P3 does not foresee the inclusion of verse within prose paragraphs. Many occurrences of verse in prose paragraphs are quoted passages, and do not present encoding difficulties. However, when the verse passage is attributable to the same narrative voice as the surrounding prose paragraph, the use of Q and QUOTE is inappropriate. The two common definitions of a paragraph consider it to be a unit of content or typography. Another approach considers the occurrence of "something else" (e.g. poetry) to constitute the end of a paragraph. While probably sufficient for most formatting functions, this does not adequately describe the conceptual structure of the passage, i.e. a group of verse lines embedded in a prose paragraph. Since the description of both the physical and the conceptual text is an important goal of the encoding, this definition was unsatisfactory. Possible solutions included: (1) using 2 P elements and an LG and linking them using JOIN or next= and prev=; (2) nesting the LG within TEXT within P; and (3) modifying the content model of P to allow LG as its direct child. The latter emerged as the most accurate and efficient solution. It is consistent with the typographical and thematic continuity of the material and thus with the most prevalent definitions of a paragraph. It facilitates the accurate description of concrete examples already found as well as unconventional ones with more heavily intermingled verse and prose. Finally, it minimizes the tagging necessary to describe the passage and its structure.