It is known that a graph decision problem can be solved in linear time over partial k -trees if the problem can be defined in Monadic Second-order (or MS) logic. MS logic allows quantification of vertex and edge subsets, with respect to which logical sentences can encode many different conditions that an input graph must satisfy. It is not always clear, however, which graph problems can be expressed in such a way. In this paper we consider problems stated as logical conditions on subsets of the vertices and nonedges of the input graph. If such a problem can be defined in MS logic (i.e., in terms of the vertices and edges of the input graph), then there is a linear-time algorithm to solve the problem over partial k -trees. This algorithm also provides a solution to some problem over the graph-theoretic complements of partial k -trees. We study several examples of these ``complement-problems.'' We introduce a variation of MS logic in which, if a graph-problem can be defined over the class of partial k -tree complements, then there is a linear-time algorithm to solve that problem over partial k -tree complements, and (equivalently) a linear-time algorithm to solve its complement-problem over partial k -trees.