Why Transliteration Systems Look the Way They Do

A transliteration of Devanagari has to represent thirty-three consonants and fourteen vowels using an alphabet with twenty-six letters, and it has to do so reversibly if scholars are to cite texts precisely. Every serious system solves the problem the same way, by adding diacritics that stand for the features the table is built on: a dot beneath a letter for retroflexion, an acute or a dot for the sibilants, a macron for vowel length, and a following h for aspiration. Once the features are marked, the mapping is one to one and a reader who knows the system can write the Devanagari back out from the Latin.

Systems that avoid diacritics, which are common in popular publishing and in the spelling of names, cannot do that: they merge the three sibilants, lose vowel length, and produce letter pairs that English readers pronounce as English. Both kinds of spelling are in circulation for most of the words this atlas covers, which is why the same term appears in several forms in the general literature. Nothing in this is a fact about the letter itself, whose value is the voiceless aspirated dental and whose place is second in the fourth class, but it is the main practical difficulty a reader meets when moving between English sources and the script, and it is worth stating plainly.