LINGUIST List 14.1963

Sat Jul 19 2003

Disc: Are new language classifications necessary?

Editor for this issue: Karen Milligan <>


  1. Ante Aikio, Re: Are new language classifications necessary?

Message 1: Re: Are new language classifications necessary?

Date: Fri, 18 Jul 2003 21:24:38 +0300 (EEST)
From: Ante Aikio <>
Subject: Re: Are new language classifications necessary?

Dear list members:

This is an answer to Yuri Tambovtsev's invitation (Linguist 14.1914)
to other linguists to share their opinions on Angela Marcantionio's
book 'The Uralic Language Family: Facts, Myths, and Statistics'
(Publications of the Philological Society 35; 2002). In the book the
author argues that the assumption of the Uralic language family is not
based on any solid scientific evidence and can thus be shown to be a
'myth'. This reply is sent directly to the list because the issue may
interest other list members, too.

 The work in question involves a number of serious problems, and
because of this my comments will regrettably not be positive. Only
brief remarks can be presented here; more detailed critiques are
available in Linguistica Uralica 1 / 2003 (by Merlijn de Smit) and on
the web at (by
Prof. Johanna Laakso; forthcoming in Finnisch-Ugrische Forschungen). I
have also written a review of this book, which will be published in
Word; several other critiques are also forthcoming.

 In brief, Angela Marcantonio's attempt to undermine the genetic
affinity of the Uralic languages can be characterized as unconvincing,
to say the least. The following fatal weaknesses in her argumentation
can be singled out:

Flaw #1: Insufficient competence in diachronic linguistics.
The author is apparently not fully acquainted with the terminology and
theoretical concepts of comparative linguistics. This is immediately
revealed by the incorrect usage of basic terms, an example of which is
the title of the sixth chapter, "Reconstructing the sound structure
and lexicon of the Uralic family tree" (sic!). Such mistakes do not
seem to result from inadvertence, because the argumentation is also
hampered by more fundamental misunderstandings related to central
theoretical issues.

Some examples:
a) The author constantly mistakes 'sound correspondences' for 'sound
laws' (i.e. 'sound rules'). Furthermore, she does not seem to fully
understand that 'correspondences' and 'laws' normally apply to
individual phonological segments and can be conditioned by
environmental factors. Thus, combinations of two distinct vowel
correspondences applying for different syllabic positions are
incorrectly interpreted as 'vowel rules' (chapter 4 passim).

 b) The central concepts of 'regularity' and 'irregularity' are
not understood. E.g., it is claimed that sound changes which have
affected only certain dialects of a language but not others are "not
regular" because "they are in the process of diffusing"
(p. 111). Presumably, a confusion between 'lexical diffusion' and
'areal diffusion' underlies this strange statement. In another passage
the author discusses a "Neo-grammarian principle" according to which
"borrowed words can be identified because... they are mostly
'irregular' " (sic!). This non-existent principle, then, is supposedly
"contradicted by modern research" (p. 155; no reference is
provided). Apparently, the author is not acquainted with the basic
methods of etymological research.

 c) The author considers it significant that certain sound changes
in Uralic languages are 'shared' with non-Uralic languages, e.g. the
shifts *s- > zero and *p- > f- in Hungarian and Manchu (p. 112). (One
can only wonder why a link between Hungarian and Germanic is not
suggested on the basis of the 'shared' change *p- > f-). It is
self-evident that such parallels can arise through chance or areal
contact, and as regards genetic relationship, they imply nothing.

Flaw #2: Defective knowledge of Uralic data and research in Uralic

 The author's command of issues related to Uralic studies appears to
be no better than her competence in historical linguistics in general. I
will provide three examples.

 a) The presented account of the history of the Uralic studies is
both incomplete and misleading. E.g., the significance of the lexical
and morphological comparison between Hungarian and Saami (Lappish) by
J. Sajnovics (1770) is not properly acknowledged. Moreover, the
distinction between the historical origins and the modern foundations
of Uralic linguistics is not taken into account. Thus, the presented
critique of J. Budenz's studies is of little relevance, because the
shortcomings of these early works are already known to every modern
Uralist. The author also suggests that political and ideological
motives (including e.g. communist sympathies) significantly
contributed to the eventual acceptance of the "Uralic paradigm"; this
appeal to a 'conspiracy theory' adds a pseudoscientific flavor to the

 b) The author stresses that she compares reconstructed forms to
primary lexical data from the attested languages, whereas previous
studies on Uralic historical phonology "have been primarily concerned
with reconstructed forms only" (p. 104), i.e. "reconstructions (rather
than actual attested forms) are usually compared against other
reconstructions" (p. 70). This statement is false: most fundamental
studies on Uralic historical phonology list cognates in the present
day languages, and the studies that explicitly operate only with
reconstructions make reference to other works where the attested forms
are listed. It is difficult to understand how the author has come to
this false belief that Uralic reconstructions are 'hanging in the air'
- does she actually think that linguists just invented the forms out
of their own heads?

 c) After a very superficial discussion on the position of Saami
(Lappish), it is concluded that the relationship of Saami to the rest
of Uralic is uncertain. Thus, the author reveals her amazing ignorance
of the actual linguistic data involved in Uralic comparisons. Saami
happens to share a huge amount of cognate morphology with Finnic,
including case suffixes, possessive suffixes, person, number, mood,
and nominal form markers on verbs, and a wide array of derivative
suffixes. As if this were not enough to prove a genetic relationship,
there are also some 400-500 cognate lexemes displaying regular sound
correspondence, including basic vocabulary such as body parts, kinship
terms, basic objects of the natural world, verbs for elementary
actions, and numerals. Anyone can check this data in basic
handbooks. The existence of these pervasive parallels is, however,
left unmentioned in 'The Uralic Language Family'. I can only see two
possible reasons for this: the author either deliberately ignores
evidence that contradicts her claims, or simply does not know what she
is talking about.

