Seed dispersal by animals is a critical interaction in ecosystems worldwide, benefiting plants by moving seeds away from their parents to high-quality microsites, protecting them from predators, and enabling colonization of new areas (Howe and Smallwood 1982; Schupp et al. 2010). Globally, more than 11,000 plant species rely on myrmecochory, or seed dispersal by ants, as their primary dispersal mechanism, with temperate forest ecosystems being hotspots of myrmecochore diversity (Beattie and Culver 1981; Handel et al. 1981; Lengyel et al. 2009, 2010). Myrmecochore plants have adaptations that facilitate interactions with ants, producing seeds with an elaiosome, which is a lipid-rich appendage that serves as food for ants (Fischer et al. 2008; Boieiro et al. 2012). Myrmecochory is sensitive to anthropogenically driven change, with ant populations, plant communities, and their interactions easily altered by disturbances or environmental shifts (Prior et al. 2015; Warren et al. 2017). Contemporary temperate forests experience widespread disturbances, including historical clearing and habitat fragmentation, leading to the creation of edge habitat (Flinn and Vellend 2005). Disturbance impacts on myrmecochory likely affect forest understory resilience, or the ability of understory plant communities to persist, passively recover, or actively recover in altered environments (Arias et al. 2023; Buono et al. 2023).

Myrmecochores are a significant component of the understory in forests of eastern North America (NA), with up to 30–40% of herbaceous plant species being myrmecochores in some locations (Beattie and Culver 1981; Handel et al. 1981). Myrmecochory in eastern NA forests is dominated by ants in the genus Aphaenogaster (Ness et al. 2009), which rapidly discover seeds released from fruits and transport them to their shallow forest-floor or log nests (Morales and Heithaus 1998; Ness et al. 2009; Lubertazzi 2012; King et al. 2018). After removing elaiosomes, ants deposit intact seeds inside or just outside their nests (Ruhren and Dudash 1996; Morales and Heithaus 1998; Ness et al. 2009), at distances that prevent overcrowding (Pudlo et al. 1980; Kalisz et al. 1999), and to locations hidden from predators and likely conducive for establishment (Culver and Beattie 1978; Smallwood 1982; Ruhren and Dudash 1996). Aphaenogaster ants seemingly persist in forests without myrmecochores, while populations of myrmecochores are less clumped and more abundant in the presence of ants, suggesting that their populations are enhanced by seed dispersal (Mitchell et al. 2002; Bronstein et al. 2006; Ness et al. 2009; Canner et al. 2012; Meadley-Dunphy et al. 2020).

Northeast NA has undergone significant landscape changes over the last few centuries (Williams 1989; Flinn and Vellend 2005; Thompson et al. 2013). Most contemporary forests are secondary growth that have regenerated from historical clearing for agriculture or timber and differ from intact remnant forests in soil characteristics, size, and connectedness (Foster et al. 1998; Kupfer and Kirsch 1998; Dupouey et al. 2002). Contemporary forests are highly fragmented, creating a patchwork of small and isolated forests surrounded by landscapes used by humans, including for agriculture (Drummond and Loveland 2010; Perry et al. 2022), residential development (Hall et al. 2002; Mockrin et al. 2013), and power line clearings (Eldegard et al. 2015; Richardson et al. 2017). Forest fragmentation negatively impacts biotic communities by altering abundance, distributions, and composition of organisms (Fahrig 2003; Hansen et al. 2005; Larsen et al. 2008; Allen et al. 2013; Nicholson et al. 2017), including diminished myrmecochore populations (Bellemare et al. 2002; Mitchell et al. 2002; Sorrells and Warren 2011; Griffiths and McGee 2018). Biotic interactions are also affected by forest fragmentation, including leading to direct or indirect decreases in mutualist interactions, such as seed-dispersing ants and myrmecochores (Mitchell et al. 2002; Warren et al. 2015b; Parker et al. 2021; Buono et al. 2023; Arias et al. 2023; Siegel et al. 2024).

Myrmecochory is vulnerable to forest fragmentation because ants move seeds over short distances, and isolation prevents myrmecochores from recolonizing new patches after previous disturbances (Buono et al. 2023). Additionally, fragmentation creates edge habitat, which causes further challenges to myrmecochory (Mitchell et al. 2002; Parker et al. 2021). Edge habitat differs from forest interiors in its greater exposure to surrounding conditions, known as “edge effects” (Murcia 1995), which can sometimes extend into forest interiors (Lindenmayer and Fischer 2013; Haddad et al. 2015). Forest edges receive more direct sunlight from open canopies, which influences soil moisture, temperature, and vegetation structure, including increases in invasive vegetation (Kapos 1989; Williams-Linera 1990; Brothers and Spingarn 1992; Matlack 1993). Matlack (1993) found abiotic edge effects to extend 50 meters (m) into the forest, while vegetation responses can extend deeper, up to 100 m towards the interior (de Paula et al. 2016). Forest organisms may be susceptible to altered edge conditions, with responses depending on the species’ sensitivity to change and the extent to which habitat conditions are altered by exposure to the surrounding landscape (Saunders et al. 1991; Bernaschini et al. 2020). Myrmecochory is sensitive to disturbance, including direct impacts on myrmecochores or indirect impacts through changes in seed dispersal via altered Aphaenogaster abundance, distribution, and behavior, or changes to antagonist interactions with seeds (Ness, 2004; Rodriguez-Cabal et al. 2012; Warren et al. 2015a, b, 2017; Prior et al. 2015; Buono et al. 2023).

