Peter T. Daniels wrote:

> Richard Wordingham wrote:
> [...]
> > Do marks expressly introduced to create new letters at will count as
> > diacritics? The nukta in Devanagari and the prime in modern Hebrew
> > are such marks.
>
> It has been suggested (I forget by whom) that the <h> in the English
> digraphs ch, sh, th is a diacritic.

But, if the "¨" in Swedish "ä" or the "-" in "G" are not diacritics (and I
agree they aren't), then also the "h" in English "sh" shouldn't be.

All the marks mentioned above are functional in making something ("ä", "G",
"sh") different from some other thing ("a", "C", "s"), but the point is that
this difference is not *systematic*. I mean: the difference between "s" and
"sh" is not the same as the difference between "t" and "th", so one cannot
say *what* "h" would be a diacritic for.

On the other hand, in the context of German morphology, the difference
between "ä" and "a" is the same as the difference between "ö" and "o" or "ü"
and "u". That's why, I think, we can call German "¨" a diacritic: because
there is a systematic difference which is consistently indicated by this
mark, so one can say that German "¨" is the diacritic *for* umlaut.

According to this characterization, even Hebrew geresh (as used in
trasliterating foreign words) and Indic nukta are not diacritics: just
graphic elements which recur in newly created letters.

--
Marco