Spatial semidiscretization of the linear stochastic wave equation with additive noise, using the spectral Galerkin method, is studied. Optimal order error estimate for the deterministic problem is obtained under minimal regularity assumption, that is used to prove the optimal rate of strong convergence O(h^\beta), \beta \in [0,2]. Comparing with the result of the finite element approximation, the spectral Galerkin method enjoys higher convergence rate due to lower regularity requirement. The theory is illustrated with numerical examples in one and two spatial dimensions
