LINGUIST List 6.1144

Tue Aug 22 1995

Sum: V-initial languages

Editor for this issue: T. Daniel Seely <>


  1. "Larry Trask", Sum: V-initial languages

Message 1: Sum: V-initial languages

Date: Tue, 22 Aug 1995 12:38:48 Sum: V-initial languages
From: "Larry Trask" <>
Subject: Sum: V-initial languages

Several weeks ago I posted a query in connection with the exceedingly
frequent vowel-initial lexical items of Basque, in which perhaps 50%
of native nouns and adjectives are V-initial (I exclude verbs, since
ancient verbs almost invariably show a prefix */e-/ in their
non-finite forms.) I asked whether other languages (apart from the
initial-dropping languages of Australia) show a comparable frequency
of initial vowels and, if so, whether a historical explanation is
known. I was particularly interested in hearing about possible cases
of the lenition and loss of initial voiceless plosives, an explanation
proposed for an ancient stage of Basque by Andre' Martinet.

One respondent queried whether there was adequate statistical evidence
to assert that the Basque case is indeed unusual. I know of no such
evidence, and indeed one of my purposes was to try to find out whether
Basque is really as unusual as is commonly believed by vasconists.

The following languages were cited in the responses.

POLYNESIAN: The Proto-Austronesian consonant system has undergone
considerable attrition in the Polynesian languages, which today
exhibit between eight and twelve consonants apiece. In Hawaiian, in
which Proto-Polynesian */h/ and */?/ have been categorically lost,
V-initial words appear to be rather common. But I could not find any
explicit discussion of this, and my impression of such Hawaiian texts
as I could find is that the proportion of V-initial words probably
does not approach 40%, though I am ready to be corrected on this.

NIGER-CONGO: Initial vowels are frequent in a number of West African
languages which are not closely related; the initial vowels generally
appear to be of morphological origin. There is uncertainty as to
whether the vowels are fossilized noun-class prefixes (the most
popular explanation) or the residue of derivational processes, such as
for making agentive nouns out of verbs. In the Cross River languages
of Nigeria, most nouns are V-initial; in some of these languages,
virtually all nouns are. Here it seems clear that the initial Vs are
fossilizations of ancient noun-class prefixes, in most cases derived
from original CV- prefixes by loss of the C. In at least some of
these languages, most verb forms are also V-initial, but only because
the presence of fully functional agreement markers. Benue-Congo
languages generally lack a distinct class of adjectives.

In some Benue-Congo languages, and especially in some Bantu languages,
CV- prefixes have instead been augmented for morphological reasons to
yield VCV- prefixes, again producing frequent initial vowels.

Ancient noun-class prefixes appear to be recoverable to varying
extents in many Niger-Congo languages, especially in Bantu.
Importantly, the original semantic values of the prefixes can often be
determined, though this is not always the case.

The Kwa language (Bini) Edo is unusual in that every single noun must
begin with a vowel, so that even C-initial loans from English have
vowels prefixed, apparently purely for morpheme-structure reasons.

Yoruba is noteworthy in that it has around 45% of V-initial words,
even though four of its ten vowels cannot appear initially.

CATALAN: A rough estimate suggests that Catalan words are about 32%
V-initial, rather more than one might expect in a language with 7
vowels and 22 consonants, and more, I suspect, than are typically
found in Romance languages. I know of no particular reason for this.

ARAWAN (Amazonia): Many of these languages have frequent V-initial
words. It is suggested (but not established) that these derive from
the loss of initial glottals. In at least some of these languages,
V-initial words differ from C-initial words in that the first group
may never bear stress on the first syllable.

BERBER: Masculine-gender nouns regularly take a prefix /a-/, while
feminine nouns take /ta-/.

MUSKOGEAN: These have VC- prefixes on V-initial words but CV- prefixes
on C-initial words. It is not clear to me what this means overall.

SIOUXAN: These seem to have undergone heavy reduction of initials, not
entirely unlike the Australian initial-dropping languages, but the
results have been complex. Siouxan languages particularly show the
lenition of initial /p[h]/ > /f/ > /h/, and possibly also some cases
of /t[h]/ > /h/.

GERMANIC: The change /k[h]/ > /x/ > /h/ is well attested here, but
not, as a rule, exclusively in word-initial position (though sometimes
in syllable-initial position). Lenition of voiceless plosives
generally is widespread in Germanic.

On the basis of this admittedly unscientific sample, I would therefore
suggest the following conclusions:

(1) Languages in which 40% or more of nouns are V-initial are not
exceedingly rare, but they do not appear to be at all common. The
majority of the examples come from Niger-Congo.

(2) In Niger-Congo, the initial vowels derive chiefly from
morphological sources, probably entirely so apart from the reduction
of CV- prefixes to V- in some languages. Fossilized noun-class
prefixes appear to be the most widely accepted origin in most cases,
and such prefixes appear to be certain in some cases, in which the
semantic value of the prefixes is still recoverable. But other
morphological processes may be responsible in some other cases.

(3) Outside of Niger-Congo, there appears to be little evidence of
morphological origins for initial vowels, and loss of initial
consonants is more usually invoked. The consonants invoked are,
unsurprisingly, more likely to be glottals than oral obstruents.

(4) Except perhaps in Siouxan, there is little or no evidence for the
systematic loss of initial voiceless plosives.

So what happened in Basque? I still don't know. The American
long-ranger John Bengtson has for years been defending the "fossilized
noun-class prefix" view of the Basque initial vowels, for reasons of
his own, but there seems to be no trace of any semantic correlations
with the initial vowels, and in fact the frequency of each of the five
Basque vowels appears to be about the same initially as elsewhere:
a > e > i > o > u.

My money is therefore still on the systematic loss of certain initial
consonants, but I'll be very surprised if I ever have to pay out or
get to collect.

My thanks to Maile Rehbock, Bruce Connell, Max Wheeler, Dan Everett,
John Koontz, David Stampe, Mark Liberman, Herbert Stahlke, and Andrew
Carstairs-McCarthy for their helpful responses.

Larry Trask
University of Sussex
Brighton BN1 9QH
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue