LINGUIST List 5.1006

Mon 19 Sep 1994

Disc: Comparative method

Editor for this issue: <>


  1. Johanna Nichols, Re: 5.936 Sum: Time ceiling on the comparative method, Case grammar

Message 1: Re: 5.936 Sum: Time ceiling on the comparative method, Case grammar

Date: Tue, 6 Sep 1994 13:47:57 -Re: 5.936 Sum: Time ceiling on the comparative method, Case grammar
From: Johanna Nichols <>
Subject: Re: 5.936 Sum: Time ceiling on the comparative method, Case grammar

Alexis Manaster-Ramer, writing on Aug. 31 about upper limits on the
comparative method, says:

>In her book, Nichols (1992) gives 8000 as the limit in one place
>(p. 6) but 10,000 in another (p. 184). In neither of these places
>does she give any argument or reference. In a third place, she
>gives no number, but does cite two references (one of which I have
>yet to check, the other being unpublished):
> The notion of a cut-off point beyond [read: up to] which the
> standard comparative method applies, and up to (but not
> beyond) which we have regular systematic sound correspondences
> and a substantial etymological base, is articulated in similar
> terms in a number of works (e.g., Austerlitz 1980; Jacobsen
> 1989....), but available language surveys rarely identify this
> cutoff point....

[See his posting for his references.]

There seem to be three criticisms here: inconsistency about ceiling cited, no
argument, no references. I actually gave what I thought was an adequate
characterization of the evidence and argument early on:

"But the comparative method does not apply at time depths much greater than
about 8000 years (this is the conventional age of Afroasiatic, which seems
to represent the upper limit of detectability by traditional historical
method) ..." (pp. 2-3)

The limit of 10,000 is mentioned this way: "Since the comparative method
is valid only for time depths up to about 10,000 years..." (184)

"Not much greater than about 8000" and "up to about 10,000" aren't really
different quantities, given that both are rounded and qualified as
approximate. The passage quoted by Manaster Ramer (from my p. 265), and
the others he mentions, deal with sample design and typological research
method, and weren't intended to describe comparative-historical method in
detail. The two references I cited in don't purport to compute the limits
of the comparative method and weren't offered as references on that
question. Both are surveys of genetic groupings in North America (+
elsewhere) that require a controlled time depth and use the deepest
groupings uncontroversially identified by the standard comparative method
as that time depth.

To my knowledge, everyone who has cited an upper limit on the comparative
method has arrived at it in the same way. The oldest uncontroversial
genetic groupings are ones like Indo-European, Uralic, Austronesian, etc.,
all of which are in the ballpark of 6000 years old. Afroasiatic is said to
be ca. 8000 years old, and seems to be generally uncontroversial as a
genetic grouping although there has been debate over whether and to what
extent it admits identification of regular sound correspondences and
reconstruction. Groupings appreciably older than these have been proposed
-- traditional Nostratic, Hokan, etc. -- but their status as genetic
groupings is not uncontroversial (though there seems to be little doubt
about their basic plausibility). The age beyond which the standard
comparative method ceases to yield usable results is the age beyond which
one ceases to get uncontroversial genetic groupings. (An uncontroversial
genetic grouping is one for which comparativists can assemble or have
assembled evidence that unambiguously identifies a unique individual
protolanguage and thereby establishes a particular line of genetic descent.
 Controversial groupings are ones for which that evidence has not been or
cannot be assembled. This doesn't mean that one can't find lexical
resemblances and even correspondences in a controversial grouping; it
merely means that, in seeking them, one is applying to an unproven grouping
the method developed for reconstruction in known families.) Whether
linguists cite 7000, 8000, or 10,000 years as the upper limit on the
comparative method depends on such things as what families they have in
mind, what ages they assume for them, and how round a number they are
aiming for. It's all the same technique.

Glottochronology has about the same limit and for the same reason: gradual
loss of the shared elements that define genetic relatedness.

If the comparative method worked well and reliably up to (say) 12,000
years, then there would be language families which we could easily see were
twice as old as Indo-European and which were as widely accepted by
historical linguists as Indo-European is. There aren't such groupings; at
that time depth all we have is candidates, hypotheses, and debated
groupings. Ergo, the comparative method fades out somewhere after the age
of Indo-European. It fades out gradually: after about 6000 years we lose
(uncontroversial) reconstructibility but can still demonstrate relatedness
(as with Afroasiatic); after about 8000 years we can't demonstrate
relatedness but can still get general consensus on likely candidates (e.g.
some or all of traditional Nostratic as the most likely sisters for

If dating techniques or the received ages of uncontroversial families
change, then estimates of the upper limit of the comparative method's
efficacy will also change, since the estimates are just upward roundings of
the received ages.
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue