Earthworms (Lumbricus terrestris) affect plant seedling recruitment and microhabitat heterogeneity


†Author to whom correspondence should be addressed. E-mail:


  • 1The effects of the anecic earthworm Lumbricus terrestris L. on plant seedling recruitment and spatial aggregation were investigated in a microcosm glasshouse experiment by varying plant seed size (small and large); functional groups (grasses, legumes, herbs); plant species diversity (1, 3, 6); and plant functional group diversity (1, 3).
  • 2Generally, earthworms buried seeds quickly irrespective of seed size and species. Secondary seed dispersal (phase II dispersal) by earthworms affected plant community composition depending mainly on seed size but less on plant functional group identity and diversity: small-seeded species were repressed whereas large-seeded were promoted.
  • 3Although, in general, recruitment of seedlings was less in the presence of L. terrestris, recruited seedlings benefited from establishing in the vicinity of earthworm burrows. The strong aggregation of plants in the vicinity of earthworm burrows resulted in plant communities with a more heterogeneous small-scale architecture. Earthworm burrows and middens acted as an important regeneration niche for emergent seedlings by reducing microsite and nutrient limitations.
  • 4In conclusion, seed dispersal, seed burial, seedling recruitment, and the spatial distribution of seedlings of plant species of different functional groups and with a wide range of seed size are strongly affected by L. terrestris, and this probably affects plant community composition.


The species diversity of plant communities results from the dynamics of plant mortality and seedling recruitment of new arrivals in the regeneration niche, sensu Grubb (1977). Understanding the mechanisms that drive seedling recruitment is therefore essential for understanding plant community establishment and maintenance. The trade-off between the number and size of seeds, and plant traits affecting seed dispersal, are major factors driving seedling recruitment (Harper 1977; Chambers & MacMahon 1994; Jakobsson & Eriksson 2000). Large-seeded species are known to have a number of advantages during establishment (ability to emerge from greater depths, higher seedling size, enhanced ability to tolerate defoliation), but have to overcome the cost of seed predation above ground, whereas small-seeded species compensate through high numbers and increased persistence in the seed bank (Harper 1977; Westoby, Jurado & Leishman 1992; Fenner & Thompson 2005). However, the establishment of seedlings also depends strongly on local processes such as small-scale disturbances (Grime 1973; Grubb 1977). There is evidence that after phase I dispersal of seeds, which includes displacement of seeds from the parent to the soil surface, earthworms play an important role in phase II dispersal (secondary seed dispersal) – the subsequent displacement of seeds on the soil surface or burial in the soil (Grant 1983; Van der Reest & Rogaar 1988; Thompson, Green & Jewels 1994; Willems & Huijsmans 1994; Decaens et al. 2003). Selective ingestion and digestion of seeds (McRill & Sagar 1973; Shumway & Koide 1994); downward or upward seed transport (Hurka & Haase 1982; Grant 1983); acceleration (Ayanlaja et al. 2001); or delaying seed germination (Grant 1983; Decaens et al. 2001) are the main mechanisms by which earthworms affect germination and recruitment of seedlings. Most studies on earthworm–seed interrelationships have concentrated on the effects of earthworms on the soil seed bank, vertical dispersion, and viability of seeds in the seed bank; little is known about how earthworms may influence seedling recruitment and the composition of plant communities (Grant 1983; Pierce, Roggero & Tipping 1994; Willems & Huijsmans 1994).

Large surface litter-feeding earthworms (anecic species, Bouché 1977), such as Lumbricus terrestris L. (Lumbricidae), are dominant components of decomposer communities in virtually all non-acidic agricultural and forest ecosystems, including pastures and meadows of the temperate zone. Lumbricus terrestris is a peregrine (invasive) species which has been spread through European agricultural practices virtually all over the world, as well as into pristine ecosystems previously devoid of earthworm species (Bohlen et al. 2004). Earthworms, in particular anecic species, function as ecosystem engineers, modifying the physical structure of soils by changing soil aggregation, soil porosity, and the distribution and abundance of micro-organisms and other soil invertebrates (Wickenbrock & Heisler 1997; Maraun et al. 1999; Tiunov & Scheu 1999).

