On a class of “impossible” figures: A new language for a new analysis

Abstract Multibar figures provide a class of illusions that have been used by psychologists as well as by artists like Escher. For scientific purposes one needs as precise a definition and analysis as possible for these figures. A first attempt at this has been made by Cowan (1974) with the theory of braids. The present work points out some semantic and technical problems not dealt with by Cowan, and proposes a new language for the study of multibar figures. This language is shown to be incomplete, and some further problems are reviewed.