Two main approaches have been employed to study changes in invasion-promoting traits over invasion-history gradients, with advantages and drawbacks to both. In the first approach, plant traits are compared amongst different invasive species with varying residence time within a region (e.g. Hawkes (2007); Iacarella et al. (2015); Sheppard and Schurr (2019); Sheppard and Brendel (2021)). Such an interspecific approach can provide a generalised overview across multiple species and requires only general knowledge on residence time of species within a region, such as the earliest report of introduction time (e.g. at the country level). However, results of multi-species experiments might be confounded due to historic biases in the types of introductions (Sheppard and Brendel 2021) or due to discrepancies between the time of introduction to certain regions and the age of the sampled populations (but see Iacarella et al. (2015)). In the second approach, plant traits are compared within a species across different populations along gradients of invasion history (e.g. Lankau et al. (2009); Gruntman et al. (2017); Tabassum and Leishman (2018)). This intraspecific approach requires knowledge on population age (a chronosequence approach) or the identity of source populations and invasion trajectories (i.e. distance to source populations), as well as information on the possibility of multiple introductions. Regardless of the approach used, the study of divergence in invasion-promoting traits is best examined in controlled common garden experiments, which expose plants with varying invasion histories to the same environmental conditions. Such studies can reveal inherent variations in trait levels and exclude variations due to plastic responses to field conditions at different sites.

Two additional approaches should be noted due to the alternative advantages they offer to the study of trait evolution along invasion-history gradients. The first approach is the use of herbaria collections, which can provide historical samples of invasive plants (reviewed in: Meineke et al. (2018); Lang et al. (2019)). Such collections can be compared at the intraspecific level to identify divergence over time since introduction in physiological and morphological traits, such as herbivore defence compounds or plant size (Zangerl and Berenbaum 2005; Buswell et al. 2011). Herbaria records can, thus, be an important tool in the study of invasion history. The second approach is the use of selection gradients (Lande and Arnold 1983) on invasion-promoting traits, which are measured in the field in introduced populations with different invasion histories. Such field measurements of selection can provide important knowledge on the adaptive relevance of a focal trait in different populations along invasion gradients, but, to date, it has mainly been applied to study variations between native and introduced populations (e.g. Franks et al. (2008); O’Donnell and Pigliucci (2010); Colautti and Lau (2016)). However, as for other studies using field-collected measurements or samples, both approaches cannot exclude plastic responses to field conditions as an explanation for trait variation rather than evolutionary change (excluding herbaria studies that examine the genetic makeup of plants: Vandepitte et al. (2014)).

The following sections provide a review of the hypotheses suggested to explain the effect of time on the evolution of different invasion-promoting traits, including defence, growth, competitive ability and dispersal ability, focusing on studies that examined them under common garden conditions.

One of the most well-studied hypotheses to explain the success of invasive plants is the evolution of increased competitive ability hypothesis (EICA) (Blossey and Notzold 1995). This hypothesis proposes that plants in their introduced range experience reduced damage by natural enemies, such as pathogens and specialist herbivores, which selects for reduced allocation to defence traits and a consequent increase in resources available for growth and competitive ability (Blossey and Notzold 1995; Bossdorf et al. 2005). Alternatively, the shifting defence hypothesis (SDH) proposes that the loss of specialist enemies at the introduced range can result in an evolutionary shift from investment in defence against specialists (digestibility reducers such as trichomes and tannins) to defence against generalists (toxins such as glucosinolates and alkaloids) (Müller-Schärer et al. 2004; Joshi and Vrieling 2005).

Both the EICA and the SDH assume that invasive plants experience release from their specialist enemies at the introduced range. This attenuating selection pressure is not likely to change with time since introduction, except due to the unintentional introduction of specialist herbivores or pathogens (e.g. Zangerl and Berenbaum (2005); Wan et al. (2019)) or their use as biocontrol agents (Stastny and Sargent 2017). Nevertheless, several studies have shown that enemy pressure on invasive plants increases over time. In particular, more established populations have been shown to experience greater colonisation rates and attacks from local herbivores (Hawkes 2007; Brändle et al. 2008; Dostál et al. 2013; Harvey et al. 2013; Schultheis et al. 2015; Gruntman et al. 2017; but see Carpenter and Cappuccino (2005)), increased pathogen accumulation (Siemann et al. 2006; Hawkes 2007; Mitchell et al. 2010; Flory and Clay 2013) and increased negative soil feedback (Diez et al. 2010). These results imply that compared to range-edge populations, older populations might incur greater attacks even from generalist enemies. Such increased enemy load could be attributed to increases in the density and abundance of invasive plants in older populations, which could increase their attraction as hosts (Agrawal et al. 2006).

