Data gathered from large-scale assessments in the Ontario Assessment Instrument Pool (OAIP) are examined. Implications for science instruction are to be found at the item level; the items should not involve more than two or three steps if the responses are to be informative. Items are collected and linked to provincial curriculum guidelines. These are given a preliminary trial in a few schools. When a few thousand items have passed the screening trials and reasonable coverage of the curriculum has been attained, many of the items are used in a field trial. An item sampling design is used to administer many of the items to a representative sample of students from the target population. Items were discussed in a set if they were coded to a topic and involved common content or technique. The sets were analyzed and assigned to one or more of seven categories according to the implications drawn. The categories are described and discussed. (DWH)