LINGUIST List 5.1111

Wed 12 Oct 1994

Misc: Comparative method, "go+verb", "linguist", Thanks

Editor for this issue: <>


  1. Jacques Guy, Limits of the comparative method
  2. Matthew Dryer, Comparative method
  3. amy uhrbach, go + verb
  4. Arnold Zwicky, GO + verb
  5. , RE: 5.1087 Linguist
  6. , Thanks for your response

Message 1: Limits of the comparative method

Date: Wed, 12 Oct 1994 16:48:17 Limits of the comparative method
From: Jacques Guy <>
Subject: Limits of the comparative method

Alice Faber:
> I agree with Alex Manaster-Ramer that there is IN PRINCIPLE no limitation to
> the depth reachable by the comparative method.

This "in principle" worries me deeply. I might say there is IN PRINCIPLE no
upper limit to earth population. But what does that mean? The mass of the
earth being finite, and humans having mass, human population is finite.
Therefore there is an upper limit, even though it is variable, and
unknowable. Likewise the depth reachable by the comparative method, as
Alice Faber is about to argue:

> Let's
> assume that you need a sufficiently large number of cognates for each
> correspondence set that you establish.

True assumption, but already problems arise, barely solvable, too.
How large is sufficiently large? At least larger than the number
of spurious cognates due to chance resemblances. The number of chance
resemblances is itself a function of

1. phonological decay.
2. the number of languages compared.
3. the size of the vocabularies used.
4. the range of semantic shifts allowed when looking for cognation.

I am particularly concerned at (1): phonological decay. Phonological
decay from what? Usually, from a reconstructed protolanguage.
Its reconstruction is seldom certain (an understatement, I should
think), and depends on more recent data, which itself can be
reconstructions. Thus, the further back the reconstruction,
the greater chances of errors, some of them due to ... chance
resemblances! I hope you see now what worries me. Unless one
pays close attention to the effects of chance resemblances
at every stage, one is certain to see their effects cumulate,
swamping true cognates, and distorting the reconstruction
process so that what you eventually end up with is an
artifact of a sloppy procedure.

> Given infinitely large corpora for the various Afroasiatic languages
> and for languages in families to which Afroasiatic might be related,
> of course we could reconstruct back earlier than 10,000-15,000 years
> ago.

I am not even sure of that. We are certain to have a proportion of chance
resemblances, whatever the sizes of the corpora. Therefore, the reconstruction
is limited to the point when when false cognates outnumber true ones.
Remember that, although we can estimate, very roughly, the number of
false cognates which can be expected to crop up, we cannot tell them
apart from true cognates. Therefore, it seems to me that there is an
upper limit to reconstruction, whatever the sizes of the corpora.
That we do not know how to compute this limit, that it is certainly
variable, a function of intractable factors, does not mean that it
is not there.

> But we don't have infinitely large
> corpora; in fact much subgrouping within Afroasiatic is based on word-lists
> less than 1,000 words long.

It is certain that the shorter the word lists, the less can be known about
the past, since the only evidence we have for it is the "transmission
errors" (innovations, if you prefer), and one list item can only record
one transmission error, even though it may have undergone many.

In conclusion, there is no upper limit that we can calculate, but
an upper limit there certainly is, even with infinite-length wordlists.
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue

Message 2: Comparative method

Date: Tue, 11 Oct 1994 23:58:21 Comparative method
From: Matthew Dryer <>
Subject: Comparative method

