Welsh alphabet

The modern Welsh alphabet ("yr wyddor") contains 28 letters, of which eight are digraphs:

Diacritics

Welsh makes use of a number of diacritics.

The circumflex is used to mark long vowels. Thus "â, ê, î, ô, û, ŵ, ŷ" are always long, but "a, e, i, o, u, w, y" are not necessarily short. Not all long vowels are marked with a circumflex. A useful rule of thumb is that they are used particularly in monosyllabic words where the vowel is followed by "-l", "-n" or "-r". There are many exceptions to this, however.

The grave accent is sometimes used to mark vowels that should be short, when a long vowel would normally be expected, e.g. " _cy. pas" IPA|/pɑːs/ (a cough), " _cy. pàs" IPA|/pas/ (a pass/permit or a lift in a car); " _cy. mwg" IPA|/muːɡ/ (smoke), " _cy. mẁg" IPA|/mʊɡ/ (a mug).

The acute accent is sometimes used to mark a stressed final syllable in a polysyllabic word. Thus the words " _cy. gwacáu" (to empty) and " _cy. dicléin" (decline) have final stress. However, not all polysyllabic words with final stress are marked with the acute accent ("Cymraeg" "Welsh", for example, is written with none).

The diaeresis indicates that a vowel letter is to be pronounced fully, not as a semivowel, e.g. " _cy. copïo" (to copy) — pronounced IPA|/kɔˈpiːɔ/, not *IPA|/ˈkɔpjɔ/.

The grave and acute accents in particular are very often omitted in casual writing, and the same is true to a lesser extent of the diaeresis. The circumflex, however, is usually included. Accented vowels are not considered distinct letters.

Predicting vowel length from orthography

As mentioned above, vowels marked with the circumflex are always long, and those marked with the grave accent are always short. If a vowel is not marked with a diacritic, its length must be determined by its environment.

An unmarked vowel is long:
*in a stressed monosyllabic word when no consonant follows, e.g. " _cy. da" IPA|/dɑː/ (good)
*before "b", "ch", "d", "dd", "g", "f", "ff", "s", "th", e.g. " _cy. mab" IPA|/mɑːb/ (son), " _cy. hoff" IPA|/hoːf/ (favourite), " _cy. peth" IPA|/peːθ/ (thing)
*before "l", "n", "r" (in the case of "i", "u"), e.g. " _cy. sgil" IPA|/sɡiːl/ ("behind), " _cy. llun" IPA|/ɬɨːn/ (picture), " _cy. hir" IPA|/hiːr/ (long)
*in Northern dialects, before clusters of two consonants when the first one is "ll" or "s", e.g. " _cy. gwallt" IPA|/ɡwɑːɬt/ (hair), " _cy. tyst" IPA|/tɨːst/ (witness)

An unmarked vowel is short:
*in an unstressed (proclitic) word, e.g. " _cy. a" IPA|/a/ (and)
*before "p", "t", "c", "m", "ng", e.g. " _cy. cam" IPA|/kam/ (step), " _cy. llong" IPA|/ɬɔŋ/ (ship)
*before "l", "n", "r" (in the case of "a", "e", "o", "w", "y"), e.g. " _cy. tal" IPA|/tal/ (tall), " _cy. llen" IPA|/ɬɛn/ (curtain), " _cy. ffwr" IPA|/fʊr/ (fur)
*in Southern dialects, before clusters of two consonants, e.g. " _cy. sant" IPA|/sant/ (saint), " _cy. gwallt" IPA|/ɡwaɬt/ (hair), " _cy. tyst" IPA|/tɪst/ (witness)
*in Northern dialects, before clusters of two consonants when the first one is "n" or "r", e.g. " _cy. sant" IPA|/sant/ (saint), " _cy. perth" IPA|/pɛrθ/ (hedge)
*in Northern dialects, in any syllable that is not both stressed and word-final
*in Southern dialects, in any unstressed syllable

References