Forest herbs occurred in all hedgerow types: remnant hedgerows attached to forest, regenerated hedgerows attached to forest, and isolated remnant hedgerows. At least 25 taxa occurred in each type. The most frequent forest herbs in hedgerows were Circaea, Smilacina and Podophyllum, all animal-dispersed. Of the 17 taxa that were in forests but not in hedgerows, almost all were in only one or two forest samples; only two taxa (Prenanthes and Hepatica) were sufficiently frequent in forests that their absence suggests poor ability to colonize or persist in hedgerows.
Such broad representation of forest herbs in hedgerows, even in regenerated hedgerows less than 60 years old, was surprising, especially given the narrow widths of our hedgerows (most < 7 m). Minimum widths of 8 or 12 m were proposed as necessary for supporting populations of forest herbs in New Jersey hedgerows (Forman & Godron 1986; Burel & Baudry 1990; Forman 1991). Two forest interior species, Asarum canadense and Thalictrum dioicum, that were notably absent from hedgerows (Ð 5 m wide) in Quebec, Canada (Boutin & Jobin 1998), in our study were present in one hedgerow each. Fritz & Merriam (1993) reported that the hedgerow environment in Ontario, Canada, was not suitable for the growth of experimental transplants of two forest herbs (Sanguinaria canadensis and Carex brunnescens) and suggested that existing populations of these and other species in hedgerows were forest relicts. Forest herbs in hedgerows in England were often sparse, except in remnant ancient hedges (Pollard et al. 1974; Peterken & Game 1981).
Our hedgerows, in Tompkins County, New York, differed in several ways from those studied in more heavily disturbed, agricultural landscapes. The majority of the hedgerows we surveyed (25 of 32) were directly connected to remnant forests, which contained many herbs. While these forests have had tree cutting and other disturbance in the past, they were not old-field stands or tree plantations. One potential reason for the relative richness of forest herbs, in the 25 attached hedgerows, was that distance from remnant forest was ≤ 90 m. Most of the hedgerows we sampled (21) were themselves remnant and thus likely to contain relict populations of forest herbs. And the 11 regenerated hedgerows were not planted, and contained mature trees, unlike heavily-managed shrub hedges (Boatman et al. 1994), which would be less likely to support forest herbs. Even the isolated hedgerow samples were within 0.1–0.5 km of old forest, in contrast to regions of Europe with little remnant forest (e.g. Grashof-Bokdam 1997). Time may also play a role; major clearing for agriculture began < 200 years ago in Tompkins County, and much of the land has already been abandoned and reverted to forest (Smith et al. 1993), in contrast to the ancient agricultural lands of Europe and the more open field landscape of New Jersey. Thus the fragmented forest ecosystem of central New York is more likely to have retained both nearby source populations of forest herbs and the forest animals that disperse their seeds.
Hedgerows as habitat corridors
Both colonization and the ability to maintain populations are important aspects of corridor function. The frequency of forest herbs in our regenerated attached hedgerows provides strong evidence for colonization. Their richness and abundance were surprising, as colonization of the regenerated hedgerows appears to have happened in a relatively short time; the sites were treeless fence lines in 1938. Much of the colonization probably occurred in the last 20–25 years, after the habitat became forest-like. The possibility of a similar delay due to habitat preferences of the plant and/or an animal dispersal agent has been suggested for the delayed invasion of post-agricultural stands by the forest tree Fagus grandifolia (Marks 1995).
Colonization was likely to have been from source populations in the adjacent forest, because compositional similarity for regenerated hedgerows was significantly associated with the adjacent forest when compared to random hedgerow/forest pairs. Animal-dispersed forest herbs were frequent in the regenerated hedgerows; interestingly, however, all dispersal modes were represented. In post-agricultural regenerated forests in north-eastern USA, Matlack (1994a) also found that forest herbs with ingested seeds were frequent, suggesting good dispersal ability.
If hedgerows are corridors for the movement of forest species, forest herb frequency should decrease with distance from source (Burel & Baudry 1990; Brunet & Von Oheimb 1998). The decreases with distance from attached forest, which we found for both richness and similarity of species composition, further support the concept that the forests were sources of seed input. This was true both in regenerated hedgerows and in attached remnants, indicating that colonization can be important even in remnant hedgerows. Two of the five most common forest herbs, the vertebrate-dispersed Arisaema triphyllum and Smilacina racemosa, declined in frequency and abundance with distance from attached forest. Interestingly, in the sampled segments farthest from forest (60–90 m) in our regenerated hedgerows, which were most likely to show dispersal limitation, taxa with short-distance dispersal modes (ant, ballistic and none) were not underrepresented when compared with the proportions of taxa with long- and short-distance modes in the attached forests.