Increased enemy pressure with time since introduction might re-select for increased allocation to defence traits in plants and particularly against generalist herbivores (Fig. 1). However, we are aware of only three studies to date that examined divergence in defence-related traits along invasion history gradients under common garden conditions. Gruntman et al. (2017) found that Impatiens glandulifera plants from younger populations within the invasive range have reduced herbivore resistance against a generalist herbivore, coupled with reduced production of secondary defence compounds, to an extent similar to those of native populations. In contrast, Siemann et al. (2006) found no effect of population age on Sapium sebiferum survival in common gardens exposed to local herbivores, even though herbivory levels increased in older populations. Similarly, Harms and Walter (2021) found no effect of population age on herbivore defence against a generalist herbivore in the invasive plant Butomus umbellatus. Additional studies are, therefore, required to provide a general understanding of the effect of invasion history on the evolution of enemy defence in invasive plants. Moreover, such future studies should compare their relative allocation to defence against specialist vs. generalist enemies across populations with different invasion histories.

Figure 1.

Schematic represenation of the predicted effect of population age on divergence in different invasion-promoting traits, including positive effects (light green), negative effects (dark green) and no effect (grey).

Invasive plants are commonly associated with increases in growth rate and size at their introduced compared to native range (Pyšek and Richardson 2008; Van Kleunen et al. 2010). While such increases in growth-related traits can be a result of plastic responses to the environment, they might also be attributed to evolutionary change. The latter idea was proposed in the EICA hypothesis, which predicts that, due to an allocation trade-off, the decrease in defence experienced by plants at the introduced range is likely to result in an evolutionary increased allocation to traits related to growth (Blossey and Notzold 1995; Bossdorf et al. 2005; Joshi and Vrieling 2005). The same defence-growth trade-off might govern the evolution of growth traits within the introduced range of invasive plants. Specifically, if as suggested above, older and more established populations incur more enemy pressure that selects for increased investment in defence, these populations are also predicted to evolve decreased allocation to growth (Fig. 1).

We found seven common garden studies that explicitly investigated the effect of invasion history on the evolution of plant growth and their results provide contrasting patterns. Kilkenny and Galloway (2013) and Evans et al. (2013) provide support for the hypothesised negative effect of invasion history on plant size in studies on the invasive plant Lonicera japonica, where plants from core populations within the introduced range were smaller than their conspecifics from edge populations. In contrast, Wan et al. (2018) found that Plantago lanceolata plants from populations with a long invasion history were larger and more fecund compared to populations at the invasion front and similar results were shown by Fenesi and Botta-Dukát (2012) in Ambrosia artemisiifolia and VanWallendael et al. (2018) in Reynoutria japonica. Finally, Monty and Mahy (2009) found no effect of population age on biomass production of Senecio inaequidens and Tabassum and Leishman (2018) similarly found no effect of distance to source population on the height of Gladiolus gueinzii.

A lack of consistent results regarding the effect of invasion history on plant size might reflect the variety of selection pressures that likely act on such a fundamental life-history trait. For example, local climate across elevational and latitudinal gradients, as well as levels of primary productivity, might also change along invasion routes and exert strong selection on plant size and growth rate (Colautti et al. 2009; Monty and Mahy 2009). Thus, although the defence-growth trade-off might be key to the invasive success of many plant species, other factors can contribute to divergence in growth traits within the introduced range.

