37.8 24.83*** 29, 26.7 2.57** 26, 104 0.74
Mean seed weight 154 1, 31.3 8.20** 29, 25.6 2.10* 25, 98 1.18
Harvest biomass 184 1, 29.0 45.43*** 29, 26.0 2.35* 26, 127 2.26**
WW 1996 Total leaf width 206 2, 41.2 0.67 19, 38.7 1.32 37, 147 0.96
Fruits 32 2, 8.1 0.57 13, 6.7 0.39 8, 8 0.86
Seeds 32 2, 8.1 0.77 13, 7.2 0.43 8, 8 1.40
Mean seed weight 32 2, 8.0 0.05 13, 7.9 0.52 8, 8 8.34**
1997 Total leaf width 167 2, 38.3 0.37 19, 37.2 1.55 32, 113 1.37
Fruits 86 2, 27.8 1.14 18, 26.5 1.48 24, 41 1.6
Seeds 86 2, 27.9 1.76 18, 26.6 1.51 24, 41 1.58
Mean seed weight 83 2, 38 0.73 18, 38 1.09 24, 38 0.73
1998 Total leaf width 159 2, 33.9 0.02 19, 32.67 2.14* 29, 108 1.86*
Fruits 106 2, 24.2 3.01 18, 23.2 0.62 22, 63 2.17**
Seeds 105 2, 24.2 3.54* 18, 23.1 1.1 22, 62 2.09*
Mean seed weight 101 2, 59 2.45 18, 59 2.07* 21, 59 1.18
1999 Total leaf width 156 2, 31.3 2.63 19, 30.7 2.13* 28, 106 2.24**
Fruits 55 2, 16.7 0.79 15, 13.0 2.13 15, 22 1.59
Seeds 55 2, 16.3 0.83 15, 13.5 2.58* 15, 22 2.06
Mean seed weight 55 2, 17.2 1.17 15, 12.4 1.17 15, 22 1.22
2000 Total leaf width 148 2, 30.9 5.40** 19, 29.9 3.16** 27, 99 1.79*
Fruits 114 2, 27.2 2.00 19, 28.8 0.85 24, 68 1.57
Seeds 116 2, 26.8 1.16 19, 27.4 2.32* 25, 69 3.32***
Mean seed weight 115 2, 69.0 2.38 19, 69.0 0.87 24, 69 1.51
Harvest biomass 137 2, 30.3 6.28** 18, 28.7 1.77 26, 90 1.67*
365
absent treatment than in the present and removal
treatments (Fig. 5d, Table 2). In 2001 there was a
marginal (P =0.0628) effect of treatment on flower
number; the only pairwise difference was greater flower
number in removed versus present treatment individuals.
Cumulative seed production over the 4-year period
differed significantly among treatments at both sites. At
GW individuals in the removal treatment averaged twice
the seed production as those in the present treatment
(Fig. 6c), with treatments differing significantly in Poisson
mean but not in zero-inflation probability (Table 5). At
WW plants in the absent and removal treatments averaged
1.6× as many seeds as those in the present treatment
(Fig. 6d). All pair-wise comparisons of treatment Poisson
means were significantly different, but treatments did not
differ in zero-inflation probability (Table 5).
Viola pubescens
Survival of V. pubescens was not significantly affected by
treatment at either site (Fig. 1e, f).
At GW surviving individuals in the removal treatment
were larger than those in the present treatment, averaging
significantly more leaves in every year but 2000 (Fig. 2e,
Table 3), when there was a significant treatment×block
interaction. At harvest, individuals in the removal treat-
ment averaged more than twice the biomass as those in the
present treatment (Fig. 3e, Table 3). At WW individuals in
the removal and absent treatments averaged significantly
more leaves than those in the present treatment in 1999
and 2002, and those in the removal treatment averaged
significantly more than those in the present treatment in
2001 (Fig. 2f, Table 3). Harvest biomass in the removal
and absent treatments was significantly greater than that in
the present treatment (Fig. 3f, Table 3).
Treatment did not affect the proportion of survivors at
GW that flowered in the first year of study (1998:
Fig. 4 The proportion of live
individuals that flowered in each
treatment for each year for Alli-
um burdickii at a GW and b
WW, Thalictrum thalictroides at
c GW and d WW, and Viola
pubescens at e GW and f WW.
Different letters indicate signif-
icant differences (P <0.05)
among treatments based on
likelihood ratio chi-square tests
(see text for details)
366
G
2
=0.46), but flowering was significantly higher in the
removal treatment in 1999 (G
2
=7.42, P =0.0065), 2000
(G
2
=8.44, P =0.0037), and 2001 (G
2
=8.93, P =0.0028),
and marginally higher in 2002 (G
2
=3.84, P =0.0501, all df
=1) (Fig. 4e). At WW, treatment did not affect flowering in
1998 (G
2
=1.99, df =2), but did in the 4 subsequent years
(Fig. 4f). In 1999 flowering was lower in the present
treatment than in the removal (G
2
=18.94, P <0.0001) and
absent treatments (G
2
=8.06, P =0.0045); the same pattern
occurred in 2002 (present vs removal: G
2
=4.30, P
=0.0381; present vs absent: G
2
=4.67, P =0.0307). In
2000 and 2001 significantly more plants flowered in the
removal treatment than in the present treatment (2000:
G
2
=7.00, P =0.0082; 2001: G
2
=9.45, P =0.0021).
