Appendices / Supplementary Material

Methods M1 – Tests of regional increases and decreases. We classified species as increasing or decreasingin abundance using G-tests and replicated tests for goodness of fit (Sokal and Rohlf 1995). These non-parametric tests do not require balanced sample sizes.The single-classification G-statistic tests for changes in the abundance of particular species within each site. Across sites,the Gtotal statistic tests the significance of changes in frequency in either direction while the Gpooled statistic tests for overall net changes in quadrat frequency. The Gpooled statistic has high power for detecting systematic regional shifts even when shifts within sites lack significance.We classified species as “increasers” or “decreasers” only if their shifts in abundance were large and consistent (i.e., both Gtotal and Gpooledremained significant after Holm’s P-value correction for multiple hypothesis tests - Holm 1979).Under these criteria, 53 herbaceous species significantly decreasedand 17 increased.

Methods M2. We used ArcGIS and recent digital orthophotos to digitize houses and roads within a 5km buffer centered on each site (see Rogers et al. 2009 for detailed methods). Land cover was estimated from data derived from 30m resolution Landsat TM images from 1992 and 1993(Wisconsin DNR, 1998). Within the 5km buffers, we classified land into urban cover (impermeable surface), agriculture (activelycropped fields), grasslands (including fallow fields androad edges), or forest (all forest types combined). For each cover class we calculated ten class-based metrics, including total area, percentage of the buffer, number of patches, mean patch size, patch size standard deviation (SD), total edge, mean shape (perimeter/minimum possible perimeter) index, mean nearest neighbor, nearest neighbor SD, and mean proximity index (a gravity-based measure of patch separation). We also measured the size, shape, and proximity of eachfocal forest stand.To reduce dimensionality, we first classified the landscape variables into five subsets, saving the first eigenvector from a principal components analysis (PCA) for each subset. This yielded five synthetic variables: road and house density (Road-House PCA), urban landscape class (Urban PCA), forest landscape class (Forest PCA), agriculture landscape class (Agriculture PCA), and forest patch metrics (Patch PCA). Because the Urban and Road-House PCA variables were highly correlated (r = 0.92,P< 0.0001), we use only the Urban PCA variable. Correlations among the other variable pairs were less than 0.40, except Patch and Forest PCA (r = 0.76). We then used Spearman’s ρ to test correlations between invader-native effect sizes and the landscape and invasive abundance variables. Where these were significant, we further analyzed landscape variables to determine which had the most effect.

TableS1-Site-level associations. Nominally significant (P< 0.05) across-site nonparametric correlations (Spearmen’s ρ) and Association Sizes (from C-score analyses) between the abundances of thethree most abundant invasive exotic species and 70 native forest herb species including tests of all pairs that co-occurred in at least 3of the 94 sites (see Supplementary Fig. 1S). (Note: the threshold used in the paper was six such sites.) Such correlations occurred with nine of the 17 natives that increased over the past 50+ years and 22 of the 53 natives that declined. Note that average association sizes with both increaser and decreaser native species are negative, and that Alliaria generally had a greater effect on deceasers whereas the invasive shrubs had greater mean effects on native increasers. Boldface valuesremained significant atp< 0.05 after Holm’s correction. Increasing species tested in addition to those listed: Asarum canadense, Aster lateriflorus, Circaea lutetiana, Eupatorium rugosum, Festuca subverticillata, Hackelia virginiana, ,, Polygonumvirginianum. Decreasing species tested in addition to those listed: Actaea rubra, Adiantumpedatum, Agrimonia gryposepala, Anemone americana,Amphicarpaeabracteata, Aster macrophyllus,Astersagittifolius, Botrychium virginianum, Brachyelytrum erectum, , Dioscorea villosa, Desmodiumglutinosum, Elymus hystrix, Fragaria virginiana, Galium boreale, Galium concinnum, Geranium maculatum, , Parietaria pensylvanica, Polygonatum biflorum, Potentilla simplex, Prenanthes alba, Pyrola elliptica,Ranunculushispidus,Smilacinaracemosa, Smilacina stellata, Solidago canadensis, Trillium grandiflorum, and Trillium recurvatum.