Flaw #3. Invalid application of statistical methods.
 One of the main aims of the book is to undermine the Uralic affinity
with the help of 'statistical methods'. Applying binomial analysis,
the author attempts to prove that the cognate sets presented as
evidence of the Uralic affinity turn out to be 'statistically
insignificant', i.e., that they can (at least for the most part) be
explained as a result of chance. However, the presented tables,
graphs, and statistical calculations are flawed to the extent that
they are completely invalid. I will mention three serious problems in
the method employed:

 a) In chapter 4 the 'statistical insignificance' of Uralic vowel
correspondences is claimed to be demonstrated. However, the
statistical analysis is not in fact based on vowel correspondences at
all, but on the *combined* correspondences of the first and second
syllable vowels - which are incorrectly interpreted as 'vowel rules'
(see Flaw #1 above).

 b) On the basis of this flawed analysis the author gets an excuse
to ignore vowel correspondences altogether in the next statistical
test she performs (chapter 5). Furthermore, she also discards a number
of consonant correspondences on obscure grounds "for the purpose of
the analysis". Thus, a cognate set such as [Finnish /suoni/ 'vein',
Saami /suotna/, Mordvin /san/, Hungarian /�n/] is reduced to a
"one-consonant match" (p. 141), even though it involves as many as
four recurring sound correspondences: 1) s/s/s/zero; 2) uo/uo/a/�; 3)
n/tn/n/n; 4) i/a/zero/zero.

 c) The statistical calculations and the premises involved in
conducting them are often not clearly explained - indeed, the author
herself admits that her analysis involves "a certain amount of
personal intuition and subjectivity" (p. 141). To give an example of
what this 'subjectivity' means, in chapter 5 the author claims to have
tested her results against a "control case which is equivalent to
random words", which is provided nowhere in the book. It is difficult,
occasionally perhaps impossible, to check the author's analysis of the
data - a grave mistake in a scientific study.

Flaw #4. Inconsistency.

 The overall nature of the argumentation is disturbingly
inconsistent. The author sets extremely high standards for evidence
that is presented in support of the Uralic affinity. At the same time,
however, she posits competing hypotheses and makes references to
contradicting claims, no matter how spurious evidence they are based
upon. Random and superficial similarities of word stems and
inflectional endings between Uralic and other language families are
referred to as "flawless", "convincing", "striking", etc. Occasionally
this line of argumentation leads to sheer absurdity: e.g., the old and
duly forgotten comparison between Finnish /Suomi/ 'Finland' and the
toponym /Sumi/ in the Altai region is referred to as a "credible,
straightforward parallel" (p. 253).

Flaw #5. Misleading and erroneous quotes and references.

 The most annoying feature of the book is the frequent occurrence
of misleading references. Often these seem to result from
misunderstanding or careless reading of the work referred to. However,
occasionally this misquotation is so pervasive that it is difficult to
see it as accidental, unless the author was totally blinded by her own
preconceptions. A revealing example is the table which is supposed to
paraphrase Juha Janhunen's comments on various kinds of "identified
problems" involved in twelve Uralic cognate sets (p. 72). A checking
of the original paper (Journal de la Soci�t� Finno-Ougrienne 77; 1981)
reveals that almost all the remarks are either cited out of context or
downright misquoted. I will give two examples: a) A comparison between
two words of identical meaning ('to rain') is said to involve
"semantic difficulties" (sic!); Janhunen says no such thing. b)
Another Proto-Uralic etymology is doubted on the grounds that,
according to Janhunen, "exact semantic reconstruction [is]
impossible". This quote is verbatim, but presented out of context.
Janhunen found no flaw in the comparison, which is based on words
meaning 'sun, day' on the one hand and 'warmth' on the other - a
perfectly transparent semantic match. The idea that a very close
semantic match would not suffice in the absence of an "exact semantic
reconstruction" is absurd, and it is even stranger that Marcantonio
should require such exactness from an Uralic etymology, as the
etymological comparisons she herself advocates are both semantically
and phonologically entirely inexact (see Flaw #4 above).

The errors discussed above are only meant to serve as examples of the
many problems that plague Angela Marcantonio's 'The Uralic Language
Family: Facts, Myths, and Statistics'. Hopefully, they suffice to
demonstrate that the author's claim of the unrelatedness of the Uralic
languages is rejectable. For more thorough critiques, I refer to the
reviews published and forthcoming (see above).

 Marcantonio's work does, however, provoke an important
question. One can only wonder how this book passed a peer review
process and came to be published in a respected series such as
'Publications of the Philological Society'. There are other scientific
disciplines which employ strict referee practices, and this has only
contributed to their progress. Why should it be any different in


Ante Aikio
University of Oulu, Finland
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue