Volume 107, Issue 10 p. 1366-1374
Research Article
Free Access

Life history variation in an invasive plant is associated with climate and recent colonization of a specialist herbivore

Sophie S. Duncan

Sophie S. Duncan

Department of Geography and Biodiversity Research Centre, University of British Columbia, 217-1984 West Mall, Vancouver, BC, V6T 1Z2 Canada

Search for more papers by this author
Jennifer L. Williams

Corresponding Author

Jennifer L. Williams

Department of Geography and Biodiversity Research Centre, University of British Columbia, 217-1984 West Mall, Vancouver, BC, V6T 1Z2 Canada

Author for correspondence (e-mail: [email protected])

Search for more papers by this author
First published: 11 September 2020



Spatial variation in selective pressures can lead to intraspecific variation in life history, favoring some life histories and constraining others depending on the vulnerability of life stages. We examined how spatial variation in herbivory and climate influences flowering size and the occurrence of semelparity (reproducing once) versus iteroparity (reproducing multiple times) in the introduced range of an invasive plant, houndstongue (Cynoglossum officinale). Houndstongue is a short-lived, semelparous perennial in its native range. In its introduced range, we previously documented increased rates of iteroparity and a higher median threshold flowering size compared to the native range. We hypothesized that the recent introduction of a specialist biocontrol insect (a root-boring weevil, Mogulones crucifer) would decrease threshold flowering size and reduce the proportion of iteroparous plants because M. crucifer preferentially attacks large individuals and may reduce overwinter survival.


We surveyed 24 sites across the northwestern United States to quantify the frequency of semelparity versus iteroparity and to estimate weevil abundance and used demographic data collected from six sites to estimate median threshold flowering size.


We found that sites with greater winter precipitation and no weevils had a greater proportion of iteroparous plants. Sites with higher weevil attack had a lower median threshold flowering size.


The variation in frequency of flowering and threshold flowering size that we documented in North American houndstongue populations, and the relationships between this variation and herbivory and climate, provide evidence for how selective pressures covary with the life histories of invasive plants.

Variation in life histories within and across taxa is driven by variation in constraints and selection pressures, and individuals within and across species may respond differently to the same trade-offs (Young, 1990; Roff, 1992; Lesica and Young, 2005; Wilbur and Rudolf, 2006; Flatt and Heyland, 2011). Within species, optimal life histories can vary given spatial variation in selective pressures (Hesse et al., 2008; Phillips et al., 2010; Kim and Donohue, 2011). Invasion dynamics can provide a significant window into how variation in selective pressures can impact plant life histories (Maron and Vila, 2001; Wolfe, 2002), as individuals encounter different climates and competitors and leave behind specialist enemies, including herbivores, from their native ranges (Müller-Schärer et al., 2004).

All organisms must delay reproduction after they are born. In plants, waiting to reproduce at a larger size increases fitness through higher seed production, farther seed dispersal, and greater attractiveness to pollinators (Wesselingh and de Jong, 1995; Sletvold, 2002). However, flowering at a larger size also has a potential cost: waiting to grow larger might mean an individual dies before reproducing (Wesselingh et al., 1997). Thus, flowering size depends on the likelihood of a plant surviving to reproduce and the fitness gains associated with reproducing at a larger size (Metcalf et al., 2003; Koons et al., 2008).

Both natural selection and physiological constraints also influence the optimal number of reproductive events. While less common than perennial plants that flower in multiple years (iteroparous), semelparous plants can live for one (annual) to many years before flowering, after which they die. Life history theory predicts that semelparity should be favored in conditions of low adult (and relatively higher juvenile) survivorship, and iteroparity for the converse conditions (Charnov and Schaffer, 1973). Like threshold reproductive sizes, semelparity and iteroparity can vary within and across taxa (e.g., Young and Augspurger, 1991; Williams, 2009).

