Abstract. A family {M p } p∈R d of cylindrical measures is constructed on the space of functions ( = trajectories) f : [0, +∞) → R d such that for every t ≥ 0 the formularepresents a solution of the Cauchy problem(with respect to the required function ψ : The integral kernels ("Green's functions") of the corresponding solution operators, which can be approximated (using Trotter's formula) by integrals of finite multiplicity of the expressions explicitly defined by the ingredients of the original equation, are (matrix-valued) transition measures that give cylindrical measures M p similarly to the way Markov transition probabilities give the distribution of a Markov process.A new method using matrix-valued transition measures is applied in this paper, leading to a stronger result for the Dirac equation than those that follow under assumptions analogous to ours from results in previous papers. In [5], uniqueness of solutions is not claimed; in [11] and [12], functional integrals are understood only as generalized integrals; in [13] (and in earlier articles by Ktitarev), the authors proceed under assumptions that are more restrictive when applied to our situation. This method can also be adapted as a generalization, to the case of infinite-dimensional algebras, of values of the coefficients (generally speaking, they depend on the "space" variables) occurring on the right-hand side of the evolution equation; in particular, for the second-order super-differential equations presented in [5], the results of which do not overlap those described below, and in which, in order to construct a functional Poisson distribution analogous to the measure M 0 , the authors use not finite-dimensional distributions (= Trotter approximations) but the Dyson series analogous to that used in [1] to solve the Schrödinger equation. Our method goes back to the similarity indicated in [14] between the properties of complex