Previous work in NA temperate forests has found that myrmecochore presence and abundance, along with seed dispersal, are limited in forest edges (Jules & Rathke, 1999; Ness, 2004; Ness and Morin 2008; Warren et al. 2015b; Parker et al. 2021). However, the mechanisms underlying changes in myrmecochore presence or abundance, and in seed dispersal, at forest edges have not been fully explored, particularly in the context of variation in surrounding land-use type. Myrmecochores may be directly affected by abiotic edge conditions, as their distributions are determined by soil conditions, including soil moisture and pH (Mayer et al. 2005; Warren et al. 2010; Griffiths and McGee 2018), or indirectly affected by increased competitive interactions with faster growing shade-intolerant plants, including invasive understory plants or shrubs (Coblentz 1990; Cronk and Fuller 1995; Vitousek et al. 1996; Westbrooks 1998).

Myrmecochores may also be affected if dispersal by Aphenogaster presence or abundance is altered (Mitchell et al. 2002; Bronstein et al. 2006; Ness et al. 2009). Aphaenogaster prefer low to moderate soil moisture (Warren et al. 2010; King et al. 2018), and they are sensitive to temperature (Smallwood 1982; Wittman et al. 2010; Lubertazzi 2012) and to other ants, including non-native ant species (Zettler et al. 2001; Carney et al. 2003; Rodriguez-Cabal et al. 2012; Warren et al. 2015a). Aphaenogaster are less abundant at forest edges than in the interiors, often with reduced dispersal at edges or in disturbed areas (Pudlo et al. 1980; Ness and Morin 2008; Banschbach et al. 2012; Parker et al. 2021). Also, their nests are often located in forest interiors with more conducive nesting conditions (Ness, 2004; Warren et al. 2015b), which could affect myrmecochore distribution away from forest edges (Jules 1998; Jules and Rathcke 1999; Ness and Morin 2008; Warren et al. 2015b; Parker et al. 2021).

Antagonistic interactions with seeds could also be higher, limiting myrmecochory at forest edges. Rodents, some other ant species, and an invasive slug are known antagonists to myrmecochory in northeast NA. Rodents are granivores, and Ness and Morin (2008) found increased rodent pressure at forest edges. In NA, ants other than Aphaenogaster infrequently interact with seeds, disperse them ineffectively, or damage them (Ness et al. 2009). Other ants can reduce dispersal by directly negatively interacting with seeds or indirectly by reducing the abundance or dispersal of Aphenaogaster. Invasive ant species can negatively affect the dispersal of ant-dispersed seeds in NA forests (Ness, 2004; Ness et al. 2004; Ness and Bronstein 2004; Rodriguez-Cabal et al. 2012; except see Prior et al. 2015), with Ness (2004) finding increased invasive ants at forest edges that antagonistically interact with seeds. An invasive slug, Arion sp., is commonly found in northeastern NA forests and effectively removes elaiosomes from myrmecochorous seeds, causing seed damage and reducing dispersal by Aphaenogaster (Meadley-Dunphy et al. 2016). Several studies have found higher slug abundance and damage in disturbed forests of northeast NA (Meadley-Dunphy et al. 2016; Kiel et al. 2020; Buono et al. 2023; Tan et al. 2025), including in forest edges (Parker et al. 2021).

Surrounding landscape types create distinct conditions at forest edges, and edge-effect intensity is related to the contrast between the forest and the surrounding landscape (Matlack 1993; Mesquita et al. 1999; Noreika and Kotze 2012). Residential landscapes, for example, are often highly different from neighboring forests due to impervious surface cover (Yang et al. 2021), with edges often shifting towards urbanized plant communities (Airola and Buchholz 1984), decreased tree diversity (Polyakov et al. 2005), and altered hydrology (Ehrenfeld and Schneider 1991). In agricultural edges, chemical inputs and physical disturbance from tillage may remain post-abandonment and continue to permeate nearby forest patches (Koerner et al. 1997; Duncan et al. 2008; de Jong et al. 2008; Dyer 2010; Didham et al. 2015; Wyngaard et al. 2016; Peddle et al. 2024). Agricultural surrounding landscapes also promote the presence of grasses and exotic plant species at edges and alter tree diversity (Boutin and Jobin 1998; Fridley et al. 2009; Ribeiro et al. 2019). Other landscape types, such as more natural areas with minimal management, are less conspicuous. In the northeast, common low-maintenance surrounding landscape types that create edge habitats include power line clearings that transect forested areas and old fields that were once managed. Low-maintenance landscapes have higher plant species richness than other surrounding landscape types and may be less contrasting with forests than residential or agricultural development (Çoban et al. 2019). However, they still alter forest communities and cause stress for wildlife through decreased suitable habitat and increased predation (Chasko and Gates 1982; Eldegard et al. 2015; Richardson et al. 2017). Given that myrmecochory is sensitive to environmental conditions (Warren et al. 2017), myrmecochorous partners and processes are expected to vary across forests depending on the surrounding landscape type (Parker et al. 2021).

