Available online at www.sciencedirect.

com

European Journal of Soil Biology 43 (2007) S86eS91

http://www.elsevier.com/locate/ejsobi

Original article
Selective vertical seed transport by earthworms: Implications
for the diversity of grassland ecosystems
Johann G. Zaller*, Nina Saxler
Institute of Organic Agriculture, University of Bonn, 53115 Bonn, Germany
Available online 21 September 2007

Abstract

Earthworms are suggested to play an important role for the plant diversity of grassland ecosystems. Here we tested whether (i)
Lumbricus terrestris L. selectively feeds on seeds of grassland species, (ii) feeding patterns match with seed species present in sur-
face casts in a permanent grassland, and (iii) grassland plant seeds deposited in different soil depths are transported by earthworms.
Food choice experiments with 10 plant species (3 grass spp., 4 non-leguminous herb spp., 3 leguminous spp.) showed that earth-
worms significantly selected between the offered seeds and generally preferred herbaceous species over grass species. Seed species
in grassland surface casts did not correlate with seed species preferred by earthworms in the feeding experiment. Mesocosm ex-
periments with L. terrestris in sterilized soil where 3500 seeds m2 of Dactylis glomerata (grass), Taraxacum officinale, Rumex
obtusifolius (non-leguminous herbs) and Trifolium repens (legume) were deposited in 2, 10, 20 and 30 cm depth showed a signif-
icantly different species-specific transport of the seeds both downward and upward the soil profile. Dependent on the initial depo-
sition depth between 26% (seeds initially at 10 cm depth) and 56% (20 cm depth) of the seeds were transported by earthworms.
After 90 days on average 585 germinable seeds m2 of D. glomerata, 94 seeds m2 of T. officinale, 38 seeds m2 of R. obtusifolius
but no seeds of T. repens were transported from deeper soil layers to the soil surface. Results suggest that seed herbivory and seed
transport are important mechanisms by which earthworms can selectively alter the diversity of grassland ecosystems.
Ó 2007 Elsevier Masson SAS. All rights reserved.

Keywords: Earthworm activity; Grassland diversity; Seed herbivory; Seed transport

1. Introduction in less diverse communities [1]. Plant species have

Evidence is increasing that earthworm activity and
grassland plant diversity are closely linked. Declining
plant species diversity has been shown to reduce the
size (biomass and number) of earthworm communities
[13,21] most likely due to declining fine root biomass
* Corresponding author. Present address: Institute of Zoology, Uni-
versity of Natural Resources and Applied Life Sciences Vienna, Gre-
gor Mendel Strasse 33, A-1180 Vienna, Austria. Fax: þ43 1 47654
3203.
E-mail address: johann.zaller@boku.ac.at (J.G. Zaller).
also been shown to differ in their degree of association
with nutrient-rich earthworm surface casts leading to
a higher growth of species more frequently associated
with casts (e.g., graminoid species) than those less fre-
quently associated with casts (e.g., herbaceous species;
19). These species-specific relationships between plants
and casts together with the effects of earthworms on soil
seedbank dynamics via seed transport [4,5,15,16], seed
burial [2,4,14e16] and seed herbivory [6,12] are sug-
gested to influence the structure and diversity of grass-
land ecosystems. While earthworm activity in grassland

