Categorization as a function of position of subordinate instance and superordinate class (e.g. cat-mammal vs. mammal-cat) was investigated in six experiments. Participant's task was to answer (by pressing yes/no key) whether an instance (e.g. cat) and a class (e.g. mammal) are categorically congruent. In Experiments 1, 3 an 5 instance was presented as the first stimulus in a pair, while in Experiments 2, 4 and 6 it was the second stimulus in a pair. In Experiments 1 and 2 instance was presented within a blocked design, while class was varied (e.g. mammal/insect), while in Experiments 3 and 4 the order was reversed, i.e. class was blocked while instances varied with respect to category (e.g. cat/bee). In Experiments 5 and 6 both instances and classes varied. The outcome of the six experiments indicates that categorization latency depends on order of presentation (instance/class vs. class/instance) and class (mammal vs. insect). Categorization time was slower if the first stimulus in a pair denotes a class. Conspicuous deviation was observed for the class of insects where congruent instance/class pairs were categorized slower than non-congruent pairs. There was also a three-way interaction (class x instance: class congruency x order of presentation).