bwald bwald at HUMnet.UCLA.EDU
Sun Mar 7 21:31:56 UTC 1999

----------------------------Original message----------------------------
I read Sally's message on the evidence for the evolution of Ma'a with great
Frankly, I did not remember exactly what Sally's position was when I
"mooted" the
direction of borrowing between Ma'a and "Bantu", e.g., Derek Nurse, an
energetic and prolific Bantu scholar with a deep interest in historical
linguistics expressed
a similar conclusion, i.e., borrowing out of, not into Bantu.  Sally's note
helps clarify
in my mind what criteria might be used to determine direction of borrowing,
in fact, I
think the details of the circumstances clarify what is meant by "direction"
of borrowing,
and are fundamentally more important than the notion of "direction" in the
context of
*fluent* (Bloomfield's "intimate") bilingualism.

She writes:
 (1) It's not a case of borrowing from a non-Bantu lg. into a Bantu lg., because
>if it were the basic vocabulary would be mostly Bantu, not Cushitic.

This is one criterion for direction, assumed on the basis of most common
and well-known
cases, e.g.,  French in English, Arabic in Swahili, etc.   Again, the
process of formation
of Mbugu is more interesting than extending the criterion to less clear
cases, as if it were
simply a matter of definition, devoid of social implications.  Hence she
gives further
arguments, e.g.,

>(2) It's not a case of shift from a non-Bantu lg. to a Bantu lg.,
>because again the basic vocabulary would be mainly Bantu, and besides,
>if there were enough imperfect learning to cause a significant residue
>of non-Bantu words even in the basic vocabulary, then there ought to
>be more syntactic & phonological interference, because shift-induced
>interference always affects the syntax & phonology most; but there is
>no distortion at all in the Bantu grammar of Ma'a.

Agreed.  "Imperfect learning" (of Bantu) is not at all an issue.  That much
I intended
in what I wrote.  Ma'a speakers are fluent in a Bantu language.  They are,
it seems,
fluent bilinguals in at least one Bantu language (usually Shambaa or Pare) AND
something else called Ma'a or Mbugu.  That was a crucial point to my
"mooting" of
direction of borrowing.  The case is clearly distinct from language
shifting, where
what is "borrowed" (actually "retained" = non-Bantu vocabulary) is
equivalent to
a "substratum", i.e., untranslated or untranslatable lexical items
usually because of cultural attachments, as when Mexican Americans retain
 "nina" (< madrina) for "godmother", and many more obvious cases).

[Actually, "fluency" in Ma'a may turn out to be a difficult concept.]

Sally's (3) and (4) can be ruled out for the same reason; they involve
"imperfect learning",
which is not the case.

>(5) it's not a case of relexification into a Bantu lg., a la Media
>Lengua (Quechua grammar + Spanish vocabulary), because there's not
>enough lexical replacement, and again, the basic vocabulary is mostly

This point is not so clear to me.  I mean I get it that it's not complete
but it seems to me there may be a continuum of degree of relexification
(depending on
what?).  Therefore, the rest of what Sally says is also important.

But in addition, the contacts between Ma'a and others have
>been with Bantu-speaking groups for a very long time; this
>can be determined by the oral histories of the Pare and the Shambaa.
>That is, there haven't been any intimate contacts with Cushitic
>speakers recently enough to account for the present and documented
>states of the Ma'a language.  And that's even aside from the fact
>that Maarten Mous has established that the non-Bantu Ma'a lexicon *can't
>possibly* be all from one Cushitic lg.: some of it is Southern
>Cushitic, some from another branch of Cushitic, and a large chunk of
>it comes from Masai (with whom the Ma'a speakers haven't been in
>contact any time lately).  (The question of what Ma'a was like
>*before* Bantuization is another and an intriguing story -- but
>alas, evidence of any solid kind is lacking, and probably always will
>be lacking.)