We intend to explain variation in the effects of edges on myrmecochory by uncovering the influence of different surrounding landscape (SL) types at forest edges and interiors, and how these effects are mediated by habitat conditions. At forest edges and interiors surrounded by common SL types (low-maintenance, agricultural, and residential) in northeast temperate forests, we observed organisms and interactions involved in myrmecochory, including myrmecochore plant presence, Aphaenogaster abundance, abundance of antagonists, seed dispersal, and seed damage. We also assessed environmental conditions that may influence organisms and interactions, including abiotic soil conditions, non-myrmecochore vegetation, habitat for ground invertebrates, such as log availability, and canopy cover, and we determined relationships between these conditions and myrmecochores, Aphaenogaster, and seed dispersal. We predict that (i) forest edges will limit myrmecochorous partners, increase antagonists, hinder seed dispersal, and increase seed damage compared to interiors, (ii) low-maintenance, agricultural, and residential SLs will have different influences on myrmecochorous partners and interactions in the entire forest patch and at edges, and (iii) these impacts on myrmecochory will be mediated by variation in environmental conditions.

Study sites and transects

We conducted our study at nine eastern deciduous forest sites in south-central New York and northern Pennsylvania from early June to late July of 2024, while myrmecochores were present and Aphaenogaster were active. We selected three forest sites with low-maintenance surrounding landscapes (SLs), three with agricultural SLs, and three with residential SLs (Online Resource 1, Fig. S1, Table S1). Sites within each category were spatially distributed to minimize confounding regional factors and encompass an accurate representation of the region (Online Resource 1, Fig. S2). The landscape types were first identified from Google Earth aerial imagery and then selected after field visits to confirm SL edges. Low-maintenance SL sites consisted of a pipeline clearing, a power line clearing, and an isolated section of an old field. These clearings were thinner and more diverse in vegetation than the other two SL types, and they were intended to represent the least contrasting landscape relative to forests. Agricultural SL sites included an active crop field and two pastures. Residential SL types ranged from a condo-style complex to a low-density, suburban neighborhood. Some forest sites had multiple SL types, and we identified the sampling locations based on proximity to the SL of interest.

Our sites had a minimum patch size of 90 acres of contiguous forest (Online Resource 1, Table S1) to ensure an appropriate distance between edge and interior transects. Matlack (1993) found edge effects on microenvironmental variation to extend 50 meters (m) from the landscape boundary and myrmecochory has previously been measured > 50 m from the edge to represent forest interior conditions (Ness and Morin 2008; Parker et al. 2021; Buono et al. 2023). Edge transects were placed within 30 m of the boundary and extended 50 m into the forest. Interior transects were placed at least 100 m (< 200 m) away from the end of the edge transects and extended 50 m in length. The transect replicates were placed between 30 and 50 m apart and did not cross each other (Online Resource 1, Fig. S1).

The start and end of all transects were marked, and along each 50 m transect, we delineated 5 m sections to indicate the boundaries of plots. Plots began at the start of each transect, were 25 m² in area, and replicates were spaced 5 m away from each other. Along the right side of the transect, we created five plots for sampling vegetation and ant habitat (Type 1 plots). Five adjacent and identical plots on the left side of the transect were used to sample ground invertebrates and to measure seed dispersal and soil abiotic variables (Type 2 plots) (Online Resource 1, Fig. S3).

Environmental surveys and sampling

We recorded data for three vegetation layers along transects. Within 1 m² quadrats in all four corners of Type 1 plots, we recorded the percent cover of understory plants (ferns, grasses, forbs, and tree seedlings) and shrubs based on visual estimates of the amount of ground covered by each vegetation type out of 100% of the quadrat. We also measured the percent cover of movable rocks, and logs (> 3 inches wide), as these are habitat for Aphaenogaster (Lubertazzi 2012; King et al. 2018) and other invertebrates. Quadrat measurements were averaged together to represent percent cover at the plot level. We also recorded the occurrence of invasive forbs, invasive shrubs, and myrmecochores within the plot (presence = 1, absence = 0). Canopy openness was measured using a densiometer at the plot level, and tree diameter at breast height (DBH) was measured for all trees in plots with stems > 10 cm and combined for total DBH. Type 1 plots at all sites were sampled one time between June 3rd and June 21 st, when myrmecochores in the region could be identified by their flowers or leaves.

