reversal of merger

Wed Apr 8 11:47:39 UTC 1998

----------------------------Original message----------------------------
So far I've received few, but interesting, responses to my query about
reversal of merger, and I'm still interested in hearing more.  I'll
post a summary in a few days.

Meanwhile, several people have enquired about the unusual case
discussed by Michelena, and so I've decided it's worth posting this.
It's certainly interesting; it's the only such case I know about; and
it's admittedly buried in the specialist literature.

The change in question is one that occurred in a dialect of Basque not
many centuries ago.  First, here is the principal reference:

Luis Michelena (1957), `Las antiguas consonantes vascas', in
D. Catalan (ed.), Miscelanea Homenaje a Andre Martinet, La Laguna:
Universidad de La Laguna, pp. 113-157. Reprinted in L. Michelena
(1988), Sobre Historia de la Lengua Vasca, J. A. Lakarra (ed.),
Donostia/San Sebastian: Anejos del Anuario del Seminario de Filologia
Vasca "Julio de Urqiojo" 10, vol. I, pp. 166-189.

The relevant passage is on pp. 119-120 in the original, and on p. 169
in the reprint.

The change is also discussed in Michelena's big book:

Luis Michelena (1961) Fonetica Historica Vasca, San Sebastian:
Publicaciones del Seminario de Filologia Vasca "Julio de Urquijo".
Second expanded edition (1977).

However, the account given here is less explicit than that in the
earlier article.

There is an account of the matter in my own book:

R. L. Trask (1997), The History of Basque, London: Routledge,
pp. 156-157.

First, some remarks on Basque orthography.

<z> = English /s/ (laminal)
<s> = Castilian Spanish /s/ (apical)
<tz> and <ts> are the corresponding affricates
<x> = English <sh>
<tx> = English <ch>

<tt>, <n~> and <ll> all represent various palatal consonants (plosive,
nasal, lateral).

Now, Basque has long had a set of palatal and palato-alveolar
consonants, which I will here refer to collectively as the "palatal"
consonants.  But these things formerly occupied, and to some extent
today still occupy, a special place in the phoneme system: they
*never* occur in the basic forms of lexical items, but only in
"expressive" variants of those items.  An expressive variant may be a
simple diminutive ("little X"), a hypocoristic ("X-ie"), or merely a
variant indicating that the speaker regards X as something familiar,
comfortable, intimate or rather sweet.

Here are a few examples:

<gozo> `sweet' > <goxo>
<gizon> `man' > <gixon>
<sardina> `sardine' > <xardina>
<sagu> `mouse' > <xagu>
<zakur> `dog' > <xakur> or <txakur>
<labur> `short' > <llabur> or <txabur>
<nabar> `many-colored' > <n~abar>
<tapa> `step' > <ttapa>
<tipi> ~ <tiki> `small' > <ttipi> ~ <txiki>
<tu> `saliva' > <ttu> or <txu>
<zuri> `white' > <xuri> or <txuri>
<Peru> `Peter' > <Pello>
<Martin> `Martin' > <Txartin> or <Matxin>
<tente> `erect' > <ttentte>
<inurri> `ant' > <txinurri>
<zezen> `bull' > <xexen>
<zoko> `corner' > <xoko> or <txoko>
<Domingo> `Dominic' > <Txomin>
<Santiago> `James' > <Xanti>
<Jose Maria> (male name) > <Joxe Mari> or <Txema>

OK.  Originally, such expressive formations provided the only
occurrences in the language of the palatal consonants (ignoring one
complication which is not relevant here).  But then loan words began
to be taken over with instances of these sounds in the basic forms of
lexical items.  Examples:

<baxera> `dishes' < Gascon <bache`re>
<axola> `care, attention' < Gascon
<Xina>, <Txina> `China' < French <Chine>, Spanish <China>
<xaboi> `soap' < medieval Spanish
<txapela> `beret' < Occitan
<pitxer> `pitcher' < Spanish

The language thus came to have a number of lexical items with palatal
consonants in their basic forms.