Modification of the physical structure of soil by creating and modifying microsites may function as a small-scale disturbance which is known to affect plant recruitment and therefore, potentially, plant community structure (Grime 1973; Connell 1978; Fox 1979). Earthworm casts and burrows probably form an important regeneration niche for plant seedlings (Crawley 1992), so it is surprising that, to our knowledge, there has been no study investigating how recruitment of seedlings of different seed sizes or plant functional groups is affected by the presence of earthworms. There is evidence that earthworms can affect plant community structure by promoting certain functional groups of plants via decomposition processes (Kreuzer et al. 2004; Wurst, Langel & Scheu 2005). On the other hand, earthworms are affected by the composition and diversity of plant communities (Spehn et al. 2000; Milcu et al. 2006). Yet it remains to be explored whether effects of earthworms on plant performance start with changes in seed germination and seedling recruitment.

We set up a microcosm greenhouse experiment to evaluate if L. terrestris affects seed germination and seedling recruitment success (defined here as the percentage of seeds sown that gave rise to emerged seedlings) of plant species through promotion and/or repression of certain species, depending on seed size and seed functional group identity (grasses, herbs, legumes) and diversity. We also investigated whether the spatial distribution, and consequently the small-scale architecture, of recruited seedlings is affected by L. terrestris.

Materials and methods

experimental setup

The experiment was set up in microcosms consisting of PVC tubes (inner diameter 16 cm, height 38 cm) covered by a 1-mm mesh at the bottom to prevent L. terrestris from escaping but allow drainage of water. The soil (pH 8·1, carbon content 4·6%, C : N ratio 15·7) was taken from the south-eastern edge of the Jena Experiment, a large biodiversity field experiment (Thuringia, Germany; Roscher et al. 2004). A total of 90 microcosms were filled with 5·5 kg sieved (1 cm) and homogenized soil and placed in a temperature-controlled greenhouse at a day/night regime of 16/8 h and 20/16 ± 2 °C. Before starting the experiment, the microcosms were watered regularly for a month and germinating weeds (unwanted plants from the seedbank) were removed. One gram of dried litter (2·53% N, C : N ratio 17·3, dried at 60 °C and cut into pieces ≈3 cm long), collected near the Jena Experiment study site and consisting mainly of grass leaves, was placed on top of the soil prior to the addition of earthworms and seeds to simulate natural conditions. Two adult L. terrestris (average fresh weight 4·2 ± 0·94 g) were introduced in half the microcosms, creating two treatments (control and earthworms). Within each treatment seeds of six plant species were sown, consisting of two seed-size classes (small and large) and three functional groups (grasses, herbs and legumes) (Table 1). The species were selected from the species pool of the Jena Experiment (Roscher et al. 2004) and sown as monocultures and two mixtures with three species (one mixture with the small seed-size species, one with the large seed-size species), in a factorial and fully randomized design with five replicates per species composition (Table 1). An additional mixture with all six species was also set up to check for the effects of species diversity. A constant number of 54 seeds per microcosm were added; in the three- and six-plant-species mixtures, the number of seeds per plant species was 18 and nine, respectively. Seeds were purchased from C. Apples Wilde Samen GmbH, Darmstadt, Germany. Vicia cracca was scarified mechanically to stimulate germination.

Table 1.  Setup of experiment varying plant species number (1, 3, 6); functional group number (1, 3) and identity (grasses, G; herbs, H; legumes, L); and seed size (small, large) in a factorial design with and without Lumbricus terrestris (five replicates each)
MixtureDiversitySpecies compositionFunctional groupSeed size (mm)
  • Seed size average was estimated by directly measuring 10 seeds per species.

  • G, Poa pratensis L., Festuca pratensis Huds.; H, Bellis perennis L., Tragopogon pratensis L.; L, Trifolium repens L., Vicia cracca L.

