A data processing apparatus is operable to identify one of a plurality of code words present in a watermarked version of a material item. The marked version is formed by combining each of a plurality of parts of a code word with one of a plurality of units from which the material item is comprised. The apparatus comprises a recovery processor operable to recover at least one part of the code word from a corresponding unit of the marked material item, and a correlator. The correlator is operable to generate for the marked material unit, a dependent correlation value for the part of the code word recovered from the material unit and the corresponding part of at least one of the re-generated code words from the set. A detector is operable to determine whether at least one of the code words is present in the marked material item from the dependent correlation value for the part of the code word exceeding a predetermined threshold. The data processor may detect the presence of the code word with improved probability, in particular when parts of the material have been corrupted.