A comment on a conjecture of N. Wiener (Q1003792)

Let \(X_t\) be a stationary process and le \(\mathcal{B}_t=\mathcal{B}\{X_s,s<t\}\) be the \(\sigma\)-filed generated by the r.v.'s \(X_s, s<t\). \textit{N. Wiener} [Nonlinear problems in random theory. Chapman and Hall (1958; Zbl 0121.12302)] posed the question of under what circumstances a stationary process \(X_t\) could have a one-sided representation \(X_t=f(\xi_t,\xi_{t-1},\cdots)\), in terms of iid r.v.'s. It was conjectured there that the necessary and sufficient condition for such a representation to hold was that the backward tail field \(\mathcal{B}_{-\infty}=\cap_t\mathcal{B}_t={\emptyset,\Omega}\) be trivial. This note shows that there are stationary sequences \(X_t\) with trivial tail field that cannot have such a one-sided representation in terms of iid r.v.'s.