We took soil measurements at the center of Type 2 plots at one time (late June-late July). We measured soil pH and moisture using a soil probe (Kelway HB-2 Soil Acidity and Moisture Tester, Pacific Star Corporation, Texas) and temperature using a soil thermometer, both pressed about 5 inches into the soil. We also took one soil sample per transect to evaluate soil moisture, organic matter (SOM), and carbon and nitrogen content. We chose a representative location away from the base of trees and within 15 m of the closest end to the forest edge. We cleared litter and dug < 10 cm into the soil to represent the organic and topsoil layers. Soil sample processing details are provided in Text S1 (Online Resource 1).

Invertebrate sampling and seed dispersal

Seed depots were placed in the center of Type 2 plots once per site on non-rainy days between June 27th and July 29th (1 depot per plot, 5 depots per transect, 30 depots per site). Fruit dehiscence, seed drop, and dispersal occur from late May through July for myrmecochores in this region (Gordon et al. 2019; Tan et al. 2025). Depots were constructed with an index card placed on a flat part of the ground, a tray (petri dish) placed on the index card, eight myrmecochorous seeds with intact elaiosomes inside the tray, and a wire mesh cage (7 × 12 × 12 cm with 1.3 cm² openings) secured with gardening staples. Seeds of Sanguinaria canadensis were sourced from a local park (Vestal, NY) with permission and frozen at − 80 °C to prevent establishment. Previously frozen seeds are no less preferred by Aphaenogaster (Servigne and Detrain 2008). The wire cages were used to exclude larger seed predators (rodents), so that only the seed-dispersal abilities of Aphaenogaster and damage by antagonistic invertebrates were represented. While rodents can have large effects on seeds (Ness and Morin 2008; Kwit et al. 2012), previous work in this area suggests interactions are infrequent (Meadley-Dunphy et al. 2020; Parker et al. 2021; Buono et al. 2023).

Eight seeds were placed in depots in the morning, when fruits likely drop (Ohkawara et al. 1997; Turnbull & Culver 1983), and depots were evaluated 24 h later to determine how many seeds were removed and how many seeds were damaged by slugs. Most dispersal of myrmecochore seeds occurs via Aphaenogaster and happens within 24 h (Prior et al. 2015; Meadley-Dunphy et al. 2020), because Aphaenogaster are active throughout the day and night, whereas slugs are largely active at night (Tan et al. 2025). Seed damage was recorded when elaiosomes were absent on the remaining seeds, and previous research shows that complete elaiosome removal is due to Arion sp. slugs (Meadley-Dunphy et al. 2016; Parker et al. 2021; Buono et al. 2023; Tan et al. 2025).

Pitfall traps were placed in two diagonal corners of Type 2 plots (10 traps per transect; 60 traps per site; 360 traps across all sites) to measure Aphaenogaster and other ground invertebrate abundance. Traps were placed in the morning on non-rainy days between June 27th and July 29th, when Aphaenogaster is active (Gordon et al. 2019), and they were collected 24 h later. Pitfall traps were made from 266mL, clear, plastic cups (diameter ~ 9.2 cm), dug into the soil, flush with the ground, and filled (~ 1 inch) with soapy water. We placed a wire mesh cover on top of pitfall traps with 1.3 cm² gaps for invertebrates to fall through and to keep larger organisms out. Contents from both pitfall traps from a plot went into one vial with ethanol. Aphaenogaster and Arion sp. slugs from pitfall traps were identified and counted, and all non-Aphaenogaster ants were counted as “other ants”.

Data analysis

All analyses were performed in R (Version 4.3.2). We performed linear mixed effects models (LME; “nlme” package; Pinheiro and Bates 2000; Pinheiro et al. 2023) with SL type (low-maintenance, agricultural, residential) and sampling location (interior, edge) as fixed factors and site as a random effect for normally distributed variables (canopy gaps, total understory cover, other understory (forbs and tree seedlings) cover, soil moisture (probe and sample), soil pH, soil temperature, soil carbon, and soil nitrogen). Some of these variables were log-transformed before running models to meet assumptions of normality (Online Resource 1, Table S2). Generalized linear mixed models (GLMMs; “glmmTMB” package; Brooks et al. 2017) were performed for other variables, with a Beta distribution for percent (proportion) variables (grass, ferns, SOM, logs, and rocks), a Gamma distribution for total DBH, a zero-inflated Poisson distribution (truncated) for seed variables, and a negative binomial distribution for abundance variables (total number of Aphaenogaster, other ants, and slugs per plot). Myrmecochore and invasive plant occurrences were assessed using G-tests (“DescTools” package; Signorell 2024).

Next, we performed a correlation analysis (“corrplot” package; Wei and Simko 2021) for all variables across all sites to identify environmental variables that correlated with myrmecochory variables to inform choices for which variables to include in a path analysis (Online Resource 1, Fig. S4). We included environmental variables that were significantly correlated with any myrmecochory variables in two principal component analyses (PCA) to produce composite vegetation and soil variables, respectively (Online Resource 1, Fig. S6, Table S3).

