Induced herbivore resistance in seaweeds: a meta-analysis


Gunilla B. Toth (tel. +46 526686 14; fax +46 526686 07; e-mail


  • 1Terrestrial plants can sense and respond to herbivory, which may lead to increased resistance towards further grazing if the responses have negative effects on the preference and/or performance of the herbivores. Marine plants (seaweeds) are exposed to a considerable grazing pressure by herbivores ranging from large, mobile fishes and sea urchins to small, sedentary crustaceans and molluscs. The number of investigations studying induced resistance in seaweeds has increased during the last decade, but empirical results are conflicting.
  • 2We performed a categorical meta-analysis to evaluate statistically the overall seaweed responses to damage or damage-related cues, and factors that may explain the observed variation in inducible seaweed resistance to herbivores.
  • 3We found a highly significant overall effect of damage on induced seaweed resistance to further herbivory. Division of the studies into different categories showed that brown and green, but not red, seaweeds induce significant resistance to further grazing in response to grazing by small crustaceans and gastropods, but not in response to large gastropods and sea urchins. The seaweeds showed stronger responses when exposed to damage for 11–20 days than in shorter or longer experiments.
  • 4Seaweeds are very important both as habitat and food for a wide range of marine animals. Our findings contribute importantly to the general ecological understanding of marine plant–herbivore interactions by showing that induced resistance in seaweeds is more common than previously assumed. Many recent marine investigations included in this study have not put emphasis on the ecological relevance and underlying mechanisms of the investigated plant–herbivore interactions. We suggest that the scientific value of future investigations concerning induced defences in marine algae would benefit from formulating more advanced and/or complex hypotheses including the genetic and biochemical mechanisms, cost and constraints of damage-induced civilian and defensive seaweed responses, as well as the effects of these responses on herbivores and other organisms/trophic levels, and on community structure and functioning.


Vascular plants are known to induce a variety of phenotypic changes in response to damage by herbivores, including alterations of plant chemical composition, phenology, morphology, photosynthesis and growth (Karban & Baldwin 1997). Induced plant responses may lead to increased resistance towards further herbivory if the changes have negative effects on the preference and/or performance of the herbivores (Karban & Baldwin 1997). Herbivore-induced responses and resistance are found in a large number of terrestrial plant species (e.g. see reviews by Karban & Baldwin 1997; Agrawal 2005), and parts of this extensive literature have recently been synthesized and evaluated using meta-analysis (e.g. Koricheva 2002; Koricheva et al. 2004; Nykänen & Koricheva 2004; Leimu & Koricheva 2006). For example, Nykänen & Koricheva (2004) used meta-analysis to show that natural or simulated damage tends to reduce plant growth, increase photosynthetic rate and increase the concentration of phenolics in woody plant species. Furthermore, damaged plants negatively affected the growth rate and pupal weight of insect herbivores, while survival and consumption were not affected (Nykänen & Koricheva 2004). In contrast to the extensive literature on terrestrial plants, there are relatively few published examples of herbivore-induced responses in marine macroalgae (seaweeds). The first example of induced responses in a seaweed species was published in the late 1980s (Van Alstyne 1988) and since then the number of studies has increased progressively (see supplementary Appendix S1). However, empirical evidence is mixed and sometimes conflicting. It appears that herbivore-induced responses are not universally expressed by all seaweed species under all circumstances. There may be several explanations for these varying results, e.g. specific traits associated with the seaweed and/or herbivore species involved in the interactions (Amsler 2001), or differences between experimental protocols. To our knowledge, there are no previously published narrative reviews or meta-analyses that specifically aim at synthesizing all current literature on induced resistance in seaweeds. Meta-analysis is an excellent complement to traditional narrative reviews because it objectively and quantitatively summarizes the findings from previously published studies, especially when results are mixed or conflicting (e.g. Gurevitch & Hedges 1993; Rosenberg et al. 2000; but see Discussions in Maestre et al. 2005, 2006; Lortie & Callaway 2006). Furthermore, by dividing studies into different categories, a meta-analysis can be used as an observational tool to explore patterns within a data set (Rosenberg et al. 2000).

In general, there may be a number of categorical variables that are important to explain the variation in herbivore-induced resistance observed in seaweeds. For example, seaweeds are a large and taxonomically heterogeneous group of organisms, which is commonly classified into three divisions according to their primary or secondary photosynthetic pigments (Chlorophyta, Rhodophyta and Phaeophyta; or green, red and brown seaweeds). There is considerable ecological and physiological variation among different species within these three divisions, including, for example, growth rate and morphology, reproductive strategies, and production of bioactive secondary metabolites that may function to deter herbivory (e.g. Lobban & Harrison 1994; McClintock & Baker 2001). Furthermore, different parts within an individual plant can have different physiological constraints and/or fitness values, and may thus show different responses (e.g. Toth et al. 2005). Differences in induced responses may also be due to characteristics associated with the herbivore species consuming the seaweeds. It has been suggested that inducible defences should evolve in seaweeds in response to grazing by small sedentary herbivores (mesoherbivores) that use individual seaweeds both as food and habitat, because large and mobile herbivores will not stay on a plant long enough to suffer the induced responses (Hay 1996). Furthermore, plants will not have time to induce resistance if herbivores are large enough to consume the entire plant in a short feeding bout. Large, mobile herbivores are instead hypothesized to select for constitutive or activated/preformed defences, i.e. defences or defence-precursors that, once produced, are present continuously within the plants (see Karban & Baldwin 1997 for definitions of induced, constitutive and activated/preformed defences). Furthermore, grazing pressure is generally considered to be higher and more dominated by large herbivores on tropical coral reefs than in temperate habitats (Hay 1996), and therefore induced resistance may be more prevalent in temperate than in tropical seaweed species.

Other possible explanations for the conflicting empirical evidence of induced responses in seaweeds may be the use of varying experimental procedures, including different inducing cues, time scale of the experiments and bioassay procedures to evaluate resistance. For example, mechanical simulations of herbivory (artificial damage) are sometimes used as a substitute for natural herbivory, but can be poor mimics of true grazing in both terrestrial and marine plant–herbivore interactions (e.g. Baldwin 1990; Pavia & Toth 2000). Furthermore, some studies have shown that seaweeds can induce resistance in response to air- or water-borne chemical signals (e.g. Toth & Pavia 2000; Arnold et al. 2001), whereas others have failed to detect a response to such cues (e.g. Taylor et al. 2002). The time scale of the experiments is probably crucial for the detection of induced chemical responses, because activation of biosynthetic pathways that lead to production of herbivore-deterrent secondary metabolites may take several days (e.g. Karban & Baldwin 1997). Furthermore, the results from bioassays using different types of food (e.g. live plants or artificial diets) may be very different depending on the nature of the induced response. For example, a change in morphology or nutritional status of a plant will not be detected in a bioassay using artificial food with extracted secondary metabolites. Moreover, the experimental set up of the bioassay may influence the probability of detecting induced resistance, because changes in consumption may be more easily detected when herbivores can choose between control and treated food (i.e. two- or multiple-choice preference bioassays) than when offered only one type of diet (i.e. no-choice consumption bioassays) where preference can be masked by compensatory feeding (e.g. Cruz-Rivera & Hay 2003; Prince et al. 2006).

The aim of this paper is to present a quantitative, as well as a qualitative, review of the ecological literature on inducible responses in seaweeds and their effects on herbivore consumption. Categorical meta-analyses were used to evaluate statistically the overall responses of seaweeds to different inducing cues, as well as factors that may explain the observed variation in inducible seaweed responses and resistance to herbivores. Furthermore, because hypothesis-driven fields in ecology, such as the study of damage-induced plant resistance, tend to favour early publication of supportive results (Jennions & Møller 2001; Nykänen & Koricheva 2004), we performed a cumulative meta-analysis to investigate possible temporal trends in the effect size of induced seaweed responses. We specifically hypothesized: (i) that if there is no bias towards publication of studies with confirmative results, there would be a stabilization of, and a decrease in the variation around, the mean effect size over time; (ii) that seaweeds from different divisions would respond differently to inducing cues; (iii) that there would be differences in induced responses depending on which plant part was studied; (iv) that grazing by different herbivore groups would induce different responses; (v) that a lower amount of damage would induce a stronger response than a higher amount of damage; (vi) that temperate seaweed species induce a larger response than tropical seaweeds; (vii) that true herbivory and natural chemical cues would induce stronger responses than artificial damage; (viii) that medium- and long-duration experiments would more frequently result in significant induced responses than short experiments; (ix) that there would be a difference in the responses reported from studies using different types of bioassay food (i.e. live plants or artificial diets); and (x) that studies evaluating induced resistance through two- or multiple-choice preference bioassays would report stronger responses than studies using no-choice consumption bioassays. The results from the meta-analysis are used to evaluate and discuss differences and/or similarities among inducible defences in marine and terrestrial habitats. Furthermore, we suggest research efforts that we think will aid the future development and progress of studies on herbivore-induced resistance in aquatic macroalgae.

