1We tested the enemy release hypothesis for invasiveness using field surveys of herbivory on 39 exotic and 30 native plant species growing in natural areas near Ottawa, Canada, and found that exotics suffered less herbivory than natives.
2For the 39 introduced species, we also tested relationships between herbivory, invasiveness and time since introduction to North America. Highly invasive plants had significantly less herbivory than plants ranked as less invasive. Recently arrived plants also tended to be more invasive; however, there was no relationship between time since introduction and herbivory.
3Release from herbivory may be key to the success of highly aggressive invaders. Low herbivory may also indicate that a plant possesses potent defensive chemicals that are novel to North America, which may confer resistance to pathogens or enable allelopathy in addition to deterring herbivorous insects.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
The environmental damage imposed by exotic plants is substantial (Vitousek 1990). However, out of the thousands of non-native species that now occur in North America, only a small proportion has become invasive in natural communities (Williamson & Fitter 1996). The success of these highly invasive exotics is often attributed to a release from the natural enemies that are thought to keep the plants in check in their native ranges (Elton 1958; Crawley 1987), the enemy release hypothesis. Studies comparing herbivores or pathogens on exotic plants in their native and introduced ranges have generally supported this hypothesis (Wolfe 2002; Mitchell & Power 2003). However, studies comparing damage to exotics and their native congeners or confamilials, which argue that exotic plants that escape their enemies enjoy an advantage relative to the native plants against which they are competing, sometimes show greater herbivory on exotics than on native congeners (Keane & Crawley 2002 and references therein; Agrawal & Kotanen 2003).
Successful biological control programmes, in which herbivores have been introduced to control invasive weeds, are consistent with the enemy release hypothesis. Examples such as the control of tansy ragwort Senecio jacobaea following introduction of the cinnabar moth Tyria jacobaeae and other herbivores (McEvoy et al. 1991) and control of the floating fern Salvinia molesta by the weevil Cyrtobagous salviniae (Room 1990) indicate that lack of herbivory was key to those plants’ initial invasiveness. Because of the expense involved in screening control agents for potential non-target effects, biological control is a solution reserved for only the most problematic plants, especially those able to invade intact communities such as rangelands and natural areas. For these highly invasive plants, release from enemies may be an important mechanism driving their success. However, introduced plants vary substantially in their impact on native communities (Ortega & Pearson 2005) and the inclusion of less invasive weeds in studies testing the enemy release hypothesis may obscure the importance of herbivore release for a few highly invasive species.
Herbivore damage to a non-native plant depends on the availability of herbivores that recognize the new plant as food and can cope with its defences. It would be surprising if, through time, native herbivores did not incorporate exotic plants into their diets. Indeed, the accumulation of a herbivore fauna can be rapid, as Strong et al. (1977) have shown for sugarcane, which was introduced throughout the tropics. Although area planted, and not time since introduction, was the primary predictor of herbivore species richness on sugarcane (Strong et al. 1977), we might nevertheless expect that plants that arrived in North America in the 1600s might now support a larger herbivore fauna and suffer more herbivore damage than those introduced more recently.
In the present study, we examined herbivore damage to the leaves of 30 native species, as well as 39 exotic species differing in degree of invasiveness and date of arrival in North America. We predicted that the exotics, as a group, would suffer less herbivory; however, given the antecedents in the literature, we expected that this difference would be small. More importantly, we predicted that among the exotics, the most invasive plants would be those that had recently arrived in North America and that suffered the least leaf herbivory.
