Ecological Archives E095-004-A1

Laura Radville, Liahna Gonda-King, Sara Gómez, Ian Kaplan, Evan L. Preisser. 2014. Are exotic herbivores better competitors? A meta-analysis. Ecology 95:30–36.

Appendix A. Details of the identification and selection of studies and statistical analyses used to determine if exotic herbivores are better competitors.


Identification and selection of studies: To identify suitable papers, we first used the Web of Science database to search for studies citing either Denno et al. (1995) or Kaplan and Denno (2007). We examined all of these studies both for their suitability and for pertinent studies in their ‘literature cited’ section. Second, we used the Web of Science database to perform keyword searches with all combinations of the terms used by Kaplan and Denno (2007). These phrases included: interspecific competition, insect herbivore, induced plant defense, induced plant resistance, plant-mediated interactions, and exploitative competition. We checked the citation lists of the resulting papers and examined any potentially-relevant publications for inclusion in the database. Third, we used the Web of Science database to search for papers citing any of the publications found in the first two steps.

Data collection: We defined the control for each interaction as the focal herbivore’s performance in the absence of a competitor, and the treatment as the focal herbivore’s performance in the presence of a competitor. If a study provided data on multiple herbivore pairs, or on the same herbivore interaction in substantively different conditions (high vs. low densities, competition on different host plant species, etc.), we treated these studies as independent data points in order to address the range of possible conditions affecting the strength of competition. If a study provided multiple measurements of a given response variable over time, we chose the median time point.

We defined spatial separation as occurring when herbivores were unable to feed in the exact same location. For example, herbivores that exclusively feed on fruits are spatially separated from herbivores that feed exclusively on roots. Additionally, herbivores that were caged on different leaves of the same plant were considered spatially separated. We defined temporal separation as occurring when the competing herbivore was introduced or settled naturally on the host plant prior to the arrival of the focal herbivore.

All lines from agricultural systems were excluded when assessing impacts of coevolutionary history between host plant and focal herbivore. Since the herbivore native/exotic status and plant–insect coevolutionary history were rarely reported in the papers themselves, we conducted extensive literature searches into the native ranges of both herbivores and their shared host plant in order to locate this information. We obtained all native range data from peer-reviewed publications or government websites. We defined plant–insect pairs as sharing a coevolutionary history if (A) they are both native to the study area; or (B) if they are both exotic but share a native range. Plant–insect pairs did not share a coevolutionary history if (A) one is exotic and the other is native; or (B) if they are both exotic but have different native ranges.

Statistical analyses: In order to reduce the error inherent in small-sample-size analyses, comparisons with fewer than five observations in each group were excluded. Additionally, 12 observations with effect sizes greater than three and less than negative three were excluded as outliers.

With random effects models, there is assumed to be random variation in effect sizes between studies, rather than a true effect size across all studies. This random component of variation in effect size is integrated into the summary statistics (Rosenberg et al. 2000). Random effects models were used in this study because we compared studies across a variety of ecosystems and species pairs, and it is more realistic to assume variability in effects sizes than to assume that the effect size is the same for all species pairs.

We assessed publication bias in our data set by performing Rosenthal’s fail-safe test (Rosenberg et al. 2000) and by visually assessing weighted histograms, funnel plots, and normal quantile plots of the variables. In order to examine bias of certain species pairs being represented more often than others, the same analyses were run using a more conservative, truncated database. In the truncated database, one data point was randomly selected per pair of competing species per response variable. For each response variable, this produced a smaller database in which each species pair was equally represented.

Literature cited

Rosenberg, M., D. Adams, and J. Gurevitch. 2000. MetaWin: Statistical software for meta-analysis (v. 2.1.4). Sinauer Associates, Sunderland, Massachusetts, USA.

[Back to E095-004]