11P. pratensisGSmall (1·9 × 1)
21B. perennisHSmall (1·2 × 0·2)
31T. repensLSmall (1·2 × 1·2)
41F. pratensisGLarge (6 × 1·3)
51T. pratensisHLarge (12 × 1·3)
61V. craccaLLarge (2·8 × 2·8)
73P. pratensis, B. perennis, T. repensGHLSmall
83F. pratensis, T. pratensis, V. craccaGHLLarge
96P. pratensis, B. perennis, T. repens, F. pratensis, T. pratensis, V. craccaGHLSmall + large

sampling and analytical procedure

During the first 3 weeks of the experiment, the number of germinated seeds was counted twice a week in order to detect treatment effects on seed germination. The experiment lasted 7 weeks, and at the end of the experiment established plants were counted, harvested separately and dried at 60 °C for 3 days. The amount of weeds germinating during the experiment was low (0·6 ± 0·8 individuals per microcosm) and all germinated weeds were counted and picked. At harvest, the position of each individual plant in the microcosm was recorded on a transparent map to assess the spatial pattern of the established seedlings. The scion image program (Scion Corporation, was used to read the Cartesian co-ordinates of each plant individual from the transparent maps.

anova as part of the GLM procedure in sas 8 (SAS Institute, Cary, FL, USA) was used to analyse, in a hierarchical order (sum of squares type I, SS1), the effects of L. terrestris, seed size, functional group diversity, functional group identity (grasses, G; herbs, H; legumes, L as three dichotomus factors and interactions) on arcsine-transformed percentages of the total established plant individuals per microcosm in the six monocultures and in the three-species mixtures. In a subsequent anova (SS1) we investigated how the presence of L. terrestris and plant diversity affected the recruitment of seedlings of different species (as arcsine-transformed percentage of seedlings established per species) in the monocultures and in the three- and six-plant-species mixtures. Likewise, changes in individual biomass of individual plant species were analysed. For the paired monocultures (with and without L. terrestris), repeated-measures anova was used to test for time effects on seed germination. The mixtures with three and six plant species were analysed by individual manovas (sum of squares type III, SS3) to evaluate the effect of L. terrestris on plant species dominance. Inspection the data for homogeneity of variance prior to anova (Levene's test) suggested that preconditions for performing anova were matched.

Plant co-ordinates were used to analyse effects of L. terrestris on the spatial distribution of individual plant species. For this, the circular microcosm area was divided into 18 sections by three concentric rings, each divided into six sectors. To assess the association between earthworm burrows and established seedlings, the co-occurrence of plants and burrows (the number of sections where both earthworm burrows and plants were present) in the 18 sections of each microcosm, averaged over identical replicates, was then compared with the frequency distribution of average co-occurrences resulting from 999 randomized Monte Carlo simulations where the observed number of plants and burrows was randomly and independently reallocated within each pot. In addition, an index of aggregation was calculated per microcosm to characterize the degree of spatial aggregation within microcosms. The variance/mean ratio of observed numbers of plants in each sub-area was not directly applicable as an index of dispersion, because the 18 sections were of different size. For each sub-area we therefore multiplied the number of observed plant individuals by the ratio between average and individual sub-area size and used the variance/mean ratio of these area-corrected values (plants per 11·17 cm2) as an index of aggregation (the expected value of this index under a random pattern depends on the size distribution of the sub-areas and equals 1·533 in our case). Due to the greatly skewed distribution of the aggregation index, the effect of L. terrestris and seed size was then assessed using the Mann–Whitney U-test.


seed burial and germination

In less then 48 h, 95% of the added seeds were buried in earthworm burrows, irrespective of seed size, plant functional group and plant species, or ingested. As a result, germination of seeds was the limiting factor for seedling recruitment in this system and germination success coincided exactly with later seedling recruitment success (see Seedling recruitment).