Climatic variation within a species’ range can create circumstances that favor different life histories across the range. While delaying reproduction can expose an organism to heightened risk in some circumstances, in others, delay can increase fitness and spread the risk across years (Tuljapurkar and Wiener, 2000). One way that individuals can limit the risks of reproducing in unfavorable environments, such as drought, is through conservative bet-hedging, defined as a reduction in individual fitness from the maximum potential in each year to reduce fitness variation across years, and thus maximize long-term fitness (Philippi and Seger, 1989; Venable, 2007; Childs et al., 2010). For species with flexible life histories, such as those that can be semelparous or iteroparous, life histories can vary across populations with spatial or temporal variation in climate (Young, 1990; Young and Augspurger, 1991; Kim and Donohue, 2012). For example, higher rates of semelparity were observed with drought in populations of two closely related Lobelia species, hypothesized to be due to reduced adult survivorship under drought conditions (Young, 1990; Young and Augspurger, 1991).

Extrinsic factors like predation and herbivory also can influence life histories (Reznick, 1982; Bastianelli et al., 2017), shaping how, when, and how successfully the prey organism can reproduce, if herbivores target some life stages or traits over others (Reznick, 1982; Day et al., 2002; Metcalf et al., 2009). Which life stages or traits are favored can translate into population level shifts in the size and timing of reproduction (Day et al., 2002). For example, an herbivore that favors larger plants and reduces stored resources, could result in a decrease in the threshold flowering size of plants in that population because the potential benefit of waiting to flower at a larger size no longer leads to greater fitness, (Prins et al., 1992; Wesselingh et al., 1997; Brys et al., 2011). In addition, decreased adult survival due to herbivory can facilitate a shift toward semelparity (Klinkhamer et al., 1997; Metcalf et al., 2003).

For invasive plants, the introduction of biological control creates an ideal scenario to examine the influence of herbivory on life history strategies, as invasive organisms reencounter selective pressures from enemies they escaped (Müller-Schärer et al., 2004). In particular, native specialist herbivores can have a strong influence on the plants they target, decreasing abundance, contributing to defoliation, and reducing survival (Maron and Vila, 2001). In addition to the rapid evolution that can occur after the introduction of an invasive species (Sargent et al., 2017), capturing the life history traits of populations as they reassociate with specialist herbivores from the native range provides the opportunity to study rapid evolution that might result from the introduction of a novel selective pressure (Stastny and Sargent, 2017).

Cynoglossum officinale (Boraginaceae) is an invasive plant that has lived in the absence of specialist herbivores for more than 150 years, since its introduction to North America from Eurasia in the mid-19th century. Although C. officinale (houndstongue) is capable of iteroparity, in its native range it is primarily semelparous (Klinkhamer et al., 1997; Williams, 2009). In its introduced range, we previously observed populations with a substantial fraction of individuals exhibiting iteroparity (mean = 18.9%), and individual survival, growth, and size at flowering were also greater compared to the native range (Williams, 2009). One of its specialist herbivores, a root-boring weevil, Mogulones crucifer, was introduced as a biocontrol in Canada in 1997 and has now moved south into the U.S. The current distribution of weevils in the introduced range provides a unique opportunity to examine life history variation during the critical period when M. crucifer has not yet arrived at all populations.

We examined how life history traits of C. officinale varied across the introduced range in association with climate and herbivory by M. crucifer, focusing on threshold flowering size and frequency of iteroparity, with field surveys and demographic data. Our study addressed three questions: (1) How does the frequency of iteroparity vary in the introduced range in association with herbivory and climate? (2) How does median threshold flowering size vary in association with herbivory? (3) How does the frequency of flowering influence lifetime fecundity? We define median threshold flowering size as the size at which >50% of plants flower (Wesselingh et al., 1997; Metcalf et al., 2003). We hypothesized that iteroparity and median threshold flowering size would be lower at sites with the recently introduced specialist insect because it is known to target larger individuals (Wesselingh et al., 1997; Müller-Schärer et al., 2004; Williams, 2009). We also expected to find increased semelparity in drier regions based on research demonstrating that drier conditions favor semelparity over iteroparity (Young, 1990; Kim and Donohue, 2012). Finally, we expected that individuals that flowered in consecutive years (iteroparous) would sacrifice reproduction in the first year compared to semelparous individuals to achieve higher total lifetime fecundity, as was found previously (Williams, 2009).


