Most researchers working to develop a comprehension process fail to distinguish the different levels often recognized by linguists. Some, for example, have attempted to incorporate a composite form such as 'has' into one terminal symbol without addressing its internal complexity. The present study in contrast distinguishes different levels. The context-sensitive dispute is thus resolved by viewing complex structures as composites of multiple context-free layers that are interrelated by context-sensitive interfaces. In the design of a comprehension system, this study recognizes the essential structural differences between the linguistic information and the computer procedure. Thus, the linguistic information is not incorporated into the program itself. Examples are illustrated from Korean and English.