Survey respondents are often asked to report their frequency of activity for particular categories of events or objects. In order to answer "How many magazines did you purchase last month?" one must determine which publications qualify as magazines and report a number for all of those items but no others. We conducted an experiment to explore how the abstractness of a category (Basic Level or Superordinate product categories) and the distinctiveness of its members (different products or the same one presented multiple times) affect the strategies and accuracy for frequency estimates. The results suggest that people count recalled products (an Enumeration strategy) if the products are distinctive, but rely on a non-numerical sense of magnitude (a General Impression strategy) when the products are not distinct. Superordinate product categories lead to underestimates, regardless of strategy, as do Basic Level categories when people enumerate. However, estimates based on general impressions appear to produce overestimates for Basic Level categories. We discuss the results in terms of data collection procedures.