This paper provides recommendations on experimental design for early-tier laboratory studies used in risk assessments to evaluate potential adverse impacts of arthropod-resistant genetically engineered (GE) plants on non-target arthropods (NTAs). While we rely heavily on the currently used proteins from Bacillus thuringiensis (Bt) in this discussion, the concepts apply to other arthropod-active proteins. A risk may exist if the newly acquired trait of the GE plant has adverse effects on NTAs when they are exposed to the arthropod-active protein. Typically, the risk assessment follows a tiered approach that starts with laboratory studies under worst-case exposure conditions; such studies have a high ability to detect adverse effects on non-target species. Clear guidance on how such data are produced in laboratory studies assists the product developers and risk assessors. The studies should be reproducible and test clearly defined risk hypotheses. These properties contribute to the robustness of, and confidence in, environmental risk assessments for GE plants. Data from NTA studies, collected during the analysis phase of an environmental risk assessment, are critical to the outcome of the assessment and ultimately the decision taken by regulatory authorities on the release of a GE plant. Confidence in the results of early-tier laboratory studies is a precondition for the acceptance of data across regulatory jurisdictions and should encourage agencies to share useful information and thus avoid redundant testing.