Next, we performed a path analysis (“lavaan” and “semPlot” packages; Rosseel 2012; Epskamp 2022) to uncover relationships among environmental variables and myrmecochory variables. We finalized the most logical flow of a path and considered the analytical power of the number of replicates (n = 264) when determining the maximum number of relationships to include. Some plots were excluded due to missing data for some variables. We included fewer than 17 relationships in the path analysis to maintain at least 15 replicates per parameter. This 15:1 threshold ratio exceeds the recommended minimum of 10 observations per parameter, and it permits a stable and reliable model structure (Grace 2006). We structured the paths to reflect the collinearity (represented by non-directional arrows) among seed interaction variables, dispersal, and damage, which are not independent because ant and slug activity overlap and damage has been shown to reduce dispersal (Meadley Dunphy et al. 2016). We also included collinear paths among environmental and vegetation variables that we predict affect myrmecochory (VegPC1, SoilPC1, and VegPC2). We then created direct paths to test our hypotheses that environmental and vegetative variation affects myrmecochory partners (Aphaenogaster and myrmecochores) and antagonists (slugs), and variation in those organisms affects seed dispersal and damage, which may also affect myrmecochore occurrence. We included slugs because they are the strongest interacting antagonist with seeds that we have observed in this region (Meadley-Dunphy et al. 2016; Parker et al. 2021; Buono et al. 2023; Tan et al. 2025), but did not include other ants to limit the number of variables and retain analytical power in the path.

Variation across surrounding landscape types and location

Organisms involved in myrmecochory varied in response to the surrounding land use (SL) type and sampling location (interior, edge) (Fig. 1, a-d). Forest edges had fewer Aphaenogaster (P = 0.005), and fewer occurrences of myrmecochores (P < 0.001), with no myrmecochores present at agricultural edges, although SL type was not statistically significant (Online Resource 1, Table S2). Viola sp. were the most dominant myrmecochores found in transect plots, followed by Trillium sp., and Uvularia sp. (Online Resource 1, Table S1). Slug abundance was highest at agricultural edges (interaction: P < 0.001), and other ants were highest at agricultural and residential edges (interaction: P < 0.001). Seed dispersal and seed damage did not vary among SL type or sampling location significantly, but there were trends for higher dispersal and lower damage at residential edges and high damage at agricultural edges (Fig. 1, e, f).

Vegetation structure (Fig. 2) and soil characteristics (Fig. 3) varied depending on SL type and location, while ant habitat, canopy cover, and tree DBH did not (Online Resource 1, Fig. S5). Most vegetation variables were higher at edges, and particularly at agricultural edges. Total understory cover (forbs, saplings, grasses, ferns) (interaction: P = 0.002), grass cover (P = 0.003), and other understory cover (forbs, saplings) (P < 0.001) were all highest in agricultural SL sites, particularly at edges (Online Resource 1, Table S2 for statistics). Invasive forb occurrences were also highest at agricultural SL sites and lowest at residential SL sites (P < 0.001), with a trend for higher occurrences at some edge types. Invasive shrubs were most common at agricultural edges and least common near residential edges (P < 0.001), with more occurrences at edges than interiors overall (P < 0.001). Fern cover had the opposite pattern to all other vegetation, with edges having the largest negative effect on cover at agricultural SL types (P < 0.01; Fig. 2c).

Edges had drier soils than interiors (P < 0.001; Fig. 3a), except edges near low-maintenance SL type (P < 0.001). Edges also had more basic soil pH than interiors, with no differences among SL types (P = 0.008; Fig. 3b). SOM, C and N were lower at edges near agriculture and higher at edges near low-maintenance SL types compared to interiors (P < 0.001; Fig. 3d-f).

Influence of environmental structure on myrmecochory

VegPC1 represented total understory (loading = 0.95) and other understory (loading = 0.74), VegPC2 represented ferns (0.69), low grass (−0.61), and low invasive forbs (and shrubs) (−0.57). SoilPC1 represented moisture (0.90), SOM (0.76), and low pH (−0.83) (Online Resource 1, Table S3). We did not include SoilPC2 in the path analysis as it had no significant relationships with myrmecochory variables, to retain analytical power.

In the path analysis, VegPC1, representing total understory and other understory cover (Online Resource 1, Fig. S6, Table S3), had a significant positive, moderate-strength relationship (β = 0.464) with slug abundance (Fig. 4; Online Resource 1, Table S4). VegPC2, representing fern cover, low grass cover, and low invasive forbs had a low to moderate positive relationship with Aphaenogaster (β = 0.280). SoilPC1 (high soil moisture (sample), SOM, low pH), had a weak positive relationship with myrmecochores (β = 0.154) and a weak negative relationship with Aphaenogaster (β = −0.211). SoilPC1 covaried with VegPC1 (β = 0.416) and VegPC2 (β = 0.432). Seed damage and dispersal had a moderate, negative covariance (β = −0.491) with each other. Slugs did not have a significant relationship with seed damage, but there was a significant negative path from seed damage to myrmecochores (β = −0.260). Aphaenogaster had a weak, positive relationship (β = 0.193) with seed dispersal, but seed dispersal did not have a significant relationship with myrmecochores.

