In we developed a Viterbi algorithm for efficient detection of M-ary phase-shift keyed (MPSK) sequences received over the additive, white, Gaussian noise (AWGN) channel with an unknown carrier phase. Its performance approaches that of coherent detection, but the observation interval for forming the decision metric is fixed and chosen based on prior statistical knowledge of the carrier phase characteristics. Thus, the metric is non-adaptive, and cannot be optimized when a priori statistical knowledge of the carrier phase is not available. The work here renders it adaptive by developing a recursive metric that is adapted on-line based on the received signal, without prior knowledge of the carrier phase characteristics. An adaptive filter is used to generate the phasor reference on-line, by extending that in to sequence detection for general M-ary quadrature amplitude modulation (MQAM). Comparison between the proposed adaptive, recursive metric and those in the literature is provided. Simulation results using a random-walk carrier phase model show the superior performance of the adaptive sequence detector over the original nonadaptive sequence detector of.