Conspecific and heterospecific plant–soil feedbacks influence survivorship and growth of temperate tree seedlings


  • Sarah McCarthy-Neumann,

    Corresponding author
      Correspondence author. School of Natural Resources and Environment, University of Michigan, Ann Arbor, MI 48109, USA. E-mail:
    Search for more papers by this author
  • Richard K. Kobe

    1. Department of Forestry and Graduate Program in Ecology, Evolutionary Biology and Behaviour, Michigan State University, East Lansing, MI 48824, USA
    Search for more papers by this author

Correspondence author. School of Natural Resources and Environment, University of Michigan, Ann Arbor, MI 48109, USA. E-mail:


1. The Janzen–Connell (J–C) Model proposes that host-specific enemies could maintain high tree species diversity by reducing seedling performance near conspecific adults. An implicit, but untested assumption of the J–C Model is that negative conspecific feedbacks would promote replacement by heterospecific seedlings.

2. In a glasshouse experiment, we tested plant–soil feedbacks as a J–C mechanism in four temperate tree species. We assessed effects of conspecific- relative to heterospecific-cultured soil extracts on seedling survival, total mass and performance for each focal species. To test the implicit assumption of replacement by heterospecific seedlings, we also compared relative performance of conspecific versus heterospecific seedlings grown with soil extract cultured by a particular tree species. We also tested whether soil microbes caused these plant–soil feedbacks and whether low irradiance increased seedling vulnerability to pathogens.

3. When grown with conspecific versus heterospecific soil extract, Acer rubrum mass decreased, Quercus rubra mass increased and Fraxinus americana increased survival. Conspecific extract reduced Acer saccharum mass in low light but increased it in high light. To integrate survival and growth, we examined seedling performance [(mean total mass × mean survival time)/(days of experiment)] at low and high light. In conspecific versus heterospecific soil extract, seedling performance was lower in two, higher in four and neutral in 18 of 24 cases, suggesting no advantage to dispersing away from conspecifics. Based on relative seedling performance within a soil extract, conspecific seedlings were disadvantaged in two, favoured in three and neutral in 19 of 24 cases relative to heterospecific seedlings.

4. Species pairwise interactions of soil modification and seedling performance were chemically mediated, occurring regardless of sterilization. Microbes lacked host specificity and reduced performance regardless of extract source. Additionally, microbial factors reduced seedling performance for Q. rubra regardless of light availability, and for A. rubrum and F. americana only in high light.

5.Synthesis. These chemical-mediated plant–soil feedbacks probably influence community dynamics, but are inconsistent with the J–C Model. Even when a species’ seedlings responded more negatively to conspecific than heterospecific soil, heterospecific seedlings were not necessarily favoured in that species’ soil, precluding heterospecific replacement as an explanation for coexistence.


Understanding the replacement and coexistence of tree species is a central focus in forest community ecology. One of the most influential hypotheses put forth to explain the maintenance of species diversity was by Janzen (1970) and Connell (1971); they hypothesized that competitive exclusion could be precluded by host-specific enemies that reduce seed and/or seedling survivorship near conspecific adults and/or at high conspecific densities, favouring the establishment of heterospecific individuals. Although originally proposed as an explanation of the maintenance of tree species diversity in tropical forests, non-competitive distance- or density-dependent (NCDD) mortality could be an important driver of community dynamics in temperate forests as well (Packer & Clay 2000; Hille Ris Lambers, Clark & Beckage 2002; Yamazaki, Iwamoto & Seiwa 2009).

Soil pathogens could be effective agents of NCDD mortality and/or reduced growth because many of them show host specialization, short generation times, high fecundity, long persistence in soil and more limited dispersal than their hosts (Gilbert 2002). These characteristics could enhance the potential for ‘culturing’ the local soil microbial community by resident plant species, leading to a potential negative feedback for plant species that cultured the microbes (van der Putten, Van Dijk & Peters 1993; Mills & Bever 1998; Klironomos 2002; Bezemer et al. 2006; Casper & Castelli 2007; Kardol et al. 2007).

As argued by McCarthy-Neumann & Kobe (2010), decreased growth or survivorship of a given species in conspecific relative to heterospecific-cultured soils will not necessarily promote species coexistence. Rather, within a given area influenced by a mature tree, a heterospecific seedling would have to perform better than a conspecific seedling, which is an implicit but untested assumption of the J–C Model. Testing for the effects of conspecific versus heterospecific-cultured soils requires differentiating species among heterospecific adults and not simply lumping all heterospecifics into a far-distance category (Augspurger 1984; Augspurger & Kelly 1984; Hood, Swaine & Mason 2004; Bell, Freckleton & Lewis 2006).

The importance of irradiance on seedling performance is well established, especially across broad gradients in light availability (Pacala et al. 1996; Kobe 1999). However, NCDD may also interact with irradiance in influencing seedling performance. For example, seedlings of many tropical species (Augspurger & Kelly 1984; Kitajima & Augspurger 1989; Hood, Swaine & Mason 2004) experience higher disease-related mortality at low than high light. Results for temperate tree species are mixed (O’Hanlon-Manners & Kotanen 2004, 2006), however, the interaction between light availability and disease on temperate tree seedling survival and growth has not been investigated.

