A simple decentralized stochastic control problem is considered where the nonclassical nature of the information pattern is induced by the uncertainty on the information transmission in the system. This is, in fact, a reformulation of the Witsenhausen counterexample, where the first station is allowed to send its information to the second station through a noisy channel. Nonconvexity of the problem in this new formulation has been established, and it is shown how this formulation relates to a classical problem and the Witsenhausen problem, respectively, when the transmission noise intensity goes to zero or infinity. Assuming small transmission noise intensity, we then use an asymptotic approach in order to find an approximated cost. A necessary condition for asymptotically optimal strategies has been obtained using a variational approach, and it is shown that the linear strategies, with slightly different coefficients than the noiseless transmission case, satisfy the necessary condition.