Despite concerted efforts by science educators to understand patterns of evolutionary reasoning in science students and teachers, the vast majority of evolution education studies have failed to carefully consider or control for item feature effects in knowledge measurement. Our study explores whether robust contextualization patterns emerge within particular evolutionary reasoning contexts, and the implications of these patterns for instruction, assessment, and models of cognition. We test four hypotheses regarding item feature effects on undergraduate biology majors' evolutionary reasoning using a sample of 1,200 open response explanations of evolutionary change across items differing in context and scale but standardized by taxon and trait. Evolutionary explanations were atomized into a series of scientific and naïve biological elements and tallied among prompts and their features. We documented clear, significant, and predictable item feature effects on evolutionary explanations. Tasks involving evolutionary trait loss elicited a significantly greater number of naïve biological elements than evolutionary trait gain tasks in all contexts, including: within species comparisons, between species comparisons, animal prompts, and plant prompts. Tasks involving between species evolutionary comparisons, regardless of gain or loss, animal or plant, always produced significantly more naïve biological explanatory elements than within species comparisons. For items prompting explanation of trait gain, the use of the core concepts of natural selection were not influenced by the hierarchical level of the task (within or between species). Explanations of trait gain were also the least sensitive to scale and context. Core concepts of natural selection were always deployed less frequently in cases of evolutionary trait loss (within and between species, in animals and plants). We discuss a series of implications of these findings for curriculum, instruction, and assessment. © 2010 Wiley Periodicals, Inc., Inc. J Res Sci Teach 48: 237–256, 2011