Most approaches for recovering objects from procedural code are exclusively
based on static information. These approaches have the advantage to build on
information easily available. However, with code that is not built around
very strong and ubiquitous data structures, substantial portions of this code
cannot be clearly assigned to one of the objects the object-oriented system
is made of.

Here, we discuss an approach where the uncertainties that necessarily appear
with purely data-structure driven approaches can be reduced by establishing
in addition to the static object model a dynamic object model. The merits
of the approach extend object recovery though. We consider the creation
of event-state diagrams also useful for software redocumentation as
well as for general program understanding.