The WWP uses <hi> to encode letters or words which are renditionally distinct for reasons which are either decorative or apparently accidental, rather than linguistically motivated. The most frequent examples of this usage are wrong-font letters (individual characters in a different font within a word), or for words highlighted at the beginning of each paragraph as a decorative feature; parts of words which are highlighted solely because of a line break (as on title pages, where a word may span a line break and the two lines may be in different fonts); and individual words on a title page which are highlighted for decorative or display reasons.
The <hi> element is also used to encode the varying rendition of punctuation or words within a phrase-level element. For instance, if a <persName> element surrounds an entire phrase, some of which is highlighted, <hi> should be used to encode these renditional shifts. In general, <hi> is used to designate highlighting which does not convey linguistic content; thus the renditional characteristics recorded on <hi> could be suppressed on output without thereby losing any indicators of textual meaning.
We do not use <hi> as a holding tank for elements awaiting further thought. This is a terminal tag. Words whose encoding is uncertain should be encoded with <unknown>.
Where individual letters within a word are intentionally highlighted to convey a special meaning (as for instance in the prophetic tracts of Eleanor Davies), we encode them using <emph>. (See 062 on emphasis.)
The WWP has added a type= attribute for <hi>. The only value currently specified for this attribute is “dic”, which is used to designate distinctive initial capitals (capital letters which are decorated or oversized).