Spearman’s ρ / Association Size
Alliaria / Lonicera / Rhamnus / Alliaria / Lonicera / Rhamnus
Native increasers:
Arisaematriphyllum / -2.09
Athyriumfilix-femina / -2.11
Caulophyllumthalictroides / -2.66
Cryptotaeniacanadensis / -2.71
Dryopteriscarthusiana / -2.61
Geumcanadense / 1.98
Laporteacanadensis / -2.69 / -3.38
Pileapumila / 0.48
Ranunculusrecurvatus / 0.49 / -3.05
Solidagoflexicaulis / -2.11
Mean: / -1.58 / -2.39 / -2.92
Native decreasers:
Anemonequinquefolia / -0.46 / 1.88
Apocynumandrosaemifolium / -2.29
Aralia racemosa / 0.72 / -2.01
Aralia nudicaulis / 0.67 / -3.10
Desmodiumnudiflorum / -2.97
Carexalbursina / -2.66
Galiumtriflorum / -3.82
Helianthusstrumosus / -2.61
Maianthemum canadense / 0.62 / 0.86
Mitelladiphylla / -2.74
Osmorhizaclaytonii / -1.94
Osmunda claytoniana / 0.83 / 0.64
Phloxdivaricata / -2.20 / -2.48
Phrymaleptostachya / -2.85
Podophyllum peltatum / 1.87 / -2.49
Polygonatumpubescens / -2.09 / -2.10
Pteridiumaquilinum / -4.43
Sanguinariacanadensis / -3.21
Saniculamarilandica / -4.63
Solidago ulmifolia / 0.77
Thalictrumdioicum / -0.43 / -2.47
Thalictrumthalictroides / -2.38
Trilliumflexipes / -2.62
Uvulariagrandiflora / -2.42
Violapubescens / -2.14
Viola sororia / -0.49
Mean: / -2.60 / -1.30 / -2.50
Ratio: / 1.65 / 0.54 / 0.86

Table S2 - Quadrat level associations. Association sizes within 1m2 quadrats pooled over all sites between each of the three invasive species and 13 of the 17 native species that increased over the past 50+ years and 41of the 53 natives that decreased. Only nominally significant associations are shown. Bolded values are significant (p < 0.05) after Holm corrections for multiple comparisons. Summary tallies from here are presented in Table 1.

Increasers / Alliaria / Lonicera / Rhamnus / Decreasers (cont) / Alliaria / Lonicera / Rhamnus
Arisaema triphyllum / -2.55 / 2.96 / Dioscorea villosa / -2.15
Asarum canadense / -2.01 / Galium concinnum / 4.35
Aster lateriflorus / 2.69 / 3.39 / Galium triflorum / 3.04 / 2.61
Athyrium filix-femina / -3.58 / Geranium maculatum / 1.87 / -2.46 / -4.22
Cryptotaenia canadensis / 4.28 / 2.34 / Maianthemum canadense / 1.93
Eupatorium rugosum / 2.66 / Mitella diphylla / -1.97
Festuca subverticillata / 2.26 / 3.51 / Osmunda claytoniana / -2.34
Geum canadense / 6.29 / Parietaria pensylvanica / 2.45 / 3.13
Pilea pumila / 2.34 / -2.71 / 4.18 / Phryma leptostachya / -6.95 / 3.60
Ranunculus recurvatus / 2.11 / Podophyllum peltatum / 1.79 / -2.05 / 3.01
Solidago flexicaulis / 3.25 / Polygonatum biflorum / -2.99 / 2.27
Decreasers / Ranunculus hispidus / 4.38
Actaea rubra / -2.62 / Sanicula marilandica / 3.85
Adiantum pedatum / -3.58 / Smilacina racemosa / 4.96 / 2.93
Agrimonia gryposepala / 5.62 / Solidago canadensis / 2.28
Amphicarpaea bracteata / -2.12 / Smilacina stellata / 2.60
Anemone americana / -2.20 / -4.40 / Solidago ulmifolia / 2.44
Anemone quinquefolia / -2.44 / Thalictrum dioicum / -2.18
Apocynum androsaemifolium / 3.26 / Thalictrum thalictroides
Aster sagittifolius / 2.50 / Trillium flexipes / -2.06
Botrychium virginianum / 3.22 / Uvularia grandiflora / -3.47
Brachyelytrum erectum / -2.18 / Viola pubescens / 4.05 / -2.16
Desmodium glutinosum / 3.93 / 2.91 / Viola sororia / 3.84 / -1.95

Figure S1. Funnel graph showing how variation in the magnitude of nonparametric Spearman rank correlations (rho) between the abundances of pairs of exotic and native species declines with increases in sample size (the number of sites where the two species co-occurred). The fitted regression line is not significant (t = -1.81, p = 0.073).

Figure S2. Distributions of effect sizes for C-pooled values of association between the three invasive species and the 70 native species. a) effect sizes between invasives and native declining species (N = 159, mean = -0.542, S.E. = 0.103); b) effect sizes between invasives and native increasing species (N = 51, mean = -0.872, S.E. = 0.163). The means of both distributions are significantly less than zero (P < 0.001) but do not differ from each other (t-test: P = 0.11). Compare to Fig. 2c.

a)

b)

1

S -