The numbers of cleistogamous (CL) flowers and fruits
exceeded the numbers of chasmogamous (CH) flowers
and fruits each year at both sites (Miller 2001, unpublished
data). At GW, the numbers of CH flowers and fruits, CL
flowers and fruits, as well as the numbers of total flowers
and fruits per reproductive individual, were all signifi-
cantly greater for individuals in the removal treatment in
1999, 2000, and 2001 (Fig. 5e, Tables 3, 4). In 2002, only
CH flower counts were complete at the time of harvest;
these did not differ significantly between treatments
(Table 4). At WW there were no significant treatment
effects on any of these flower or fruit variables, except the
number of CH fruits was marginally (P =0.055) affected in
1999, when individuals in the absent treatment averaged
twice as many CH fruits as those in the present treatment,
with the removal treatment intermediate (Fig. 5f, Tables 3,
4).
At GW the number of seeds per fruit was significantly
higher in the removal treatment in 1999 (F =8.65, df =1,
105, P =0.004), but not in 2000 (F =0.23, df =1, 141).
Average seed weight did not differ between treatments
either year (1999: F =0.03, df =1, 44; 2000 CH fruits:
F=0.09, df =1, 15; 2000 CL fruits: F =0.00, df =1, 36). At
WW there was no treatment effect on seeds/fruit in 1999
(F =0.42, df =2, 95) or 2000 (CH fruits: F =0.40, df =2, 11;
CL fruits: F =0.92, df =2, 270). Treatments did not differ
Fig. 5 Mean (+SD) reproduc-
tive output per flowering indi-
vidual in each treatment for each
year at GW (a, c, e) and WW ( b,
d, f). Reproductive output was
quantified as number of seeds
for Allium burdickii (a, b) and
Thalictrum thalictroides (c, d)
and number of fruits for Viola
pubescens (e, f). Different letters
indicate significantly different
(P<0.05) treatment means based
on analysis of variance and, for
WW, Bonferroni multiple com-
parison test (Tables 1, 2, 3). The
asterisk indicates an overall
treatment effect, but no signifi-
cant pairwise differences
367
in average seed weight in 1999 (F =1.35, df =2, 45), or for
seeds in CH fruits in 2000 (F =3.05, df =2, 9). However,
there was a marginally significant treatment effect on
average seed weight of CL fruits in 2000 (F =3.09, df =2,
59, P =0.0531); CL fruits in the absent and removal
treatments had seeds that averaged 1.8 mg while those in
the present treatment averaged 1.4 mg.
Treatments affected cumulative CH and CL fruit
production similarly. At GW, individuals growing where
L. maackii was removed averaged more than 4× as many
CH fruits and more than 5× as many CL fruits as those
growing where L. maackii was present (Fig. 6e). The
removal treatment had a significantly lower zero-inflation
probability for CL fruits, and significantly higher Poisson
means for both CH and CL fruits (Table 5). At WW,
individuals in the removal treatment averaged nearly twice
as many CH and CL fruits as those in the present
treatment, with the absent treatment intermediate (Fig. 6f).
Compared to the present treatment, the removal treatment
had significantly lower zero-inflation probabilities for both
CH and CL fruit counts (Table 5). Furthermore, the
Poisson mean for CL fruits was significantly higher in the
removal treatment than in the present or absent treatments
(Table 5).
Discussion
We found that the invasive shrub, Lonicera maackii,
significantly reduced the growth and reproduction, but not
survival, of all three species of perennial herbs. Effects
appear to be cumulative, as treatments frequently did not
differ in the early years of the study, and became stronger
over time.
Survivorship
The lack of a negative effect of Lonicera maackii on the
survival of any of these three understory species (Fig. 1)
suggests that perennial herbs are somewhat less sensitive
to competition from this invasive shrub than are some
Fig. 6 Mean (+SD) cumulative
reproductive output of an indi-
vidual in each treatment at GW
(a, c, e) and WW (b, d, f) over
the course of the experiment for
Allium burdickii (1996–2000),
Thalictrum thalictroides (1997–
2000), and Viola pubescens
(1998–2001). For V. pubescens,
chasmogamous (CH) and cleis-
togamous (CL) fruits are distin-
guished, but standard deviations
are for total fruits. Significance
tests for treatment effects are
reported in Table 5. Significance
of treatment effects on V. pub-
escens are not indicated on this
graph, because separate tests
were carried out on CH and CL
fruits (see Table 5)
368
forest annuals (Gould and Gorchov 2000) and tree
seedlings (Gorchov and Trisel 2003). This might be
attributable to the stored reserves in underground per-
ennating organs. If the herbs had been exposed to
herbivory by mammals, a loss of leaf tissue would have
exacerbated the negative impacts of L. maackii, resulting
in higher mortality. Thus our exclusion of mammals may
have prevented the manifestation of a treatment effect on
survival.
Growth
As predicted, L. maackii negatively affected growth of all
three herb species in at least one, if not several, years over
the course of the experiment. Effects were generally
stronger, and manifest sooner, at the more disturbed site
(GW). Significant treatment effects usually emerged after
one or two years, then persisted and expanded in
subsequent years.
L. maackii treatment affected growth of A. burdickii
and V. pubescens more consistently than growth of T.
thalictroides (Fig. 2). This is not entirely consistent with
our hypothesis that L. maackii has a stronger effect on
growth of spring ephemerals than full-season perennials.
The similarity of responses of the first two species is
surprising given that A. burdickii is probably less tolerant
of shade, based on Rothstein and Zak’s(2001) finding that
the closely related and phenologically similar A. tricoccum
has a higher light compensation point than V. pubescens
and the semi-evergreen Tiarella cordifolia. Growth of T.
thalictroides differed more between stands than among
treatments (Fig. 2c, d). The very rapid growth of this herb
at GW, combined with high mortality at both sites (Fig. 1c,
d), suggests its life history strategy differs from those of
the other two herbs; it readily takes advantage of high
resource levels, such as those at GW (particularly where L.
maackii was removed), but has reduced ability to persist
and grow under the low resource conditions in the more
mature WW.