Soil pathogens are only one of many possible agents for creating plant–soil feedbacks (Ehrenfeld, Ravit & Elgersma 2005). For instance, the presence of a particular plant species could be associated with formation of mycorrhizal networks (Booth 2004), production of allelochemicals (Stinson et al. 2006), alterations to soil structure (Rillig, Wright & Eviner 2002) and nutrient availability (Finzi, Canham & Van Breemen 1998a; Finzi, Van Breemen & Canham 1998b). A particular species could modify soil to the detriment or benefit of conspecific or heterospecific seedlings (Bezemer et al. 2006). Lumping all heterospecific species into a single category (e.g. Augspurger & Kelly 1984; Packer & Clay 2000; Hood, Swaine & Mason 2004) could obscure complex relationships among plant species mediated through soil biota and chemistry.

This study’s purpose was to examine seedling mortality and growth responses of four temperate tree species to: the species of adult tree that cultured the soil (‘source’ effects), presence of microbial pathogens and light level; we also conducted a parallel experiment for tropical species, which enabled direct comparisons between temperate and tropical tree seedlings (see McCarthy-Neumann & Kobe 2010). We used extracts filtered from soils cultured by different adult tree species to test the net effect of soil biota and chemical factors. Extracts contained bacteria and pathogenic fungal spores < 20 μm and water-soluble chemicals. Specifically, we tested: (H1) soil extracts cultured by conspecific adults reduce seedling survival and/or growth more than soils cultured by heterospecific adults; (H2) heterospecific seedling survival and/or growth are favoured relative to conspecific seedlings in soil extracts cultured by a particular tree species; (H3) sterilization of soil extract decreases negative effects of conspecific-cultured soils because of elimination of soil pathogens; and (H4) higher irradiance reduces negative effects of soil microorganisms. Following convention, we compare within-species seedling performance in conspecific and heterospecific soils (H1), but we also introduce a novel comparison: the relative performance of conspecific versus heterospecific seedlings in a given soil source, analogous to seedlings contesting a space occupied by a mature tree (H2). The latter is important in evaluating the potential for NCDD processes to promote species coexistence through reciprocal replacement.

Materials and methods

Overview of methods

The majority of negative feedback studies take soil from a common pit and ‘culture’ it with different species of seedlings (van der Putten, Van Dijk & Peters 1993; Bever 1994; Mills & Bever 1998; Klironomos 2002) to differentiate the microbial community; studies looking at distance-dependent effects typically collect soil at varying distances from adults (Packer & Clay 2000; Hood, Swaine & Mason 2004). To investigate plant–soil feedbacks as a mechanism for NCDD, we collected soil near adults of each of our study species, further cultured the soil in the glasshouse with seedlings, extracted microbial inoculum from cultured soil and inoculated experimental seedlings with these soil extract treatments (Appendix S1 in Supporting Information).

Soil culturing by conspecific versus heterospecific adults (H1) and conspecific versus heterospecific seedling response to a given extract (H2)

We used four dominant tree species (Acer rubrum, Acer saccharum, Fraxinus americana and Quercus rubra) that co-occur on moraines in Michigan, USA, for this study (Host & Pregitzer 1992). To test effects of conspecific versus heterospecific culturing on seedling performance (H1) and conspecific versus heterospecific seedling response to a given extract (H2), we collected two soil cores (7.5 cm diameter × 25 cm depth) within 1 m from the bole of three trees of each of the four species for a total of 96 soil cores (four species of seedling × four species of adult culturing × three trees × two samples per tree). Soil was collected in November 2004 when microorganisms were likely to be present as spores, enabling higher microbe survival during transport and cold storage than in summer samples. Sampled trees were randomly selected among adults with a diameter at 1.37 m height ≥ 75th percentile for that species, located in four mapped stands on moraines in Manistee National Forest, MI, USA. To minimize the potential for multi-species soil culturing, we took soil under trees that were at least two crown diameters away from adults of other species. Sampling locations for culturing by heterospecific adults was subject to the additional criterion that no conspecific adults were closer than 20 m, except a 10-m distance was used for A. saccharum because of its high local abundance. Soil was stored for c. 2 weeks at 4 °C until seeds were available for planting in intact soil cores.

Culturing of field soil by seedlings

To test effects of seedling density and species on soil culturing (and the subsequent performance of experimental seedlings), the field soils were further cultured for 14 weeks in a glasshouse at 2% full sun at two seedling densities (see Appendix S2 for detailed rational). However, both species and density of seedlings culturing soil did not alter responses of experimental seedlings to soil culturing by adult trees, so data were pooled across these treatments for analysis (see Appendix S2 for details); henceforth, we disregard the seedling culturing treatment and focus on adult soil source, sterilization and light effects. Acer rubrum and F. americana seeds were from Sheffield’s Seed Co. (Locke, NY, USA) and A. saccharum and Q. rubra seeds from Wisconsin Department of Natural Resources (Hayward, WI, USA). Seeds were surface-sterilized (0.6% NaOCl solution) and weighed prior to planting.

Sterilization effect on pathogen infection (H3)

Before obtaining microbial extracts, roots of the culturing seedlings were cut and mixed with the soil. Above-ground seedling portions were discarded. Microorganisms < 20 μm were extracted from cultured soil (cultured by adult trees in the field and then seedlings in the glasshouse) using a wet-sieving method adapted from Klironomos (2002). For each extraction, 50 g of soil was blended with 250 mL of tap water for 30 s. The liquid suspension was washed through 250, 45 and 20-μm analytical sieves with tap water, keeping the extract to ≤ 1 L. Sieves were cleaned ultrasonically for 5 min between each extraction, which minimized, if not eliminated, contamination between treatments. To test for microbial effects on seedling performance (H3), pots planted with germinating seeds were amended with autoclaved (30 min at 121 °C) versus unsterilized soil extract. Sterile soil extracts taken from individual adult trees were combined, resulting in 16 treatments (four species of seedling × four species of adult culturing); for the non-sterile soil extracts, all culturing treatments (e.g. four species of seedling × four species of adult culturing × three trees × two samples per tree) were kept separate throughout the experiment.

In a previous glasshouse experiment under similar conditions (S. McCarthy-Neumann & R. K. Kobe, unpublished data), arbuscular mycorrhizas (AMF) enhanced seedling mortality, which may hav arisen from the additional carbon cost of establishing a mycorrhizal network in the glasshouse pots compared with field conditions. Therefore, we used microbial extracts to assess plant–soil feedback instead of whole soil because we were able to isolate the pathogen and saprobic fraction (filtrate < 20 μm) for the treatments while eliminating AMF spores (collected on the 45-μm sieve). In addition, by using microbial extracts we are able to investigate the relative effect of seedling density on feedbacks without confounding seedling competition; i.e., seedlings in the experiment were not subject to competition but the soils for the extractions had different histories of seedling densities.

Planting methods and irradiance effect on pathogen infection (H4)

Seeds with newly emerged radicles were planted into pots (7.5 cm diameter × 25 cm depth) filled with 1:4 mixture of sterilized field soil and commercial peat moss (Fafard Mix #2; Conrad Fafard Inc., Agawam, MA, USA). Field soil was collected from a common pit in a mixed Fagus–Acer stand at Michigan State University’s Tree Research Centre and was autoclaved for 1 h at 121 °C, followed by 2-day incubation and a second autoclaving. Lethal temperatures (≥ 121 °C) were confirmed at the centre of each soil bag. Each experimental seedling received 100 mL of non-sterilized or sterilized extract. To test for irradiance effects (H3), seedlings were grown at two light levels (2% vs. 22% full sun). Light levels were designed to mimic endpoints in the range of light conditions from understorey to treefall gaps encountered in northerner Michigan forests (Schreeg, Kobe & Walters 2001). To achieve the desired light levels, we used an inner layer of black shade cloth with an outer layer of reflective knitted poly-aluminium shade cloth to minimize heat build-up.

Experimental treatments, seedling measurements and harvesting

To summarize, experimental treatments consisted of species of adult-culturing soil, sterilization and irradiance level. The 1668 seedlings were randomly assigned to eight benches (six for low and two for high light) and were allocated among treatments as follows. Under low light: 4 species of seedlings × [(1 treatment of soil cultured by conspecific adults × (60 replicates in the microbial treatment + 10 replicates in the sterile microbial treatment)) + (3 treatments of soil cultured by adults of other study species × (60 replicates in the microbial treatment + 10 replicates in the sterile microbial treatment)) + 12 replicates in tap water] for a subtotal of 1168 seedlings. Under high light: 4 species of seedling × [(1 treatment of soil cultured by conspecific adults × (30 replicates in the microbial treatment + 10 replicates in the sterile microbial treatment)) + (3 treatments of soil cultured by adults of the other study species × (15 replicates in the microbial treatment + 10 replicates in the sterile microbial treatment)) + 10 replicates in tap water] for a subtotal of 500 seedlings (see Appendix S1 for diagrammatic representation of experimental design).

Emergence and survival were censused thrice weekly and seedlings were watered (∼50 mL of deionized water) by hand twice weekly for 12 weeks (March–June 2005). We assigned date of death as the first census where total leaf or stem tissue necrosis was encountered, at which time seedlings were harvested. To determine effectiveness of extract sterilization and isolate soil pathogens, roots were surface-sterilized for 1 min with 0.6% NaOCl and cross-sections from the edge of the disease lesion were plated on water agar amended with Ampicillin. To determine mass, we harvested seedlings surviving to the end of the experiment, divided seedlings into organ fractions and oven-dried living tissue at 70 °C to constant mass.