Materials and methods

the data base

The data base was assembled by keyword searches in the ISI Web of Science data base covering papers published between 1986 and 2007, and by searching the cited literature in the obtained papers and in recent reviews (e.g. McClintock & Baker 2001; Paul & Puglisi 2004; Pohnert 2004). The following criteria had to be met in order for a publication to be included in the analysis: (i) the inducing treatment had an appropriate control that was not induced during the experiment, (ii) the paper included a presentation of the means, some measure of the variance and the sample sizes for the control and the treated experimental groups, and (iii) quantitative measures of the effect of induced responses on herbivore consumption were included. On these grounds, papers measuring concentrations of secondary metabolites (e.g. polyphenolics) or responses (e.g. oxidative bursts or gene expression) without reference to, or tests of, their herbivore-deterrent effects were not included in the analysis. Furthermore, studies investigating activated/preformed defences were not included, because activated defences can be regarded as a special case of constitutive defence, i.e. the compounds or precursors are produced constitutively and are activated by tissue damage (cf. Karban & Baldwin 1997). The final data base consisted of 18 studies published between 1988 and 2007 (Appendix S1). The experiments included in the analysis were conducted with a total of 41 seaweed and 18 herbivore species (Table S1), and contained 127 measurements of herbivore consumption on control and damaged seaweeds.

response variables

Induced plant resistance can be evaluated through bioassays measuring the effect of induced plant responses on either the feeding behaviour or the performance of the herbivores. In contrast to the terrestrial literature, where effects on herbivore performance are often measured, induced resistance in seaweeds is usually evaluated through herbivore feeding experiments. To our knowledge, there is only one study investigating the effect of induced seaweed responses on herbivore performance traits (growth and reproduction; Toth et al. 2005), and therefore our analysis consisted only of studies where herbivore consumption was quantified.

Similar criteria to those used in the study on induced responses in woody plants by Nykänen & Koricheva (2004) were applied to the selection of individual measurements in the present study. When several measurements were made on the same individual (e.g. after different time intervals or in different plant-parts), only the result showing the largest difference between the control and the treated seaweed was used. However, if the experiment was designed so that treatments were applied to, or measurements taken from, different seaweed individuals, all measurements were included in the analysis because the dependence between samples was not considered to be severe (cf. Nykänen & Koricheva 2004). When damage-induced responses were studied at different nutrient or light levels, we only included the control measurements or the measurements that were as close to natural levels as possible.

explanatory variables

In order to test if there were differences among particular groups of measurements included in the data base, we categorized the studies according to the following explanatory variables. (i) Publication year. (ii) Seaweed division – the seaweeds were divided into brown (Phaeophyceae), green (Chlorophyceae) and red (Rhodophyceae) (see Table S1 for species names). (3) Seaweed plant-part – studies where whole plants were used, or where it was not evident from which plant-part a measurement was taken, were excluded from this analysis. Seaweed plants were divided into meristematic, non-meristematic and reproductive tissues. However, measurements of herbivore consumption on reproductive tissues were too few and were not included in the analysis. (iv) Herbivore group – the herbivores used in the induction experiments were divided into different groups (isopods, amphipods, small gastropods, large gastropods and sea urchins) based on their taxonomy (see Table S1 for species names). (v) Amount of damage – a surprisingly high number of the studies (39%) did not report on the extent of damage inflicted on the seaweeds during the induction experiments. However, the studies that quantified the extent of damage were divided into two classes based on the amount of tissue removed (lower ≤ 20% and higher ≥ 21% of initial wet or dry weight or tissue area). (vi) Latitude – the seaweeds used in the experiments were divided into temperate (66.5–23.5°N, and 23.5–66.5°S) or tropical (23.5°N−23.5°S) species based on the latitude where the study was performed. (vii) Inducing cue – the treatments used to induce seaweeds were true grazing by natural herbivores, artificial damage and unidentified chemical cues from grazed seaweeds. (viii) Time scale of induction experiments – the induction experiments were divided into three classes based on the number of days that seaweeds were exposed to grazers (short = 1–10 days, medium = 11–20 days, long = 21–120 days). (ix) Type of bioassay food – three different food types were used in the consumption or preference assays following the induction experiments. These were live seaweed plants (live seaweeds), artificial food with dried, homogenized seaweed material [artificial food (seaweeds)] and artificial food with different types of metabolites extracted from the seaweeds [artificial food (extract)]. (x) Bioassay set-up – the studies were categorized according to the experimental set-up of the bioassay used to evaluate induced resistance. Choice assays (two- or multiple-choice) measured the herbivore feeding preference by allowing herbivores to choose between control and induced food, while in no-choice assays herbivore consumption was measured by forcing the herbivore to feed from either control or treated food.


Hedge's d was used as a measure of the effect size and was calculated for each individual measurement as the difference between the mean of the experimental treatment and the control divided by their pooled standard deviation and multiplied by a correction term to reduce bias from small sample sizes (Gurevitch & Hedges 1993). Consequently, a negative effect size indicates induced resistance. All analyses were performed using the computer program MetaWin 2.0 (Rosenberg et al. 2000). Visual data exploration was performed using normal quantile plots (Wang & Bushman 1998) in order to check (i) that the data were normally distributed (data should fall on an approximately straight line and within the 95% confidence limits), (ii) whether studies came from multiple populations (the regression curve is nonlinear, or the slope is ≠ 1) and (iii) for publication bias (the curve is non-linear or has gaps where data are missing). Rosenthal's fail-safe number was also calculated to test for publication bias in the data base (Rosenthal 1979). This number indicates the number of studies with zero effect that has to be added to the data base in order to change the result from significant to non-significant. If this number is sufficiently high (> 5n + 10, where n is the number of measurements), the results can be considered as robust with regard to publication bias (Rosenthal 1979).

Possible temporal trends in reported results for induced resistance were tested with cumulative meta-analysis. In order to test the effect of different explanatory variables, data were analysed using mixed-effects models of meta-analysis (Rosenberg et al. 2000). The reported results are from a re-sampling test generated from 4999 iterations with bootstrap values of the 95% confidence intervals. The magnitude of the effect size was considered significant when the confidence interval did not include zero (Gurevitch & Hedges 1993), and the effects of the explanatory variables were evaluated using heterogeneity statistics (Q). When a significant effect of an explanatory variable was detected, we examined the association between that variable and the other variables using chi-square tests of independence (cf. Nykänen & Koricheva 2004). However, in most of the tests the expected frequency was < 5 in > 20% of the contrasts, which makes the chi-square test of independence inappropriate. Furthermore, this indicates that our data set was highly heterogeneous and that the different explanatory variables tested may have been confounded. We could not split the data set and test the effect of each variable on each level of the other variables (cf. Nykänen & Koricheva 2004), because the data set became too small. However, we have tried to consider this bias in our discussion of the results.


The normal quantile plots for induced seaweed resistance (Fig. 1) showed approximately linear relationships between the standardized effect size and the normal quantiles, indicating that data were normally distributed. However, the ‘bumps’ on the linear regression curve, and the fact that the slopes of the regression lines were ≠ 1 (k = 1.4, Fig. 1), indicate that the data are from more than one population. No gaps were observed in the linear regression curves (Fig. 1), indicating that publication bias is not a problem with the present data sets. Furthermore, Rosenthal's fail-safe number for the total data set was larger (2060) than the critical value (645). Therefore, the number of non-significant, unpublished studies needed in order to change the results of the meta-analyses from significant to non-significant was sufficiently high to conclude that the observed results can be treated as a reliable estimate of the true effect size.

