A graphemics parallel in every way to phonemics is rendered infeasible by several familiar considerations, such as (i) the partial dependence of graphies on phonemic form, (ii) the fact that graphic systems are of many different kinds, while all phonemic systems are of essentially the same kind, (iii) the relative artificiality of graphic systems. One consequence of the first consideration is that the term allograph is ambiguous. According to one linguist\ "the graphic shape of an allograph is dependent on its graphic surroundings". On the other hand according to the authors2 of a recent study of OlEnglish graphies in relation to phonemics, two graphs are allographs of the same grapheme if they represent the same phoneme, whether or not they stand in relations of complementary distribution, as they confess that the two "allographs" with which they are concerned do not (despite a general tendency in this direction). Presumably also for them, graphic resemblance would not count as a criterion; for instance the two forms of Greek sigma would be regarded as allographs whether or not one held there to be some similarity. 

A graphemics parallel in every way to phonemics is rendered infeasible by several familiar considerations, such as (i) the partial dependence of graphies on phonemic form, (ii) the fact that graphic systems are of many different kinds, while all phonemic systems are of essentially the same kind, (iii) the relative artificiality of graphic systems. One consequence of the first consideration is that the term allograph is ambiguous. According to one linguist\ "the graphic shape of an allograph is dependent on its graphic surroundings". On the other hand according to the authors2 of a recent study of OlEnglish graphies in relation to phonemics, two graphs are allographs of the same grapheme if they represent the same phoneme, whether or not they stand in relations of complementary distribution, as they confess that the two "allographs" with which they are concerned do not (despite a general tendency in this direction). Presumably also for them, graphic resemblance would not count as a criterion; for instance the two forms of Greek sigma would be regarded as allographs whether or not one held there to be some similarity.