The lack of a significant treatment effect on total leaf
production of V. pubescens at both sites in 2000 was
unexpected, given the presence of a significant treatment
effect the previous year (Fig. 2e, f). At GW, individuals
were about the same size in 2000 as they were in 1999, but
L. maackii reduced the size of individuals in the present
Table 2 Two-factor mixed
GLM analysis of variance re-
sults for size and reproduction
variables measured for Thalic-
trum thalictroides individuals
grown in different treatments
and blocks (see Table 1 for
details). All dependent variables
were log-transformed, as indi-
cated by italics. Seed counts at
WW in 1997 were not analyzed
due to low sample sizes. Seed
counts in 2001 were not ana-
lyzed because plants were har-
vested before all seeds were
mature
* P=<0.05
** P=<0.01
*** P=<0.001
§
P=<0.07
Site Year Dependent variable n Treatment Block Treatment×Block
df F df F df F
GW 1997 Leaflets 147 1, 38.3 0.14 28, 22.5 3.05** 24, 93 0.46
Seeds 19 1, 2.0 0.93 12, 1.7 0.27 2, 3 2.07
1998 Leaflets 113 1, 29.3 1.72 27, 21.0 1.53 20, 64 0.67
Flowers 60 1, 11.9 3.80 23, 12.0 1.38 9, 26 0.78
Seeds 60 1, 10.1 0.12 23, 10.1 0.75 9, 26 2.04
1999 Leaflets 105 1, 21.9 6.67* 26, 17.5 1.01 18, 59 1.25
Flowers 87 1, 19.1 12.13** 26, 16.8 0.99 17, 42 2.44**
Seeds 87 1, 18.9 10.85** 26, 16.8 1.26 17, 42 2.76**
2000 Leaflets 98 1, 25.8 24.70*** 26, 16.1 1.89 17, 53 0.66
Flowers 87 1, 18.9 20.71*** 25, 14.1 0.70 15, 45 1.26
Seeds 87 1, 18.7 10.24** 25, 14.1 0.89 15, 45 1.32
2001 Flowers 85 1, 14.9 15.71** 26, 11.7 0.66 13, 44 2.03*
Harvest biomass 89 1, 15.4 18.71*** 27, 13.6 0.55 14, 46 2.84**
Tubers 86 1, 15.6 8.11* 27, 13.1 0.77 14, 43 2.07*
Mass per tuber 86 1, 43.0 12.56*** 27, 43 0.86 14, 43 1.04
WW 1997 Leaflets 144 2, 35.0 3.36* 19, 33.5 1.50 28, 94 0.97
1998 Leaflets 102 2, 31.6 0.03 17, 21.3 2.05
§
23, 59 0.68
Flowers 44 2, 13.9 1.68 16, 8.8 1.06 13, 12 0.97
Seeds 44 2, 14.2 1.97 16, 7.8 1.02 13, 12 0.78
1999 Leaflets 91 2, 28.2 1.02 17, 20.5 1.90 22, 49 1.31
Flowers 60 2, 19.4 2.24 17, 16.6 1.05 16, 24 0.94
Seeds 60 2, 18.1 1.93 17, 16.4 0.93 16, 24 1.53
2000 Leaflets 84 2, 28.5 1.77 17, 18.7 2.13
§
21, 43 1.05
Flowers 61 2, 19.9 5.01* 17, 9.1 1.20 14, 27 0.62
Seeds 61 2, 20.2 3.87* 17, 8.9 1.39 14, 27 0.59
2001 Flowers 63 2, 18.5 3.23
§
17, 16.0 1.30 16, 27 1.28
Harvest biomass 74 2, 22.9 2.20 17, 18.7 2.56* 19, 35 1.43
Tubers 73 2, 21.8 0.84 17, 18.7 1.52 19, 34 1.77
Mass per tuber 73 2, 34.0 7.91** 17, 34.0 1.67 19, 34 1.06
369
treatment only in some blocks (treatment×block interac-
tion). A significant interaction was also detected that year
at WW, but in addition, plants were smaller in 2000
compared to 1999. Despite their reduced size, individuals
produced more flowers and fruits, which suggests that in
2000, resources were allocated to reproduction rather than
growth.
Fecundity
As with growth, reproduction was affected by L. maackii
earlier and more strongly at GW, and effects on T.
thalictroides were less consistent than effects on the other
two species (Figs. 4, 5). L. maackii reduced both
proportion flowering and fruits/seeds per flowering
individual in V. pubescens and A. burdickii, but only
seeds per flowering individual in T. thalictroides. It is not
clear why the response of the first two species was similar,
given that they have contrasting leaf and reproductive
Table 3 Two-factor mixed
GLM analysis of variance re-
sults for size and reproduction
variables measured for Viola
pubescens individuals grown in
different treatments and blocks
(see Table 1 for details). All
variables were log-transformed
as indicated by italics. Flower
and fruit counts at GW in 1998
were not analyzed due to small
sample size
* P=<0.05
** P=<0.01
*** P=<0.001
§
P=<0.06
&
P=0.065 but removal treat-
ment significantly greater than
present treatment in Bonferroni
tests.
