Tag clouds have become a frequently used interaction technique in the web. Recently several approaches to present tag clouds with the tags semantically clustered have been proposed. However, it remains unclear whether the expected gains in performance and advantages in interaction actually can be realized as no empirical evaluations of such approaches are available yet. In this paper we describe a series of experiments designed to evaluate the effects of semantic versus alphabetical and random arrangements of tags in tag clouds. The results of our work indicate that semantically clustered tag clouds can provide improvements over random layouts in specific search tasks and that they tend to increase the attention towards tags in small fonts compared to other layouts. Also, semantically structured tag clouds were preferred by about half of the users for general search tasks. Tag cloud layout does not seem to influence the ability to remember tags. Author Keywords Tag clouds, folksonomy,...