All of this is important, and VERY interesting.  One might suppose that Ma'a
continues a practice/tradition that has been carried on by its speakers
through a
number of (fluent) bilingual contact situations, each one of which is
reflected in
Ma'a vocabulary, presumably as distinct historical strata of contact.  Note
that there
is also Bantu vocabulary in Ma'a, according to Sally's criterion for
excluding it as
a case similar to Media Lengua.  Intriguing is the possibility that as Ma'a
has evolved,
it "always" proceeded by dragging its lexical baggage into the grammar of
the "new"
language, and then rejected most of that grammatical matrix as it moved on
to the next
situation -- e.g., that at one time Ma'a vocabulary was embedded in a
Maasai grammar,
left behind more easily with the loss of bilingualism in Maasai than its
Maasai vocabulary.
This is speculation to be rejected as unfounded if no evidence of this can
be found.
I am only suggesting it more abstractly as a possible process to account
for what is currently
(or recently) observable, and the accretion of lexical strata in Mbugu.
Thus, the Maasai
vocabulary may have come into Ma'a some other way without a period of
bilingualism (i.e., it never adopted Maasai grammar) -- it could be various
contacts with
different varieties of Cushitic that first started Mbugu off on its path --
where grammar
(and some lexicon) would not be so strikingly different at first.  (Where
grammar is not
that different, lexicon is most important to act as an ethnic symbol -- and
that is typical
even where grammar is quite different.)

>   Finally, there *was* some active non-Bantu grammar in Ma'a, both
>in the earliest good attestations ca. 1930 and in comparative material
>that shows chronological layering of Bantuization changes (e.g. Cushitic
>suffixes added *after* the Bantu-induced opening of word-final syllables).

This gets us into the issue of "inflectional" vs. "derivational" morphemes,
the latter may be more susceptible to borrowing, i.e., they are more easily
as "lexical" material.  This could stand further clarification before being
accepted as
evidence for a particular direction of borrowing.  All the grammatical
morphemes at
issue seem to be complete syllables and cause no phonological irregularities.

>Strikingly, the few Cushitic grammatical features that remained in the
>earliest materials are now gone, to judge by the most recent fieldwork
>(by Maarten Mous).  So what we can observe in the documentary record
>is the loss of the last relics of Cushitic grammar.  Even in the
>earliest records, the only Cushitic grammatical features that remain
>are things that fit well into Bantu typology: a causative suffix,
>pronominal possessive suffixes that aren't all that unlike the
>Bantu possessive formation, etc.  ...With one exception: a non-Bantu
>(probably Cushitic) collective suffix appears on a few nouns, sometimes
>alone and sometimes *with* a Bantu plural prefix, to indicate plurality,
>in the earlier materials.

Right.  That's the collective suffix -no, to which sometimes the Bantu
collective prefix
ma- is optionally added, e.g., (ma-)[lhare-no] "a group of clouds", cf.
ma-book-s reported
by Scotton I think for Sotho as an example of retaining English plural
inflection in using the
English *word* in a Sotho context.  Marking such nouns for a specific noun
class might be
difficult to resist in speech (as opposed to in citation forms) since they
might have to govern
a concordial form in speech, esp. a concordial subject-marker prefixed to
the verb -- which
would necessarily be Bantu, either for Sotho or Ma'a.

As a separate matter, I thought that all verb derivation
(even on a non-Bantu root) was Bantu in Ma'a, e.g., causative -ija. (Bantu
causatives have
various forms. -ija < *ed-i-a seems most likely, where *ed- is originally
the applied suffix
to which causative -i- is added and causes a sound change, now simply
morphologised as
-ija in Ma'a.  In contrast, Bantu languages always have at least a few
irregular causative
formations for certain verbs. Hence, it would be interesting if Ma'a used
the -ija causative
with a Bantu verb root where it is not used in Pare or Shambaa, but I doubt
that would happen.)

Also, though the phonology (even in the
>non-Bantu portions of the lexicon) has also been Bantuized, there is
>still one un-local-Bantu Cushitic phoneme, a lateral fricative; according
>to Mous, Ma'a speakers insert this even into Bantu words, by way of
>emphasizing the differentness of their special in-group language.

I guess that's the lateral fricative spelled "lh".  I could imagine Welsh
speakers playfully
doing that in English as well; "take me to the llake".  (The segment is
pre-Bantu in
the general area, and occurs in other non-Bantu languages, e.g., Sandawe, a
language classified as Khoisan.)

>   One other comment: Wald says that nobody has suggested dismissing
>Ma'a `as some kind of (in-group) "slang"': wrong.  This is what Mous
>says it now is, and he has good evidence.  (Mous and I differ, though,
>on how it got that way.)

Yes.  I thought I was making it up to exaggerate my point, but then Dixon's
comment on
Mous's work was called to my attention.  At least Dixon's characterisation
of what Mous was
saying struck me as going pretty much in the direction of my point.  So
maybe this is the
"end of the line" for Ma'a, and that it is going to end up similarly to the
way Ian Hancock
represents the Romani spoken in England, which does seem to be
superficially similar to Media Lengua, i.e., English grammar but Romani
vocabulary (probably with various stratified accretions).

>   A general point can be made here: the routes by which even the
>most exotic mixtures emerge aren't necessarily undiscoverable; if
>we have enough information, we can often figure out how mixed languages
>got that way.  Sorting out the effects of contact isn't different
>in that respect from sorting out the effects of internally-motivated
>change.  "Enough information" is of course the crucial requisite.

Except that enough information in this context seems to include enough
social information,
which also was part of my point in reconstructing and separating out
"compressed" strata
when we do not know the social situation which produced those strata (in
the event that
they are strata and not a single "pure" -- unlikely by everything we know
-- "rock-bottom"
stratum that validates even deeper reconstruction under the "single
language" hypothesis.

I am grateful for Sally's scenario, which I think will help me remember
what she means
in saying the direction of borrowing was FROM Bantu TO Ma'a (Mbugu).  It is Ma'a
that continually shows the pressures of language shift due to bilingualism,
but survives
(though "scathed" by replacement of "original" vocabulary and grammar) as
its speakers
move on to a new bilingual contact situation (including from one Bantu
language to another).

I still have reservations about how to clearly distinguish the bilingual
processes involved
 from the Media Lengua or British Romani type of process -- I expect closer
to British
Romani, where again it might be supposed that there was a gradual (?)
grammatical loss
of the ethnic language with retention of much of the ethnic lexicon,
including pronouns
 and other "basic" vocabulary  (pronouns = a CLOSED lexical set, cf.
affixes during a specific period of time).  Presumably Media Lengua did not
evolve this way
but was a "sudden" flushing of Qechua vocabulary in  favor of Spanish,
while retaining
the Qechua grammar as the "host" for relexification.  Of crucial
importance, I'm sure, is that
Media Lengua flushes out the *ethnic* language, NOT the NON-ethnic
language.  Thus,
different social circumstances were involved in such a process, e.g., Mbugu
and British
Romani seem to imply a situation originating in intimate bilingualism and
some kind of
NOMADISM (nomadism leading to a particular solution in preserving ethnicity
  -- a life-style most likely nearing an end for Ma'a speakers  -- unlike
the Maasai they do not
have a "reserve", i.e., a reservation).

In any case, I completely agree with Sally that we must understand the
processes involved
in all cases of language "mixture" (and that that concept conflates a
number of different
evolutionary possibilities), if we are ever going to be able to apply that
knowledge plausibly
to megalo-reconstructions in order to separate out deep strata and resolve
the gnawing
problems of "borrowing" vs. "genetic" inheritance.  I stand by for further
clarification, esp
what social circumstances account for Media Lengua, but also for further
details of the
considerations I raised for Ma'a. Media Lengua is clearly different from
Mbugu or Romani
because it is the "ethnic" lexicon which is lost in Media Lengua, while it
is the NON-ethnic
lexicon that is flushed out in Mbugu.  (I have also heard some question
raised about the
"stability" of Media Lengua -- though I'm not quite sure what that means --
other than that
some think that it may be a transient phenomenon -- but does it necessarily
have to be?)
Without being able to distinguish what is "ethnic" and what is not, e.g.,
in a deep reconstruction,
we would have a hell of a time distinguishing the two types.  Thanks to
scholars like Sally,
and many others, e.g., Muysen for ML, we know there are at least two types
of Lexicon1 + Grammar2.

P.S.  In view of the clarification that Sally has given, I'm glad that my
original message chose
Mbugu/Ma'a rather than the equally thorny problem for deep reconstruction
given the type
Gumperz described for Kupwar, a converged grammar and three distinct
(by different means, perhaps, Dixon describes a similar situation for some
parts of Australia,
where, it seems, speakers have a large number of synonyms for basic words,
which comes
in handy when someone dies and one of the basic words that sounds like
their name must be
avoided for a while; such an avoidance custom also in some Bantu-speaking
areas; unclear
in Dixon's cursory discussion was whether the synonyms are perceived as
part of a single
language or organised into different over-all sets, as in the case of
Kupwar, whether or not
each set is associated with a different language label -- I'm sure he
discusses that more fully
in some other work that has not yet come to my attention).  The issue for
deep reconstruction
remains the social inferences we make when combining families into deeper
families.  This
remains the persistent point of deep reconstruction, as far as I can tell,
and has been since the
origin of the genetic hypothesis.  When we hear that, say, Afrasian and IE,
or whatever, are
"related", we immediately think ah! that means there was once a "single"
culture that later
fragmented into various known cultures, i.e., to the extent that we do not
necessarily assume
that the current speakers are themselves relevantly biologically related in
any way to the
members of that single culture.

More information about the Histling mailing list