Site Year Dependent variable n Treatment Block Treatment×Block
df F df F df F
GW 1998 Leaves 161 1,33.7 7.97** 29, 28.3 1.04 28, 102 1.53
1999 Leaves 138 1, 32 15.49*** 29, 26.4 2.09* 26, 81 1.43
CL flowers 90 1, 24.8 25.99*** 27, 21.8 1.18 21, 40 1.28
Total flowers 90 1, 25.7 33.70*** 27, 22.0 1.41 21,40 1.02
CL fruits 90 1, 25.9 32.75*** 27, 22.0 1.42 21, 40 0.99
Total fruits 90 1, 27.0 40.88*** 27, 22.2 1.73 21, 40 0.82
2000 Leaves 100 1, 18.7 1.11 26, 16.3 1.11 16, 56 2.41**
CL flowers 53 1, 8.4 10.20* 19, 7.7 0.60 8, 24 3.57**
Total flowers 53 1, 8.5 11.42** 19, 7.6 0.85 8, 24 2.65*
CL fruits 53 1, 8.4 12.19** 19, 7.7 0.70 8, 24 3.44**
Total fruits 53 1, 8.6 13.83** 19, 7.7 1.00 8, 24 2.31
§
2001 Leaves 90 1, 24.4 25.01
§
27, 14.2 3.11* 15, 46 0.51
CL flowers 53 1, 9.6 17.90** 20, 8.9 1.38 9, 22 2.32
§
Total flowers 53 1, 9.9 25.71*** 20, 8.9 2.02 9, 22 1.54
CL fruits 53 1, 10.3 31.78*** 20, 8.9 1.48 9, 22 1.12
Total fruits 53 1, 10.3 29.52*** 20, 8.9 2.02 9, 22 1.05
2002 Leaves 86 1, 20.0 13.69** 27, 12.1 2.20 14, 43 0.76
Harvest biomass 86 1, 19.6 12.48** 27, 12.2 1.48 14, 43 0.82
WW 1998 Leaves 212 2, 39.5 0.11 19, 38.6 1.75 38, 152 1.58*
CL flowers 34 2, 9.1 1.47 16, 8.4 0.82 9, 6 2.40
Total flowers 34 2, 9.1 1.39 16, 8.5 0.74 9, 6 3.26
CL fruits 34 2, 9.1 1.34 16, 8.3 0.83 9, 6 2.09
Total fruits 34 2, 9.1 1.38 16, 8.4 0.66 9, 6 2.45
1999 Leaves 194 2, 41.9 9.32*** 19, 39.6 3.05** 38, 134 1.16
CL flowers 98 2, 33.9 1.16 19, 31.5 2.56** 29, 47 1.74*
Total flowers 98 2, 35.3 2.00 19, 32.2 2.69** 29, 47 1.37
CL fruits 98 2, 36.1 1.61 19, 32.6 2.22* 29, 47 1.21
Total fruits 98 2, 37.5 2.30 19, 33.2 2.25* 29, 47 1.02
2000 Leaves 175 2, 39.3 2.37 19, 38.1 2.18* 37, 116 2.37***
CL flowers 77 2, 28.7 0.88 17, 21.0 1.84 21, 36 0.86
Total flowers 77 2, 29.7 0.94 17, 21.1 1.87 21, 36 0.76
CL fruits 77 2, 26.2 0.08 17, 23.3 0.98 21, 36 1.16
Total fruits 77 2, 30.2 0.90 17, 21.1 1.78 21, 36 0.72
2001 Leaves 169 2, 39.2 2.93
&
19, 36.9 1.76 36, 111 2.02**
CL flowers 125 2, 38.3 1.71 19, 34.0 1.54 32, 71 1.75*
Total flowers 125 2, 37.7 1.65 19, 33.8 1.64 32, 71 1.95**
CL fruits 125 2, 37.0 1.72 19, 33.6 1.45 32, 71 2.19**
Total fruits 125 2, 35.9 1.60 19, 33.2 1.53 32, 71 2.82***
2002 Leaves 163 2, 39.5 4.28* 19, 36.2 1.08 35, 106 1.56*
Harvest biomass 160 2, 38.6 4.93* 19, 36.0 1.92* 35, 103 1.91**
370
phenologies; A. burdickii produces a single inflorescence
as leaves are senescing whereas V. pubescens continues to
produce CL flowers and fruits throughout the growing
season.
Phenology provides a possible explanation for why seed
mass in A. burdickii was more consistently reduced by L.
maackii than seed mass of V. pubescens (Tables 1, 3); for
A. burdickii seed mass is the only variable that can be
adjusted in response to resource levels late in the season,
whereas in V. pubescens the response may be manifest in
reduced growth, CL flower production, or seeds per
flower.
Because V. pubescens produces CH and CL flowers on
the same individual, each individual is subject to the costs
and benefits associated with producing both flower types.
