[An-lang] Proto-Dialect chains

Andy Pawley apawley at coombs.anu.edu.au
Fri Jul 18 03:42:30 UTC 2003

In response to comments by Isidore Dyen, John Terrell and Piet
Lincoln on the dialect chain issue:

Yes, Lynch and I were discussing more fundametal comparative issues.
John's query about a convenient nomenclature was not the basic

My jumping off point was a long-standing interest in using
subgrouping and dialect geography to draw inferences about where
earlier stages of a set of related languages were spoken and about
the nature of regional variation within such earlier stages.

We start with the well-known fact that regional variation in
languages is (in varying degrees) a problem for defining subgroups
and for reconstructing earlier stages. I had objected to certain
instances in which linguists had applied the term 'subgroup', without
qualification, to some cases where the distribution of innovations
was pretty messy.

Let us contrast two hypothetical situations where one might be
tempted to say that within a language family X, languages A-D form a
subgroup.  Let us also make the simplifying assumption that we know
which features are innovations of A-D as opposed to retentions from
proto X.

(1) The ideal subgroup. Languages A-D share a large body of
innovations apart from all other languages in family X.  All the
innovations are present in all members of A-D. Furthermore, there are
no innovations linking any subset of A-D. Finally, all four languages
are geographically far apart.  Such a pattern strongly indicates that
A-D stem from an earlier stage, proto A-D, in which there was
relatively little regional variation, and that each of A-D became
geographically separated at about the same time. In these ideal
circumstances, if a cognate set is shared by any two or more members
of A-D it can confidently be attributed to proto A-D.

(2). A messier situation. Languages P-S are linked only by many
overlapping innovations, such that certain innovations are shared by
P-Q, others by Q-R, others by R-S, and still others by P-R or Q-S.
Furthermore, P-S are geographically contiguous. Here we must conclude
that P-S diverged gradually from an old dialect chain, in which P was
next to Q, Q was next to R and R was next to S, in much the same
geographic relations as exists between the contemporary languages,
and that there was a long period during which innovations spread
along the dialect chain. At some stage there may have been a more
homogeneous language ancestral to P-S but we cannot tell this from
the available evidence.

Do we call P-S a subgroup?  Put another way, can we speak of P-S as
sharing a proto-language? More specifically, if we are trying to
reconstruct proto PS, what status do we give to those elements that
are shared only by a subset of P-S, say a cognate set found only in M
and N?

Actual cases may resemble types (1) and (2) more or less closely.
Malcolm Ross, has used the name 'innovation-defined subgroups' for
sets of languages that come close to type (1). Sets that show a
pattern resembling (2) he calls 'innovation-linked subgroups'.

I have some some reservations about using 'subgroup' for sets of
languages ,linked only by overlapping innovations. Certainly it is
the case that some reconstructions associated with type (2)
situations will represent different dialect regions and different
times. I think we should make a point of describing/labelling such
reconstructions in a way that shows their problematic status.

Isidore Dyen comments:

>  I see little point to trying to specify the type of membership of
>a protolanguage in its nomenclature. Ths is better handled in a list
>with indications of the interrelation of the members in a family tree
>that allows for multiple simultaneous branchings that indicate that we
>are unable to distinguish the order of the branchings because the
necessary data are lacking and in many cases will never be accessible.

There are certainly recalcitrant cases that fit the description in
Dyen's last sentence. But I think the type (2) situation, of
overlapping innovations that show a clear geographic pattern, is
somewhat different. Here we have no reason to believe that there was
an order of branching within P-S. Instead there was a gradual in situ
disintegration of a dialect chain.

Sometimes careful study of the data, combining the methods of dialect
geography with the Comparative Method, enables us to work out the
relative chronology and direction of the spread of particular
innovations along a dialect chain. There are many such cases
presented in Paul Geraghty 1983 book The History of the Fijian
Languages, which is surely the most brilliant application of dialect
geography to a part of the Austronesian family.

Andy Pawley

>The impression I have is that thre are two different things involved.
>The first is the the hypothesis of a prolanguage and the second is the
>membership in terms of languages and dialects in the subgroup emanating
>from that language. The protlanguage is in any case a dialect chain. No
>two speakers of the same language have identical idiolects (speech-
>types). A dialect is a collection of idiolects that share the same
>trait or collection of traits. A language is a collection of dialects
>(or idiolects)held together by a chain of peirs of mutually
>intelligible idiolects. A dialect-chain is a grouping of dialects that
>constitute a language. As a language your 'Ancestral..' is a
>protolanguage with a dialect membership that we can vaguely, if at all,
>see by way of reconstruction, not by imposing a shape to it, except as
>the procedures of reconstruction imply them. The kind of dialect chain
>that constituted the protolanguage is hard to infer from the data
>available without very detailed comparisons along the lines that you
>are engaged in and I hope you will continue to pursus. But as you can
>see, I see little point to trying to specify the ty[pe of membership of
>a protolanguage in its nomenclature. Ths is better handled in a list
>with indications of the interrelation of the members in a family tree
>that allows for multiple simultaneous branchings that indicate that we
>are unable to distinguish the order of the branchings because the
>necessary data are lacking and in many cases will never be accessible.
>I hope you will not take offense with my dealing with what I think of
>as the fubdamentals of comparative work, but i do believe that your
>problem as you presented it was best dealt with from a review of those
>fundamentals. Cordially and with best wishes for continued discussions.
>Content-type: multipart/alternative;
>  boundary="Boundary_(ID_FOsoyfDtB+6wNaT0bY0Rxg)"
>Content-type: text/plain
>Andrew Pawley and I have been having a little private e-discussion, but felt
>it might be better to come on-list to get a wider spread of ideas.
>The query relates to the nomenclature used for protolanguages and families
>as opposed to proto-dialect chains and linkages. One might refer to the
>Polynesian subgroup, for example, and to Proto-Polynesian, as implying a
>(fairly) homogeneous and uniform single ancestral language. Forms labelled
>PPn *xyz would thus represent forms which, to the best of our knowledge,
>were part of that protolanguage.
>My concern is with dialect-chains. There was probably a Southern Oceanic
>linkage, an dialect-chain ancestral to the languages of Vanuatu and New
>Caledonia. To use the same nomenclature for this proto-dialect chain - i.e.,
>Proto-Southern Oceanic - tends to imply that it is of the same nature as a
>more homogeneous protolanguage, which it is not.
>Nevertheless, it seems to me that it would be useful to have some fairly
>tight and neat way of referring to such ancestors, and not by some
>long-winded expression like "Ancestral Southern Oceanic dialect-chain" or
>some such, especially since one can legitimately also make reconstructions
>for such a "language", with the proviso that they are less secure and were
>probably more subject to internal variation than those made for
>My suggestion to Andy was to use lower-case p: proto-Southern Oceanic and
>pSOc as opposed tpo Proto-Polynesian and PPn. I felt this would be a
>sufficient indication of difference, and yet at the same time brief,
>succinct, and similar enough to existing conventions.
>Does anyone have any reactions to such a convention, or more importantly any
>idea on what others might have used?
>John Lynch
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.linguistlist.org/pipermail/an-lang/attachments/20030718/548561e5/attachment.htm>
-------------- next part --------------
An-lang mailing list
An-lang at anu.edu.au

More information about the An-lang mailing list