The correcting properties of concatenated codes with parallel decoding over an additive channel are investigated. The ith inner decoder's output is a codeword if the Euclidean distance between the received vector and some codeword is less than Delta /sub i/ and an erasure otherwise. The outer decoders correct errors and erasures. The error-correcting capability, which is taken to be the minimum length of any noise vector that can cause an error, is obtained for a bank of z inner and outer decoders as a function of the thresholds used. The set of thresholds that maximize the error-correcting capability is also found. It is shown that for a small number of branches, the error-correcting capability is nearly as large as any decoder. >