Figure 1.

Data exploration and publication bias. Standardized effect size vs. standard normal distribution (normal quantile) for herbivore consumption of control and damaged seaweeds. Data are normally distributed when the points fall within the confidence intervals (dotted lines). The slope of the linear regression line (black solid line) shows standard deviation of the data; a slope ≠ 1 (grey solid line) indicates that the studies are from more than one population. Publication bias is indicated by gaps in the plot or very nonlinear curves.

Among the experiments investigating induced seaweed resistance towards herbivory, about 91% (116 of 127 comparisons) were published after 2000. Cumulative meta-analysis revealed a large change in the magnitude of induced resistance between the late 1980s and the beginning of the present century (Fig. 2). Early studies conducted in the late 1980s reported strong significant effects of induced seaweed responses on herbivore consumption (i.e. d < 0). This effect decreased markedly during the 1990s, but remained strong and statistically significant. The cumulative effect size continued to increase during the early 2000s and currently the magnitude of the effect of induced seaweed responses on herbivore consumption is moderate but has remained statistically significant (Fig. 2).

Figure 2.

Cumulative meta-analysis showing temporal changes in cumulative effect sizes (Hedge's d) and 95% confidence intervals for herbivore consumption of control and damaged seaweeds. The analysis at the bottom of the graph presents the results for the first two measurements. Subsequent measurments were added in chronological order so that the analysis at the top of the graph illustrates the result of all measurements included in the data base. Figures within the graphs represent the first measurement that publication year, and measurements from the same year were added in random order. Calculations are based on fixed effects.

There was an overall moderate, but significant, negative effect of induced plant responses on herbivore consumption [dT = −0.489, 95% CI (−0.653; −0.324), n = 127], clearly showing that induced resistance is present in seaweeds. Partitioning the variation of the data according to models with different explanatory variables revealed that variances among effect sizes could, in part, be explained by the taxonomic affiliation of the seaweeds and herbivores used as model organisms, and also by the time scale of the experiment (Table 1). However, the seaweed plant-parts, the amount of damage inflicted on the seaweeds, the latitude where the experiments were performed, the inducing cues, the bioassay food types and the experimental set-up used to evaluate induced responses did not significantly explain the variance among effect sizes (Table 1). Furthermore, there was still a large proportion of the variance in the data set that was not explained by any of the above variables (see QW values in Table 1), indicating that there are other, yet unidentified, variables that affect induced resistance in seaweeds.

Table 1.  Meta-analysis of categorical data. The between-group heterogeneity (QB), within-group heterogeneity (QW) and significance levels (P) when heterogeneity was tested against a χ2-distribution. d.f. = degrees of freedom. Bold numbers indicate P < 0.05. Data on mean effect sizes and 95% confidence intervals for the different classes within the explanatory variables are shown in Fig. 3
Explanatory variableQB (d.f.)PQW (d.f.)P
Seaweed group 6.30 (2)0.043197.16 (124)< 0.001
Plant-part 2.96 (1)0.086130.22 (76)< 0.001
Herbivore group12.74 (4)0.013178.94 (114)< 0.001
Amount of damage 0.257 (1)0.623 35.47 (33)0.308
Latitude 2.672 (1)0.102199.30 (125)< 0.001
Inducing cue 1.637 (2)0.441197.93 (124)< 0.001
Time scale 9.70 (2)0.008194.12 (124)< 0.001
Bioassay food 0.89 (2)0.642199.24 (124)< 0.001
Bioassay set-up 1.24 (1)0.265200.34 (125)< 0.001

There was a significant difference in the effect size between seaweeds from different divisions (Fig. 3a). Brown and green seaweeds induced responses that affected herbivore feeding, but no significant induced resistance was found in red seaweeds. However, most experiments using green and red seaweeds as model organisms investigated induced resistance in vegetative tissues in response to crustacean grazing for a medium or short period of time. Both non-meristematic and meristematic seaweed parts induced significant resistance towards further grazing, but there was no significant difference in the effect size between the two groups (Fig. 3b). Seaweeds responded differently when exposed to grazing by different herbivore species, as shown by the significant difference in effect size between different herbivore groups (Fig. 3c). All mesoherbivores (isopods, amphipods and small gastropods) induced a significant resistance in the seaweeds, but the larger herbivores (large gastropods and sea urchins) did not (Fig. 3c). However, small gastropods and sea urchins were only used to induce brown seaweeds. Both a moderate and a high amount of damage affected the subsequent herbivore consumption negatively (Fig. 3d), and both temperate and tropical seaweeds induced significant resistance towards further herbivory in response to herbivore cues (Fig. 3e). Surprisingly, artificial damage, chemical cues and direct herbivore grazing all resulted in induced seaweed resistance (Fig. 3f). Seaweeds exposed to inducing cues for 11–20 days had a significantly stronger negative effect on herbivore consumption than seaweeds subjected to shorter or longer exposure (Fig. 3g). However, most experiments conducted for more than 21 days were performed using temperate brown seaweeds that were directly grazed by amphipods, which may confound the results. The type of food used to evaluate induced resistance in the consumption or preference bioassays did not affect the outcome of the experiments, i.e. there was no significant difference in effect sizes of studies using live algae or different artificial foods in the preference bioassays (Fig. 3h). Furthermore, there was no statistically significant difference in the effect size when herbivore feeding was evaluated through choice or no-choice bioassays (Fig. 3i).

Figure 3.

Induced resistance. The effect of different: (a) seaweed divisions (brown, green and red seaweeds); (b) plant-parts (non-meristematic and meristematic tissue); (c) herbivores (isopods, amphipods, small gastropods, large gastropods and sea urchins); (d) amount of damage (lower and higher); (e) latitudes (temperate and tropical species); (f) inducing cues (artificial damage, chemical cues and direct herbivore grazing); (g) time scale of the experiments (short, medium and long); (h) bioassay food (live seaweeds, and artificial diets with seaweeds or extracts); and (i) bioassay set-up (choice and no-choice) on the mean effect size (Hedge's d) of herbivore consumption on control and damaged seaweeds. A significantly negative effect size indicates a lower consumption on induced seaweeds (i.e. induced resistance). Error bars show bootstrap values of 95% confidence intervals, and numbers above bars represent sample sizes. Data presented in (b)–(d) are based on a subset of studies (see text for explanation).


The results from the meta-analysis clearly show that seaweeds can respond to herbivory by inducing resistance to further grazing. Apart from a decrease in effect size in the late 1980s and the early 1990s, the cumulative meta-analyses did not reveal any dramatic temporal changes in the magnitude and direction of reported induced seaweed resistance. Previous meta-analyses on studies with ecological and evolutionary hypotheses have also found small but significant negative relationships between the strength of research findings and publication year, indicating that the early publications tend to show stronger effects than later publications (Jennions & Møller 2001). Moreover, the decrease in cumulative effect size with time for induced resistance in seaweeds may, at least in part, be due to a ‘bandwagon’ effect (cf. Poulin 2000), because the early studies on induced plant responses and resistance in marine systems (e.g. Van Alstyne 1988; Cronin & Hay 1996; Pavia & Toth 2000) were highly influenced by the literature for terrestrial plants, where this phenomenon had been well described. Alternatively, early studies may have chosen model species where the probability of a significant treatment effect was high. Therefore, early marine corroborative studies may have been easier to publish than investigations showing non-significant results. Nevertheless, the normal quantile plots and Rosenthal's fail-safe number, as well as the tendency for a stabilization of, and a decrease in the variation around, the mean effect with time, indicate that a bias towards publication of significant results was not a large problem for the studies included in the data base (cf. Nykänen & Koricheva 2004). However, the data base did not contain enough measurements to allow statistical testing of possible associations between the explanatory variables (cf. Nykänen & Koricheva 2004). Comparing the number of studies in different categories clearly showed that all the significant explanatory variables were correlated to each other or to non-significant explanatory variables. Consequently, the induced resistance patterns revealed in the present study must be interpreted cautiously, because the explanatory variables were confounded. However, the results presented are still useful because they represent an objective overall summary of the present knowledge, and clearly identify several areas where more research is needed in order to increase our understanding of induced resistance in seaweeds.

