Most major conceptualisations of product involvement locate it on the level of the product category. The aim of the present research was to assess whether this is empirically valid, or whether an alternative conceptualisation of involvement - located on the level of the product attributes - provides a better explanation of observed variations in product involvement. An experimental design was used that allowed the separation of category-level variation in involvement from attribute-level variation in involvement. The design was tested in four super-categories of consumer products (clothing, foods, household appliances, alcoholic beverages) and on two involvement dimensions (interest and importance). The results indicate that in three of the four super-categories the predominant source of variation in product involvement were the attributes of the products, not the product category they belong to. Only for foods, the traditional conceptualisation of involvement held. The findings suggest that parts of previous research on involvement may have to be re-interpreted and that future research should adopt more complex conceptualisations of involvement. From a managerial point of view, this research indicates that a clear focus on product attributes may increase the effectiveness of marketing communication and thus increase customer attraction.