Study system

Cynoglossum officinale L. (Boraginaceae) is a monocarpic (semelparous) perennial, that like many plants labeled biennial, can sometimes survive flowering to reproduce in a subsequent year. In its introduced range, we previously showed that houndstongue has higher rates of iteroparity (mean percentage iteroparous plants ± SE: 18.9% ± 13.1%; range: 2%–45%) than in its native range where iteroparity rarely occurs (mean percentage of iteroparous plants ± SE: 0.99% ± 2.98%) (Williams, 2009). Houndstongue grows as a vegetative rosette during its first year and requires vernalization to flower (Wesselingh et al., 1997). Plants can either bolt in spring and flower in early summer in the second year or delay flowering to a subsequent year, with primordial inflorescences appearing in vegetative plants by late summer of the year before flowering (Wesselingh et al., 1997; de Jong et al., 1998). Plants indicate the potential for iteroparity by retaining a vegetative rosette connected to the taproot simultaneous to flowering. This side rosette remains vegetative throughout the spring and summer flowering season and creates the capacity for a plant to flower the subsequent year.

In the native range of houndstongue, adults of a specialist root-boring weevil, Mogulones crucifer Pallas (=Ceutorhynchus cruciger Herbst, =Mogulones cruciger Herbst, Coleoptera: Curculionidae) consume leaves, and larval development occurs in the root. Adult weevils typically feed on aboveground plant tissue and preferentially select larger plants to lay eggs in the roots during the fall and spring (Prins et al., 1992; Schwarzlaender, 1997). Larvae feed on the roots during development and then emerge to pupate in the soil from mid-June through mid-October in the native range (Schwarzlaender, 1997). Larvae can decimate root tissue, inhibiting houndstongue’s ability to conduct water and nutrients (Catton et al., 2016). Because weevils target large individuals that have already converted meristems from vegetative to flowering, they do not affect individual growth, but can significantly reduce seed production (Williams et al., 2010), and the decimation of root tissue likely eliminates the advantage of iteroparity should individuals retain a side rosette (Williams, 2009).

In Canada, M. crucifer was released in 1997 as biocontrol agent. Since its introduction, houndstongue populations in Canada have almost disappeared (Catton et al., 2016). While M. crucifer is an approved biocontrol in Canada, concerns for other native members of the Boraginaceae family have prevented approval in the United States. Despite lack of official approval, weevils have started to move south over the border toward populations of houndstongue in the northwestern United States. Additionally, M. crucifer has likely been illegally introduced without government approval, further expanding the range.

Field surveys to quantify variation in extent of iteroparity and weevil abundance

To quantify variation in the extent of iteroparity, we surveyed flowering plants at 24 sites in Washington (WA), Idaho (ID), and Montana (MT) in 2018 (site details in Appendix S1; data collection details in Appendix S2). Sites were haphazardly located, with assistance from local managers and ranchers. Criteria for site selection were at least 50 flowering plants within a 50-m radius and no herbicide spray treatment in the past few years. In addition, we avoided sites on riverbanks or in wetlands so as not to introduce differences in site hydrology as a confounding factor. We attempted to revisit nine sites from the previous study (Williams, 2009). At the four sites where populations persisted and that we were able to access, we resurveyed even if there were fewer than 50 flowering plants. We included the four revisited sites in our analyses.

At each of the 24 sites, we surveyed all flowering plants within a 50 × 2 m transect, haphazardly placed in the middle of the population. For each plant within the transect, we documented the presence or absence of a side rosette. The presence of the side rosette indicated the potential for iteroparity.

We also conducted root dissections to confirm weevil presence on the first 15 flowering plants in a 15 × 1 m area of the transect and to assess a measure of weevil attack rate if weevils were present. We quantified weevil attack by determining how many plants had weevils present of those surveyed. If weevil larvae were found on at least one plant we stopped dissections after 15 plants. If no weevil larvae were present, we conducted 20 root dissections. We conducted all larval assessments before mid-July to ensure that larvae had not matured and exited the roots. At the 11 sites where we dissected roots from 20 plants, we only found weevil larvae at one of the sites in the five additional plants, and where weevils were present, 47.9% of plants sampled, on average, had weevil larvae in their roots. Thus, where we classify sites as having no weevils detected, it is unlikely, although still possible, that weevils were present, but at much lower density than sites classified as having weevils.

