Is "satem" a gradient?

[petegray]

There are two potential gradients:  sibilation and merger.

A sibilation gradient certainly exists, between "centum" languages showing
one reflex of *k' or *k, and one of *kw (eg Latin), and an ideal "satem"
language showing one reflex of *k', and one of *k or *kw (eg Indic, almost).
This gradient falls entirely within the group that merge original *k and
*kw, and is based on the degree to which all original *k' have maintained
their palatalisation, rather than merging with the common reflex of *k and
*kw (Baltic - Slavic - Iranian - Indic).

Armenian and the Luwian group of Anatolian do not fit this gradient, because
in these languages it is not a question of how far original *k' sibilates.

Armenian sibilates all *k', keeps all *k plain, merging them with *kw before
a back vowel, and differently sibilates *kw (*gw, *gwh) before an original
front vowel.    Because of the sibilation and the merger of most *k and *kw,
it could be called a satem language which maintains the distinction of some
*kw from *k by sibilation.

The Luwian group sibilates *k', and, like all the Anatolian languages,
maintains the difference between *k and *kw.    Because *k and *kw are never
merged, it could be called a centum language group which maintains the
distinction of *k' from *k by sibilation.

Thus on a merger gradient, we would have only Armenian - all the others show
complete merger or complete separation of *k and *kw.

Peter