Chemical analysis of soil extracts

To test for potential chemical differences, we measured exchangeable base cations (Ca, K and Mg), total organic C, total N, C:N ratios and protein-precipitable phenolics in extracts for each soil source (species of adult culturing). Extracts were stored for 2 years at 4 °C before analysis; the delay occurred because testing for chemical effects was not in our original research plan. Soil extracts were filtered with Whatman #2 papers and exchangeable base cations (Ca, K and Mg) were measured using a Perkin–Elmer Optima 2100 DV Optical Emission Spectrometer (Perkin Elmer, Norwalk, CT, USA). For total organic C, total N and C:N analysis, soil extracts were filtered through Whatman GF/F papers prior to being analysed with a TOC-V CPN Total Organic Carbon Analyzer equipped with a total nitrogen measuring unit (Shimadzu Co., Kyoto, Japan). We measured protein-precipitable phenolics with a spectrophotometer, following Makkar, Dawra & Singh (1988).

Statistical analysis of experimental seedlings

Seedling survival times were analysed with survival analysis (spss v. 14.0; SPSS Inc, Chicago, IL, USA) and included both pre- and post-emergence stages (pre-emergence mortality date was estimated as mean emergence date for that species). Cox proportional hazards regression (Cox & Oakes 1984) tested relative effects of soil source (species culturing soil), sterilization, light availability and initial seed mass on survival times.

We tested treatment effects (adult soil source, sterilization and irradiance level) and their interactions on seedling total mass with a split-plot ancova, split for light, using bench as a blocking factor (spss). Because seed mass can influence seedling size, estimated dried embryo mass (based on regressions of dry to fresh mass from c. 30 randomly selected seeds for each species) was a covariate. Data were ln-transformed when errors were not normal. We ran full models (main treatments, bench, covariate and interactions) for each species to test the assumption that covariate effects were independent of treatment effects; interaction terms were removed when P > 0.05. If either terms for bench, covariate or interaction between main treatments had P > 0.25 (Bancroft 1964), then the highest-order term with the highest P-value was removed and the analysis was run with a reduced model. This process was repeated until all terms with P > 0.25 were removed. Adjusted means were compared when the covariate was retained and raw means when the model reduced to anova. When main effect of soil source was significant (P < 0.05), a Holm adjustment was used to compare the conspecific with each of the three heterospecific adult soil sources for each species. Comparisons with a P-value ≤ 0.10 after adjustment are highlighted in this study. Since the primary interest of this study was to investigate plant–soil feedbacks as they relate to the J–C Model, we did not compare the relative effect among the three heterospecific-cultured soil extracts on seedling performance for a particular species, but only the relative effect of each of the heterospecific-cultured soil extracts versus conspecific-cultured soil extracts on seedling performance for a particular species. Differences between treatment means were assessed and P values calculated through degree of overlap in 95% confidence intervals (Austin & Hux 2002).

To summarize reciprocal effects, we compared percentage difference in integrated seedling performance [(mean total mass × mean survival time)/(days of experiment)] for each species pair in soil extract (combined sterile and non-sterile treatments) relative to tap water in high and low light. Our metric of seedling performance encapsulates the race among species from seedling to canopy tree in a space that was previously occupied and cultured by another canopy tree. The winning species for a given set of conditions will attain the largest size among the species that have at least one surviving individual (Kobe 1999). In addition, to determine relative effects of microbial versus chemical factors, we also calculated percentage difference in integrated seedling performance for each species pair as: adult-culturing through chemical effects alone (=seedling performance in sterile extract/seedling performance in tap water) and adult-culturing through soil microorganisms alone [=(seedling performance in non-sterile extract − seedling performance in sterile extract)/seedling performance in tap water]. We estimated 95% confidence intervals for integrated seedling performance metrics by bootstrapping 3000 data sets for each species (sampling with replacement) and analysing each data set as described above using R (R Development Core Team 2008).


Conspecific and heterospecific culturing had similar influence on seedling performance (H1)

Fraxinus americana seedlings had greater survival times in soil extract cultured by conspecific than Q. rubra adults (Fig. 1a, Table 1). Soil source did not affect survival time for any other study species (Appendix S3).

Figure 1.

 Survival curves for Fraxinus americana seedlings by (a) soil source (tree species culturing soil: Qr = Quercus rubra, Ar = Acer rubrum, As = Acer saccharum and Fa = F. americana) and (b) soil microbial treatment (sterile vs. non-sterile extract). Survival curves end at 58 days since no F. americana seedlings died after that date.

Table 1.   List of significant responses: (a) of seedlings to conspecific versus heterospecific-cultured soil extracts and (b) among conspecific versus heterospecific seedlings to soil extract cultured by a given species. Significance shown as follows: *P ≤ 0.10; **P ≤ 0.05; ***P ≤ 0.01; ****P ≤ 0.001
SpeciesMortality hazard†Total mass‡Integrated seedling performance‡
  1. Data condensed from Appendices S3 and S4 and Table 2. P values for multiple comparisons have been adjusted with a Holm adjustment.

  2. †Hazard ratios > 1 indicate reduction in days to mortality of study species.

  3. ‡Total mass and integrated seedling performance are the negative (−) or positive (+) percent difference in seedling response between (a) conspecific versus heterospecific tree soil sources and (b) conspecific versus heterospecific seedlings.