Because CL flowers are less costly (Schemske 1978;
Waller 1979) their production is thought to be more
resistant to environmental variability (Schemske 1978);
thus one might predict less sensitive to competition from
an invasive. However, we found that both CH and CL
Table 4 Analysis of Deviance
tables from generalized linear
models on chasmogamous (CH)
flower and fruit numbers per
flowering V. pubescens indivi-
dual. These variables could not
be analyzed by ANOVA due to
non-normal distributions. P va-
lues of significant chi-square
tests are in bold . Samples sizes
were too low in 1998 for
statistical analyses
Site Year Dependent variable Source Deviance df Chi-Square Pr>Chisq
GW 1999 CH Flowers Intercept 51.07
Treatment 43.78 1 23.63 <.0001
Block 11.11 23 105.85 <.0001
CH Fruits Intercept 135.60
Treatment 113.04 1 22.32 <0.0001
Block 60.66 27 51.81 0.0028
2000 CH Flowers Intercept 76.06
Treatment 67.97 1 7.61 0.0058
Block 34.01 19 31.96 0.0316
CH Fruits Intercept 67.51
Treatment 62.26 1 4.74 0.0294
Block 35.35 19 24.36 0.1827
2001 CH Flowers Intercept 118.67
Treatment 99.73 1 12.43 0.0004
Block 47.24 20 34.44 0.0233
CH Fruits Intercept 92.93
Treatment 82.09 1 9.23 0.0024
Block 36.42 20 38.87 0.0069
2002 CH Flowers Intercept 78.38
Treatment 74.82 1 2.55 0.1103
Block 39.05 24 25.65 0.3713
WW 1999 CH Flowers Intercept 35.98
Treatment 35.39 2 1.13 0.5680
Block 24.66 19 20.45 0.3678
CH Fruits Intercept 136.69
Treatment 129.22 2 5.80 0.0550
Block 99.11 19 23.40 0.2204
2000 CH Flowers Intercept 3.52
Treatment 3.47 2 1.59 0.4520
Block 0.48 13 105.40 0.0001
CH Fruits Intercept 9.42
Treatment 9.34 2 0.24 0.8885
Block 5.99 13 9.52 0.7329
2001 CH Flowers Intercept 174.48
Treatment 174.01 2 0.36 0.8367
Block 137.76 19 27.11 0.1022
CH Fruits Intercept 169.87
Treatment 169.21 2 0.50 0.7802
Block 136.28 19 24.89 0.1643
2002 CH Flowers Intercept 83.40
Treatment 82.38 2 1.91 0.3851
Block 55.45 19 50.02 0.0001
371
reproduction were reduced by L. maackii (Table 4), similar
to Matilla and Salonen’s(1995) finding that both CL and
CH flower production in V. mirabilis were reduced in low
light. Each year, and in each treatment, CL flower and fruit
production exceeded CH reproduction, similar to patterns
reported for V. sororia (Solbrig et al. 1980) and Oxalis
acetosella (Berg and Redbo-Torstensson 1998). The
predictable seasonal shift from CH to CL flower
production we observed in V. pubescens suggests the
transition is controlled by daylength rather than resources,
as it is in other Viola spp. (Evans 1956; Mayers and Lord
1983).
L. maackii suppressed the long-term reproductive
success of all three perennial herb species at both sites
(Fig. 6). For A. burdickii and T. thalictroides, our ZIP
analysis finding that treatments did not differ in zero-
inflation probability but did differ in Poisson mean
(Table 5) indicates that, at least over a 5-year period, L.
maackii does not reduce the likelihood of reproduction,
but does reduce fecundity of reproductive individuals.
In contrast, for cumulative production of CL fruits in V.
pubescens, both parameters were significantly affected by
treatment, indicating that L. maackii both reduced the
likelihood of reproducing and reduced CL fruit production
of reproductive plants. Although treatment effects on CH
fruit production were similar in magnitude to those on CL
fruit production, only the zero-inflation probability was
significantly affected at WW, and only the Poisson mean
at GW. We believe the lack of statistical significance in
both parameters is more likely due to the consistently
lower levels of CH fruiting, rather than qualitative
differences in how L. maackii affects CH versus CL
reproduction.
Consequences for populations
For all three understory perennial herbs, L. maackii
reduced growth and fecundity, but not adult survivorship.
For herbaceous perennials in general, population growth
rates are most sensitive to rates of individual growth
(progression) and stasis (survival within same stage class)
and less sensitive to recruitment of seeds and seedlings,
clonal growth, and retrogression, based on elasticity values
of transition matrices (Silvertown et al.49">1993). There-
fore, the reductions in individual growth caused by L.
Table 5 Model-generated estimates of zero inflation probability
and Poisson mean of cumulative fruit and seed counts for each L.
maackii treatment for three perennial herb species. Significance tests
and coefficient of determination calculation are described in
Materials and methods
Site Species Response
variable
n Treatment Mean Zero-inflation probability Poisson mean Coefficient of
deter-mination
Parameter
estimate
F
a
Parameter
estimate
F
a
GW Allium Seeds 234 Present 35.68 0.04 1.19 37.26 576.75*** 0.92
Removal 58.64 0.02 59.67
Thalictrum Seeds 147 Present 73.63 0.25 3.38 97.67 1357.84*** 1.00
Removal 114.59 0.39 187.92
Viola CH fruits 161 Present 0.46 0.51 1.48 0.94 19.40*** 0.26
Removal 2.01 0.36 3.14
Viola CL fruits 161 Present 2.63 0.41 5.47* 4.48 321.39*** 0.93
Removal 13.50 0.24 17.7
WW Allium Seeds 205 Present 30.50 0.19 1.50 37.62a 138.91*** 0.76
Absent 36.88 0.31 53.64b
Removal 42.51 0.28 59.33c
Thalictrum Seeds 145 Present 22.65 0.48 1.30 43.42a 123.92*** 0.83
Absent 36.86 0.55 81.47c
Removal 37.67 0.39 61.34b
Viola CH fruits 212 Present 0.89 0.5b 3.05* 1.77 1.92 0.07
Absent 1.55 0.39ab 2.52
Removal 1.59 0.22a 2.05
Viola CL fruits 212 Present 4.64 0.43b 5.32** 8.14a 10.69*** 0.14
Absent 6.68 0.26ab 9.03a
Removal 8.93 0.18a 10.93b
a
F tests for GW have df=1, for WW df=2. For WW, treatments with different letters were significantly different based on pairwise
comparisons with Bonferroni adjustment
* P<0.05
** P<0.01
*** P<0.001
372
maackii that we documented for all three herb species are
likely to significantly reduce population growth rates, and
may be more important than the reductions in fecundity or
the lack of effects on survival. We can make more specific
inferences for A. burdickii because stage transition matri-
ces for A. tricoccum, which has a very similar life history,
have been parameterized (Nault and Gagnon 1993). Their
elasticity analysis indicates that growth and vegetative
reproduction of larger ramets had the greatest effect on
population growth rate. Since our study of A. burdickii
incorporated vegetative reproduction within growth, the
reductions in growth of established ramets caused by L.
maackii is likely to significantly reduce population growth
rate.
Mechanisms
This experiment did not test the mechanism(s) by which L.
maackii reduces growth and reproduction of native herbs,
but it does provide some insights. If allelopathy were
important, demographic rates of the herbs in the L.
maackii absent treatment at WW would be higher than in
the removal treatment, where allelopathic toxins could
have accumulated in the soil before the shrubs were
removed. However, we rarely detected higher demo-
graphic rates in the absent treatment (only T. thalictroides
flowers and seeds per flowering plant in 2000 and
cumulative seed production), suggesting allelopathy was
unimportant. In fact, demographic rates in the absent
treatment were frequently intermediate between those in
the present and removal treatments, and in some cases (%
of T. thalictroides fruiting in 2000, cumulative seed
production of Allium, and cumulative CL fruit production
of Viola) were significantly lower than in the removal
treatment. We hypothesize that lower rates in the absent
versus the removal treatment are attributable to greater
competition with established herbs in and near the plots, as
the removal plots probably initially had less herb biomass
due to suppression by L. maackii.
Competition is the most likely mechanism by which L.
maackii alters herb demography. Even if the community is
not resource limited upon introduction (Davis et al. 2000),
this shrub probably competes with native herbs for
resources shortly after it colonizes. Light is frequently
the limiting resource for understory herbs. The shady
conditions caused by a dense L. maackii understory begin
as early as March and continue through November,
significantly reducing light for herbs that depend on the
period before canopy leaf-out for much or most of their
annual carbon gain. Consistent with this light competition
hypothesis, herbs in the removal treatment consistently
had higher demographic rates at GW, the site with lower
tree basal area and a more open canopy.
In this more disturbed stand (GW) L. maackii had a
greater impact on these herbs, consistent with the hypoth-
esis that disturbance favors invasives by altering the
selection regime that would otherwise favor locally-
adapted native species (Byers 2002), for example by
changing the seasonal pattern of light availability. We do
not think the greater impact at GW was due to this site’s
higher density of L. maackii or larger plot size, because
impacts at both sites were presumably due primarily to the
large shrub in the plot center.
Microsite factors also influenced the effect of L. maackii
on these herbs. Significant block effects were prevalent in
the analyses, which suggest blocks differed in environ-
mental factors such as light penetration or soil moisture
that affected the herbs. The significant treatment×block
interactions for some of the demographic variables may
also have been due to differences in light or soil moisture.
In wetter or more shaded blocks, the herbs would have
benefited from L. maackii removal due to increased sun
exposure, whereas in gaps or drier blocks, herbs might
have been water-stressed by increased light. Although
there was some year-to-year variation in treatment effects,
suggesting some interaction with weather; in no year did
L. maackii have a significant positive effect on any
demographic parameter of any of the study species.
Conclusions
It is possible that the negative effects of L. maackii on the
growth and reproduction of these herbs is no stronger than
that which would be caused by a comparable biomass of
native shrubs. However, native shrubs are very sparse at
both study sites and in other stands in this region; e.g. in
an old-growth stand at nearby Hueston Woods State Park,
the dominant shrub, Lindera benzoin, has a density of only
0.0066 stems >1 cm dbh per m
2
(Foré et al. 1997), two
orders of magnitude less than L. maackii at GW and WW.
Thus, regardless of whether L. maackii has a greater per-
capita or per-gram competitive effect than native shrubs,
its high density means that its invasion subjects perennial
herbs to a level of shrub competition not experienced in
undisturbed forests.
The fact that impacts on the spring ephemeral, A.
tricoccum, were similar to impacts on the two full-season
herbs, fails to support our hypothesis that impacts correlate
with a critical overlap in phenology. However, full season
herbs such as V. pubescens fix most of their carbon before
canopy closure (Rothstein and Zak 2001), and thus might
be as sensitive to L. maackii shading as spring ephemerals.
Based on our original hypothesis, we predict that summer
herbs, those that grow primarily after canopy leaf-out, are
less impacted by early leafing invasive shrubs, than any of
the herbs included in this study.
Except for Gould and Gorchov (2000), this is one of the
first studies to show experimentally the direct effects of an
invasive plant on native herbs, and to assess the cumu-
lative effects over five years. Our findings that effects of L.
maackii on herb growth and reproduction were generally
not detectable until the second or later year of the study
highlight the need for multiple-year studies to assess the
effects of invasive species on native species.
Populations of native perennials are clearly at risk from
L. maackii and other invasive plants that expand early and
373
shade the forest floor. This risk could be overlooked by
short-term studies of herb cover or density if, as shown in
this study, the invader does not increase mortality of adult
herbs. However, reductions in growth and reproduction of
individual herbs, such as those caused by L. maackii, will
likely reduce native herb population sizes over time. Even
if native plants respond slowly, precautions should be
made for combating invasive plants before negative
impacts are realized.