To examine how variability in climate contributes to the extent of iteroparity, we identified climate variables relevant to the lifecycle of houndstongue (spring precipitation, summer precipitation, autumn precipitation, winter precipitation, annual precipitation, winter temperature, summer temperature, and annual temperature) a priori due to the role of vernalization in the winter and the importance of water for seedling success and survival (de Jong and Klinkhamer, 1988, 1989; Wesselingh and de Jong, 1995; Wesselingh et al., 1997). Since plants can only flower starting in the second year, climate data from the year before flowering provide insight into the growing conditions that influenced the initiation of flowering. We sourced climate data for each site from 2017 (year before surveys) and the 1981–2010 normals from ClimateWNA.com (Wang et al., 2016).

Demographic data to estimate median threshold flowering size and fecundity advantage of iteroparity

To quantify the influence of weevil presence on the size-dependent probability of flowering, we collected demographic data at six of the 24 field sites from 2017 to 2018. These six sites were located in Washington (western portion of all sites sampled). We tagged the first 50 vegetative rosettes with at least three leaves in half of the 50 × 2 m transect described above (now 50 × 1 m transect) in July and August 2017. For each plant, we measured the length of the longest leaf and counted the number of leaves. These size measures were previously shown to correlate highly with biomass and predict survival and probability of flowering (Williams, 2009). In 2018, we returned to these sites to follow the fates of previously tagged vegetative plants, recording whether they flowered and measured their fecundity by counting the number of cymes. Cymes are the inflorescences on each stalk, each containing 10 to 35 flowers, which ultimately produce up to four seeds per flower, making the number of cymes a useful estimate of fecundity (de Jong et al., 1990).

To compare total fecundity of semelparous and iteroparous plants, for each flowering plant in the 50 × 2 m transect at the six demography sites, we counted the number of inflorescences in 2017 and recorded whether it had a side rosette (an indicator of the potential for iteroparity). We tagged each potentially iteroparous plant and in 2018 relocated plants and recorded whether they survived to flower a second year and if they did, how many inflorescences they produced. We also counted inflorescences on 10 semelparous individuals in 2018. Thus, we have data on the number of inflorescences for semelparous plants in 2017 and 2018 and for iteroparous individuals in two consecutive years.

Data analyses

To examine the contributions of weevil presence and climate and weather, including precipitation and temperature, to variation in the probability of iteroparity across the introduced range, we used generalized linear models (GLMs) with a binomial error distribution. We used the number of potentially iteroparous plants (referred to as iteroparous plants) at each site offset by the total as the response variable and weighted each observation by the number of plants sampled to account for differences in population size across sites. We used a nested approach to first determine which of the climate variables best explained variation in the probability of iteroparity. We compared those models to one that tested for an effect of weevil presence without any climate variables. We compared and ranked models based on Akaike’s information criterion (AIC) (Burnham and Anderson, 2002). We then added weevil presence to the top-ranked climate model and to models that could not be distinguished from the top model (ΔAIC < 2) to test for a contribution of both climate and herbivory. To determine whether winter precipitation in the year before flowering (2016–2017) differed between the western (sites in WA and ID) and eastern (sites in MT) regions of our study area (split by the Rocky Mountains), we used a t-test.

To calculate how probability of flowering depends on vegetative size in the previous year and weevil attack and thus estimate the median threshold flowering size, we used demographic data collected from the six WA sites in 2017 and 2018. We first calculated rosette size as log (number of leaves × maximum leaf length) for vegetative plants tagged in 2017. Then, using a generalized linear model with a binomial error distribution, we modeled the probability of flowering in 2018 as a function of 2017 rosette size (as in Williams, 2009) and weevil attack. We included an interaction between rosette size and weevil attack, thus allowing weevil attack to influence both the slope and intercept of the model. We classified the six demography sites into two groups: high weevil attack (weevils present in more than ten of 15 dissected stems) and low weevil attack (weevils present in only one or zero of 20 dissected stems). Three sites qualified as having high weevil attack and three sites as low attack; data from the three sites in each category were pooled. Finally, we calculated median threshold flowering size using the coefficients from the probability of flowering model (median threshold flowering size = −intercept/slope).

