You may change as many things as you like with any kind of the three classes mentioned above. The sequence of processing is like this: The window is checked from top to bottom for lines that have been changed. A line even is regarded as changed if there only was some character replaced by itself. The kind of change is concluded from the position of the line in the window, from its previous and its current content.
A line that was changed and in which there was no part of the workfiles (i.e. beyond the data zone or within the data zone above the first line used or beneath the last line used), is always interpreted as a command.
For any other changed line the following rule is valid: If the number area has the same content as it had before, a direct change was made. If an accepted prefix command is identified in the changed number area, the change is recognized as a prefix command. In advance to this, the data area, which may have been changed, is treated as a direct change. Contrastingly, if there is no valid prefix command in the number area, the whole line is treated as a command.
If the explanations above are too abstract for you, you should carefully work through the examples in the chapter First Steps.