Acknowledgements This research was supported by State of Ohio
Academic Challenge Grants administered by the Miami University
Department of Botany, the Miami University Summer Scholars
Program, and by W.J. and J.W. Hagedorn. We thank the Miami
University Natural Areas Committee and Thomas Gregg for access
to field sites. We are very grateful to Andrew Gould for establishing
plots and constructing exclosures, to Karen Doersam, Katie Dowell,
Joe Liszewski, and Andrew Ertley for the season of field work each
contributed, and to K. Summerville, M. Edelen, H. Richards, S.
Beiting, E. Notman, and R. Hrenko for additional field work. Mike
Hughes provided substantial guidance on statistical analyses and
wrote the SAS code for the ZIP analyses; John Bailer, Anne
Bartuszevige, and Bryan Endress also provided input on statistical
analyses. Major portions of this work were completed as part of
Kara Miller’s master’s degree at Miami University. We thank A.
Bartuszevige, B. Endress, M. Vincent, N. Smith-Huerta, and two
anonymous reviewers for helpful comments on earlier drafts.
References
Allison PD (1995) Survival analysis using the SAS system: a
practical guide. SAS Institute, Cary, N.C.
Ballard HE (1994) Violets of Michigan. Mich Bot 33:131–199
Berg H, Redbo-Torstensson P (1998) Cleistogamy as a bet-hedging
strategy in Oxalis acetosella, a perennial herb. J Ecol 86:491–
500
Berg H, Redbo-Torstensson P (1999) Offspring performance in three
cleistogamous Viola species. Plant Ecol 86:49–58
Böhning D, Dietz E, Schlattmann P (1997) Zero-inflated count
models and their applications in public health and social
science. In: Rost J, Langeheine R (eds) Applications of latent
trait and latent class models in the social sciences. Waxmann,
Münster, pp 333–344
Byers JE (2002) Impact of non-indigenous species on natives
enhanced by anthropogenic alteration of selection regimes.
Oikos 97:449–458
Collier MH, Vankat JL, Hughes MR (2002) Diminished plant
richness and abundance below Lonicera maackii, an invasive
shrub. Am Midl Nat 147:60–71
Cronk QCB, Fuller JL (1995) Plant invaders. Chapman and Hall,
London
Culver DC, Beattie AJ (1978) Myrmecochory in Viola: dynamics of
seed-ant interactions in some West Virginia species. J Ecol
66:53–72
D’Antonio CM, Kark S (2002) Impacts and extent of biotic
invasions in terrestrial ecosystems. Trends Ecol Evol 17:202–
204
Davis MA (2003) Biotic globalization: does competition from
introduced species threaten biodiversity? BioScience 53:481–
489
Davis MA, Grime JP, Thompson K (2000) Fluctuating resources in
plant communities: a general theory of invasibility. J Ecol
88:528–534
Dillenburg LR, Whigham DF, Teramura AH, Forseth IN (1993)
Effects of below-and above-ground competition from the vines
Lonicera japonica and Parthenocissus quinquefolia on the
growth of the tree host Liquidambar styraciflua. Oecol 93:48–
54
Evans LT (1956) Chasmogamic flowering in Viola palustris L.
Nature 178:1301
Foré SA, Vankat JL, Schaefer RL (1997) Temporal variation in the
woody understory of an old-growth Fagus-Acer forest and
implications for overstory recruitment. J Veg Sci 8:607–614
Fox GA (1993) Failure-time analysis: emergence, flowering,
survivorship, and other waiting times. In: Scheiner SM,
Gurevitch J (eds) Design and analysis of ecological experi-
ments. Chapman and Hall, New York, pp 253–289
Gleason HA, Cronquist AC (1991) Manual of vascular plants of
northeastern United States and adjacent Canada, 2nd edn. The
New York Botanical Garden, New York
Gorchov DL, Trisel DE (2003) Competitive effects of the invasive
shrub, Lonicera maackii (Rupr.) Herder (Caprifoliaceae), on the
growth and survival of native tree seedlings. Plant Ecol
166:13–24
Gould AMA, Gorchov DL (2000) Effects of the exotic invasive
shrub
Lonicera maackii, Amur Honeysuckle, on survival and
fecundity of native forest annual herbs. Am Midl Nat 144:36–
50
Guisan A, Theurillat JP (2000) Equilibrium modeling of alpine plant
distribution: how far can we go? Phytocoenologia 30:353–384
Hubbell SP (2001) The unified neutral theory of biodiversity and
biogeography. Princeton University Press, New Jersey
Huenneke LF, Thomson JK (1995) Potential interference between a
threatened endemic thistle and an invasive nonnative plant.
Conserv Biol 9:416–425
Hutchinson TF, Vankat JL (1997) Invasibility and effects of Amur
Honeysuckle in southwestern Ohio forests. Conserv Biol
11:1117–1124
Jones AG (1979) A study of wild leek, and the recognition of Allium
burdickii (Liliaceae). Syst Bot 4:29–43
Lubbers AE, Christensen NL (1986) Intraseasonal variation in seed
production among flowers and plants of Thalictrum thalic-
troides (Ranunculaceae). Am J Bot 73:190–203
Luken JO (1988) Population structure and biomass allocation of the
naturalized shrub Lonicera maackii (Rupr.) Maxim. in forest
and open habitats. Am Midl Nat 119:258–267
Luken JO, Thieret JW (1995) Amur Honeysuckle (Lonicera
maackii; Caprifoliaceae): its ascent, decline, and fall. Sida
16:479–503
Luken JO, Tholemeier TC, Kunkel BA, Kuddes LM (1995) Branch
architecture plasticity of Amur Honeysuckle (Lonicera maackii
(Rupr.) Herder): Initial response in extreme light environments.