Plants were sampled from natural areas within the Ottawa-Gatineau region of Canada from late May to September 2003. Once a population of a given plant was discovered, we haphazardly chose 20 individuals by pointing with a metre stick while averting our eyes. One leaf was taken randomly from each plant. Leaves were brought back to the laboratory to be scanned for damage. Herbivore abundance can be variable in space and time (e.g. Root & Cappuccino 1992), so sampling several populations of each species would have been ideal. However, because the unit of replication for our analyses was the species, we traded spatial coverage within species for wider taxonomic coverage. Our compromise was to try to find two populations of each species; however, some of the less common plants were sampled from one site only. In total, 69 species from 15 plant families were sampled, consisting of 39 exotic and 30 native species (Appendix S1 in Supplementary Material). Each family was represented by at least one native and one exotic species. Because plants have different phenologies, we collected leaves when the plants began to produce seeds, with the exception of coltsfoot Tussilago farfara, which flowers before the leaves emerge. Sampling date for each species was recorded as a modified Julian date, with 1 May as the starting date. If the two populations were not sampled on the same date, the average date was used.
Areas of damage on each leaf (holes made by chewing herbivores, as well as mines and galls) were identified and labelled with a fine-point felt-tip marker. The leaves were then scanned using imaging software (Scion Image Beta 4.02, Scion Corporation, Frederick, Maryland, available online at http://www.scioncorp.com) to measure total leaf area and total area damaged. For leaves with damage along the edge, approximate areas were drawn relying on the symmetry of the leaf to estimate how the leaf may have looked before damage. The average percentage damage over all leaves was calculated for each species.
For the introduced species, dates of introduction were obtained from a variety of sources (Appendix S1). For 28 species, we found references to dates in the primary literature, or in books that summarized the primary literature and herbarium records (e.g. Rousseau 1968; Leighton 1970, 1976). For the remaining species we found dates on Internet websites sponsored by universities, governmental organizations and non-governmental organizations involved in weed control, plant conservation or restoration. When only an approximate time-frame was reported for an introduction, a date corresponding to the middle of the period was assigned. For example, species introduced in ‘the 1600s’ were assigned 1650; those from ‘the late 1700s’ were assigned 1775. We verified that species with dates prior to 1850 were recorded in Gray's Manual of Botany, second edition (Gray 1856), the earliest edition to which we had access (the first edition was published in 1848). Four species did not appear in Gray (1856) despite claims on websites that they had been introduced in the 1600s or 1700s. These were omitted from analyses using date as a predictor variable. All species reported in the literature as having been introduced after 1850 were absent from Gray (1856).
Invasiveness indices were derived for the exotic plants based on their inclusion and ranks in lists found on the websites of state, provincial and regional governmental organizations (Appendix S2 in Supplementary Material) in the north-eastern United States and eastern Canada. This approach is similar to that taken by Mitchell & Power (2003), with the exception that we avoided using state noxious-weed lists, which tend to include mainly agricultural weeds as opposed to natural-areas weeds. Fourteen lists were found. Two indices of invasiveness were calculated: the number of lists on which each species occurred and the average rank of each species on the lists on which it occurred. Most of the lists provided a three-, four- or five-tiered ranking system. We recalibrated all lists so that the rankings ranged from 1 (least invasive) to 3 (most invasive). We did not use any lists that did not differentiate between levels of invasiveness. Species not appearing on any list were assigned a rank of zero. One list, that of the USDA Forest Service Eastern Branch (USDA Forest Service 2003), included plants from the entire geographical region we were considering. We used this list as a third, binary index of invasiveness (a species was on it or not). Also, as the Forest Service list was divided into four invasiveness categories, we created another binary variable that considered a species as invasive if it appeared in the top two invasiveness categories (high and moderate) and not invasive if it appeared on that list as a low-level invasive or was absent from the list.
Difference in herbivory between natives and exotics was examined using anova. Sampling date and family (as a random factor) were then added to the model. Percentage herbivory was arcsine-transformed prior to analysis.
Linear regression was used to determine the relationship between time since introduction of the exotic species and arcsine-transformed herbivore damage. Time since introduction and percentage herbivore damage were then used as the independent variables in linear regressions with the two continuous indices of invasiveness (occurrence and average rank), as determined from the lists, as dependent variables. Multiple regression was used to investigate the combined effects of time since introduction and herbivore damage. Logistic regression was used to assess the relationship between the same two continuous independent variables and two binary invasiveness variables (whether or not included in the Forest Service list and high/low ranking in this list).