We assessed how fecundity differs between iteroparous and semelparous plants at the six demography sites by quantifying how the total number of inflorescences a plant produced was influenced by life history. We calculated the total number of inflorescences per plant from both years of flowering for iteroparous plants that survived and flowered a second year. We then compared the number of inflorescences produced in the first year of flowering for iteroparous plants (2017) to the total number for both years of flowering for iteroparous plants (2017 + 2018), and to the number of inflorescences semelparous plants produced each year (2017 or 2018). Using linear mixed-effects models (with the package lme4), we conducted a likelihood ratio test to determine whether the number of inflorescences (log-transformed to meet model assumptions) differed significantly among iteroparous and semelparous plants by comparing an intercept only model to one with life history as a fixed effect (Bates et al., 2014). We included site as a random effect to account for variation among sites in both models. We then conducted pairwise comparisons using the emmeans package to calculate the estimated marginal means to determine which groups differed significantly (Lenth et al., 2019). We used R version 3.4.1 (R Foundation for Statistical Computing, Vienna, Austria) for all analyses.


Iteroparity varied substantially across the 24 sites surveyed in the introduced range (mean percentage of iteroparous plants ± SE: 12.9% ± 2.7%, range: 0–42.3%) (Fig. 1). Only one site had no iteroparity detected, and at five sites, more than 30% of the plants were iteroparous. Of the climate variables examined, average winter precipitation from the prior growing season (December 2016, January 2017, February 2017) was the best predictor of variation in the proportion of iteroparous plants across sites (Fig. 2, Table 1; Appendix S3). Sites with higher winter precipitation had a greater proportion of iteroparity (Fig. 2). Precipitation also varied by region: sites in or east of the Rocky Mountains had significantly lower winter precipitation (t19= −2.087, P = 0.05). Including weevil presence improved the model from the null (ΔAIC = 3.1, Appendix S3), and including weevil presence in the model with average winter precipitation from the prior growing season (2017) significantly improved the overall model fit (ΔAIC = 12.9, Appendix S3). Sites with weevils present had a significantly lower proportion of iteroparous plants than those where no weevils were detected (Fig. 2, Table 1).

Details are in the caption following the image
Map of 24 sites in introduced range of Cynoglossum officinale surveyed with proportion of semelparity (grey) and iteroparity (purple) indicated at each site. Semelparity refers to plants that flower once before dying; iteroparity refers to plants that have the capacity to flower in more than 1 year.
Details are in the caption following the image
Relationship between frequency of iteroparity (flowering more than once) in 2018 and winter precipitation in year before initiation of flowering (2016–2017) at 24 sites across the introduced range of Cynoglossum officinale. Lines show best-fit model for proportion of iteroparous plants at sites with weevils present or absent.
Table 1. Coefficients for a generalized linear model for the effect of weevil presence and 2017 winter precipitation (mm of rainfall from December 2016 to February 2017) on the proportion of iteroparity in 2018 (weighted by sample size at each site), N = 24 sites.
Proportion of iteroparity ~ 2017 winter precipitation + Weevil presence Estimate SE Z P
Intercept 0.0278 0.0241 –1.154 0.247
Weevil presence 0.0594 0.0156 3.810 <0.001
2017 winter precipitation 0.0006 0.0001 5.547 <0.001

We found that the probability of flowering increased with increasing rosette size (Fig. 3, Table 2). Sites with high weevil attack (2/3 of plants sampled had weevil larvae) had significantly lower probability of flowering compared to those with low weevil attack, and the median threshold flowering size shifted to a smaller size (Table 2). Median threshold flowering size was 40% larger at sites with low weevil attack compared to those with high weevil attack.

Details are in the caption following the image
Relationship between rosette size, quantified as log (number of leaves × length in centimeters of longest leaf), in 2017 and probability of flowering in 2018 at six sites (see Appendix S1). Lines show best-fit model for probability of flowering for sites with high compared to low weevil attack. Vertical lines indicate median threshold flowering size. Data were binned into eight equal segments for display (black circles were plants from sites with high weevil attack (N = 86), and gray circles, those with low attack (N = 127)).
Table 2. Coefficients from a generalized linear model for the effect of rosette size and weevil attack (high vs. low) on probability of flowering (N = 213 plants).
Probability of flowering ~ Rosette size × weevil attack (high vs. low) Estimate SE Z P
Intercept –10.086 1.846 –5.465 <0.001
Rosette size 2.040 0.393 5.190 <0.001
Weevil attack 5.310 2.382 2.229 0.026
Rosette size × weevil attack –1.000 0.506 –1.977 0.048