1164-5563/$ - see front matter Ó 2007 Elsevier Masson SAS. All rights reserved.
doi:10.1016/j.ejsobi.2007.08.010
 J.G. Zaller, N. Saxler / European Journal of Soil Biology 43 (2007) S86eS91
is suggested to affect plant diversity it may not necessar-
ily affect the aboveground biomass production, indicat-
ing that apparent earthworm-induced increases in plant
nutrient availability might not be sufficient to promote
the growth of perennial plant species in native grass-
lands [20].
The aim of this study is (i) to quantify the degree of
selective ingestion of seeds from grassland species in
the laboratory, (ii) to assess whether feeding patterns
match with abundances of seed species in surface casts
collected in a grassland, and (iii) to quantify the ability
of earthworms to transport seeds placed in different soil
depths using mesocosms.
2. Materials and methods
2.1. Experiment to test seed herbivory
Twenty-five seeds of the following species were of-
fered to two adult similar-sized individuals of L. terrestris
(average fresh mass 4.4 g, 24 hour starvation period prior
to use) in plastic trays (16  12  5 cm): Alopecurus
pratensis (average seed size: 5.5  2.3 mm), Arrhenathe-
rum elatius (8.5  2.0 mm), Dactylis glomerata
(5.5  1.3 mm)dgrasses; Plantago lanceolata (2.8 
1.3 mm), Rumex obtusifolius (2.3  1.2 mm), Sangui-
sorba officinalis (2.9  1.4 mm), Taraxacum officinale
(3.8  1.0 mm)dnon-leguminous herbs; Trifolium pre-
tense (1.9  1.3 mm), T. repens (1.3  1.0 mm), Vicia
cracca (2.3  3.2 mm)dleguminous herbs. These spe-
cies are frequently abundant in permanent grasslands,
vary in their size and shape and aredwith the exception
of R. obtusifoliusdconsidered as species with a high
feeding value for cattle. Seeds of two plant species
were simultaneously placed in one feeding tray; addition-
ally species were recombined to avoid a possible bias by
testing similar species pairs (n ¼ 10). Trays were lined
with moist filter paper at the bottom and stored in growth
chambers in darkness at 10  C and at 60% relative
humidity. Number of seeds removed relative to number
of seeds offered was assessed after 24 h.
2.2. Seeds in earthworm surface casts from the field
In a permanent grassland of the organic research farm
of the University of Bonn, Germany (65 m a.s.l.; 7 170
E, 50 480 N), surface casts were collected on 10 marked
50  50 cm areas over a period of 5 weeks starting in
November 2003. Collected casts were spread in plastic
trays and the germination capacity monitored over six
months in a greenhouse. Some seedlings were addition-
ally cultivated in pots until identification was possible.
S87
2.3. Experiments to test vertical seed transport
by earthworms
To test the effect of earthworms on the vertical
movement of seeds in the soil, 24 opaque polyethylene
tubes (40 cm high, 15 cm diameter) were filled with
sterilized substrate (steamed for 24 h at 100  C) consist-
ing of field soil (Fluvisol), peat moss, sand and chopped
rye straw in the ratio 4:2:1:1, respectively. Soil in the
tubes was compacted to approximate field bulk densi-
ties (c 1.4 g cm3). All mesocosms were covered with
a fine mesh at both ends to prevent earthworms from
escaping. Soil moisture content in the mesocosms was
stabilized by keeping the cylinders in plastic trays filled
with water. During the filling of the soil into tubes, 62
seeds (equivalent to about 3500 seeds m2) of the grass
D. glomerata non-leguminous herbs R. obtusifolius and
T. officinale and the leguminous herb T. repens were de-
posited at four depths (2, 10, 20 and 30 cm). All four
species were simultaneously present in a mesocosm,
but placed at different depths to facilitate tracking of
the seeds after earthworm activity (n ¼ 6). All seed ma-
terial was obtained from a professional supplier that
guaranteed high germination rates (Rieger-Hofmann
GmbH, Blaufelden-Raboldshausen, Germany).
After filling, 4 adult individuals of L. terrestris (aver-
age fresh mass 4.6 g) were introduced, resembling
a density of about 200 earthworms m2, which is
slightly higher than the ambient density in permanent
grassland of this location. The mesocosms were ran-
domly placed on benches of an unheated greenhouse
under natural light conditions between December
2003 and March 2004. Surface casts were collected
daily from the top of the soil in the mesocosms and in-
vestigated for germinable seeds.
After 90 days, the mesocosms were destructively har-
vested by carefully pressing the soil column out of the
tube and cutting the soil in 4e8 cm thick slices. Slices
had different thickness because we wanted accurately
sample the layers where seeds were initially deposited.
Seed numbers were standardized to account for different
thickness. The soil slices were spread in thin layers on
plastic trays placed in the greenhouse in order to facili-
tate germination of seeds present in these slices.
2.4. Statistical analyses
We used a GLM-ANOVA model in SAS (vers. 8.02
for Windows, SAS Inc., Cary, NC, USA) to analyse
the data. Therefore, for each depth of seed placement
a two-way ANOVA with the factors Species (spp.) and
Location depth (LD) and their interactions was used.
S88 J.G. Zaller, N. Saxler / European Journal of Soil Biology 43 (2007) S86eS91

