Re: Initial 's' in Brittonic

From: richardwordingham
Message: 14204
Date: 2002-08-02

--- In cybalist@..., guto rhys <gutorhys@...> wrote:
> Can anyone explain why Brittonic retained initial 's' in some words
while mutating it to 'h-' in others. I am aware that this change is
dated to the Roman occupation in Britain as words borrowed from Latin
retain the 's' (`'saeth' - arrow, 'sych' - dry etc.). It seems not to
be related to the nature of the following vowel as we have 'saith' -
seven < *se- and 'hy' strong/proud/rude < seg-. Do similar situations
exist in other IE languages?

Yes, similar situations do exist.

In early Modern English, the pronunciations of 'ea' and 'ee' merged.
However, 'steak', 'break' and 'great' are exceptions, merging
with 'ai' and long 'a'. The 'r' in 'break' and 'great' has been
blamed for the divergence, but it is not enough in itself -
consider 'bream', 'breathe', 'grease' and 'greave'. It appears to be
due to dialect mixture - at one time the speech of London merged 'ea'
with 'ai'. I think (European) dialect mixture may explain contrasts
such as `Ronald Reagan' - like `ai' - but `Donald Reagan' -
like `ee' - in the same US administration.

In Modern German, quite a few Low German forms (as diagnosed by their
initial consonant) have displaced the inherited High German forms.
Unfortunately, I can't think of any examples at the moment.

In Latin, intervocal PIE bh and dh did not always yield /b/ or /d/;
there are forms with f instead. An example is 'rufus' and 'ruber',
both meaning 'red' and deriving from the PIE stem h1reudH. (VCr
counts as intervocalic - see .) A closely
related Italic dialect (Faliscan?) always had /f/ between vowels.

These all appear to be genuine cases of dialect mixture, which can be
comparative linguistics for "I don't know". In each case, the
aberrant form from another dialect was consistent with the phonology
of the borrowing dialect. (Latin had intervocal /f/ in compounds,
e.g. 'refero' from 'fero', and in reduplicated perfects - e.g.
fefelli from fallo.)

Brythonic had two parallel changes: s- > h-, and st- > s-. Thus, IF
some dialects did not lenite s- to h-, they could lend forms
retaining s- to dialects that had lenited them. The recipient
dialect could interpret the s- as being the sound it had developed
from st-. Having said this, I wonder if there may be a reason for
preferring saith to *haith for 'seven' if the choice were made
available by dialect mixture. Other discussions on this list about
this sound change, e.g. in pre-Iranian, have proposed a path s > ch >
h, where 'ch' denotes the guttural fricative. (Thus the evolution
of sw- simply stops at chw-, which cluster is retained in some
dialects of Persian.) What would the form of the cluster in the word
for 'seven' be? Would it still be -ct-, or would it already have
been -cht-? How would counting proceed? 'Six, seven' would be
something like *chwech checht- with the the regular forms, not a
pleasant sequence! I can imagine *secht- being preferred to *checht-
. (I am using a dash because I do not know how the word would have
ended.) Also, *chwech (asterisked because it might have actually
have been *chwechs or *chwecch rather than already being chwech) and
*checht- MAY have been inconveniently similar.

At this point, a nasty thought occurs to me. If *chwech and *checht-
were inconveniently similar, could the Latin form 'septem' have been
borrowed? As Latin 'captivus' gave Gallo-Roman 'cactivus', (from
which we get English 'caitiff',) Latin 'septem' would have been given
a form like *sechtem, possibly indistinguishable from the form in a
hypothetical non-leniting Brythonic dialect. (I don't think it is
useful in this context to note that Latin is a non-leniting Italo-
Celtic dialect!) I called the thought `nasty' because we do not like
to think of Indo-European languages borrowing numbers. It has been
observed that numbers are unusually conservative in Indo-European

I wonder if there may be a pattern to the retention of `s-'.
Welsh `saeth' and `sych' need not have been borrowed late (assuming
indeed that they are not cognate). I suspect we have a pattern sV- >
hV- except before ch! (Each sC-combination is a law unto itself - sw-
> chw-, sr- > ffr-, st- > s-, sk- > ysg, sl- > ll, though sl- and sw-
are similar to sV.)

What is the trajectory of `saeth'? Is it (ignoring permutations)

sagitta > sagetta (a-affectation) > sagetha > saghetha (lenition) >
sagheth (final vowel loss) > saeth


sagitta > sagta (syncopation) > sacta > sachta > sacht (vowel loss) >
sachth > saith, and then respelt? (I never got to grips with Welsh

If the latter, we may be seeing preference for over (c)
when given the choice!

If the former, we could try saying sV- > hV- except before guttural
fricatives, but hy < seg- is an exception to that exception. To say,
on the basis of the initial consonant, that `saeth' but not `sych'
was borrowed late would be to cheat.

Is there any mileage in this suggestion of dissimilation? I have
very little data on Welsh, so I cannot test this hypothesis myself.
By the nature of the mechanism, it may be that many exceptions have a
common pattern that is shared with non-exceptions, like `break'
and `great' in English. There may also be residual exceptions,
like `steak'.

The dialect mixture argument would be much stronger if we could show
there were dialects that did not undergo the change. However, a
conservative dialect might function as a non-leniting dialect before
it ultimately underwent the change. A poor example that comes to
mind is the fate of Germanic sk- in English. Old English softened it
(ultimately yielding Modern English sh-), but then hardened word
forms were re-introduced from or under the influence of Viking
settlers, yielding word such as `skin', `skirt' and the myriad words
in `scr-'. However, Scandinavian (or at least, Norwegian and
Swedish) have now softened sk- before front vowels. I am uncertain
of the validity of this example; the similar Scandinavian and English
sound changes may be unrelated.

Does s- > h- apply to the Brythonic dialects of Scotland? (I am
embarrassed to say I cannot remember the name of the Gaelic-
influenced Brythonic dialect that long survived in Galloway.)

Languages also undergo sporadic sound changes. For example, the
change of intervocalic dH to Latin b referred to above, in detail the
process dH > *θ > *ð > *v > b (or dH > *θ > *f > *v > b - is ther=
any evidence to tell between them?), where *v is NOT the sound
written <v> in Latin, did not always occur. There are instance where
the change went dH > *θ > *ð > d, i.e. the change from dental
fricative to labial fricative was not universal intervocally.
Examples taken from include
gradus `step' from PIE *gHredH and foedus `league' and fides `trust'
from PIE *bHeidH. The change *θ > *f looks like a sporadic change
that occurred often enough to have become a sound law for initial
consonants and, by Piotr's account, before /r/. It will have become
a sound law once pre-Latin no longer had /θ/. /ð/ was finally
eliminated by merger with /d/.

However, I do not think s- > h- in Brythonic should be viewed as
originally sporadic, for Brythonic did not have an /h/ so that a
contrast of s- and h- would distinguish words. I believe the change
would have affected all words before the initial /h/ (or /ch/) would
have been associated with the initial consonant of what was once -ct-
rather than with the /s/ of sk-. A partially comparable case of non-
association is the loss of /s/ before voiceless consonants in French,
e.g. fenestre > fenêtre. The intermediate sound is believed to have
been [h], but there is no evidence that it was associated with the
still-pronounced hache aspiré, as in 'la haie'. On the other hand,
the loss of both instances of [h] in French could well have occurred
on a word by word basis.