uliar phonetic features as I have mentioned could have evolved
independently in neighboring groups of languages.

[Footnote 168: Ugro-Finnic and Turkish (Tartar)]

[Footnote 169: Probably, in Sweet's terminology, high-back (or, better,
between back and "mixed" positions)-narrow-unrounded. It generally
corresponds to an Indo-European long _u_.]

[Footnote 170: There seem to be analogous or partly analogous sounds in
certain languages of the Caucasus.]

How are we to explain these and hundreds of similar phonetic
convergences? In particular cases we may really be dealing with archaic
similarities due to a genetic relationship that it is beyond our present
power to demonstrate. But this interpretation will not get us far. It
must be ruled entirely out of court, for instance, in two of the three
European examples I have instanced; both nasalized vowels and the Slavic
"yeri" are demonstrably of secondary origin in Indo-European. However we
envisage the process in detail, we cannot avoid the inference that there
is a tendency for speech sounds or certain distinctive manners of
articulation to spread over a continuous area in somewhat the same way
that elements of culture ray out from a geographical center. We may
suppose that individual variations arising at linguistic
borderlands--whether by the unconscious suggestive influence of foreign
speech habits or by the actual transfer of foreign sounds into the
speech of bilingual individuals--have gradually been incorporated into
the phonetic drift of a language. So long as its main phonetic concern
is the preservation of its sound patterning, not of its sounds as such,
there is really no reason why a language may not unconsciously
assimilate foreign sounds that have succeeded in worming their way into
its gamut of individual variations, provided always that these new
variations (or reinforced old variations) are in the direction of the
native drift.

A simple illustration will throw light on this conception. Let us
suppose that two neighboring and unrelated languages, A and B, each
possess voiceless _l_-sounds (compare Welsh _ll_). We surmise that this
is not an accident. Perhaps comparative study reveals the fact that in
language A the voiceless _l_-sounds correspond to a sibilant series in
other related languages, that an old alternation _s_: _sh_ has been
shifted to the new alternation _l_ (voiceless): _s_.[171] Does it follow
that the voiceless _l_ of language B has had