Let B-mu nu(a) (a = 1,..., N) be a system of N free two-form gauge fie
lds, with field strengths H-mu nu rho(a) = 3 partial derivative([mu)B(
nu rho)(a)] and free action S-0 equal to (-1/12) integral d(n)x g(ab)H
(mu nu rho)(alpha)H(b mu nu rho) (n greater than or equal to 4). It is
shown that in n > 4 dimensions, the only consistent local interaction
s that can be added to the free action are given by functions of the f
ield strength components and their derivatives (and the Chem-Simons fo
rms in 5 mod 3 dimensions). These interactions do not modify the gauge
invariance B-mu nu(a) --> B-mu nu(a) + partial derivative[(mu)Lambda(
nu)] of the free theory. By contrast, there exist in n = 4 dimensions
consistent interactions that deform the gauge symmetry of the free the
ory in a non trivial way. These consistent interactions are uniquely g
iven by the well-known Freedman-Townsend vertex. The method of proof u
ses the cohomological techniques developed recently in the Yang-Mills
context to establish theorems on the structure of renormalized gauge-i
nvariant operators.