The results from the categorical meta-analysis showed a significant difference in induced resistance between different seaweed divisions. Brown and green, but not red, seaweeds induced resistance towards further herbivory, and the response was strongest in brown seaweeds. Brown seaweeds contain variable levels of phlorotannins (polyphenolics), which can be induced by both direct grazing and chemical cues (e.g. Pavia & Toth 2000; Toth & Pavia 2000; Borell et al. 2004) and which have been shown to deter herbivores (e.g. Gieselman & McConnell 1981; Pavia & Toth 2000). In contrast, no herbivore-induced deterrent metabolites have been identified to date in green or red seaweeds. Furthermore, a significant difference in induced seaweed resistance was found after grazing by different herbivore groups. Resistance was induced by small gastropods and crustaceans, but not by large gastropods and sea urchins. These results are in accordance with the hypothesis that inducible defences should evolve in seaweeds in response to grazing by mesoherbivores, rather than large and mobile herbivores (Hay 1996). Moreover, there was a significantly stronger effect of induced responses on herbivore consumption when seaweeds had been exposed to damage for 11–20 days in comparison with shorter experiments. This is in accordance with the hypothesis that induced resistance may take several days to express (Karban & Baldwin 1997). However, when herbivory was continued for more than 20 days, induced resistance was again less pronounced. The conditions in long-term laboratory experiments may not be ecologically realistic because herbivores are forced to continue feeding on the plants even if they induce responses that would result in decreased herbivore pressure in short-term experiments or in a more natural field situation.

We did not detect any significant differences in induced resistance between different seaweed tissues, latitudes, type or amount of damage, or the type of food or experimental set-up used in the bioassays. The explanatory variables accounted for a small part of the total heterogeneity in the data set, implying that additional models to those tested in the present study should have been explored. In a similar meta-analysis of studies on damage-induced changes in terrestrial woody plants, Nykänen & Koricheva (2004) examined the overall effect of host plant damage on subsequent herbivore consumption. In addition, they tested the explanatory effect of plant growth rate (fast and slow), and the time interval between the induced responses and their possible effects on herbivore consumption (rapid or delayed induced resistance), on the total heterogeneity of their data set. In contrast to our results, they could not detect a significant overall negative effect of plant damage on herbivore consumption. However, consumption was significantly decreased on fast-growing plants after damage, whereas no change could be detected on slow-growing plants. Furthermore, no difference between rapid and delayed induced responses was detected. Objective categorization of plant growth rate is difficult, and Nykänen & Koricheva (2004) did not explicitly define a clear boundary between the categories used in their study. Instead, they categorized whether a plant species was a fast- or a slow-grower based upon data in the original studies or in reviews. Such information was not available in most of the studies included in our data base, and therefore we chose to omit seaweed grow rate as an explanatory variable in the analysis. However, differences in growth rate, as well as in other seaweed traits such as morphology, generation time or reproductive strategy, could be important in explaining the heterogeneity in the data set. Furthermore, we are not aware of any studies of delayed induced resistance in seaweeds, and therefore this explanatory variable was not possible to include in the analysis.

The publications included in the data base (Appendix S1) can be divided into one of the following groups based on their tested hypotheses: (i) studies screening a number of different seaweed species for the presence or absence of induced resistance in response to different herbivores; (ii) studies testing the interactive effects of herbivory and other factors such as resources (e.g. nutrient availability and light) or stressors (e.g. emersion or UV-B radiation) on induced seaweed resistance; (iii) studies investigating the nature of the inducing cues (e.g. water- or air-borne chemical cues, mechanical damage); and (iv) studies testing hypotheses derived from theoretical models formulated to explain spatial or temporal variation in plant resistance (e.g. the optimal defence model or the carbon-nutrient balance model). All of these approaches can be valuable in elucidating the importance of, and mechanisms behind, induced resistance for seaweed fitness traits. However, compared with the terrestrial literature [where current research on induced responses addresses major concepts in ecology and evolutionary biology including (i) impact of induced resistance on herbivore competition and communities, (ii) costs involved in expressing resistance and (iii) the macroevolution of induced plant responses (Agrawal 2005)], recent marine studies have focused on rather basic and descriptive hypotheses concerning the presence or absence of induced resistance in seaweeds. We believe that marine studies should, from now on, address more advanced hypotheses, similar to those presented in the terrestrial literature (cf. Stamp 2003; Agrawal 2005). Screening experiments can still be valuable first tools to identify marine model systems that can later be used to test more complex hypotheses regarding the ecology and evolution behind induced resistance, or as a tool to test the specific hypotheses derived from general patterns. For example, screening experiments could be used to investigate if induced resistance is dependent on seaweed growth rate, morphology, reproductive strategy or other life-history traits. Obviously, the choice of model organisms in these studies must still be firmly established in observations of ecologically relevant seaweed–herbivore interactions.

