A linear code over \(\text{GF}(q)\) is called divisible if the weights of its codewords have a common divisor \(\Delta > 1\). If \(\Delta\) is relatively prime to \(q\), the code is essentially a \(\Delta\)-fold replicated code. So the case of interest is when the characteristic \(p\) divides \(\Delta\). The author gives a new, character-free proof for his bound on the dimension of such a divisible code (originally proved in [\textit{H. N. Ward}, IEEE Trans. Inf. Theory 38, 191-194 (1992; Zbl 0746.94021)], by a character-theoretic argument) and discusses several applications, for instance to type I self-dual codes.