Seed herbivory was compared using a t-test followed by
least-significance difference mean comparisons after
TukeyeKramer.
3. Results
Earthworms selectively fed on offered grassland
seeds (Fig. 1). With a consumption of 40.5% of the of-
fered seeds, D. glomerata was the most preferred grass
species. Of the non-leguminous herbs, 59% of the seeds
of T. officinale, 51% of S. officinalis but only 18.5% of
R. obtusifolius and 13% of P. lanceolata seeds were
eaten (Fig. 1). Seeds of T. repens were the most pre-
ferred leguminous herbs (55% seeds eaten), the least
preferred seeds were those of V. cracca (2% seeds
eaten; Fig. 1). During 5 weeks 503 germinable seeds
of 32 species could be found in surface casts in a perma-
nent grassland (data not shown).
Relative number of seeds eaten (%)
0 20 40 60
Alopecurus d
P < 0.001
Arrhenatherum d 4. Discussion
Averaged across species, more than 90% of the in-
serted seeds could be recovered at the end of the meso-
cosm experiment. The number of seeds transported by
earthworms varied significantly between species and
was dependent on how deep the seeds were originally
deposited (Fig. 2aed). When seeds were placed in
2 cm depth, 73% of the seeds could be found at this
depth at the end of the experiment while the remaining
seeds of the four species were similarly transported
across the soil profile; however, seeds were only trans-
ported downward and no seeds could be found on the
surface (Fig. 2a). When seeds were originally placed
in 10 cm depth, 74% of the seeds remained in this layer
while the other seeds were transported by earthworms
in species-specific numbers in different depths upwards
and downwards (Fig. 2b). Of seeds originally deposited
in 20 cm depth, 44% of the seeds could be found at this
depth while the remaining seeds were transported spe-
cifically for each species but equally across depths
(Fig. 2c). When seeds were originally placed in 30 cm
80 100
depth, 53% could still be found in this layer while sig-
nificantly different numbers were transported between
species, also different numbers of seeds could be found
in different depths (Fig. 2d).

Generally, L. terrestris preferred seeds of non-

Dactylis ab leguminous (T. officinale, S. officinale) and leguminous
species (T. repens) over seeds of grasses. Feeding pref-
Plantago cd erences did not show any clear relationships with seed
size indicating that other factors than size (e.g. taste,
nutritional value and chemical parameters [3]) play an
Rumex c important role. Our finding that V. cracca with globe-
shaped seeds of about 3 mm in size was hardly
consumed indicates that seeds were too large for the
Sanguisorba a
earthworms to swallow (see also [8]) and is in line
with reports that L. terrestris is not able to feed on
Taraxacum a particles >2 mm in size [12]. Generally, our result
underlines the suggestion that particles are eaten by
earthworms only until a certain dimension while other
Trifolium prat. bc characteristics like seed size, shape, texture, and taste
are important in determining palatability of smaller par-
ticles [4,10,11]. Feeding rates for T. officinale are simi-
Trifolium rep. a
lar [11] but those for the legumes T. pratensis (30%) and
T. repens (55%) are much higher than those reported in
Vicia d other studies [8,9]. This discrepancy might be explained
by (1) the use of differently sized earthworms with dif-
ferent preferences in these studies or (2) by using differ-
Fig. 1. Seed consumption by L. terrestris in laboratory field trials.
See text for full names of plant species. P-values derived from t-tests, ent approaches to determine food choice by either
different letters indicate significant differences between species at offering one or more species simultaneously. It has
P < 0.05 (Tukey LSD test). Means  SE, n ¼ 10. also to be noted, that selective feeding by earthworms
 Number of germinable seeds m-2
0 1000 2000 0 1000 2000 0 1000 2000 0 1000 2000

(a) (b) (c) (d)

Surface

J.G. Zaller, N. Saxler / European Journal of Soil Biology 43 (2007) S86eS91

Seed placement

0-8

Seed placement

8 - 12

Spp: P = 0.005
LD: P = 0.002
Soil layer (cm)

12 - 18 Spp * LD: P = 0.406

Dactylis Seed placement

18 - 22 Rumex
Taraxacum
Trifolium
22 - 28

Seed placement
28 - 32 Spp: P = 0.178 Spp: P < 0.001 Spp: P = 0.005
LD: P = 0.253 LD: P = 0.021 LD: P = 0.154
Spp * LD: P = 0.915 Spp * LD: P = 0.023 Spp * LD: P = 0.824

32 - 40

Fig. 2. Seed distribution of D. glomerata, R. obtusifolius, T. officinalis, and T. repens in mesocosms after 90 days of earthworm activity. Before earthworms were added to mesocosms, 3500
seeds m2 of each species were placed at 2, 10, 20 and 30 cm depth. Bars only indicate seeds that were transported by earthworms, seeds that remained at their initial deposition depth are not
shown. P-values from two-way ANOVAs with the factors species (Spp), location depth (LD) and their interaction (Spp * LD). Mean  SE, n ¼ 6.

