Abstract.A variety of universal algebras is called deductive if every subquasivariety is a variety. The following results are obtained: (1) The variety of modules of an Artinian ring is deductive if and only if the ring is the direct sum of matrix rings over local rings, in which the maximal ideal is principal as a left and right ideal. (2) A directly representable variety of finite type is deductive if and only if either (i) it is equationally complete, or (ii) every algebra has an idempotent element, and a ring constructed from the variety is of the form (1) above. This paper initiates a study of varieties of universal algebras with the property that every subquasivariety is a variety. We call such varieties deductive. In §2 it is proved that the variety of left modules of a left Artinian ring R is deductive if and only if R is the direct sum of matrix rings over local rings in which the maximal ideal is principal as a left and right ideal. In §3 we consider varieties of universal algebras of finite type that are directly representable. A variety is directly representable if it is generated by a finite algebra and has, up to isomorphism, only finitely many directly indecomposable finite algebras. Such a variety is deductive if and only if either (i) it is equationally complete, or (ii) every algebra has an idempotent element and a certain ring, constructed from the variety, has the properties listed above.Deductive varieties were first considered by those studying the varieties arising in algebraic logic. (The Russian equivalent for "deductive" would be "hereditarily structurally complete, in the finitary sense.") Roughly speaking, subvarieties correspond to new axioms added to the logic under consideration, while subquasivarieties correspond to new rules of inference. Thus, in the logic associated with a deductive variety, every rule of inference can be replaced by a set of logical axioms. Some other results on deductive varieties can be found in Tsitkin [19] on varieties of Hey ting algebras, and in Igosin [10] on some deductive varieties of lattices. This paper got its start with the consideration of discriminator varieties, which falls under the heading of algebraic logic. From there the material in §3 developed, reducing the problem of characterizing those directly representable varieties that are deductive to the corresponding problem for finite rings.