The simple linear regressions were first performed on the data set for all 39 exotic species (or 35 when the independent variable was time since introduction). Because some of the families were represented by only a single species in the exotics data set, an appropriate test of the effect of family could not be performed. On the other hand, including more than one species from the better-represented families might inappropriately inflate the degrees of freedom and overestimate the strength of the relationships. To circumvent this problem, we subsampled the main data set by randomly choosing one species from each family, running correlation analyses on the four variables of interest (time since introduction, arcsine-transformed damage and the two continuous invasiveness indices), and then repeating this procedure 40 times. Mean and standard error of the correlation coefficients are reported.
All analyses were performed using JMP-IN version 5.1 (SAS Institute, Cary, NC, USA).
Exotic species experienced significantly lower leaf herbivory than native species although the proportion of the variance explained by plant origin was low (Fig. 1; F[1,67] = 5.02, P = 0.028; = 0.056). Plant family added no predictive value to the model ( = 0.057). Likewise, sampling date contributed little to the model ( = 0.08). Contrary to one of our predictions, plants that had been introduced earlier did not suffer greater herbivory (Fig. 2; F[1,33] = 0.269, P = 0.607; = 0.008).
Plants that were ranked as more invasive suffered less damage than weak invaders (Fig. 3a; F[1,37] = 22.59, P < 0.0001; = 0.362). Similarly, plants that occurred on more lists also suffered less damage than those occurring on few lists or none (Fig. 3b; F[1,37] = 7.20, P = 0.011; = 0.140). These relationships were not driven by the outlying point to the far right on the damage axis, which represents Erysimum cheiranthoides; excluding this species still resulted in a significant relationship between rank and damage (F[1,36] = 11.24, P = 0.002; = 0.217) as well as between occurrences on the lists and damage (F[1,36] = 5.68, P = 0.023; = 0.112). In logistic regressions, mean percentage herbivore damage was a significant predictor of both inclusion in the Forest Service list and high ranking in that list (inclusion in list, χ2 = 9.962, d.f. = 1, P = 0.0016, R2 = 0.271; high ranking in list, χ2 = 4.842, d.f. = 1, P = 0.028, R2 = 0.122).
Plants introduced more recently were marginally more invasive than earlier arrivals (Fig. 4a; F[1,33] = 3.13, P = 0.086; = 0.059) and they appeared on more lists (Fig. 4b; F[1,33] = 5.16, P = 0.030; = 0.109). Time since introduction was not a significant predictor of inclusion on the Forest Service list (logistic regression, χ2 = 0.104, d.f. = 1, P = 0.747, R2 = 0.003); however, it was a weak predictor of high vs. low ranking on that list (logistic regression, χ2 = 3.08, d.f. = 1, P = 0.079, R2 = 0.078).
In a multiple regression, both damage and time since introduction were significant predictors of invasiveness ranking (herbivore damage, F[1,33] = 24.19, P < 0.0001; time since introduction, F[1,33] = 4.93, P = 0.034; model, F[2,33] = 14.29, P < 0.0001, = 0.447). Herbivore damage and time since introduction were also significant predictors of occurrences on the lists (herbivore damage, F[1,33] = 7.80, P = 0.008; time since introduction, F[1,33]= 5.99, P = 0.020; model, F[2,33] = 7.06, P = 0.003; = 0.263).
Analysis of the subsampled data sets, in which one species was randomly chosen from each family, revealed that the significant negative relationship between damage and invasiveness observed in the analysis of all 39 species was not driven solely by the over-representation of certain highly invasive families whose species suffer little herbivory. All 40 of the correlations between arcsine-transformed damage and invasiveness rank were negative, and 70% were significant at α = 0.05. The mean of the correlation coefficients was −0.546 (± 0.025). The relationships between damage and occurrences on the lists were less strong (mean r = −0.332 ± 0.147) and only three were significant at α = 0.05, although all were negative. Time since introduction and damage were never significantly correlated (mean r = 0.064 ± 0.178). Time since introduction was always negatively correlated with occurrences on the lists (mean r =−0.309 ± 0.175); however, only 20% of the correlations were significant. Likewise, time since introduction was always negatively correlated with invasiveness rank (mean r = −0.3336 ± 0.119), although only 7.5% of these correlations were significant at α = 0.05.
