Skinner's definition of verbal behavior, with its brief and refined versions, has recently become a point of controversy among behavior analysts. Some of the arguments presented in this controversy might be based on a misreading of Skinner's (1957a) writings. An examination of Skinner's correspondence with editors of scientific journals shows his sophisticated mastery of English and his knowledge of contemporary approaches of linguistics, and might help to settle the meaning of the passages involved in the controversy. A more precise definition of verbal behavior, deduced from Skinner's distinction between verbal and nonverbal operants, is suggested, and a possible reason why Skinner did not define verbal behavior in the terms proposed by this alternative definition is discussed. The alternative definition is more compatible with a functional approach to behavior and highlights what is specific to verbal behavior by pointing to the conventions of the verbal community. Some possible consequences of adopting this alternative definition are described. (Contains 1 table and 9 footnotes.)