Fecundity varied significantly across semelparous and iteroparous plants (urn:x-wiley:00029122:media:ajb21531:ajb21531-math-0001 = 39.97, P < 0.001). Iteroparous plants in their first year of reproduction (2017) had significantly lower fecundity compared to semelparous plants from 2017 or 2018 (Fig. 4, Table 3). If iteroparous plants survived to a second year they produced on average, over 45% more inflorescences (cymes) over 2 years than semelparous individuals produced in one year. However, this difference was not significant (Fig. 4, Table 3). In addition, across all sites, only 30.6% of iteroparous plants survived to reproduce in 2018.

Details are in the caption following the image
Fecundity (measured as number of inflorescences, or cymes) of iteroparous and semelparous plants in 2017 and 2018. Gray boxes indicate semelparous plants. Line dividing the middle of the boxes indicates median number of inflorescences, boxes indicate 2nd and 3rd quartiles, while whiskers indicate first and fourth quartiles, with the end of each whisker indicating the minimum and maximum number of inflorescences respectively. Different letters indicate groups significantly different from pairwise comparisons; see Table 3.
Table 3. Pairwise comparisons of the estimated marginal means for average cyme production (log-transformed) for iteroparous and semelparous plants in 2017 and 2018 (N = 556 plants). Bolded comparisons are significant. Contrast estimates specify the difference (and SE) of the log-transformed means between the two groups in each pairwise comparison. Where the estimate is negative, the second group was larger than the first. Box plots in Figure 4.
Pairwise comparison Contrast estimate SE P
2017 Iteroparous plants - (2017 + 2018 Iteroparous) plants 1.145 0.260 0.001
2017 Iteroparous plants - 2017 Semelparous plants 0.646 0.115 <0.001
2017 Iteroparous plants - 2018 Semelparous plants 0.744 0.144 <0.001
(2017 + 2018 Iteroparous plants) - 2017 Semelparous plants 0.499 0.244 0.172
(2017 + 2018 Iteroparous plants) - 2018 Semelparous plants 0.401 0.258 0.404
2017 Semelparous plants - 2018 Semelparous plants –0.097 0.108 0.805


Both flowering size and frequency of flowering are life history traits that can vary within a species in response to spatial and temporal variation in selective pressures such as herbivory and climate. In surveys across the introduced range of the invasive plant Cynoglossum officinale, we found that median threshold flowering size and the frequency of iteroparity varied substantially. Both variation in the presence of a recently introduced biological control insect (a root-boring weevil), and climate, in particular winter precipitation, provide insight into the life history patterns we observed.

Variation in iteroparity across the introduced range

The frequency of iteroparity we observed indicates that iteroparity continues to persist in the introduced range, although at somewhat lower rates than observed previously at a subset of the sites in 2004–2005 (mean percentage of 12.9% in current study compared to 18.9% in Williams, 2009), and still notably higher than the very rare observations of plants flowering in more than 1 year in the native range (Williams, 2009). Across the 24 sites surveyed in 2018, sites with increased precipitation and no weevils detected had a higher frequency of iteroparity indicating that wetter conditions and absence of the specialist herbivore favored iteroparity. Our findings confirmed our hypothesis that drier conditions and weevil presence would contribute to increased semelparity based on established links between drought, herbivory, and shifts to semelparity in other species (Young, 1990; Young and Augspurger, 1991; Kim and Donohue, 2012). The only documented instances of increased iteroparity in the native range occurred under “good growing conditions in a common garden experiment” in England, where 10% of flowering plants reproduced a second year (Klinkhamer et al., 1997), suggesting that adult survivorship was higher where growing conditions were favorable (Charnov and Schaffer, 1973). Similarly, in the introduced range, greater winter precipitation might create favorable spring growing conditions for vegetative rosettes by increasing moisture availability through snowmelt, and thus lead to higher adult survivorship. While previous studies have documented the importance of precipitation for houndstongue’s fitness, distribution, and seedling survival in both the native and introduced ranges (de Jong and Klinkhamer, 1988, 1989; de Jong et al., 1990; Moyer et al., 2007; Momayyezi and Upadhyaya, 2017), our findings establish a relationship between increased winter precipitation and a higher frequency of iteroparity in the introduced range.

In the native range, where the specialist weevil coexists with houndstongue populations, iteroparity is rare and plants are for the most part semelparous (Klinkhamer et al., 1997; Williams, 2009). The root damage caused by the specialist herbivore increases the risk of infection by microorganisms, possibly leading to selection against an iteroparous life history (Klinkhamer et al., 1997), because plants would be unable to survive to flower in the subsequent year. The root-boring nature of the weevil might also make houndstongue plants more susceptible to drought. Particularly in drier areas, where a healthy taproot facilitates access to water, weevil damage could limit adult survival and thus eliminate any advantage of iteroparity. Given the recent arrival of the weevil to many of these sites, the full extent of measurable life history shifts might not have occurred yet. As the weevil herbivore spreads across the introduced range and increases in density at local sites, we hypothesize that the proportion of iteroparity at sites in the introduced range will decrease from what we observed, in line with the slight decrease from the past decade (Williams, 2009) to this study. These predictions would be strengthened by further examination of other factors beyond those examined here in influencing life history variation.

We also found that houndstongue populations do not necessarily persist over time. Of the sites previously surveyed in 2004–2005, only four of the seven populations we were able to revisit persisted. Either local management activities such as spraying and hand pulling, or metapopulation dynamics similar to those observed in the native range (Van Der Meijden et al., 1992), could contribute to the disappearance of these populations. Of the four populations we resurveyed, three populations maintained similar levels of iteroparity, while one population had a dramatic decrease from 35.3% in 2005 to 3% in 2018. Since long-term demographic studies have not tracked the frequency of iteroparity across time, there is limited evidence for how often and to what extent the frequency of iteroparity changes within a population.

Our demographic data indicate that plants with the capacity for iteroparity that do not survive to flower a second year have on average reduced fitness, measured by total fecundity, compared to semelparous plants. Our results also suggest iteroparous plants that flower in sequential years could achieve higher reproductive success than semelparous plants did in 1 year as we found earlier (Williams, 2009). Iteroparity, in this case, could be a form of bet-hedging that attempts to reduce the risk associated with limiting reproduction to 1 year (Wilbur and Rudolf, 2006; Childs et al., 2010). However, the result was not significant, potentially due to the small number of plants surviving to flower twice. So while iteroparity is possibly beneficial for houndstongue, it is a risky strategy since we observed so few plants surviving to flower the subsequent year.

Influence of reintroduction of specialist herbivore on threshold flowering size

The reduction in median threshold flowering size at sites with high weevil attack supports our hypothesis that weevils can contribute to reducing flowering size. Weevils are known to preferentially attack larger plants, and thus the size-dependent nature of the herbivory selects for a smaller flowering size (Prins et al., 1992). In the case of houndstongue populations at sites with high weevil attack, the cost of waiting to reproduce and risking death while remaining vegetative outweighs the reduced fitness of producing fewer seeds sooner at a smaller flowering size. The reduction in flowering size associated with high weevil attack could indicate conservative bet-hedging, in which reduced reproductive delay prevents organisms from facing future risks (Childs et al., 2010). The difference in median threshold flowering size between sites with low and high weevil attack suggests that not only weevil presence or absence, but also the level of weevil attack can impact life histories. Because movement of the biocontrol agent has not been monitored in the United States, we do not know how long weevils have occurred at the sites we surveyed. Knowing the length of time weevils have infested a site would provide vital insight into how quickly life histories can shift in response to weevil attack, as well as provide insight into factors that might influence the rate of population growth of weevils after establishment.

While our evidence builds on previous studies about life history changes observed in the introduced range in response to enemy escape, climate, and biocontrol reintroduction (Williams, 2009; Williams et al., 2010; Catton et al., 2016), we cannot determine the underlying genetic contributions or adaptive functions from our field surveys. Previous work demonstrated both genetic and environmental contributions to observed threshold flowering size (Wesselingh et al., 1997). A previous common garden experiment, which included individuals from four sites resurveyed here, found that all plants were semelparous, regardless of origin (native or introduced, and semelparous or iteroparous parent), suggesting a large environmental contribution to iteroparity (Williams et al., 2008). A study comparing neutral genetic markers across the invasive and native ranges of houndstongue established a loss of genetic diversity in the introduced range compared to the native range and support for two source regions for populations in North America (Williams and Fishman, 2014). If native populations vary genetically in potential for iteroparity, founder effects that occurred with the introduction of houndstongue to North America might shape some of the patterns we observed.


