Purpose - To investigate how far sociocyberneticians have tried to meet the criterion of "empirical testability" in their publications. What changes in this respect have taken place in recent years? Design/methodology/ approach - Inventory of papers published in Kybernetes from 1977 to 2004, and classification of sociocybernetic papers by type, i.e. the degree to which empirical testability plays a role in the argument. Findings - The number of sociocybernetic papers has increased. The purely conceptual, theoretical or methodological papers make way for papers in which particular social systems and processes are analyzed, using the conceptual apparatus of cybernetics. However, the percentage of papers in which empirical data are used for the testing of hypotheses and models has decreased. It seems that the testability problem of sociocybernetics is unsolvable. Originality/value - These findings lead to a reflection on the way in which sociocybernetics could eventually overcome the testability problem, by more methodological sophistication, on the one hand, and a more parsimonious formulation of sociocybernetic theories, on the other hand. © Emerald Group Publising Limited.