The errors that can occur in a computer system during data reading or recording in the memory can be of different types depending upon the memory organization. They can be random bit errors or burst errors. Therefore, if high reliability is required, the use of an error correcting technique that will be able to handle both types of errors is necessary. In this study the capability of some classes of error correcting codes is analysed and their performance with both types of errors is tested. Reed Solomon and concatenated codes are examined in more detail because they are known to be among the best classes of codes. In order to evaluate the performance of these codes two well known classes of codes are used: BCH codes and Fire codes. The performance of all the codes with regard to random error correction is analysed using a binary symmetric channel model. BCH codes are shown to be more powerful for average codeword length, but as the codeword length increases RS and concatenated codes perform better than BC3 codes of the same rate of transmission. A new model for systems with burst errors is introduced with which a large variety of real systems can be simulated by choosing the appropriate distributions of burst errors. The performance of all these codes at correcting burst errors is simulated using this model. It is shown that RS codes and concatenated codes are very powerful with burst errors and can increase significantly the reliability of a signaling system incorporating these types of errors. An advantage of RS codes compared to concatenated codes is that they can be very easily implemented and can be employed efficiently for systems with any codeword length. Concatenated codes can perform better than RS codes only when very long codewords are required. / M.S.