It could be argued that decreases in richness and similarity with distance from adjacent forest were due to poor habitat quality farther into the hedgerow, rather than dispersal limitations. That hedgerows are suitable habitat, or corridors, for forest herbs is supported by the observation of individuals of some species that were seen flowering during the sampling. It is unclear from our study to what extent shade, protection from wind, and other microclimate factors are important for hedgerow forest herbs. Forman (1991) commented that no forest herb populations in a New Jersey survey extended continuously into hedgerows from the forest more than 50 m, and suggested that this corresponded to the boundary of a forest-like microclimate. However, percentage cover of forest herbs in our attached hedgerows showed no pattern with distance across 0–90 m from forest that would suggest such a difference in microclimate. The high proportion of taxa found at least once in the 60–90-m segments indicates that the hedgerows could support a variety of forest herbs.
Habitat suitability for long-term population maintenance is a second important aspect (in addition to colonization) of corridor function. Evidence for the potential long-term persistence of forest herbs in hedgerows was the richness and abundance of forest herbs in remnant hedgerows, especially the isolated samples. Since the Tompkins County area was cleared for agriculture during the 1800s (Smith et al. 1993), the isolated remnant hedgerows were at least 100 years old, and it is possible some or many populations were relicts from the original forest. The absence of certain forest herbs from regenerated hedgerows (Polygonatum spp., Caulophyllum thalictroides, Aster divaricatus and Thalictrum dioicum) suggests that these may be poor colonizers that were present in attached and isolated remnants as relict populations, although larger sample sizes are needed to verify the pattern.
Overall, for remnant and regenerated hedgerows, habitat suitability for forest herbs as measured by richness and abundance was not strongly affected by width or orientation. Most of our hedgerows were 5–7 m wide, and only one was > 11; it is likely that extremely narrow and much wider hedgerows would show an effect. This contrasts with the relationship found for width in New Jersey hedgerows (Burel & Baudry 1990). It should be noted that our sampling transects were 2-m wide along the midline of the hedgerow, in order to keep the area constant for comparisons of richness and abundance; due to species–area relations, surveys across the entire hedgerow would be more likely to find an effect of width on species number. The lack of width response for richness or abundance may be because our hedgerows were much narrower than the 10–50 m extent of edge effects, as seen in forest edges, for microclimate and forest herb abundance (Matlack 1993, 1994b). Forest edge effects do not necessarily reduce forest herb numbers; in some stands edges can be richer than forest interior (Matlack 1994b). Matlack (1994b) found both positive and negative forest edge effects for different forest species, suggesting that some may find a hedgerow habitat favourable, while others may be limited in hedgerows as much by environment as dispersal.
Comparisons among hedgerow types
Our sampling design was predicated on clear differences among the hedgerow types: remnant attached hedgerows, which presumably contained relict populations as well as more recent colonists, should have had more taxa and greater cover of forest herbs than either regenerated hedgerows (primarily colonists) or isolated remnants (primarily relicts). Surprisingly, we found that richness, abundance and similarity to forest were not significantly lower in regenerated hedgerows, nor were isolated samples particularly low in richness or abundance, relative to the attached remnants. A potential explanation for the lack of statistical differences in richness among the three types is high variability and small sample size, especially of isolated hedgerows. In addition, there are several reasons for attached remnants to be lower than expected, for regenerated hedgerows to possibly contain relict patches of forest herbs in addition to colonists, and for isolated samples to potentially have colonists as well as relicts. Moreover, richness in attached remnants will not be as great as the number from regenerated plus isolated hedgerows, because a proportion of the species colonizing attached remnants would already be present as relicts.
Relict populations might be low in some remnant hedgerows, due to tree cutting or other disturbance. Examination of aerial photographs from 1954 and 1968, in addition to the 1938 and 1980 photographs used for choosing sites, showed that more than a third of the remnants (attached and isolated) had periods when tree cover had been sparse along the sampled sections. Although those hedgerows were not particularly low in richness or abundance, an unknown number of forest herb patches may have been lost. Another potential reason for there not being significantly more herbs in remnant attached hedgerows than in regenerated ones is that within the 2-m wide sampling transect, substantial areas in remnants were often unavailable due to large tree stems and stumps, stone heaps and deer paths. However, stone piles and animal paths were also present in some regenerated hedgerows; overall there was no strong difference among the three types.
Forest herb richness in our regenerated hedgerow samples might be higher than expected due to the presence of relict patches. Because the field edges appeared to be long-standing boundaries unlikely to have ever been ploughed, the possibility of long-term persistence cannot be excluded. For relict patches to occur in regenerated hedgerows today, populations or clones would have had to persist through a prolonged period in the open. Open conditions would have begun with forest clearing sometime in the 1800s and continued until a tree canopy developed along the fence line after the 1930s, a period of 50–150 years. As the seeds of forest herbs are not generally known for their longevity, it is unlikely that a seed bank could have persisted. We presume relict patches were generally rare in regenerated hedgerows, because we restricted our sampling to species characteristic of mesophytic forest and not open fields. But given the mean of < 7 taxa, the occasional relict patch of an additional species in some regenerated hedgerows could have an effect.
Isolated samples could have had more herbs than expected if some were colonists. However, we do not know to what extent relict populations vs. colonists contributed to species richness or the abundance of herb patches in any of the hedgerow types. Since colonization should be reduced in sites isolated from forest, it may be that with larger sample sizes the tendency for isolated hedgerows to have fewer species (with one rich outlier excluded) would become significant. The lower similarity to forest for isolated hedgerows is consistent with reduced input relative to attached remnants, but low similarity may be caused as much by spatial variation in composition simply due to distance or soils as by direct effects of isolation on colonization. Although richness in isolated hedgerows was not significantly lower than in attached, a few species were absent from isolated sites. Arisaema triphyllum may be the best example of a forest herb that appears to require colonization to maintain hedgerow populations, as it was common in forests, attached remnants and regenerated attached hedgerows, but declined with distance from forest and was absent from the seven isolated remnants.
It is unclear to what extent long-distance colonization may have contributed to the isolated hedgerows, which were 0.1–0.5 km from the nearest remnant forest. Most had become connected to forest via a hedgerow network at some point since the 1930s, which could be a factor if colonization along hedgerows is important, across distances of 0.3–1 km. Among the animals that disperse forest herbs with ingested fruits are thrushes and catbirds, turtles, racoons and deer (Hoppes 1987; Matlack 1994a). These animals are generally not restricted to deep forest habitat, in fact catbirds and the mammals frequently use hedgerows, but it is not known how far they would disperse seeds. While interior forest birds may respond to agricultural land as a barrier, hedgerows as strips of forest may facilitate movement of some bird species (Forman & Godron 1986), and it is likely that in the autumn migration period forest thrushes cross open fields. Herbs dispersed by adhesion could be carried in the fur of animals using the paths that often extended down the middle of the hedgerows. Relative to ingested and adhesive fruits, and dust-like spores, seeds of forest herbs dispersed by wind, ants or no specific mechanism would be less likely to colonize over distances as long as several hundred metres (Matlack 1994a; Grashof-Bokdam 1997; Brunet & Von Oheimb 1998). However, since most populations in isolated remnants could be relicts, and regenerated hedgerow samples were within 90 m of forest, it may not be surprising that there were no significant differences in the three hedgerow types in the proportions of taxa with long- vs. short-distance dispersal modes.
Circaea (C. lutetiana = quadrisulcata, and C. alpina), with animal-dispersed bristly fruits, was the most common forest herb in all three types of hedgerow. A species of forest Circaea was also noted in hedgerows in France (Burel & Baudry 1990), and Circaea quadrisulcata occurred in post-agricultural forests unconnected to older stands in eastern USA (Matlack 1994a). Circaea lutetiana, Caulophyllum thalictroides and Maianthemum canadense were among the forest herbs found in hedgerows in an extensive survey of the flora of a variety of habitats in Quebec (Jobin et al. 1996). Clearly Circaea has the ability to use hedgerows as habitat corridors, both for colonization and establishment. It is interesting that although it is not found in open fields, Circaea showed a positive response to light and was most abundant close to edges in Matlack (1994b) forest samples, rather than being a species of deep forest interior.
In contrast to the abundance and apparent dispersal ability of Circaea was Prenanthes (P. altissima and P. alba), wind-dispersed species that were common in our forest samples but never found in the hedgerows. Prenanthes was also frequent in Matlack's (1994a) survey of old forests but only seen in one of his successional stands. The absence of Prenanthes in our remnant, as well as regenerated, hedgerows suggests that habitat factors, rather than simply dispersal, may limit Prenanthes in hedgerows and post-agricultural forests.
Further work addressing distance effects in portions of hedgerow farther from forest, and in hedgerows isolated by greater distances, is needed to determine the extent to which the hedgerows provide habitat corridors in a landscape of forest fragments. Which factors, such as poor seed dispersal, or failure to establish and survive, limit the ability of certain forest species to colonize hedgerows successfully? Sampling in developing secondary post-agricultural stands adjacent to embedded hedgerows would help show whether hedgerows provide significant source populations for recolonization, in regions where open fields are now returning to forest.