Yes, classified documents have a wide range of markings, some
of which may include the lengthy citation Arnold cites but not all,
at least on not on every page.
And there are surely some which have not been declassified which
bear markings not yet seen in public.
Special weapons docs reportedly have markings that will finger
the person who reveals them, that is, each and every access to
them is indelibly logged by ID, time and place, in such a way that
any iteration made public can be identified, though if the charges
against Mr. Lee of LANL him are true, he was able to override this
lock (but maybe not, only that the system he erased was but step
one of the actual deeper system which will not be revealed in
open court, the public charges a campaign of disinfo).
The declassified docments I've seen usually have a single word
of classification at top and bottom of the page, and may contain
additional classification markings at paragraph titles and at each
paragraph itself to distinguish various levels of classification
of the page's content.
Some initial markings are revised by later ones but left in place
with a line through them, whether incidicating upgraded classification
or downgraded.
Change in classification laws and directives have led to changes
in classification categories and older documents will show the
evolutionary markings of these changes.
Levels above Top Secret are intriguing in which "Codeword"
typologies are utilized to further compartmentalize information
and access to it, such as "UMBRA," "NOFORN-ORCON,"
"GAMMA." In some cases the codewords are classified
because their names could reveal the importance, or point to
the existence, of secret information or operations. Recent
examples of these are the Codeword docs published by
Bill Gertz in "Betrayal," a few of which we offer on the Web:
http://jya.com/rudocs-gertz.htm
The basis for some type of classifications are not revealed to those
who may have custody of the material, in which declassification
requires approval of the agency which first classified it (marked
"OADR," that is, the reason for the classification is itself classified,
perhaps at a higher level than the material itself, say, because
the material was the product of an (ongoing) intelligence gathering
mission. As exemplified in William Simpson's FBI file, publicized
recently, in which he says learned that someone was informing
on his strong crypto advocacy within IETF. (Whether true or not
in Simpson's case, it is a tactic of the TLAs to allege the use of
informants in released docs to create suspicion and paranoia.)
A curious problem is that of several nations attempting to agree
on classifiation methodologies when they wish to share
information. Not all share interpretations of the value of
the information, or do not have the sames laws and directives
governing it. In these case, protocols are written and agreed
for a particular range of information. Lack of such an encrypted
communications protocol seems to have been the reason allies
of the US could not communicate with US pilots, air controllers
and targeteers as speedily as desired. The US was reluctant
to release TS comm data to allies who was suspected of
supporting the other side. Though, that too could be disinfo.