Coded modulation schemes based on codes for QPSK modulation are directly incorporated into QAM based modulation systems, forming trellis coded QAM, to provide a practical coding structure that is both efficient in bandwidth and data reliability. Concatenated coding with QPSK based trellis coding and symbol error correcting coding is used. In an encoder (Fig. 2), an N-bit QAM constellation pattern (80) is divided into four subsets, each including N/4 symbol points of the constellation pattern. A two-bit QPSK codeword (92) is assigned to each of the four subsets (82, 84, 86, 88). A symbol to be transmitted is first encoded using an outer error correcting encoding algorithm, such as a Reed-Solomon code (12). Part of the symbol is then encoded (48) with an inner code that comprises a rate 1/2 trellis encoding algorithm to provide a QPSK codeword, which is mapped (50) with the remaining bits of the symbol to provide a modulation function, wherein the remaining bits (94) correlate the symbol with one of the symbol points included in the subset defined by the QPSK codeword. At a receiver (Fig. 3), the recovered modulation function is pruned (62) to provide a set of metrics (66) corresponding to the subsets and to provide a plurality of conditional determinations of the constellation point identified by the remaining bits. The metrics are used in a rate 1/2 trellis decoder (68) to recover a first bit that is encoded using a rate 1/2 encoding algorithm to recreate the QPSK codeword. One of a plurality of the conditional determinations is selected in response to the recreated codeword. The selected conditional determination is combined with the recovered first bit to provide a decoded output that is further decoded using a symbol error correcting algorithm such as a Reed-Solomon code (36).