This paper studies sequence modeling for prediction tasks with long range dependencies. We propose a new formulation for state space models based on learning linear dynamical systems with the spectral filtering algorithm [HSZ17]. This gives rise to a novel sequence prediction architecture we call spectral state space models. The resulting models are evaluated on synthetic dynamical systems. These evaluations support the theoretical benefits of spectral filtering for tasks requiring very long range memory.