As predicted by the enemy release hypothesis, the exotic species in our study were less damaged than their native counterparts. However, plant origin, native or exotic, explained only a small (5.6%) percentage of the variance in herbivore damage. Although some exotic species were virtually damage-free (for example, field pennycress Thlaspi arvense and pale swallow-wort Vincetoxicum rossicum), others (corn gromwell Buglossoides arvensis and wormseed wallflower Erysimum cheiranthoides) suffered more leaf damage than any of the native plants. The large variance in herbivory corresponded to our expectations; given the mixed results of previous tests of the enemy release hypothesis (Keane & Crawley 2002 and references therein; Agrawal & Kotanen 2003), we did not expect the exotics to enjoy a strong advantage in terms of escape from herbivory. Plant family explained a similarly low proportion of the variance in herbivory (5.7%), which was surprising in light of the strong phylogenetic effect on herbivory observed in previous studies (e.g. Agrawal & Kotanen 2003).
For the exotic species in our sample, low leaf herbivory was strongly correlated with invasiveness in natural communities (r2 = 0.36 for the relationship between herbivory and invasiveness ranking). Both the number of invasive-species lists on which a species appeared, as well as its average invasiveness ranking on those lists, were higher for species that experienced minimal leaf damage. Although we sampled each exotic species at only two sites, additional casual observations on these same species have indicated that the most invasive plants, Vincetoxicum rossicum and V. nigrum, Polygonum cuspidatum and Alliaria petiolata, are rarely attacked by herbivores. Likewise, we have never seen some of the less invasive plants, such as Barbarea vulgaris, without herbivore damage. The high leaf herbivory of the less invasive exotics suggests that herbivores might be contributing to the control of these plants and preventing them from becoming more invasive in natural areas.
We recognize that a plant's presence on an invasive-species list is somewhat subjective. We also acknowledge that the lists we used may not be independently derived; presence of a species on one list might prompt a neighbouring state or province to include that species as well (no species appeared on lists of states or provinces in which they did not occur, so pro-active listing is not an issue). To circumvent the possible non-independence of the lists, and the possibility of inappropriately inflating the invasiveness of plants occurring in several small neighbouring states, we also considered presence on a single list, the USDA Forest Service Eastern Region list, as indicative of invasiveness anywhere in the region under study. Whether we used presence/absence on that list as a binary response variable, or high rank on that list vs. low rank or absence, the pattern was similar to that found by tallying state and provincial lists: plants with less herbivore damage were more invasive.
Given that there was no anticipatory listing of species not yet present in a state or province, the number of lists that a plant species occurs on indicates the range over which it is successful in North America. Widespread plants are expected to support more herbivore species (Opler 1974; Strong 1974), which could possibly translate into higher herbivore damage. From this perspective, the lower levels of herbivory on plants occurring on several lists might be considered surprising. This suggests that low herbivory contributing to plant success outweighs the effect of wide geographical range driving herbivore accumulation.
The high leaf herbivory of the less invasive exotics in our study suggests that herbivores may be effective at preventing these plants from becoming highly invasive pests of natural areas. However, we have not demonstrated a link between herbivory experienced by exotic plants and any aspect of their performance, a crucial step for understanding the enemy release hypothesis. Escape from pathogens may also be important in allowing some exotic plants to flourish (Mitchell & Power 2003). The lack of herbivore damage experienced by our most invasive species suggests that these plants possess strong defensive chemicals, which may render them resistant to pathogens as well as to herbivores. Potent secondary chemicals may have allelopathic properties as well as anti-herbivore properties. For example, the sesquiterpene lactone cnicin in spotted knapweed Centaurea maculosa is lethal to non-adapted generalist herbivores (Landau et al. 1994) in addition to being allelopathic (Kelsey & Locken 1987). Two of the most invasive and least damaged plants in our study, Vincetoxicum rossicum and V. nigrum, produce compounds in their roots that have marked anti-fungal effects (M. Smith and J. T. Arnason, personal communication) in addition to herbivore-deterrent properties (N. Cappuccino, unpublished data), lending further credence to this hypothesis.