S89
S90 J.G. Zaller, N. Saxler / European Journal of Soil Biology 43 (2007) S86eS91
is probably much more pronounced in laboratory feed-
ing experiments as compared to the field, where seeds
are mixed with organic and inorganic soil particles
[11]. Overall, it remains unclear how important seed
consumption is for the nutrition of earthworms.
In the current field study, over a period of 5 weeks up
to 500 germinable seeds per square metre comprising
a total of 32 species could be found in surface casts col-
lected in permanent grassland. Assuming that earth-
worms are active for at least six months, up to 3000
seeds m2 could therefore be transported to the soil sur-
face annually. This demonstrates the importance of
casts as regeneration niche for plant communities
[8,14,19]. It has been reported that 70% of seedlings
in temperate grasslands germinate out of earthworm
casts, although the casts covered only 24e28% of the
soil surface [4]. Surprisingly, many plant species pres-
ent in casts were not present in the vegetation indicating
that also in the field earthworms transport seeds from
deeper soil layersdwhere the seedbank of a different
vegetation might have been presentdto the surface.
Overall, in permanent grassland beside earthworms
several bioturbators (e.g. gophers, moles, voles; [7])
and abiotic influences (e.g. frost) could potentially
transport seeds within the soil profile. Earthworm gut
transit have been shown to delay or stimulate seed
germination [6,11,12]; however, the fact that we could
recover more the 90% of the buried seeds by the germi-
nation method indicates that germination capacity of
the tested species was only little affected by earthworm
feeding. In this experiment earthworms transported be-
tween 26% (when seeds were deposited in 10 cm depth)
and 56% (when seeds were deposited in 20 cm depth) of
the inserted seeds upwards and downwards the soil pro-
file indicating that the earthworms were mainly active at
20 cm depth. Seeds of the grass D. glomerata were most
often, while seeds of the legume T. repens least often
transported suggesting that the seed bank dynamic of
the grass species is most influenced and that of the
legume rather unaffected by earthworms. Dactylis was
also the most preferred species in the feeding experi-
ment. In contrast, T. repens that had the smallest seeds
among the tested species was often preferred in the
feeding experiment but not often transported in the mes-
ocosms. An explanation for this could be that feeding
preferences especially for species with small seeds dif-
fer considerably between laboratory and more realistic
mesocosm experiments where various alternative food
sources were available.
For the species tested here, germination and estab-
lishment in the community is only possible if seeds
are transported from deeper zones to the surface since
all require light for their germination and would not
be able to germinate from deeper soil layers. Interest-
ingly, no seeds were transported to the soil surface
when they were originally deposited in 2 cm depth, in-
dicating that earthworms were mainly active in deeper
soil layers. Seeds that were most often transported to
the soil surface were those of D. glomerata (585, 226
and 75 seeds m2 when deposited in 10, 20 and
30 cm depth, respectively), T. officinale (94 m2 when
deposited in 20 cm depth) and R. obtusifolius (38 m2
when deposited in 30 cm depth). Especially the latter
species is a very aggressive weed in grasslands through-
out Europe [17], and it remains to be tested what role
earthworms play in the establishment of this species
and how they interact with cultural control strategies
against this species [18]. The general finding that earth-
worms transport seeds is in line with a previous study
[16]; however, the current experiment differs from the
previous one in that we used four seed species concur-
rently in mesocosms, while previously two species were