Two hypotheses were suggested to account for the evolution of competitive ability in invasive plants and can be similarly applied for divergence in competitive ability within the introduced range. First, as suggested above and following the premise of the EICA hypothesis, older populations are predicted to undergo selection for increased defence associated with decreased allocation to growth and plant size. This decrease in size is, therefore, likely to be manifested in reduced competitive ability via resource competition in older and more established populations. In contrast, populations at the invasion front are predicted to undergo selection for decreased defence and increased size and competitive ability.

The second hypothesis suggested to explain the evolution of increased competitive ability in invasive plants is the novel weapons hypothesis (NWH: Callaway and Aschehoug (2000); Callaway and Ridenour (2004)). The NWH suggests that the production of toxic allelochemicals should be selected for in introduced populations due to its enhanced negative effects on naïve native competitors compared to co-evolved ones at the native range (Callaway and Ridenour 2004; Prati and Bossdorf 2004; Abhilasha et al. 2008). Hence, this hypothesis proposes increased selection for competitive ability via interference competition in introduced populations. As for plant size and its derived competitive ability, the adaptive advantage of allelopathy might also decline with time since introduction, due to three selection pressures. First, the above-suggested need for increased production of defence compounds in older populations, might favour decreased production of allelochemicals due to a trade-off between these secondary compounds (Inderjit et al. 2011; Gruntman et al. 2017). Secondly, in older and more established populations, plants might experience a shift from competition with heterospecific to conspecific neighbours, which are usually unaffected by conspecific allelochemicals (Inderjit et al. 2011). Thirdly, with time since introduction, co-occurring native species or their mutualist soil biota might also evolve resistance to the novel allelochemicals of the invasive plants (Callaway et al. 2005; Lankau 2011; Dostál et al. 2013).

Competitive ability can be attributed to two components that were suggested to be associated with different traits (Goldberg 1990, 1996). Competitive effect, which is the ability to suppress neighbours, can be attributed to rapid resource acquisition and growth or to allelopathy; while competitive response, which is the ability to withstand competition, can be attributed to tolerance of low resource levels or to neighbour avoidance (Goldberg and Landa 1991; Cahill et al. 2005). Thus, the two aforementioned hypotheses regarding the evolution of decreased competitive ability with time since introduction via allocation to plant size or allelopathy, are mostly related to competitive effect (Fig. 1). In contrast, competitive response might be either not affected by these processes or even increase, if lower growth rate and smaller size correlate with stress tolerance (Fig. 1).

Changes in competitive ability at the introduced range over time were examined in several studies. Some of these studies used an interspecific approach and examined the competitive effect of multiple invasive species with different residence times, using either common garden experiments (Sheppard and Schurr 2019; Sheppard and Brendel 2021) or a meta-analysis approach (Iacarella et al. 2015) and their results reveal different patterns. For example, Sheppard and Brendel (2021) used a common garden experiment to study the competitive effect of 47 non-native Asteraceae species on native plants and found that species with longer residence time had stronger competitive effects. However, this study referred to plants of varying non-native status rather than strictly invasive species and the variation in their competitive effect was better explained by this status (casual vs. established neophytes, archaeophytes or native species). In contrast, Iacarella et al. (2015) performed a meta-analysis of studies that examined the competitive effect of 36 invasive species with a known residence time at the collection sites and found that competitive effect of the plants decreases with time since introduction. The results of this study provide compelling evidence for temporal shifts in the evolution of competitive ability, because they are based on introduction time of the studied population rather than the entire invaded region. However, such an interspecific approach might be confounded due to variations in competitive ability amongst species, which can be avoided by comparing competitive ability of conspecifics with different residence time.

A few common garden studies used an intraspecific approach and explored divergence in competitive ability amongst populations of the same species across invasion gradients. While some of these studies have attributed competitive ability to growth traits, such as plant height and biomass (see “Divergence in growth traits” above), we found only six studies that have explicitly examined divergence in competitive ability and all compared competitive effects on the performance of neighbours or the production of allelochemicals. Lankau et al. (2009) and Lankau (2012) provide support to the predicted decrease in competitive effect with invasion age, showing that the production of allelochemicals in invasive Alliaria petiolata declines in older populations. Similarly, Oduor et al. (2022) found that invasive Solidago canadensis plants from older populations had a lower competitive effect on native species and a greater competitive response to them, although these interactions depended on plant-soil feedbacks, suggesting that soil biota has an important role in these interactions. Evans et al. (2013) also found a reduced competitive effect in older Lonicera japonica populations under competition with conspecifics. In contrast, Huang and Peng (2016) found that the competitive effect of the invasive vine Mikania micrantha in intraspecific competition is higher in more established core populations, while Gruntman et al. (2017) found no effect of population age on allelopathic ability of invasive Impatiens glandulifera.

