Abstract
The paper concentrates on further generalizations of logic-oriented neurons as introduced in Pedrycz (1993) by furnishing them with additional numerical facilities. Some relationships between these new constructs and the concepts linguistic modifiers are studied. Furthermore we point at some essential links between fuzzy neurons and a well-known OWA aggregation operator. In sequel, we bring up new architectural aspects of fuzzy neural networks including those aimed at representing and processing uncertainty associated with the input data. A fuzzy controller is studied in this context with a main emphasis focused on its performance achieved in presence of a genuine linguistic information coming from the environment under control.