Graphemes and speech sounds

Units do not match one for one

A grapheme is a functional unit in an orthography; a glyph is a particular visible shape; a phoneme is a contrastive sound unit in a language. One grapheme can have several glyph forms, and one phoneme can be written by more than one sign or sign sequence. In English, the letters in “ship” are four visible alphabetic characters, but “sh” commonly works together to write one consonant phoneme. That observation does not turn the alphabet into a syllabary. It shows why counting letters, graphemes and sounds answers different questions.

A slash notation such as /ʃ/ describes a phoneme in a linguistic analysis. Square brackets such as [ʃ] are normally used for phonetic detail. A transliteration such as sh is a representation of the written form under a stated convention. The three may align in a simple example, but they remain different kinds of claim. Do not use a modern transliteration as proof of an ancient pronunciation.

Shape and value

The same encoded character can be drawn by several fonts, while one historical sign can have several uses. Unicode therefore distinguishes abstract characters from their rendered glyphs. A handwritten variation may matter to palaeography without creating a new character. Conversely, two signs that look alike on a damaged surface may be different characters in the local writing system. Compare repeated forms in the same witness before assigning a value.

The distinction becomes important when searching. A search for a transliterated value finds editorial readings, not every possible shape on an object. A search by code point finds encoded text, not photographs. For a reliable record keep the image location, sign identification, transliteration and linguistic interpretation linked but separate.

Check your analysis

For “ship,” list four levels: the printed shapes s-h-i-p; the writing sequence with sh treated as a sound-writing combination; the usual phonemic analysis /ʃɪp/ in a variety that pronounces it that way; and the spoken sounds of a particular speaker, which require an actual recording. The first three are analyses of a known spelling; the fourth cannot be recovered from print alone.