I think it might be useful to try to be clearer on what is involved in the
recent debate between Alexis Manaster Ramer and Johanna Nichols regarding
limits on the comparative method. That the usability of the comparative
method fades as we move back in time is relatively uncontroversial. But
the applicability of the comparative method to a given group of languages
is surely a function not only of the time depth of that group but also of
the number and diversity of languages in the group. The success of the
comparative method with Indo-European is due, not only to its time depth,
but to the number of branches and the number of languages in many of those
branches. (Of course, the availability of data on earlier languages has
had an important role as well.) If Indo-European had had fewer branches,
and ones with few languages, say just Greek, Albanian, and Armenian, then I
assume Indo-European might have been somewhat beyond the range of the
successful application of the comparative method. For a family with such a
structure, the limit of the comparative method is presumably less than 6000
years. Conversely, if a family were to have sufficient number of branches
and sufficient number of languages within each branch, there is no reason
to believe that the comparative method would not work well beyond 6000
years. As an extreme hypothetical case, imagine a family with time depth
12000 years with 100 branches each with 100 languages. It does not seem
implausible that the comparative method might work successfully for such a
group, despite its time depth. Of course, there are no families of such a
size. It is even conceivable that for practical purposes, the families
that exist in the world have properties such that the comparative method
does not work well beyond a time depth of, say, 8000 years. But such a
limit would not be a limit on the comparative method itself, but rather a
limit on the practical applicability of the comparative method to the set
of existing languages. Furthermore, it is far from clear that we are
currently in a position to conclude that the set of existing languages is
such that it does have this property. Thus even stating a practical
limitation at this point seems premature.

Matthew Dryer
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue

Message 3: go + verb

Date: Wed, 12 Oct 94 7:57:13 EDTgo + verb
From: amy uhrbach <>
Subject: go + verb

 Richard Ingham <> wrote:
> I assume it's the case that for American speakers _ go_ can't be used in
> the habitual present:
> *I go play golf on Saturdays
> Is this true?
No, this is perfectly natural to me. However, my husband thinks this is
strange, so it is somewhat dialectal. I grew up in the NYC area, and he
grew up in DC and Detroit. Who else can get these constructions?
 -Amy Uhrbach
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue

Message 4: GO + verb

Date: Wed, 12 Oct 94 08:20:47 EDGO + verb
From: Arnold Zwicky <>
Subject: GO + verb

G. K. Pullum, "Constraints on intransitive quasi-serial
verb constructions in modern colloquial English" (in B. D.
Joseph and A. M. Zwicky, eds., OSU WPL 39, 1990, _When
Verbs Collide: Papers from the 1990 Ohio State Mini-Conference
on Serial Verbs_), discusses the GO/COME + verb construction
in some detail, providing (a) a survey of the traditional
grammarians' discussions of the construction, (b) a survey
of generative grammarians' discussions of it, (c) a list
(fairly considerable) of similar but distinct constructions,
some of which have been mentioned in the LINGUIST postings on
the topic, (d) a phrase structure grammar analysis of the
construction, and (e) an account of the (*very* considerable)
dialect variation with respect to the "inflection
condition" on the construction (which bars *She goes stare(s)
a lot, and similar examples, for most speakers). [The
flow chart on p. 234 has the right shape, but the labels
on the last line are out of order; restoring them to their
right order is an interesting exercise for the reader.]

In any case, this is the place to start for any further

arnold zwicky (
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue

Message 5: RE: 5.1087 Linguist

Date: Wed, 12 Oct 1994 10:24:14 RE: 5.1087 Linguist
From: <>
Subject: RE: 5.1087 Linguist

Just a note to point out that the Joint Publications and Research
Service (JPRS), a government agency which provides translations for
other government agencies, uses the term "linguist" to refer to
someone in their organization who knows a particular language and
selects articles for translation by one of their independent
translators. So it's no surprise that the general used the word
"linguist" in the way he did.

Alan Stevens
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue

Message 6: Thanks for your response

Date: Tue, 04 Oct 1994 16:22:50 Thanks for your response
From: <>
Subject: Thanks for your response


I just want to thank all of you who responded to my request for e-mail
penpals (or as one person noted, "e-mail keyboard pals"). I will be
writing each of you individually.

I would also like to say that I have enough volunteers, so I am sorry if
you were interested and had not gotten a chance to let me know. Maybe
next semester?

Thanks again. I really appreciated the enthusiastic response I received.

Rebecca Kastendick
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue