A possible oldest example of Man'yōgana is the iron Inariyama Sword that was excavated at the Inariyama Kofun in 1968. In 1978, X-ray analysis revealed a gold-inlaid inscription consisting of at least 115 Chinese characters and this text, written in Chinese, included Japanese personal names which were supposedly phonetically written. This sword is thought to have been made in the year 辛亥年 (471 A.D. in commonly accepted theory). There is a possibility that the inscription of the Inariyama sword may be written in a version of the Chinese language used in the Korean-peninsula kingdom of Baekje.
Man'yōgana uses kanji characters for their phonetic rather than semantic qualities—in other words, they are used for their sounds and not their meanings. There was no standard system for choice of kanji; different kanji could be used to represent the same sound, the choice made on the whims of the writer. By the end of the 8th century, 970 kanji were in use to represent the 90 morae of Japanese, far more than what was needed. For example, the Man'yōshū poem 17/4025 was written as follows:
In the poem, the sounds mo (母, 毛) and shi (之, 思) are written with multiple, different characters. While all particles and most words are represented phonetically (e.g., 多太 tada, 安佐 asa), the words ji (路), umi (海) and funekaji (船梶) are rendered semantically.
In some cases, specific syllables in particular words are consistently represented by specific characters. This usage is known as Jōdai Tokushu Kanazukai. This usage has led historical linguists to conclude that certain disparate sounds in Old Japanese, consistently represented by differing sets of man'yōgana characters, may have merged since then.
In writing utilizing man'yōgana, kanji are mapped to sounds in a number of different ways, some of which were straightforward and others which are less so.