  4. NSD, no significant difference; Ar, Acer rubrum; As, Acer saccharum; Fa, Fraxinus americana; Qr, Quercus rubra.

(a) Species’ seedling response to conspecific versus heterospecific-cultured soil extracts (H1)
 Acer rubrumNSDAr vs. As (−16%*)Ar vs. Qr in low light (−15%*)
Ar vs. Qr (−18%*)
 Acer saccharumNSDAs vs. Fa in low light (−22%**)As vs. Fa in low light (−17%**)
As vs. Fa in high light (+31%***)
As vs. Qr in high light (+23%**)As vs. Fa in high light (+24%***)
 Fraxinus americanaFa vs. Qr (1.84*)NSDNSD
 Quercus rubraNSDQr vs. Ar (+12%**)Qr vs. Ar in low light (+10%**)
Qr vs. As in low light (+11%*)
Qr vs. Fa (+10%*)Qr vs. Fa in low light (+11%*)
Soil sourceIntegrated seedling performance‡  
(b) Relative response of conspecific versus heterospecific seedlings to soil extract cultured by a given species (H2) 
 Acer rubrumNSD  
 Acer saccharumAs vs. Fa in high light (+19%**)  
 Fraxinus americanaFa vs. Ar in high light (−20%**)  
Fa vs. Qr in high light (−18%**)  
 Quercus rubraQr vs. As in high light (+27%**)  
Qr vs. Fa in high light (+32%****)  

Soil source affected seedling total mass for all species except F. americana (Fig. 2, Table 1, Appendix S4), but generally, conspecific extracts did not disproportionately affect mass relative to heterospecific extracts. Acer rubrum was the only species with lower seedling mass in conspecific than heterospecific (A. saccharum and Q. rubra) soil extract. In contrast, seedlings of Q. rubra had greater mass in conspecific than in A. rubrum and F. americana soil sources. For A. saccharum seedlings, the influence of soil source was mediated by irradiance. In low light, seedling mass was lower in conspecific than F. americana soil source, whereas in high light, mean seedling mass was greater in conspecific than F. americana and Q. rubra soil sources. Sterilization of extracts had no effect on whether soil source influenced survival time and/or total mass for any species.

Figure 2.

 Total final mass by soil source (tree species culturing soil) and extract (sterile vs. non-sterile) for each study species in low and high light [(a, b) Acer rubrum (Ar), (c, d) Acer saccharum (As), (e, f) Fraxinus americana (Fa) and (g, h) Quercus rubra (Qr)]. Dotted line represents seedling mass when grown with tap water and is shown for reference only.

Comparing conspecific and heterospecific culturing on the basis of integrated seedling performance yielded similar results to total seedling mass (Tables 1 and 2). Specifically, conspecific feedbacks were relatively beneficial for Q. rubra seedlings (compared with the three heterospecifics) at low light and A. saccharum seedlings (compared with F. americana) under high light. Conspecific feedbacks were relatively detrimental for A. saccharum (vs. F. americana) and A. rubrum (vs. Q. rubra) under low light. Thus, taking the conventional approach of comparing performance in conspecific versus heterospecific soils, heterospecific soils were as likely to be detrimental as conspecific soils.

Table 2.   Reciprocal effects (percent difference in integrated seedling performance [(mean total mass × mean survival time)/(days of experiment)]) of plant–soil feedbacks relative to a tap water control for each study species integrated across extract treatment and at (a) high and (b) low irradiance levels
Species of seedling responding to soilAcer rubrumAcer saccharumFraxinus americanaQuercus rubra
  1. Effect of conspecific-cultured soil on seedlings is in bold. Bootstrap-derived 95% CI are in parentheses.

(a) High lightSpecies of adult culturing the soil
 Acer rubrum22% (−33, −12)−5% (−20, 10)−13% (−23, −3)−6% (−20, 7)
 Acer saccharum−30% (−43, −13)19% (−29, −11)−43% (−54, −33)−37% (−53, −20)
 Fraxinus americana−37% (−46, −27)−38% (−49, −27)33% (−43, −22)−42% (−54, −30)
 Quercus rubra−16% (−26, −8)−15% (−28, −5)−15% (−26, −5)10% (−21, −1)
(b) Low lightSpecies of adult culturing the soil
 Acer rubrum18% (−28, −7)−4% (−16, 9)−13% (−22, −3)−1% (−13, 13)
 Acer saccharum−17% (−28, −7)20% (−28, −11)−3% (−14, 5)−12% (−21, −4)
 Fraxinus americana−19% (−30, −6)−16% (−27, −4)8% (−21, 10)−21% (−34, −5)
 Quercus rubra−20% (−27, −14)−20% (−28, −14)−21% (−28, −14)10% (−18, −3)

Heterospecific seedling performance was more likely to be reduced than conspecific performance in soil extract cultured by a particular tree species (H2)

From the perspective of seedlings contesting space occupied and cultured by a particular canopy tree, conspecific seedlings were disadvantaged in only two and favoured in 3 out of 12 cases in high light; there were no significant differences among integrated seedling performance in low light (Tables 1 and 2). Acer rubrum seedlings tended to experience less severe feedbacks than seedlings of other species in heterospecific-cultured soils (except for F. americana cultured soil at low light).

Sterilization (H3) broadly influenced seedling performance; irradiance (H4) interacted with sterilization

Sterilization of soil extracts increased survival time and/or growth for all species, except A. saccharum (Appendices S3 and S4). Fraxinus americana seedlings had increased survival time in sterile extract across all soil sources (Fig. 1b). Sterile extract also increased total seedling mass for Q. rubra regardless of soil source or irradiance level (12%, = 0.01); Acer rubrum (23%, = 0.01) and F. americana (23%= 0.04) seedling mass increased in all sterilized soil sources but only in high light (Fig. 2). Under higher irradiance, total mass was greater for all species (Fig. 2, Appendix S4), but survival time did not vary (Appendix S3). Sterilization of extracts and efforts to minimize cross contamination were effective; 70% of dying seedlings in non-sterile treatments were infected with soil pathogens versus 6% of dying seedlings in sterile treatments. 75% of infected seedlings harboured at least one of five Fusarium morphotypes. Organ mass responses to soil source, sterilization and irradiance treatments generally reflected total mass patterns for all species (Appendices J-M in McCarthy-Neumann 2008).

Under low light, chemical effects dominated and only one microbial effect was significantly different from zero (a negative effect of Q. rubra soil source on Q. rubra seedlings); chemical effects on performance were significantly lower than zero in A. saccharum (for all extracts except from F. americana sources) and Q. rubra (for all extracts except from conspecific sources) seedlings and marginally significant for F. americana (A. rubrum and Q. rubra soil sources) (Fig. 3). Acer rubrum was not influenced by chemical effects under low or high light. Under high light, microbial effects were manifested for A. rubrum and F. americana in all soil extracts except from A. saccharum sources; A. saccharum and Q. rubra seedlings were insensitive to microbial effects. Additionally under high light, A. saccharum and F. americana were influenced by chemical effects in soil extracts from all soil sources, and chemical effects from A. rubra soil source affected Q. rubra performance.

Figure 3.

 Relationship between chemical (sterile extract/tap water) and microbial [(unsterile extract/tap water) − (sterile/tap water)] effects in soil extracts ‘cultured’ by different species of adult on seedling performance [(mean total mass × mean survival time)/(days of experiment)]) for each study species in high and low light. Ar = Acer rubrum, As = Acer saccharum, Fa = Fraxinus americana and Qr Quercus rubra. Bootstrap-derived 95% CI included. Solid line represents equal impact of both chemical and microbial factors in soil extract on seedling performance.

Chemical analysis of soil extracts

Soil extracts cultured by Q. rubra adults had lower base cation availability (combined Ca, K and Mg) than extracts cultured by A. saccharum and F. americana adults but had greater total C and C:N ratios than sources from the other three species (Appendix S5). Acer rubrum source soils also had lower base cations and greater C:N ratios than A. saccharum and F. americana source soils (Appendix S5). Total N (Appendix S5) and protein-precipitable phenolics (= 0.42, d.f. = 3,19, = 0.74) did not differ among soil sources.


Soil-mediated species-specific feedbacks between tree adults and seedlings were ubiquitous but were largely idiosyncratic and inconsistent with the Janzen–Connell (J–C) Model in several respects. First, conspecific and heterospecific feedbacks had similar prevalence and magnitude of influence among the study species, suggesting that dispersal away from conspecific adults is not necessarily advantageous for all species. Second, heterospecific seedling performance was more likely to be reduced than conspecific seedlings in a given soil extract suggesting that plant–soil feedbacks are more apt to facilitate conspecific than heterospecific recruitment into the canopy, precluding heterospecific replacement as an explanation for tree species coexistence. Third, species-specific feedbacks tended to be chemical rather than natural enemies as envisioned by the J–C Model. Finally, natural enemies were not host specific; soil microbes reduced seedling performance regardless of which tree species cultured the soil.

Conspecific and heterospecific culturing had similar influence on seedling performance (H1)

For three of four species under high light and two of four species under low light, conspecific soil sources actually had the least adverse effect on seedlings among all soil sources in terms of rank, however, not all ranks differed statistically. Many studies have shown that individual plant–soil feedback responses vary among species depending upon characteristics such as local abundance (Klironomos 2002), shade tolerance (McCarthy-Neumann & Kobe 2008), successional (Kardol et al. 2007) or exotic (Reinhart et al. 2003, 2005) status. In this study, there does not appear to be a simple explanation for the way that species responded to plant–soil feedbacks mediated by conspecific or heterospecific adults. In addition, none of the soil sources here were uniformly beneficial or detrimental to seedlings of all species and relative influences of biotic and chemical factors from a given soil source could change depending upon the species of responding seedlings. As a whole, these results support fairly specific interactions between pairs of species that are mediated through both chemical and biotic factors (similar to results by Bezemer et al. 2006 with grassland species).