Vicia cracca was the only species in which L. terrestris accelerated germination with significantly more individuals germinating in the first week than in the control (+128%, F3,6 = 4·8, P = 0·048). However, after 3 weeks the number of recruited seedlings in the control treatment equalled that in the earthworm treatment, and at the end of the experiment there were fewer plants in the presence of earthworms.

seedling recruitment

Irrespective of plant functional group and plant diversity, L. terrestris strongly reduced recruitment success from 47 to 16% (Table 2). Overall, seed size significantly affected the total number of seedlings recruited per microcosm, with large seeds having a higher recruitment success (40%) compared with small seeds (23%), whereas increasing functional group diversity slightly increased recruitment (+6%) (Table 2). Recruitment of plant species with small seeds was reduced more strongly (−23%) in the presence of L. terrestris than that of plants with large seeds (−10%) (significant L. terrestris–seed size interaction, Table 2; Fig. 1). However, some species were more affected than others: the reduction was high in P. pratensis (from 24 to 5%), B. perennis (from 56 to 6%), T. repens (from 43 to 10%) and F. pratensis (from 64 to 22%); but low in T. pratensis (from 54 to 38%) and V. cracca (from 47 to 37%) (Table 3). The recruitment and individual biomass of the grass species (P. pratensis and F. pratensis) were affected by earthworm presence and plant diversity (L. terrestris × diversity), but no consistent pattern could be observed. In contrast to seedling recruitment, L. terrestris significantly increased the above-ground biomass of individual plants at the end of the experiment in P. pratensis (+138%, F1,18 = 12·87, P = 0·005), F. pratensis (+223%, F1,18 = 18·4, P < 0·001), T. pratensis (+124%, F1,18 = 17·1, P = 0·005) and V. cracca (+48%, F1,18 = 5·9, P = 0·023), with the same increasing trend for the other species (Fig. 2; Table 2).

Table 2.  Results of anova (SS1) on the effects of Lumbricus terrestris, seed size, functional group diversity and identity of functional groups (grasses, herbs, legumes) on the arcsin-transformed percentage of total recruited seedlings in monocultures and three-species mixtures
  • ***

    , P < 0·001;

  • **

    , P < 0·01; NS, P > 0·05.

Lumbricus terrestris203·8***
Seed size 56·8***
L. terrestris × seed size  9·0**
Functional group diversity  4·5*
Grasses  4·9*
Herbs  2·9 NS
Legumes  1·0 NS
Figure 1.

Percentage recruitment of plant species as affected by seed size (small and large) and presence of Lumbricus terrestris. Error bars, ± SE.

Table 3.  Results of anova (SS1) on the effects of Lumbricus terrestris, plant species diversity (diversity) and their interaction on the recruitment of seedlings of different species (as arcsin-transformed percentages of the number of seeds added per species) and on the biomass of plant individuals at the end of the experiment from monocultures and three- and six-species mixtures
SourcedfSmall seededLarge seeded
P. pratensisB. perennisT. repensF. pratensisT. pratensisV. cracca
  • ***

    , P < 0·001;

  • **

    , P < 0·01;

  • *

    , P < 0·05; NS, P > 0·05.

  • For full species names see Table 1.

Lumbricus terrestris 129·8***12·8**209·5***3·4 NS89·5***0·4 NS114·4***18·4**11·2*17·1**12·1*3·9 NS
Diversity 2 0·3 NS18·5***  1 NS1·5 NS 1·0 NS1·6 NS  1·6 NS 0·6 1·4 NS 2·3 NS 0·7 NS0·6 NS
Diversity linear 1 0·6 NS 1·9 NS  0·4 NS0·2 NS 0·5 NS0·1 NS  2·2 NS 1·7 NS 1·5 NS 0·4 NS 0·9 NS0·2 NS
L. terrestris × diversity 2 0·9 NS19·0***  1 NS0·9 NS 0·3 NS0·3 NS  4·3* 4* 0·8 NS 0·8 NS 2·5 NS0·1 NS
Figure 2.

Mean biomass (DW) of plant individuals as affected by the presence of Lumbricus terrestris. Error bars, ± SE. For full species names see Table 1.

In the three-species mixtures, L. terrestris significantly affected the relative abundance of plant species within communities with small (Pillai's trace, F3,6 = 49·87, P < 0·001) and large seeds (F3,6 = 43·19, P < 0·001), and this was also true for the six-species mixture (F6,2 = 42·58, P = 0·023) (Fig. 3). Standardized canonical coefficients suggest that B. perennis contributed most to the change in composition of plant species in small-seed and in all six-species seed mixtures, while in the large-seed mixtures F. pratensis was the species that contributed most (Table 4), both of them being strongly depressed. Furthermore, significantly more weed seedlings established in the presence of L. terrestris (0·91 seedlings per microcosm) than in the control (0·28 seedlings per microcosm, F1,88 = 14·5, P < 0·001).

Figure 3.

Changes in plant species composition (percentage of established) as affected by presence of Lumbricus terrestris and seed size. For full species names see Table 1.

Table 4.  Standardized canonical coefficients reflecting the contribution of individual plant species to changes in plant community composition in the two mixtures with three plant species (small and large) and in the mixture with all six plant species as affected by Lumbricus terrestris
Plant speciesStandardized canonical coefficients
Three plant species with small seedsThree plant species with large seedsAll six plant species
  1. Statistically significant effects in bold.

Poa pratensis−0·12 1·42
Bellis perennis 5·08 6·47
Trifolium repens−0·23 0·58
Festuca pratensis 4·53 1·68
Tragopogon pratensis−0·47−1·80
Vicia cracca 0·76 3·65

effects on spatial distribution of recruited seedlings

Aggregation of recruited seedlings increased significantly (index of aggregation increased from 1·38 ± 1·17 to 7·57 ± 1·31) in the presence of L. terrestris (U = 6·71, P < 0·001), plants being associated with earthworm burrows except for the legumes T. repens and V. cracca (Table 5). Seed size had no significant effect on spatial aggregation (U = 0·67, P = 0·505).

Table 5.  Association between established plant seedling and burrows
MixtureDiversityFunctional groupSeed sizeAverage co-occurrencesP
  1. The average number of co-occurrence of plants and burrows in the microcosms with Lumbricus terrestris is used as a test statistic in a Monte Carlo test to compare with the frequency distribution of simulated co-occurrences resulting from 999 Monte Carlo simulations where the observed number of plants and burrows was randomly and independently reallocated within each pot. Functional group composition represented by grasses (G), herbs (H) and legumes (L). For full species names see Table 1.

1P. pratensisGSmall0·190·400·032
2B. perennisHSmall0·210·600·003
3T. repensLSmall0·710·800·235
4F. pratensisGLarge1·112·800·001
5T. pratensisHLarge2·173·600·002
6V. craccaLLarge3·373·60·245
7P. pratensis, B. perennis, T. repensGHLSmall0·3610·002
8F. pratensis, T. pratensis, V. craccaGHLSmall3·213·80·068
9P. pratensis, B. perennis, T. repens,GHLSmall + large1·642·80·002
 F. pratensis, T. pratensis, V. cracca     


In the present experiment plant seeds were buried quickly by L. terrestris (within 48 h) regardless of size, species or functional group, which is surprising. In previous studies, L. terrestris fed selectively on plant seeds of a certain size, shape and surface texture (McRill 1974, Grant 1983); selective feeding on leaf litter by L. terrestris also varies with these parameters (Satchel & Lowe 1967). We did not distinguish between seeds that were ingested and those that were pulled into the burrow and buried. As L. terrestris is unable to feed on particles with a diameter >2 mm (Shumway & Koide 1994; Tiunov & Scheu 1999), some seeds, such as those of V. cracca and T. pratensis, were too large for the earthworms to swallow.

Seed germination and seedling recruitment were generally reduced in the presence of L. terrestris, irrespective of plant species and seed size. However, plant species with small seeds were affected more strongly, presumably due to digestion or damage during passage through the gut of L. terrestris (McRill & Sagar 1973; Shumway & Koide 1994), or due to burial below the germination level. On the other hand, large seeds were not ingested, but were buried mostly in the upper 3–4 cm of soil or used, together with litter, to build up middens at the burrow entry, which are typical for anecic earthworms. The improved germination and establishment of large-seeded species in the presence of earthworms in this system was driven, in part, by the interaction between earthworms and seed size, where the large-seeded species benefited from the ability to emerge from greater depths coupled with the reduced ingestion and digestion rates by earthworms. Furthermore, earthworms are able to promote or repress certain plant species: in the six-species mixture, plant species with large seeds dominated in the presence of L. terrestris whereas small-seeded species, especially B. perennis, were detrimentally affected.

