Constraint satisfaction technology emerged from AI research. Its practical success is based on integration of sophisticated search with consistency techniques reducing the search space by removing as many as possible inconsistent values from the problem. Singleton consistency is a meta-consistency technique that reinforces other pure consistency techniques by their repeated invocation and thus it achieves even better domain pruning. The paper presents a new algorithm for singleton arc consistency that improves practical time efficiency of singleton arc consistency by applying the principles behind AC-4.