Right.  That's all background.  Now to the serious stuff.

Basque has an ancient consonant /j/ (US /y/), a palatal glide.  This
occurs in a large number of native words, such as <jaun> `lord', <jan>
`eat', and <joan> `go'.  It also occurs in a few loan words, such as
<joko> `game', from Latin <iocu(m)>.  Now, in almost all varieties of
the language, this /j/ has undergone some kind of strengthening.  The
result is different in different places, but it is always some kind of
voiced palatal obstruent (apart from the Gipuzkoan case discussed
below).  In the largest part of the country, the result was a voiced
palato-alveolar fricative, like the <zh> in <Zhivago>.  This is still
the pronunciation in some regions.

However, in much of the center of the country, and especially in the
Gipuzkoan dialect, this fricative was devoiced.  As a result, it
merged with the existing voiceless fricative <x>.  This merger still
exists today in some south-central varieties, including bits of
southern Gipuzkoa.  Here, for example, the inherited <jan> `eat' is
pronounced (and traditionally written) <xan>, and so on.

But then, at some later stage, most of Gipuzkoa underwent a further
change: the fricative <x> underwent backing to a back fricative, velar
/x/ or uvular /X/.  This unusual change appears to have been borrowed
from Castilian Spanish, which underwent the same change in the late
16th and early 17th century.  Be that as it may, the change occurred,
but here's the important bit: this backing affected *only* those
instances of <x> derived from original /j/, or present in loan words,
and did *not* affect those instances of <x> found in expressive
formations.  Hence we have an apparent reversal of merger.

So, in modern Gipuzkoan, all those words containing original /j/ are
pronounced with a back fricative.  That includes both native words
like <jaun> `lord' and <jan> `eat' and loan words like <joko> `game'.
The back fricative is also found in loan words like <bajera> `dishes'
and <jaboi> `soap', the last two contrasting with the <baxera> and
<xaboi> of other dialects.

But the original expressive <x> remains, except that it has usually
become the affricate <tx> in word-initial position.  So, for example,
the expressive form of <gozo> `sweet' is still <goxo>, and not
*<gojo>, and the expressive form of <zakur> `dog' is <txakur>, and not
*<jakur>.

Now, we might surmise that the merger never actually took place, that
the devoiced version of /j/ always remained somehow distinct from
original <x>.  But, apart from the phonetic dubiousness of this, there
are two pieces of evidence against it, as noted by Michelena.

First, in the area which has undergone devoicing of /j/ but not
backing of <x>, the merger exists today.  No variety of Basque has two
contrasting versions of <x>.

Second, and crucially, there are a very few instances in which an
original expressive <x> *has* undergone backing.  This occurs, for
example, in the expressive word <gaxo> `poor' (as in `poor fellow'),
also `sick', a word whose original non-expressive form has been lost
(as not infrequently happens).  In Gipuzkoan, the form today is
<gajo>, and not the expected *<gaxo>.  Likewise, `poor fellow' is
<gizarajo> or <gixajo>, in which the /j/ derives from an earlier
expressive <x>.

Michelena therefore draws the following conclusions.

(1) The devoiced /j/ genuinely did merge completely with the inherited
<x> in Gipuzkoan.

(2) However, when the backing of <x> was introduced, speakers were
readily able to distinguish instances of <x> bearing expressive value
from instances of <x> lacking such expressive value.  They therefore
backed *only* the second group, while leaving the first unaffected.
But <gajo> and <gizarajo>, in spite of their etymology, underwent
backing because speakers no longer perceived their <x> as having
expressive value.

So: there was an unconditioned merger, but the merger was later
reversed, because just one of the two original segments undergoing the
merger possessed a distinctive phonological role in the language,
allowing speakers to distinguish the merged segments in all but a
couple of cases.

That's the story.  I like it.

Larry Trask
COGS
University of Sussex
Brighton BN1 9QH
UK

larryt at cogs.susx.ac.uk