Error-correcting codes that can effectively encode and decode messages of distinct lengths while maintaining a constant blocklength are considered. It is known conventionally that a k-dimensional block code of length n defined over $$\texttt {GF}(q^{n})$$ is designed to encode a k-symbol user data in to an n-length codeword, resulting in a fixed-rate coding. In contrast, considering $$q=p^{\lambda }$$, this paper proposes two coding procedures (for the cases of $$\lambda =k$$ and $$\lambda =n$$) each deriving a multiple-rate code from existing channel codes defined over a composite field $$\texttt {GF}(q^{n})$$. Formally, the proposed coding schemes employ $$\lambda$$ codes $${\mathcal {C}}_{1}(\lambda , 1), {\mathcal {C}}_{2}(\lambda , 2), \ldots , {\mathcal {C}}_{\lambda }(\lambda , \lambda )$$ defined over $$\texttt {GF}(q)$$ to encode user messages of distinct lengths and incorporate variable-rate feature. Unlike traditional block codes, the derived multiple-rate codes of fixed blocklength n can be used to encode and decode user messages $$\mathbf{m}$$ of distinct lengths $$|\mathbf{m}| = 1, 2, \ldots , k, k+1, \ldots , kn$$, thereby supporting a range of information rates—inclusive of the code rates $$1/n^{2}, 2/n^{2},\ldots , k/n^{2}$$ and $$1/n, 2/n, \ldots , k/n$$ ! A simple decoding procedure to the derived multiple-rate code is also given; in that, orthogonal projectors are employed for the identification of encoded user messages of variable length.