As seeds at the soil surface are more vulnerable to predation (Chambers & MacMahon 1994; Wilby & Brown 2001) and germination might be hampered by the lack of water, seeds avoid above-ground predation by being buried. Especially larger seeds, which are known to be more prone to granivory than smaller seeds (Heske & Brown 1990), will benefit form being buried by earthworms. Moreover, seeds surviving the gut passage may find favourable environmental conditions for germination, recruitment and growth due to increased water-holding capacity and concentrations of nutrients in earthworm casts and middens, in particular nitrogen and phosphorus (James 1991; Blanchart et al. 1999). In support of this view, the above-ground biomass of individual plant seedlings growing in the presence of earthworms was increased for all species, but this was significant only in the case of P. pratensis, F. pratensis and T. pratensis. Shoot biomass also increased, due in part to reduced intra- and interspecific competition as the number of established plants was lower in the presence of L. terrestris. Burial affected germination of large-seeded V. cracca beneficially, perhaps because the seeds on the soil surface did not remain wet in the intervals between watering. Large seeds are known to be produced by large plant species (Fenner & Thompson 2005). Favouring recruitment of plant species with large seeds may therefore feed back positively to earthworms through the production of large amounts of litter.

The significant association of recruiting plants with earthworm burrows in the present study is consistent with Grant's (1983) finding that 70% of seedlings in temperate grasslands germinate out of earthworm casts, although they covered only 24–28% of the soil surface. This indicates that the physical structures created by earthworms (middens, casts and burrows), and the favourable conditions associated with them, act as important regeneration niches in grassland communities (Grubb 1977). In the field, reduced intra- and interspecific competition between seedlings, and improved nutrient and water supply in and around earthworm burrows, presumably result in high seedling biomass and improve seedling recruitment success by reducing microsite, nutrient and light limitations, particularly when competing with established plant species (Tilman 1988). Improving the competitiveness of recruited seedlings probably more than compensates for seed digestion during gut passage through earthworms, which is known to be low (McRill & Sagar 1973; Grant 1983).

Burial of seeds by anecic and other earthworms contributes significantly to the build up of the soil seed bank (Grant 1983; van der Reest & Rogaar 1988; Willems & Huijsmans 1994). Weeds (unwanted plant from the seed bank) germinating in the microcosms during a period of 4 weeks before the start of the experiment were removed. Nevertheless, at the end of the experiment the number of weeds in the microcosms with L. terrestris had increased significantly, suggesting that the earthworms translocated seeds from deeper in the soil to the soil surface, where they germinated. As the soil seed bank contributes strongly to the resilience of grassland communities, speeding up regeneration following disturbances (Uhl et al. 1981; Marks & Mohler 1985; Kalamees & Zobel 2002), earthworms play an important role in the stability of grassland systems. However, detrimental effects of earthworms on herbaceous plants and tree seedlings have also been documented (James & Cunningham 1989; Gundale 2002; Hale 2004), but these reports are restricted to habitats previously devoid of earthworms and now invaded by non-native earthworm species (Bohlen et al. 2004; Hale 2004).

In conclusion, our findings suggest that secondary seed dispersal by anecic earthworm species, such as L. terrestris and probably also other earthworm species, form a major driving force for seed germination, seedling recruitment, plant community composition and stability of grassland communities. Lumbricus terrestris affects the recruitment of seedlings, promoting or repressing certain plant species depending on seed size but less on functional group identity or diversity. Therefore, for understanding of ecological and evolutionary relationships between seed size and seedling recruitment, the role of the soil fauna needs closer consideration. Results of this study suggest that earthworm middens, casts and burrows are important regeneration niches in grassland ecosystems, thereby reducing recruitment limitation. Fast removal of seeds from the soil surface, increasing microhabitat heterogeneity and acceleration of seed-bank formation are also prominent mechanisms through which earthworms affect seed germination, seedling recruitment and, presumably, plant community composition. The role of earthworms in the formation, maintenance and succession of plant communities deserves further attention.


Financial support by the German Science Foundation is gratefully acknowledged (FOR 456; The Jena Experiment). Thanks also to T. Volovei for help in the laboratory, and to V. Temperton for valuable comments on the manuscript.