The lack of consistent results for the effect of invasion history on the competitive ability of the studied plants might be attributed to variations in competitive environments experienced by these plants. For example, as suggested above, invasive plants often experience a shift from inter- to intraspecific competition with time since introduction, which could select for different competitive strategies. Indeed, in this review, the three studies whose results support the predicted decrease in competitive effect used heterospecific neighbours, while a study that employed conspecific competitors found the opposite trend (Huang and Peng 2016). Further studies are, therefore, needed to differentiate between the effect of invasion history on inter- vs. intraspecific competitive ability.

As for other invasion-promoting traits, evolution of traits related to dispersal ability might also take place between different invasion stages within the introduced range. The most common hypothesis in this regard proposes that, during range expansion, higher dispersal ability is likely to be selected for in individuals arriving at the invasion front compared to core populations (Hargreaves and Eckert 2014; Hodgins et al. 2018) (Fig. 1).

The notion that dispersal ability should be selected for at range edges of invasive species has been suggested in several theoretical models (Travis and Dytham 2002; Phillips et al. 2008; Travis et al. 2009) and supported by several studies on invasive animal species (e.g. Hughes et al. (2003); Phillips et al. (2006); Lombaert et al. (2014); Ochocki and Miller (2017)). However, evolutionary shifts in dispersal ability at range edges of invasive plants have been relatively less studied, of which we are aware of only two studies that were conducted with seeds collected from plants grown under common garden conditions. These studies, by Monty and Mahy (2010) and Huang et al. (2015), provide support for increased dispersal ability (i.e. increased investment in seed pappus) at the invasion front of the invasive plants Senecio inaequidens and Mikania micrantha, respectively. A few additional studies examined dispersal ability in seeds that were collected directly from field populations and not from plants growing under common garden conditions, showing similar results (Tabassum and Leishman 2018; Robinson et al. 2023; but see Bartle et al. (2013)).

In summary, accumulating evidence provides support for different ways in which invasion-promoting traits such as defence, growth, competitive ability and dispersal might evolve in the introduced range over time. However, our review of the studies did not reveal consistent directions in divergence for most of the studied traits, which could be attributed to other selection pressures that might vary along invasion gradients. Moreover, existing studies that have explicitly explored trait divergence along gradients of invasion history are still very few, ranging from two to seven studies per trait, thus precluding our ability to reach generalised conclusions and highlighting the need for further studies on the subject. The aim of the following meta-analysis is to provide a more general overview on the subject.

To test for directional changes in the different traits along the invasion-history gradient, we used two literature review procedures. In the first procedure, we searched for studies explicitly investigating divergence in plant characteristics along invasion gradients at the introduced range, which included information on population ages. The literature was searched using two databases, Web of Science Core Collection (WOS) and Google Scholar. We first screened the literature in WOS (last accessed on 17 January 2023), using the search terms (chronosequence OR time-since-introduction OR invasion-history OR residence-time OR range-expansion OR colonization-history OR introduction-history) AND (plant*) AND (invasi*). We then complemented our search and screened the literature in Google Scholar (last accessed on 6 July 2022), using similar search terms.

Papers selected for the analysis had to meet the following criteria: (1) the study aimed to test the relationship between residence time of an invasive plant and at least one of the following traits: defence against herbivores (measured as, for example, the inverse of leaf damage or herbivore mass following feeding or the production of defence metabolites), plant growth (e.g. plant biomass or height), competitive effect (e.g. effects on the performance of native species or allelopathy), competitive response (e.g. performance of the invasive species under competition with native species) and dispersal ability (e.g. the ratio between the size of dispersal structures such as wing or pappus and seed mass); (2) a gradient of invasion history was explicitly reported in the paper, either as differences in time (generally in years, although papers that reported residence times at large geographical scales, such as country were not included) or as a distance from source to expanding populations; (3) the study reported the results of controlled experiments under common garden conditions, thus ensuring that variations amongst populations in the studied traits are the result of genetic differentiations rather than plastic responses to environmental conditions at the site. A total of 24 cases from 19 papers were included after meeting these criteria.

