Submitted to: Meeting Proceedings
Publication Type: Proceedings
Publication Acceptance Date: August 13, 2004
Publication Date: N/A
Technical Abstract: Critical issues that influence the ability of a field trial to detect possible effects of transgenic crops include the selection of appropriate indicator taxa, replication of treatments, plot size, and data analysis. Selection of indicator taxa is the most basic requirement for non-target field research. General criteria for selection of taxa should include likelihood of exposure or susceptibility to expressed Cry proteins, functional importance, abundance, and ecological diversity. Adequate replication is another key consideration for field tests of non-target effects because of the relationship between replication and power. Simply put, the power of a test is the probability of detecting a real difference among treatments. While power may be affected by choice of indicator groups, sampling method, or timing of sample collection, power is consistently enhanced by increasing replication. Small plots may give misleading results in part because various non-target species establish in or re-colonize disturbed areas at different rates. However, limitations of time and labor may dictate that replicated scaled-down plots within a single field be used as a substitute for field-sized replicates. For robust evaluation of mobile arthropods, a minimum plot size that is both practical and experimentally sound should be used. An appropriate statistical treatment of data ultimately determines the value of time and effort invested in field research, but suitable options depend on the specific issues to be addressed. Researchers might be interested in comparing transgenic crops to other management options based on abundance of non-target herbivores, rates of predation or parasitism, the composition of functional groups or the entire arthropod community; these comparisons also may be made at one or two specific points in time, or over a series of dates. This diversity of hypotheses makes recommending a standard approach to analysis inappropriate, but problems with common methods in non-target analysis should be considered.