Contemporary assessment practices in science education have undergone significant changes in recent decades. The basis for these changes and the resulting new assessment practices are the subject of this two-part paper. Part 1 considers the basis of assessment that, more than 25 years ago, was driven by the assumptions of decomposability and decontextualization of knowledge, resulting in a low-inference testing system, often described as "traditional." This assessment model was replaced not on account of direct criticism, but rather on account of a larger revolution-the change from behavioral to cognitive psychology, developments in the philosophy of science, and the rise of constructivism. Most notably, the study of the active cognitive processes of the individual resulted in a major emphasis on context in learning and assessment. These changes gave rise to the development of various contextual assessment methodologies in science education, for example, concept mapping assessment, performance assessment, and portfolio assessment. In Part 2, the literature relating to the assessment methods identified in Part 1 is reviewed, revealing that there is not much research that supports their validity and reliability. However, encouraging new work on selected-response tests is forming the basis for reconsideration of past criticisms of this technique. Despite the major developments in contextual assessment methodologies in science education, two important questions remain unanswered, namely, whether grades can be considered as genuine numeric quantities and whether the individual student is the appropriate unit of assessment in public accountability. Given these issues and the requirement for science assessment to satisfy the goals of the individual, the classroom, and the society, tentative recommendations are put forward addressing these parallel needs in the assessment of science learning.