Although there are limited data to test whether similar species pair wise interactions are operating in the field, the one result that we could test with field data was supported. We used seedling demography data spanning 6 years from north-western Michigan (R. K. Kobe, unpublished data) to test if F. americana seedlings had a shorter survival time with Q. rubra than conspecific source soils. Average 1-year survival of newly germinated F. americana seedlings was 15% lower when both Q. rubra and F. americana adults were within 10 m of focal seedlings than when only conspecific adults were present. The 15% reduction in survivorship was consistent with the present study’s results (Fig. 1, 10% reduction at 77 days). Lower F. americana survivorship in the presence of Q. rubra is unlikely to arise from species differences in canopy light transmission since both tree species transmit similar irradiance (Canham et al. 1994). Thus, our glasshouse results suggested testing a pattern in the field that would not have been apparent to us otherwise. Our glasshouse results may also help to explain the widely documented increase in red maple (A. rubrum) abundance throughout its natural range in the eastern United States (Fei & Steiner 2007). The traditional explanation for red maple’s expansion is fire exclusion and increased deer browse of competing species, but red maple insensitivity to soil culturing (the least affected species by rank in five of eight soil × light combinations; Table 2) may also be contributing to its increasing abundance.

Heterospecific seedling performance was more likely to be reduced than conspecific seedlings in soil extract cultured by a particular tree species (H2)

An implicit assumption of many studies investigating NCDD processes as a mechanism of species coexistence is that a given species, whose seedlings grow or survive worse in the presence of conspecific than heterospecific adults, would more probably have an adult replaced by heterospecific than conspecific seedlings. However, this assumption may not be valid if heterospecific seedlings are affected more than conspecific seedlings in a given soil source, even if the given species’ seedlings perform relatively poorly in conspecific versus heterospecific soil. For example, in this study, even though A. rubrum seedlings were most negatively affected with soil extract cultured by A. rubrum adults, both F. americana and A. saccharum seedling performance was more strongly negatively affected (under high light). Not accounting for other aspects of species life history, the expectation is that Q. rubra and A. rubrum seedlings would be least affected by A. rubrum soils and thus the most likely species to replace an A. rubrum canopy tree when it died and created a gap. Thus, knowledge of relative benefits or costs of ‘escape’ from conspecific adults on seedling recruitment (i.e. H1) does not translate into knowledge of which species’ seedlings are favoured for recruitment near a particular tree species (i.e. Hypothesis 2).

Sterilization broadly influenced seedling performance (H3)

Surprisingly, the effect of adult culturing on seedling survival time and/or growth occurred in both non-sterile and sterile soil extracts, suggesting that both chemical and biotic processes were important. If effects of adult culturing had been due only to biotic agents, then we would have expected negligible differences among soil sources for sterilized extracts. Chemical analyses of the extracts suggest that base cation availability may be partially responsible for the species-specific feedbacks between tree adults and seedlings. Exchangeable cation availability differs under canopies of different species (Finzi, Van Breemen & Canham 1998b) and manipulation of exchangeable cations can affect seedling and sapling performance (Kobe, Likens & Eagar 2002; Bigelow & Canham 2007). With limited sample sizes in this study (c. 50% of seedlings had extract analysed), however, base cation availability was not related to total mass for any species. Variation in C:N ratios among soil sources (which was largely a function of differences in total organic C) is a less likely mechanism for observed plant–soil feedbacks since N immobilization is unlikely under the observed ratios of <25:1. Chemical analysis of soil extracts were post hoc and were conducted after 2 years of cold storage. Thus, we consider these results to be tentative. Experiments linking soil chemical factors with species-specific feedbacks between tree adults and seedlings are still needed to fully ascertain which chemical factors are responsible for creating these feedbacks.

Contrary to expectation, negative effects of soil biota were not host-specific. Three of four species had greater growth or survival in sterile extract treatments, but responses were similar regardless of which adult tree species had cultured the soil. Furthermore, five distinct Fusarium morphotypes were the primary seedling mortality agents in non-sterile extracts; these morphotypes were likely to be widespread in the field since each morphotype was cultured from seedlings grown with soil extract from various adult tree sources. Acer saccharum was the one species whose seedlings were not affected by extract sterilization, consistent with a lack of pathogen effect on the germination and viability of A. saccharum seeds (O’Hanlon-Manners & Kotanen 2006). Acer saccharum’s shade tolerance, the highest among the study species, also could explain its tolerance to soil pathogens, as shade and pathogen tolerances are positively correlated across tropical species (Augspurger & Kelly 1984; McCarthy-Neumann & Kobe 2008).

Irradiance interacted with sterilization (H4)

Increased irradiance did not ameliorate the negative effect of soil microbes on seedling performance as we had predicted. Indeed, F. americana survival time and Q. rubra growth were reduced by soil microbes without regard to irradiance level and growth of A. rubrum and F. americana seedlings were reduced by soil microbes only at high light availability. Seedlings have been hypothesized to be less affected by disease in high-light environments through five mechanisms: (i) compensation for tissue lost to disease by accumulating biomass more rapidly (Augspurger 1990), (ii) reduced exposure to disease because of more rapid lignification (Augspurger 1990), (iii) unfavourable conditions (e.g. higher temperature or decreased moisture) that lower pathogen abundance (Augspurger 1990), (iv) increased mycorrhizal colonization (Gehring 2003; Gamage, Singhakumaraand & Ashton 2004), which could suppress disease (Borowicz 2001) and (v) through the seedling response to mycorrhizal colonization switching from detrimental in low light to beneficial in high light (similar to the negative seedling response in high versus low soil resource environments, Neuhauser & Fargione 2004). The first two mechanisms probably are not operating here since they would have led to lower disease effects at high light with our methodology. However, since we minimized abiotic differences between light treatments and excluded AMF spores from extracts, this study may not have adequately tested the third through fifth potential mechanisms, which may explain contradictory results with studies that allow abiotic factors (e.g. soil moisture and temperature) and/or AMF colonization to vary between irradiance levels (Augspurger 1984; Augspurger & Kelly 1984; Hood, Swaine & Mason 2004). Additionally, growth for some species (e.g. A. rubrum and F. americana) may have been so severely constrained in shade that effects of soil microbes were not manifested.

