A method of recovering a value of a symbol received through an optical communications system. A multi-bit estimate of the symbol is subdivided into a first part and a second part, each part including at least one respective bit of the estimate. A most likely value of the first part is detected. The most likely value of the first part is processed using a Forward Error Correction (FEC) decoder to generate a corrected first part value, which is used to detect a most likely value of the second part. The most likely value of the second part is then processed by the FEC decoder to generate a corrected second part, which is combined with the corrected first part to recover the value of the symbol.