In contrast to our prediction, the relationship between the amount of leaf herbivory that exotic plants suffered and the length of time they have been in North America was non-significant. This agrees with earlier published observations of herbivore accumulation rates on introduced crop plants. For example, Andow & Imura (1994) found that the number of herbivores on crop plants introduced to Japan was not related to time since introduction. Herbivore accumulation on cacao (Strong 1974) and sugarcane (Strong et al. 1977) increased rapidly to a level determined by the area planted. Strong et al. (1984) identified area planted (or achieved by natural spread) and taxonomic, morphological or phytochemical similarity to native plants as the primary determinants of herbivory on introduced plants. None of the plants in our study were morphologically unusual (all were herbaceous plants or subshrubs) and none were from families absent from the native flora. We suggest that the invasive plants experiencing little herbivory in our study may be particularly well defended with chemicals that are novel to North American herbivores, much as Callaway & Aschehoug (2000) and Vivanco et al. (2004) hypothesize that some plants become invasive by virtue of their novel allelopathic ‘weapons’.
Plants in our sample that have recently been introduced to North America occurred on more lists of invasive species and were ranked as more invasive than plants introduced earlier, although these relationships were weaker than those between herbivory and the invasiveness rankings. Because we found no relationship between herbivory and time since introduction, this high invasiveness of recent arrivals cannot be due to their escape from herbivory. It is also a counter-intuitive result, as one might expect that as a plant's range expands through time, it would appear on the lists of an increasing number of states and provinces. The higher invasiveness ranking of recently introduced exotics suggests that these plants may differ from those that arrived in North America with the first European colonists. Mack (2001) suggested that the first human immigrants to a new area tend to introduce familiar food plants and medicinal plants, whereas later generations import plants from around the globe for landscaping and erosion control. If plants introduced for erosion control have, for example, more aggressive root systems, this could lead to greater invasiveness. The impact of an invasive plant might also decrease with time if its native neighbours adapt to its presence, for example by becoming more resistant to its allelopathic root exudates. Unfortunately, we have no information on whether the non-invasive plants in our survey were at one time invasive and have become less invasive through time. A final, non-biological explanation for the relationship between time and invasiveness is simply that long-standing exotics are now accepted as part of the North American flora, whereas new arrivals are more likely to catch the attention of land managers and the public.
The ability to predict which exotic plant species are likely to become highly invasive would permit land managers to combat these plants while populations were still small and ranges limited. Many traits thought to lead to invasiveness have been identified: large leaf area, shade tolerance, small seed mass, rapid maturation, ability to fix nitrogen, clonal growth and vine-forming habit (Rejmánek & Richardson 1996; Reichard & Hamilton 1997; Daehler 1998; Goodwin et al. 1999; Smith & Knapp 2001). If lack of natural enemies in North America indeed reflects novel defensive chemistry, rather than the stochastic process of losing enemies during the move, the ability to resist non-adapted North American herbivores and pathogens is another trait that holds promise as a predictive tool. Further research into how this trait could be quantified a priori, perhaps by deriving an index of phytochemical novelty, could aid decisions about which exotic plants should be the focus of management efforts before they overtake natural communities.
We thank the National Capital Commission (Greenbelt Division) and Al Tweddle of the Friends of Petrie Island for access to collecting sites. Discussions with Andrew Simons and comments from Ragan Callaway, Anurag Agrawal, Richard Mack and anonymous reviewers greatly improved the study and the manuscript. This work was supported by the Natural Sciences and Engineering Research Council of Canada (Undergraduate Student Research Award to DC and Discovery Grant to NC).