We carefully checked whether species were defined as invasive (rather than, for instance, alien or naturalised), based on the terminology given in the specific studies as well as in the CABI compendium digital library, invasive species section (https://www.cabidigitallibrary.org/product/qi). Moreover, in several cases, in which naturalised vs. invasive ranges of introduced species were compared, only the invasive range was used in our analysis.

In the second literature review procedure, we searched for studies investigating variation in characteristics of invasive plants across populations under common garden conditions, but that did not explicitly include data on invasion history. Instead, these data were extracted from additional sources. The literature was searched using the Web of Science Core Collection (WOS) (last accessed on 18 January 2023), using the search terms (common garden OR greenhouse) AND (population OR accession) AND (plant*) AND (invasi*). Data on the location of the collection sites used in the different studies were extracted when possible. Invasion history of the populations was obtained when possible from additional papers that studied the same populations. For other cases, this information was extracted from additional sources such as the Global Biodiversity Information Facility database (https://www.gbif.org/) and CABI compendium digital library- invasive species section. Such information was extracted only for the same locations or for nearby locations at the scale of kilometres. In cases where information on the age of certain populations was missing, such populations were excluded from the analyses. Moreover, when information was given in the literature on the location of the first introduction of the invasive plant (given mostly at the local scale, for example, city), this location was used to estimate the distance from the source population with Google Earth Pro. In such cases, distance to source population was used instead of population age in the analysis. Papers selected for the analysis had to meet similar criteria as in the first literature review procedure, with the exception that information on population ages was not provided, but could be extracted from external sources, following which a gradient of invasion history was used to compare across sites.

Using the two literature review procedures, a total of 79 cases from 62 papers were included after meeting our inclusion criteria in the final dataset of the meta-analysis (see Suppl. material 1: fig. S2 for more details on paper screening and selection process).

Data on the relationship between invasion history (population age or distance from source population) and the studied traits were extracted for each of the selected study cases. When source data were not available, the data were extracted from figures using the software GetData Graph Digitizer ver. 2.26 (http://getdata-graph-digitizer.com). In studies where several treatments were applied (e.g. water or nutrients addition or different disturbance levels), only a subset of the data, representing standardised controlled conditions, was used, such as high water availability (see Suppl. material 1: table S1 for details on specific studies).

As considerable variation could be found amongst studies in the ranges of ages or distances across the studies populations, both the invasion history and measured trait data were first transformed using z-standardisation (standardised by subtracting the mean from each value and dividing by the standard deviation). A linear regression was then performed between the measured trait values and population age/distance. The standardised slope of the regression (β) was taken as the estimated effect size and the variance of the estimate of the standardised slope (SE squared) was taken as the estimated sampling variance (see Suppl. material 1: table S1). When other geographic variables were provided for the populations, such as elevation or mean annual temperature, the estimated standardised slope and variance were generated using linear model or linear mixed model analyses, with these variables as covariates (see Suppl. material 1: table S1).

The effect of invasion history on overall trait divergence (regardless of its direction) was examined with the absolute value of the estimated standardised slope (|β|), while the effect on directional changes in traits was examined with the standardised slope (β) as an effect size. For both meta-analyses, a random-effects model was used in order to combine the estimated effect sizes from the different studies. Such random-effects models allow for both variation of effect sizes amongst studies and sampling variation within studies (Koricheva and Hayes 2018). In order to estimate model parameters, a Restricted Maximum Likelihood (REML) approach was used. In order to determine means and confidence intervals for the different trait categories, trait category was used as a moderator in the models. As some studies used the same invasive plant species or different traits were sometimes measured in the same study (see Suppl. material 1: table S1), study as well as species identity were also included as random factors in the models. Effect sizes were considered significant if the 95% confidence intervals did not overlap zero. Meta-analyses were performed using the METAFOR package v.4.4 (Viechtbauer 2010) in R-Studio v.2023.9.1.494 (Posit team 2023) and R (R Core Team 2023).