studied separately at different periods of earthworm ac-
tivity. Thus, the current experiment is to our knowledge
the first attempt that tries mimicking field conditions by
studying the effects of earthworms on multi-species soil
seed bank dynamics.
Taken collectively, the results of the current study
show that via selective seed herbivory and species-spe-
cific transport of seeds within the soil seed bank earth-
worms have a great potential to affect plant community
composition. However, this has so far not been tested
experimentally.
Acknowledgements
We are grateful to Henning Riebeling, Johannes Sie-
bigteroth, Harriet Leese, Britta Staffel, Edda Schulte,
Thomas Hiersemann and Lena Saxler for their help.
References
[1] J.A. Arnone III, J.G. Zaller, E. Spehn, P.A. Niklaus, C.A. Wells,
C. Körner, Dynamics of fine root systems in intact native grass-
lands: effects of elevated atmospheric CO2, New Phytol. 147
(2000) 73e85.
[2] T. Decaens, L. Mariani, N. Betancourt, J.J. Jimenez, Seed dis-
persion by surface casting activities of earthworms in Colom-
bian grasslands, Acta Oecol. 24 (2003) 175e185.
[3] C.A. Edwards, G.W. Heath, The role of soil animals in break-
down of leaf material, in: J. Doeksen, J. van der Drift (Eds.),
Soil Organisms, North Holland Publishers, Amsterdam, The
Netherlands, 1963, pp. 76e80.
[4] J.D. Grant, The activities of earthworms and the fates of seeds,
in: J.E. Satchell (Ed.), Earthworm Ecology. From Darwin to
vermiculture, Chapman and Hall, London, 1983, pp. 107e122.
 J.G. Zaller, N. Saxler / European Journal of Soil Biology 43 (2007) S86eS91
[5] H. Hurka, R. Haase, Seed ecology of Capsella bursa-pastoris
(Cruciferae) dispersal mechanisms and the soil seed bank, Flora
172 (1982) 35e46.
[6] R.M. McRill, G.R. Sagar, Earthworms and seeds, Nature 243
(1973) 482.
[7] F.J.R. Meysman, J.J. Middelburg, C.H.R. Heip, Bioturbation:
a fresh look at Darwin’s last idea, Trends Ecol. Evol. 21
(2006) 688e695.
[8] A. Milcu, J. Schumacher, S. Scheu, Earthworms (Lumbricus
terrestris) affect plant seedling recruitment and microhabitat
heterogeneity, Funct. Ecol. 20 (2006) 261e268.
[9] R. Neilson, B. Boag, Feeding preference of some earthworm
species common to upland pastures in Scotland, Pedobiologia
47 (2003) 1e8.
[10] T.G. Piearce, Gut contents of some lumbricid earthworms,
Pedobiologia 18 (1978) 153e157.
[11] T.G. Piearce, N. Roggero, R. Tipping, Earthworms and seeds,
J. Biol. Educ. 28 (1994) 195e202.
[12] D.L. Shumway, R.T. Koide, Seed preferences of Lumbricus
terrestris L, Appl. Soil Ecol. 1 (1994) 11e15.
[13] E.M. Spehn, J. Joshi, B. Schmid, J. Alphei, C. Körner, Plant
diversity effects on soil heterotrophic activity in experimental
grassland ecosystems, Plant Soil 224 (2000) 217e230.
S91
[14] K. Thompson, A. Green, A.M. Jewels, Seeds in soil and worm
casts from a neutral grassland, Funct. Ecol. 8 (1994) 29e35.
[15] P.J. Van der Reest, H. Rogaar, The effect of earthworm activity
on the vertical distribution of plant seeds in newly reclaimed pol-
der soils in the Netherlands, Pedobiologia 31 (1988) 211e218.
[16] J.H. Willems, K.G.A. Huijsmans, Vertical seed dispersal by
earthworms: A quantitative approach, Ecography 17 (2)
(1994) 124e130.
[17] J.G. Zaller, Ecology and non-chemical control of Rumex crispus
and R. obtusifolius (Polygonaceae): a review, Weed Res. 44
(2004) 414e432.
[18] J.G. Zaller, Sheep grazing vs. cutting: regeneration and soil
nutrient exploitation of the grassland weed Rumex obtusifolius,
BioControl 51 (2006) 837e850.
[19] J.G. Zaller, J.A. Arnone III, Interactions between plant species
and earthworm casts in a calcareous grassland under elevated
CO2, Ecology 80 (1999) 873e881.
[20] J.G. Zaller, J.A. Arnone III, Earthworm and soil moisture ef-
fects on the productivity and structure of grassland communi-
ties, Soil Biol. Biochem. 31 (1999) 517e523.
[21] J.G. Zaller, J.A. Arnone III, Earthworm responses to plant
species’ loss and elevated CO2 in calcareous grassland, Plant
Soil 208 (1999) 1e8.