This paper derives a novel multiple symbols soft-decision metrics for frequency-shift keying signals which are affected by additive symmetric α -stable (S α S) noise and fading. The approximate metric, which is for the case where channel state information (phases, amplitudes, and noise dispersion parameter) is unknown is obtained based on a generalized-likelihood ratio (GLR) approach. The metric is obtained in closed form and proved to be effective. The performances of the multiple symbols soft-decision metrics are compared numerically for a turbo-coded system. The proposed multiple symbols metric provides substantial improvement over earlier single-symbol metrics.