I would like to add one thing to the discussion: There is an assumtion
that in unmarked sentences the grammatical concept of "subject", the
semantic concept of "agent" and the pragmatic concept of "topic" will
coincide. That is, the subject will also be the agent and the topic.
When they do not coincide, we have a marked sentence. If the
speaker/writer decides to topicalize some constituent other than the
subject agent ('That house I like!'), or demote the agent from subject
position, replacing it with, say, the object ('Jesus was baptized by
John') in order to make the object the topic, we have a marked sentence.
These sentences are, in normal human communication, marked because they
are structurally and cognitively more complex and less frequent. But
Michael Palmer brought up the interesting point of genre. Some genres or
discourse types are more marked than others (vis-a-vis "normal" every
day human communication), and structures that are marked in one
discourse type may well not be in another. So while Givon considers the
passive marked in normal human communication, he considers it unmarked
in typical academic discourse, mainly because of its frequency
distribution. Of course, academic discourse is a marked discourse type
(congnitively and structurally more complex and infrequent compared to
normal human communication).
This, of course, doesn't tell us which is the unmarked word order in
Greek, but it might contribute something to the discussion of what
markedness is.
Ron Ross
Department of Linguistics
University of Costa Rica
UBS Translation Consultant