Our findings support previous studies in North American temperate forests, showing that forest edges are not conducive to the persistence of myrmecochore plants (Jules & Rathke, 1999; Ness 2004; Ness and Morin 2008; Warren et al. 2015b) and generally harbor lower abundances of mutualist Aphaenogaster ants, while supporting organisms with antagonist interactions with seeds, such as slugs and non-mutualist ants (Ness 2004; Ness and Morin 2008; Banschbach et al. 2012; Parker et al. 2021). We found that the surrounding landscape type affected edges differently, but these effects did not permeate into forest interiors. Generally, the most significant edge effects on organisms involved in myrmecochory occurred in forests adjacent to agricultural landscapes, where there were no myrmecochores, low Aphaenogaster abundance, and high slug abundance. Environmental conditions, including soil conditions and vegetation, also differed between edges and interiors, with the most pronounced effects at agricultural edges. Variation in environmental conditions directly influenced the organisms involved in myrmecochory and appears to be more influential in myrmecochore limitation at edges than the lack of seed dispersal by ants. Myrmecochory is an important interaction in temperate forests, and myrmecochores are often depauperate in contemporary forests (Matlack 1994a; Bellemare et al. 2002; Mitchell et al. 2002; Sorrells and Warren 2011; Griffiths and McGee 2018; Buono et al. 2023). Our study suggests that preserving forest interiors and minimizing edge effects are important for myrmecochore persistence and improved seed dispersal, and that restoring myrmecochores should be prioritized in interiors, especially away from forest edges adjacent to agricultural areas.

Forest interiors were more intact and consistent in environmental conditions and myrmecochory than edges. Our findings support prior research by demonstrating that forest interior habitats are buffered from surrounding landscape effects (Matlack 1993; Ness and Morin 2008). Myrmecochores were more frequent in plots with higher soil moisture and SOM and lower soil pH, conditions often associated with forest interiors and consistent with previous findings of myrmecochore habitat preferences (Warren, 2010; Griffiths and McGee 2018). Soil in agricultural edges lacked these conditions, which could limit the persistence and establishment of myrmecochores. Forest edges are likely drier due to increased exposure to sunlight and wind (Riutta et al. 2012; Li et al. 2018) and may exhibit decreased acidity from nearby soil inputs and invasive plant species that alter pH (Trammell et al. 2021; Garvey et al. 2023). Despite the surrounding landscape type, forest interiors had similar myrmecochore frequencies, which were still generally low, as all of these sites are secondary forests, having regenerated from clearing in recent history. It is well known that myrmecochores are not resilient to previous forest clearing and are absent or in low abundance in regenerated forests (Bellemare et al. 2002; Mitchell et al. 2002; Griffiths and McGee 2018). The most dominant myrmecochores in the plots were Viola sp., which are not remnant forest indicator species, persisting in disturbed forests (Griffiths and McGee 2018; Buono et al. 2023). The other two main species we recorded in transects were remnant forest indicator species, Trillium sp. and Uvularia sessilifolia, which were mostly observed in interiors, and less frequently at residential and natural edges. We recorded other myrmecochores in forests (but not in our transects), including Erythronium americanum, Asarum canadense, and Hepatica sp..

Aphaenogaster were also inhibited overall at forest edges. Aphaenogaster distributions are influenced by soil moisture, preferring moderate conditions that may not always overlap with optimal, high moisture conditions for myrmecochore plants (Warren et al. 2010; King et al. 2018), which we also observed in this study, as some surrounding landscape types were more suitable for one partner than the other. Drier soil conditions may contribute to lower abundances of Aphaenogaster at edges. Aphaenogaster also had positive relationships with fern cover and negative relationships with grass and invasive forb cover. Fern cover was higher in forest interiors and is often indicative of intact, shady forests (Della and Falkenberg 2019). Forest edges are known to have increased invasive plant cover because of their openness and direct exposure to the neighboring landscape (Hansen and Clevenger 2005; Allen et al. 2013). Aphaenogaster nest under rocks, in rotting logs, in leaf litter cover, and shallow in the soil; dense grass and invasive plants at forest edges may make the ground less conducive to nesting sites by altering soil structure via root traits (Smallwood 1982; Angers and Caron 1998; Lubertazzi 2012; Nunez-Mir and McCary 2024). We didn’t find differences in log, rock or leaf litter cover between interiors and edges.

Aphaenogaster benefit myrmecochore populations through seed dispersal (Mitchell et al. 2002; Bronstein et al. 2006; Ness et al. 2009), but we did not find that seed dispersal varied among surrounding landscape types or forest locations. However, dispersal followed the general pattern of Aphaenogaster abundance, trending higher in interiors in sites surrounded by low-maintenance and agricultural landscapes, but not sites surrounded by residential landscapes, and we found a significant positive relationship between Aphaenogaster abundance and seed dispersal in the path analysis, suggesting that abundance is linked to dispersal, and both are influenced by environmental conditions. Another reason abundance and dispersal may not be linked is that dispersal is affected by slug damage, which is correlated. Thus, locations with high Aphaenogaster abundance and high slug abundance could have lower dispersal if slugs damage seeds before ants disperse them (Meadley-Dunphy et al. 2016; Tan et al. 2025).