We have identified five research areas within the field of damage-induced resistance that are, or are becoming, well developed for terrestrial plants but where published studies on seaweeds are either absent or very scarce. (i) Identification and quantification of damage-induced seaweed responses that are likely to affect the food quality for herbivores from a ‘phytocentric’ (sensu Karban & Baldwin 1997) perspective. A phytocentric perspective argues that induced responses to herbivory should be considered in the context of how a plant grows (Karban & Baldwin 1997; Nykänen & Koricheva 2004). Studies are needed that investigate induced civilian and defensive responses by measuring changes in water, protein, nutrient, carbohydrate and secondary metabolite (other than phlorotannins) content in seaweeds after damage. (ii) Effects of induced chemical defences on herbivore performance and behavioural or physiological adaptations to low-quality food (e.g. Felton 1995; Karban & Baldwin 1997; Berenbaum & Zangerl 1999; Nykänen & Koricheva 2004). As evident from the present meta-analysis, which only included measurements of how induced seaweed responses affect herbivore consumption, studies on how damage-induced seaweed responses affect herbivore growth, reproduction, survival, compensatory feeding responses and physiological adaptations are rare in marine systems. However, they are vital to understand fully the ecology and evolution of induced resistance in seaweeds. (iii) Effects of damage-induced seaweed and herbivore responses on other organisms/trophic levels, community structure and functioning (e.g. Ohgushi 2005). Although there are a number of studies on the forces that regulate trophic structure and dynamics in marine ecosystems (i.e. bottom-up vs. top-down control; Shurin et al. 2006), no previous studies have explicitly addressed whether herbivore-induced resistance in seaweeds can drive so-called trait-mediated indirect interactions (TMIIs) in the marine environment. (iv) Genetic variation, costs and limits of induced responses in seaweeds (e.g. DeWitt et al. 1998; Strauss et al. 2002; Agrawal 2005). Although a few studies have investigated genetic variation and/or costs (summarized in Dworjanyn et al. 2006) of constitutive defence production in seaweeds, similar studies for induced seaweed responses and resistance remain to be performed. (v) Genetic and biochemical mechanisms of induced responses and resistance in seaweeds (Karban & Baldwin 1997; Kessler & Baldwin 2002). Most mechanistic work in seaweeds has been performed with pathogens (bacteria or parasites) as eliciting agents (Potin et al. 1999). Clearly there is a lot more work to be done on the nature of elicitors and signalling pathways, as well as expression of defence-related genes and the metabolic changes that result in herbivore-induced seaweed resistance.


Research has shown that terrestrial plants and herbivores are involved in complex, often finely tuned, interactions and that the plants are by no means passive players in this game. Terrestrial plants can sense and respond to different herbivore cues with elaborate changes that decrease their food value for herbivores, and that may have far-reaching consequences for higher trophic levels as well as for the structure of whole communities. Marine plants are also exposed to a considerable grazing pressure by herbivores ranging from large, mobile fishes and sea urchins to small, sedentary crustaceans and molluscs. Although research concerning different aspects of inducible defences in seaweeds is still in its infancy, the present meta-analysis clearly shows that seaweeds can also sense and respond to damage-related cues to decrease further herbivory. As such, there is no logical reason to assume that marine plant–herbivore interactions should be less complex, or that seaweeds should be less elaborate in their responses towards grazing, than their terrestrial counterparts. Previous research on seaweed–herbivore interactions has generated a thorough understanding of marine ecosystems on a community level. However, marine researchers would be wise to follow the progress made by their terrestrial colleagues and formulate more conceptual hypotheses regarding the mechanisms and consequences of induced seaweed resistance in response to herbivore damage.