Understanding how the reproductive life history of C. officinale shifts in response to its environment and its reunion with its weevil specialist has implications for future management strategies. The higher frequency of iteroparity that we found across the introduced range, compared to the native range suggests that management strategies targeting individuals with the capacity for multiple flowering are important, particularly in populations with wetter winters, to reduce contributions of individuals to future generations and to the seed bank across multiple years of flowering. Reduced median threshold flowering size at sites with high weevil attack, coupled with the previously demonstrated negative effects of weevils on population growth rates, has positive implications for using M. crucifer as a biocontrol agent for houndstongue management (Williams et al., 2010; Catton et al., 2016). As weevils disperse naturally and reach higher densities in populations across the introduced range, we expect further reductions in flowering size. Future studies could elucidate how flowering size and frequency of flowering respond to colonization by weevils and increasing weevil population size with longer term monitoring at sites during the ongoing weevil invasion, which could reveal how quickly weevil invasion impacts flowering life histories.

Additionally, more invasive organisms, which formerly “escaped” their predators, are facing predator “reunification”. Just as invasive species can rapidly evolve in the absence of specialists, rapid evolution can also happen in response to the introduction of a specialist as a biocontrol (Stastny and Sargent, 2017). So far, houndstongue has not developed resistance to the specialist biocontrols introduced to Canada given that plant populations have quickly crashed in abundance after weevil introduction (Catton et al., 2016). However, in the native range, houndstongue populations persist despite weevil presence, indicating that it is possible for houndstongue and its specialist to coexist. Developing resistance to weevils does not necessarily mean that the biocontrol will become ineffective. However, in the United States where weevils randomly disperse at low-levels, weed management programs should monitor how quickly small weevil populations can reach outbreak levels and whether houndstongue populations can tolerate lower weevil attack. If M. crucifer receives approval in the United States, targeting priority populations with releases of at least 100 weevils, the smallest release size so far documented to be effective, can reduce the risk associated with houndstongue developing tolerance to low-level weevil presence (De Clerck-Floate and Wikeem, 2009).

Our findings highlight that life histories can vary substantially within a species, making cross-taxa comparisons challenging. The variation we observed within the introduced range of houndstongue thus underscores the importance of surveying extensively throughout a species range or ranges, especially for species with flexible life histories. Despite the utility of intraspecific variation for cross-taxa comparisons, the body of research documenting intraspecific variation in life history remains relatively scant (Frederiksen et al., 2005; Hesse et al., 2008; Bastianelli et al., 2017). Establishing the scope and extent of life history variation within one species will allow comparisons between species to more fully represent the variation or similarities in life history strategies among species. Finally, investigating how life histories shift in response to “reunification” between specialist herbivores from the native range and invasive plants in their introduced range provides further insight into the ways herbivory and evolution interact. As invasive species expand their ranges and encounter changes within their existing ranges, such as the introduction of biocontrols or changing climate conditions, how species respond to these changes provides vital information about the potential for adaptive evolution to shift life histories.


We thank the numerous land managers, ranchers, and others who helped locate field sites, particularly Mark Schwarzlaender and Marijka Haverhals, and Danielle Main and Emily West for assistance collecting data in the field. Amy Angert, Brett Eaton, Nina Sletvold, Jeannette Whitton, and two anonymous reviewers provided valuable comments on earlier versions of this manuscript. This research was supported by a Natural Sciences and Engineering Resource Council of Canada (NSERC) Discovery Grant to J.L.W. and the Canadian Foundation for Innovation and British Columbia Knowledge Development Fund.


    S.D. and J.W. designed the study. S.D. collected data and conducted analyses with input from J.W. S.D. wrote the first draft of the manuscript, and S.D. and J.W. contributed substantially to revisions.

    Data Availability

    Data are available from the Dryad Digital Repository at https://doi.org/10.5061/dryad.9w0vt4bc6 (Williams and Duncan, 2020).