In our path analysis, we also found no significant link between seed dispersal and myrmecochore presence, but myrmecochore presence was more strongly influenced by environmental conditions. While myrmecochory increases plant fitness by increasing population growth and distribution (Kalisz et al. 1999; Canner, 2012; Prior et al. 2015), the mutualism is not obligate; plants can persist without ants, and ants are present in forests without myrmecochores (Mitchell et al. 2002; Gorb et al. 2000; Giladi 2006; Warren et al. 2010, 2021; Bronstein et al. 2006). The path results suggest that the absence or low abundance of myrmecochores in these forests and edges may not be primarily limited by ants (Parker et al. 2021; Buono et al. 2023). Rather, myrmecochores are absent or low in abundance overall in these forests, likely due to a lack of recruitment from previous land use disturbance, and due to environmental conditions limiting these plants in disturbed forests (Koerner et al. 1997; Bellemare et al. 2002; Flinn and Vellend 2005; Dyer 2010). While we didn’t find dispersal to be significantly lower at forest edges, we did find that Aphaenogaster were less abundant, and that abundance significantly influenced dispersal in the path analysis. This suggests that for myrmecochores that can establish at edges, population growth and distribution may be limited if they do not receive the benefits of ant-mediated dispersal, including dispersal into more conducive habitats away from edges. Viola sp. was the dominant group at edges, and they are diplochorous (ballistic and ant-dispersed) (Beattie and Lyons 1975), potentially making them less reliant on high ant dispersal activity at edges.

Forest edges bordered by agriculture had the most contrasting environmental conditions and impacts on myrmecochory compared to interiors and other edge types. Edges near agricultural landscapes were the most depauperate in myrmecochore plants, with zero observations in transects. We found that agricultural landscapes create forest edge conditions that likely hinder the establishment of myrmecochore plants, with low soil moisture and SOM, high pH, and altered vegetation structure. These distinct forest edges may result from the influence of agricultural landscapes, with large open areas dominated by crop and pasture plants that may inhibit native species or facilitate invasives (Fridley et al. 2009). Also, agricultural land, and potentially adjacent edges, have likely experienced direct soil manipulation, which minimizes the water-holding capacity and carbon content of soil (Murty et al. 2002). It is important to note that agricultural practices vary drastically in northeast NA, including choices of chemical additives, soil manipulation techniques, harvested crops, and whether they have livestock. These choices contribute to different impacts on forest edges, supporting or discouraging certain types of organisms depending on the form of landscape manipulation (Boutin and Jobin 1998; Didham et al. 2015). In the agricultural edges we visited, we found high cover of grasses and invasive plants, which may compete with myrmecochores (Gehlhausen et al. 2000; Fridley et al. 2009). However, our path analysis results do not demonstrate a direct connection between vegetation structure and myrmecochores. Furthermore, time since disturbance could be affecting myrmecochore populations at agricultural edges. If these edges were recently disturbed, then the limited recovery time would dictate an early successional stage (Matlack 1994b) and potentially explain the absence of myrmecochores (Flinn and Vellend 2005).

Edges near agriculture additionally supported the highest abundances of the antagonist slug. Slug abundance was positively related to total understory cover, which was highest at edges near agriculture, suggesting that these conditions are favorable for the invasive slug. This aligns with other studies that describe Arion food preferences for decaying vegetation and their use of grass for habitat (Beyer and Saari 1978). Arion are generalists and they exploit crops as a food resource, which may be another reason for their success at agricultural edges (Hammond and Byers 2002; Hommay 2002). Additionally, agricultural practices may have aided their establishment if they were introduced via organic soil inputs. Although we did not find a link between slug abundance and seed damage in this study, other studies have shown that Arion consume elaiosomes and prevent dispersal services by Aphaenogaster (Meadley-Dunphy et al. 2016; Kiel et al. 2020), so their high abundance in forest edges near agriculture raises concern for native understory diversity in these locations.

A potential reason for the apparent lack of a relationship between slug abundance and seed damage is that we measured slugs in pitfall traps, which may not be an accurate approach to measure their abundance and attraction to food resources. We did find a negative relationship between slug damage and myrmecochore occurrence in our path. Slug damage may contribute to low myrmecochore occurrence, or both may correlate with other factors related to degraded edges that are not conducive to myrmecochores. There are likely other factors correlated to slug presence and seed damage that contribute to myrmecochore limitation, including factors in the path (soil) and factors that we did not include. Since seed damage and soil conditions did not have particularly strong relationships with myrmecochores, other factors not included in the path (e.g., land use-history) may contribute to their distribution. However, since seed damage was associated with myrmecochore absence, slugs’ potential contribution to myrmecochore persistence is worth further investigation, especially in agricultural edges where high numbers of Arion could contribute to seed damage and limit seed dispersal (Meadley-Dunphy et al. 2016; Kiel et al. 2020).

Residential edges, while still disturbed with respect to soil conditions, are somewhat more resilient, having lower slug abundance than agricultural edges and non-significant trends for higher Aphaenogaster abundance and seed dispersal and lower seed damage. Also, residential edge effects were the lowest regarding changes in vegetation compared to interiors. Myrmecochores were lower at edges than interiors, with a trend for higher myrmecochore presence at residential edges than agricultural edges. Overall, forest patches near residential landscapes had low understory cover, including low grass cover and invasive plants, compared to other site types, which may contribute to Aphaenogaster’s tolerance of this edge type (Smallwood 1982; Lubertazzi 2012). Low understory cover found both at edges and interiors near residential landscapes could be a result of vertical barriers that interrupt wind-dispersed seeds and limit recruitment in forests (Hahs and McDonnell 2007), pressure from deer browsing (Faison et al. 2016), or trampling by humans (Kuss and Hall 1991). We sampled near residential landscapes with varying levels of urbanization, and it is possible that suburban edges promote fewer forest edge disturbances than more urbanized settings because of variation in building density, paved surfaces, and human presence (Wang and Yang 2022). Residential edges also had higher soil pH than interiors, and we found that myrmecochores corresponded to lower soil pH.

Of all the edge habitats, those surrounded by low-maintenance landscapes trended towards having the most myrmecochores, suggesting these edges are also somewhat resilient. However, the dominant representative myrmecochore in plots were Viola sp., and their presence in low-maintenance edge plots may not accurately represent suitability for more sensitive myrmecochore species. Soil moisture and SOM were high at these edges and more similar to conditions in the interiors than in other edge types, suggesting these edges are likely more conducive to more sensitive myrmecochores than agricultural edges. While myrmecochores were present, low-maintenance edges had low Aphenogaster abundance. Although lower contrast than agricultural sites, low-maintenance edges had higher grasses, invasive forbs, and shrubs than interiors (and compared to residential edges), which may not be conducive to Aphaenogaster nests. Antagonist slug abundance was also lower at low-maintenance and residential edges compared to agricultural edges, potentially because of comparatively low understory cover (Beyer and Saari 1978). While edge environmental structure differed from interiors and there were limited Aphaenogaster, our results suggest that myrmecochores are more likely to establish near low-maintenance landscapes, although still being limited compared to interiors.

Conservation implications

Our findings have conservation implications for myrmecochory in contemporary, disturbed, northeastern NA temperate forests. Secondary forest interiors > 100 m away from edges supported consistent representation of myrmecochore plants, more Aphaenogaster, and fewer antagonists. As a result, our study suggests that preserving forest interiors and reducing edges allows for improved myrmecochory, regardless of the surrounding landscape type. Given that myrmecochores generally do not recover well from previous forest clearing (Bellemare et al. 2002; Mitchell et al. 2002; Sorrells and Warren 2011; Griffiths and McGee 2018), their populations need to be augmented by planting individuals or completely restored (Ruhren and Handel 2003; Ines and Prior unpublished data). Restoration or augmentation in forest interiors would allow for intact seed dispersal, promoting population establishment by increasing the speed and extent of propagule movement across the forest floor. If considering myrmecochory restoration near forest edges, approaches should differ depending on the surrounding landscape type. Edges near low-maintenance landscapes can be considered, as they are likely to respond most positively to myrmecochore restoration efforts due to low numbers of invertebrate antagonists, higher occurrences of myrmecochores, and supportive soil characteristics (high soil moisture and SOM). However, a lack of ants may slow recovery. Edges near residential landscapes may have Aphaenogaster available to disperse seeds but may not have suitable environmental conditions. Given the general lack of understory, protective measures such as fences may be required to limit disturbance from humans and deer in restoration plots (Kuss and Hall 1991; Ruhren and Handel 2003; Faison et al. 2016). In edges with low soil moisture, it would also be beneficial to add downed woody material, as this may increase Aphaenogaster nesting opportunities (Smallwood 1982; Angers and Caron 1998; Lubertazzi 2012; Nunez-Mir and McCary 2024) and soil moisture (Haskell et al. 2012), thereby increasing suitability for myrmecochore establishment (Mayer et al. 2005; Warren et al. 2010; Griffiths and McGee 2018). Restoration of forest edges near agricultural landscapes should be avoided, given the many limitations to establishing healthy myrmecochore populations.

Taken together, we found the most significant effects on organisms and processes involved in myrmecochory at forest edges, particularly at edges near agriculture, where notable environmental conditions are likely limiting. Our results corroborate findings from other studies (Watling and Orrock 2010; Noreika and Kotze 2012) that demonstrate landscape matrix types differentially influence biotic communities at forest edges, and we show that different edge effects apply to myrmecochory. Our results inform myrmecochore restoration or augmentation approaches, considering landscape context in anthropogenically disturbed northeastern temperate deciduous forests. We highlight that some land-use types are less disruptive to myrmecochory at forest edges than others, and that forest interiors are generally resilient to different surrounding land-use types, including those with the largest edge effects.