The plant–soil chemical and microbial feedbacks that we documented here may have similar impact on seedling populations as spatial heterogeneity in light availability. The latter has greater potential to affect seedling performance, but that potential may not typically be realized. For example, mean seedling mass increased by 90–3000% across species from 2% to 22% full sun, which corresponds with light levels from understorey to large treefall gaps in Michigan deciduous forests (Schreeg, Kobe & Walters 2001). However, the potential influence of spatial heterogeneity in light levels rarely is realized. Along 200-m belt transects at four moraine sites where adult-cultured soils were originally sampled, irradiance in 800 1-m2 quadrats averaged 1.5% canopy openness, ranging from 0.5% to 6% (R. K. Kobe, unpublished data). In contrast, we would expect spatial heterogeneity in soil culturing to always be present. The two strongest soil source effects were a 43% reduction in A. saccharum performance in F. americana soil under high light and a 42% reduction in F. americana performance in Q. rubra soil. Summary effects of many of these relatively small-magnitude plant–soil feedbacks could have an equal or greater impact on seedling community dynamics than the summary effects of large-magnitude seedling responses to rare episodes of enhanced irradiance.

Potential caveats

Results from this experiment may have been influenced by the soil extraction methodology used to obtain the soil source treatments. For instance, excluding the filtrate component >20 μm in soil extracts may have eliminated some pathogenic microbes, underestimating their role. Moreover, since we excluded AMF spores, our results do not necessarily reflect the net effect of the soil microbial community on seedling recruitment patterns in the field. In addition, our study did not assess plant–soil feedbacks in the presence of interspecific competition; the absence of competitors can considerably weaken negative effects of soil pathogens (Kardol et al. 2007; Petermann et al. 2008; but see Casper & Castelli 2007). Sterilization of soil extracts also may have resulted in a nutrient pulse, which would have heightened differences between sterilization treatments. A nutrient pulse arising from sterilization is consistent with increased growth for two species in sterilized extract under high light, where nutrients could constrain growth (Kobe 2006). Counter to this interpretation, the nutrient pulse from sterilizing field soil (20% of the potting medium for all seedlings) should have overwhelmed any nutrient contribution from extract sterilization. Extract culturing by mature trees of different species may be confounded with an underlying template of soil chemistry that determines tree species occurrence. However, soils were collected from a narrow range of fertility conditions where all four species occur. Finzi, Canham & Van Breemen (1998a), Finzi, Van Breemen & Canham (1998b) also documented that the study species occur across a similar range of mineral-bound nutrients (presumably uninfluenced by species occupancy) but have divergent exchangeable nutrient pools, which could be modified by species occupancy.


Adult trees exerted species-specific effects primarily on abiotic soil properties (possibly through differences in base cation availability). These plant–soil feedbacks were specific to species pairs with conspecific and heterospecific feedbacks occurring at similar strength when comparing among all species. In addition, heterospecific seedlings were not always favoured in a given soil type suggesting that plant–soil feedbacks do not facilitate reciprocal seedling establishment. Differentiating between relative benefits to seedling ‘escape’ from conspecific adults (H1) versus which species’ seedlings are favoured for recruitment near a particular species (H2) would not have been possible if we had lumped all soils cultured by heterospecific adults into one ‘far’ treatment. Contrary to expectations, soil microbes acted as generalists and effects of the microbial treatment on seedling performance manifested more strongly in high rather than low light. Plant–soil feedbacks affected seedling performance by as much as 43%, a smaller effect than from enhanced irradiance, but the summary effects of constant plant–soil feedbacks may have an equal or greater impact on community dynamics than the larger response to rare episodes of canopy gap formation. The plant–soil feedbacks documented here are not consistent with the J–C hypothesis that conspecific density dependence would constrain species from becoming sole dominants, thereby promoting species coexistence. However, our results strongly suggest that plant–soil feedbacks influence temperate community dynamics through differential effects on seedling performance via pairwise species interactions that are largely mediated through species chemical effects on soils.


We thank Mindy McDermott, Sam Tourtellot, Renee Pereault, Amanda Gevens, Jennifer Hunnell and Melissa McDermott for their invaluable assistance. We also thank Michael Walters, John Klironomos, Andrew Jarosz, David Rothstein, Tom Baribualt, Ellen Holste and two anonymous referees for comments on earlier versions of this manuscript and Andrew Finley for statistical help. This research was funded by the National Science Foundation (DEB 0235907) and the Michigan Agricultural Experiment Station.