2. The same Roman type is always to represent the same foreign letter, and the same foreign letter is always to be represented by the same Roman type.
3. Simple letters are, as a rule, to be represented by simple, compound by compound types.
4. It is not attempted to indicate the pronunciation of foreign languages, but only to represent foreign letters by Roman types, leaving the pronunciation to be learnt, as it is now, from grammars or from conversation with natives.
5. The foundation of every system of transliteration must consist of a classification of the typical sounds of human speech. Such classification may be more or less perfect, more or less minute, according to the objects in view. For ordinary purposes the classification in vowels and consonants, and of consonants again in gutturals, dentals, and labials suffices. In these three classes we distinguish hard (not-voiced) and sonant (voiced) consonants, each being liable to aspiration; nasals, sibilants, and semivowels, some of these also, being either voiced or not-voiced.
6. After having settled the typical sounds, we assign to them, as much as possible, the ordinary Roman types of the first class.
7. We then arrange in every language which possesses a richer alphabet, all remaining letters, according to their affinities, as modifications of the nearest typical letters, or as letters of the second and third class. Thus linguals in Sanskrit are treated as nearest to dentals, palatals to gutturals.
8. The manner of expressing such modifications is uniform throughout. While all typical letters of
