One of the oldest questions in software methods is: How do you break down a large system into smaller systems? We ask this question because as systems get large, it becomes difficult to understand them and the changes we make to them.
Structured methods used functional decomposition, in which the overall system was mapped as a function and broken down into sub-functions, which were broken down further into sub-sub-functions, and so forth. The functions were like the use cases in an object-oriented system in that functions represented something the system as a whole did.
Those were the days when process and data were separated. So in addition to a functional decomposition, there was also a data structure. This took second place, although some Information Engineering techniques grouped data records into subject areas and produced matrices to show how the functions and data records interacted.
It is from this viewpoint that we see the biggest change that objects have wrought. The separation of process and data is gone, and functional decomposition is gone, but the old question still remains.
One idea is to group the classes together into higher-level units. This idea, applied very loosely, appears in many object methods. In the UML, this grouping mechanism is called the package.