An error-trellis is an efficient means to accomplish maximum likelihood decoding of a linear (n,k) block code C over GF(q). It represents the coset of C that contains the symbol-by-symbol detected version of the received vector. The trellis we construct has only n-k sections and the structure of each section is determined solely by the value of an associated q-ary symbol of the syndrome. Hence, the entire set of q/sup n-k/ trellises for all the cosets of C is formed by a set of q(n-k) distinct sections. The error-trellis is invariant under permutation of coordinates and attains the vertex count of a minimal code-trellis with the same sectionalization.