In linguistics an allograph is one of the forms of a grapheme, a written character. The "letter A" is a grapheme, and individual realizations like A, a, a, and A are its allographs. A graph is a single written form. Allographs are graphs that belong to the same grapheme.

Some graphs are in free variation: the choice of typeface usually doesn't make any significant difference, nor do the variations that occur in handwriting. Other allographs are conditioned: in English and languages written like it, capital letters and italics usually occur for narrowly specific purposes (though they can also be used just for general effect).

The Greek letter sigma has a capital allograph Σ, and two lower-case allographs, σ and ς, the latter used only at the end of a word. Several Hebrew letters also have word-final allographs. In addition, the modern Hebrew handwritten alphabet is strikingly (indeed, incomprehensibly) different from the traditional printed one: allographs need not be merely minor variants. In Arabic script, which is cursive, all letters have between two and four allographs depending on their position in the word. The Roman alphabet used to have two lower-case allographs of s, the long s being abandoned about 1800.

In a case-sensitive computer language, A and a are different graphemes. In fact you could take this view in ordinary language too, and say that A and A are allographs of the capital grapheme, distinct from the lower-case one. (I find you can never be dogmatic about these definitions in linguistics.)