Since Eastman's pioneering study1), quite a few efforts have been made to clarify design problem-solving behavior. This study also takes advantage of Protocol Analysis method and its background data as many of those previous efforts did, and examines spatial design activities as a kind of problem-solving that accompanies drawing processes. The paper pays special attention to Eckersley's verbalization model2) and, by specifying its effectiveness and ineffectiveness in describing designers' verbalization during architectural design processes, proposes an improvement of the model to describe architectural design activities in -quasi?- real world. The results suggest representation-dependent model of design processes and emphasize, even in major architects' design processes, the shapes and figures already drawn on paper could trigger out the following steps toward unexpected development of their design.