Error detecting polynomial codes are usually formed by defining a correspondence between data bits and coefficients of the representative polynomial. These codes are easily implemented in hardware using shift registers; however, implementation in character-oriented processors may be too time consuming. A new class of polynomial codes is described for which a correspondence between n-bit data characters and polynomial coefficients is defined. Two particular types of these "character polynomial codes" are discussed; these may be easily implemented with hardware or with processor character manipulations. The burst error detecting ability of these two types of codes is shown to be the same as for the common "bit polynomial codes."