PARENT SESSION
Contributed Oral Session 14: Evolutionary Ecology: Theory; Modeling
Monday, August 8, 8:00 AM - 11:30 AM, Meeting Room 521 A, Level 5, Palais des congrès de Montréal
A test of the stress-gradient hypothesis using meta-analysis.
Lortie, Christopher*,1, Callaway, Ragan2, 1 York University, Toronto2 University of Montana, Missoula
ABSTRACT- Abiotic gradients and the general concept of stress have a long history of use in the ecological literature, primarily to infer the relative importance or intensity of interactions between plants. In a more recent derivation of these ideas, the stress-gradient hypothesis has been proposed which predicts that the relative importance of facilitation or positive plant interactions increases with increased stress along a hypothetical gradient. Direct empirical tests of this idea are not extensive; however, meta-analysis or a quantitative comparison of weighted effect sizes is now possible. We used the following two sources of data to test the generality and precision of the stress-gradient hypothesis: (i) a data set from a published study with 10 replicated identical manipulations globally (at high and low stress sites) and (ii) a subset of the published literature with similarly stratified within-study levels of stress. Where possible, three metrics of effect size were tested including Hedges d, the log response ratio (LRR), and the relative neighbor effect (RNE). In most cases, biomass was the response measure of interest although other measures were also explored. Trend analyses in both major data sets suggested that there was no publication bias, i.e. sample size did not predict effect size, nor significant heteroscedascity in that the variance and effect sizes across studies were not positively correlated. More importantly however, the variability associated with effect sizes at higher stress sites was significantly lower than the variance associated with effects detected at lower stress sites. This suggests that positive interactions commonly detected at higher stress sites are less variable than negative interactions occurring at lower stress sites. The use of meta-analytic techniques did not lend itself to the same general interpretation as traditional approaches, i.e. that facilitation necessarily increases with increased stress - in spite of the fact that within each study there was statistical evidence for a change in sign from negative to positive. This difference is extremely intriguing and necessarily leads us to further explore the generality and precision of the stress-gradient hypothesis. In short, the statistical generality of this idea is extremely robust while the precision is lacking.
Key words: stress, gradient, plant community, alpine