Best practice suggests the use of controlled vocabulary. Don't confuse this element with a figure caption. A caption is text primarily intended for display with an illustration. It may or may not function as a description of the illustration.
<elementSpec xmlns="http://www.tei-c.org/ns/1.0" rend="add" ident="pgDesc" module="MEI.shared"><gloss xml:lang="en">page description</gloss>
<desc xml:lang="en">Contains a brief prose description of the appearance or description
of the content of a physical page.</desc>
<content>
<zeroOrMore><choice><rng:text/>
<ref name="model.graphicPrimitiveLike"/>
<ref name="model.textComponentLike"/>
<ref name="model.annotLike"/>
<ref name="model.locrefLike"/></choice></zeroOrMore>
</content>
<remarks xml:lang="en">
<p>Best practice suggests the use of controlled vocabulary. Don't confuse this element
with a
figure caption. A caption is text primarily intended for display with an illustration.
It
may or may not function as a description of the illustration.</p>
</remarks></elementSpec>