Bull Torrey Bot Club 122:190–195
Luken JO, Kuddes LM, Tholemeier TC, Haller DM (1997)
Comparative responses of Lonicera maackii (Amur Honey-
suckle) and Lindera benzoin (Spicebush) to increased light. Am
Midl Nat 138:331–343
Mack RN, Simberloff D, Lonsdale WM, Evans H, Clout M, Bazzaz
FA (2000) Biotic invasions: causes, epidemiology, global
consequences, and control. Ecol Appl 10:689–710
Matilla T, Salonen V (1995) Reproduction of Viola mirabilis in
relation to light and nutrient availability. Can J Bot 73:1917–
1923
Mayers AM, Lord EM (1983) Comparative flower development in
the cleistogamous species Viola odorata I.: a growth rate study.
Am J Bot 70:1548–1555
McCarthy BC (1997) Response of a forest understory community to
experimental removal of an invasive nonindigenous plant
(Alliaria petiolata, Brassicaceae). In: Luken JO, Thieret JW
(eds) Assessment and management of plant invasions. Springer,
Berlin Heidelberg New York, pp 117–130
McKenzie D, Halpern CB (1999) Modeling the distributions of
shrub species in Pacific northwest forests. For Ecol Manage
114:293–307
374
McNeal DW, Jacobsen TD (2002) Allium. In: Flora of North
America Editorial Committee (eds) Flora of North America
north of Mexico, Magnoliophyta: Lilidae: Liliales and Orchi-
dales, vol 26. Oxford University Press, New York, pp 224–276
Medley KE (1997) Distribution of the non-native shrub Lonicera
maackii in Kramer Woods, Ohio. Phys Geogr 18:18–36
Meekins FF, McCarthy BC (1999) Competitive ability of Alliaria
petiolata (Garlic Mustard, Brassicaceae), an invasive, non-
indigenous forest herb. Int J Plant Sci 160:743–752
Midgley GF, Rutherford MC, Davis GW, Bosenberg JD (1992)
Photosynthetic response of heliophilous Rhus species to
environmental modification by invasive shrubs. Funct Ecol
6:334–345
Miller KE (1991) Effects of the invasive shrub, Lonicera maackii
(Amur honeysuckle), on survival, growth, and fecundity of
native understory perennial herbs in southwestern Ohio forests.
MS thesis, Miami University, Oxford, Ohio
Mitchell RJ (1993) Path analysis: pollination. In: Scheiner SM,
Gurevitch J (eds) Design and analysis of ecological experi-
ments. Chapman and Hall, New York, pp 211–231
Nagelkerke NJD (1991) A note on the general definition of
determination. Biometrika 78:691–692
Nault A, Gagnon D (1993) Ramet demography of Allium tricoccum,
a spring ephemeral, perennial forest herb. J Ecol 81:101–119
Parker IM, Simberloff D, Lonsdale WM, Goodell K, Wongam M,
Kareiva PM, Williamson MH, Von Holle B, Moyle PB, Byers
JE, Goldwasser L (1999) Impact: toward a framework for
understanding the ecological effects of invaders. Biol Inva-
sions1:3–19
Pimentel D, Lach L, Zuniga R, Morrison D (2000) Environmental
and economic costs of nonindigenous species in the United
States. Bioscience 50:53–65
Ridout MS, Demetrio CGB, Hinde JP (1998) Models for count data
with many zeros. In: Proceedings of the XIXth International
Biometrics Conference. Cape Town, South Africa, pp 179–192
Rothstein DE, Zak DR (2001) Photosynthetic adaptation and
acclimation to exploit seasonal periods of direct irradiance in
three temperate, deciduous-forest herbs. Funct Ecol 15:722–
731
SAS (1989) SAS/STAT user’s guide, version 6, 4th edn, vol 2. SAS
Institute, Cary, N.C.
Schemske DW (1978) Evolution of reproductive characteristics in
Impatiens (Balsaminaceae): the significance of cleistogamy and
chasmogamy. Ecology 59:596–613
Silvertown J, Franco M, Pisanty I, Mendoza A (1993) Comparative
plant demography—relative importance of life-cycle compo-
nents to the finite rate of increase in woody and herbaceous
perennials. J Ecol 81:465–476
Solbrig OT, Newell SJ, Kincaid DT (1980) The population biology
of the genus Viola: the demography of Viola sororia. J Ecol
68:521–546
Trisel DE (1997) The invasive shrub, Lonicera maackii (Rupr.)
Herder (Caprifoliaceae): factors contributing to its success and
its effect on native species. Ph.D. Dissertation, Miami
University, Oxford, Ohio
Trisel DE, Gorchov DL (1994) Regional distribution, ecological
impact, and leaf phenology of the invasive shrub Lonicera
maackii. Bull Ecol Soc Am 75(Supplement):231
USDA, NRCS (2001) The Plants Database, Version 3.1 (http://
plants.usda.gov). National Plant Data Center, Baton Rouge
Vitousek PM, D’Antonio CM, Loope LL, Westbrooks R (1996)
Biological invasions as global environmental change. Am Sci
84:468–478
Waller DM (1979) The relative costs of self- and cross-fertilized
seeds in Impatiens capensis (Balsaminaceae). Am J Bot
66:313–320
Witkowski ETF (1991) Growth and competition between seedlings
of Protea repens (L.) L. and the alien invasive, Acacia saligna
(Labill) Wendl. in relation to nutrient availability. Funct Ecol
5:101–110
375