Study cases selected for the meta-analyses included information on invasion history data extracted from different sources, i.e. reported in papers (n = 36 cases) or estimated from external databases (n = 43). In addition, invasion history was measured in two ways, i.e. residence time (n = 65) or distance from source populations (n = 14) (Suppl. material 1: table S1). A meta-regression model was, therefore, conducted to assess the effect of these moderators on the magnitude of effect sizes. Here, a random-effects meta-regression was performed, in which data origin and type of study were served as moderators, study as well as species identity were included as random factors and an REML approach was employed to estimate model parameters. To interpret the significance of the chosen moderators, Q_m statistics was used to test the extent of heterogeneity explained by the moderators. The meta-regression was performed using the METAFOR package v.4.4 (Viechtbauer 2010) in R-Studio v.2023.09.1.494 (Posit team 2023) and R (R Core Team 2023).

The magnitude and significance of effect sizes may affect the publication and/or visibility rates of studies (e.g. based on the impact factor of journals) (Koricheva et al. 2013). To test for a possible publication bias in cases where the overall effect sizes in our meta-analyses were found to be significant, several approaches were used. First, we tested for temporal publication bias, examining a potential correlation between effect sizes and publication year. Additionally, we checked for a potential correlation between effect sizes and the journal’s impact factor at the year of publication. We also tested for the possibility that differences in age and distance ranges amongst studies could influence effect sizes. To that end, an LMM analysis was used, in which publication year, journal’s impact factor and age/distance ranges per study served as the independent variables and study identity and the invasive plant species served as random factors. Finally, we estimated the fail-safe numbers using the Rosenberg method (Rosenberg 2005), which indicates the number of additional studies with effect size of 0 needed to reduce the significance level of the observed average effect size to α = 0.05. This analysis was conducted using the Fail-Safe Number Calculator software (https://www.rosenberglab.net/Rosenberg2005FailSafe.html). Statistical analyses, unless indicated otherwise, were performed using JMP Pro 17.1 (SAS Inst. Inc.).

The meta-regression results indicate no significant effect of the two moderators on absolute effect sizes (Q_M = 3.193, p = 0.203, residual heterogeneity Q_E = 162.94, p < 0.001). Specifically, there were no significant differences in absolute effect size between the two types of study (residence time: mean ± SE = 0.272 ± 0.041; distance: mean ± SE = 0.334 ± 0.084, p = 0.48; Suppl. material 1: fig. S3). Yet, a non-significant trend was found for the origin of data, according to which absolute effect size values of data reported in papers was slightly higher than data estimated using external databases (reported in papers: mean ± SE = 0.369 ± 0.085; estimated from external sources: mean ± SE = 0.247 ± 0.097, p = 0.075; Suppl. material 1: fig. S3).

The chosen studies were published between the years 1994–2022. When testing for temporal bias, publication year had no effect on absolute effect size (LMM results: slope ± SE = 0.0038 ± 0.0058, p = 0.51). Nevertheless, a significant effect of the journal’s impact factor was found on absolute effect size, according to which studies with lower absolute effect sizes were published in higher impact journals (LMM results: slope ± SE = -0.039 ± 0.017, p = 0.039; Suppl. material 1: fig. S4A). Moreover, a positive correlation was found between the journal’s impact factor and the study’s total sample size (Pearson’s correlation (log-log scale): r = 0.539, p < 0.001; Suppl. material 1: fig. S4B). In addition, when time/distance ranges per study was considered, no significant effect of differences in ranges on absolute effect size was observed (LMM results (log range): slope (± SE) = 0.036 ± 0.076, p = 0.63). Finally, the estimated Rosenberg’s fail-safe number (i.e. additional number of studies with an average effect size of 0 needed to reduce the significance level of the observed average effect size to α = 0.05) was 4767, suggesting our results of overall absolute effect size are robust against possible publication bias.